University of Southern California Dissertations and Theses

Understanding normative influence of neighborhoods: a multilevel approach to promoting Latinas’ cervical cancer prevention behaviors in urban ethnic communities

(USC Thesis Other)

Understanding normative influence of neighborhoods: a multilevel approach to promoting Latinas’ cervical cancer prevention behaviors in urban ethnic communities

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

UNDERSTANDING NORMATIVE INFLUENCE OF NEIGHBORHOODS:
A MULTILEVEL APPROACH TO PROMOTING LATINAS’ CERVICAL CANCER
PREVENTION BEHAVIORS IN URBAN ETHNIC COMMUNITIES

by

Nan Zhao

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Communication)

December 2015

Copyright 2015 Nan Zhao
ii

DEDICATION
To my uncle, Hong Chen (1954-2010).

iii

ACKNOWLEDGMENTS
This work was supported by the National Cancer Institute (NCI) for Barriers to Cervical
Cancer Prevention in Hispanic Women: A Multilevel Approach, an award to the University of
Southern California (R01CA155326 - Murphy/Ball-Rokeach). The content is solely the
responsibility of the authors and does not represent official views of the NCI or of the NIH.
It is hard to imagine completing this dissertation without the unbounded guidance and
support from my advisor and mentor Sheila Murphy. Sheila helped me consolidate ideas, find
resources, develop materials, and turn a few pages of research outline into a dissertation. I
learned from her that a good research paper tells a story that flows, and that less is more. The
perseverance and thoroughness that I have learned from her, both as a researcher and a person,
will continue to be an invaluable asset of my life. My special appreciation also goes to Sandra
Ball-Rokeach. From her I learned that, as a researcher, I should always love the questions I ask
rather than the answers I land with. This way, my research will always be “interesting and
personally relevant”. Working with her in the past 5 years has tremendously shaped the
trajectory of my research, and my way of seeing the dynamic relationship between people,
neighborhoods, and cities.
I would also like to thank Chih-Ping Chou for his valuable feedback for my proposal in
the early stage of the project and along the way. I keep to this day the photo of the equations he
drew on the whiteboard during our meeting, where he helped me flush out my analytic approach.
I am indefinitely grateful to Lourdes Baezconde-Garbanati, for her advice and heartfelt
encouragement she has given me over the past few years as a teacher, a mentor, and a friend.
The champion cyclist stays focused, she told me, which I will always remember.
iv

For many, there is a period of time in life where, when our family is away on another
continent, our friends are our family. Chenhong (Connie) He, for continuing to be the calming
force in my life, and for fiercely believing in me when I questioned myself. Ritesh Mehta, for
being a thoughtful friend on a very similar wavelength to “w ander” with. Li Lu and Jin Huang,
for being my endless cheerleaders and encouragers, both at Annenberg and at the finishing line
of my first half-marathon. Styles Akira, for being “Charlie”. Neta Kligler -Vilenchik, Jingfang
Liu, and Lin Zhang, for their critical minds. Prawit Thainiyom, for being a fellow “outlier” to
experiment a different path with. Joyee Chatterjee, for all the nutritious conversations we had in
the car while carpooling. Angeline (LeeAnn) Sangalang, for being my “office-buddy”. Carmen
Gonzalez, Meghan Moran, Lauren Frank, and Nancy Chen, for making my last two years on
research projects a particularly rewarding learning experience.
Words cannot fully express my gratitude towards my parents for their unbending support
and aching sacrifice, only to allow me to incrementally find my path. I can write pages after
pages about their kind souls and tireless optimism, and about how they planted in me since I had
memory, by being incomparable role models, that little kindness goes a long way. However,
what I most want to say at this moment is that, to borrow my friend Ritesh’s remarks to his
parents, thank you for letting me go, and thank you for letting me go far.
Last, but certainly not the least, I am immensely grateful to my fiancé Demetrius Martin
for his unconditional love, understanding, and encouragement that kept me sane during my last
weeks of finishing this dissertation. The moments see us.

v

TABLE OF CONTENTS

DEDICATION………………………………………………………………………………….…ii
ACKNOWLEDGEMENTS…………………………………………………………………..…..iii
LIST OF TABLES……………………………………………………………………………...viii
LIST OF FIGURES……………………………………………………………………………….x
ABSTRACT……………………………………………………………………………………...xi
CHAPTER 1 : INTRODUCTION ................................................................................................................ 1
Understanding Normative Influence in the Context of Residential Life .................................................. 1
A Multilevel Communication Infrastructure Model of Normative Perceptions ....................................... 6
The Research Context ............................................................................................................................... 8
Purpose of the Study ............................................................................................................................... 11
Organization of the Dissertation ............................................................................................................. 12
CHAPTER 2 : UNDERSTANDING THE SOURCES OF NORMATIVE INFLUENCE ON HEALTH
BEHAVIORS .............................................................................................................................................. 14
Types of Social Norms ............................................................................................................................ 16
Relationships of Perceived Norms and Health Behaviors ...................................................................... 18
Social and Behavioral Theories Exploring Normative Influence ........................................................... 20
Social Norms Theory .......................................................................................................................... 20
Theory of Normative Social Behavior ................................................................................................ 21
Integrative Model of Behavioral Prediction ........................................................................................ 22
Communication and Sources of Normative Influence ............................................................................ 25
Symbolic Environment ....................................................................................................................... 27
Social Environment ............................................................................................................................. 31
Physical Environment ......................................................................................................................... 34
Conclusion .............................................................................................................................................. 36
CHAPTER 3 : A MULTILEVEL COMMUNICATION INFRASTRUCTURE MODEL OF
NORMATIVE PERCEPTIONS ................................................................................................................. 39
Communication Research Examining Health-Related Outcomes in Ecological Context ....................... 41
Knowledge Gap Hypothesis................................................................................................................ 42
Structural Influence Model (SIM) ....................................................................................................... 43
Communication Infrastructure Theory .................................................................................................... 45
Neighborhood Storytelling Resources and Health Outcomes ............................................................. 50
vi

Communication Action Context and Health Outcomes ...................................................................... 54
A Multilevel Communication Infrastructural Model of Normative Perceptions .................................... 57
Hypotheses and Research Questions Group 1 .................................................................................... 61
Hypotheses and Research Questions Group 2 .................................................................................... 70
CHAPTER 4 : METHODOLOGY ............................................................................................................. 78
Data Collection ....................................................................................................................................... 78
Individual-Level Data Collection ........................................................................................................ 78
Neighborhood-level data collection .................................................................................................... 79
Measures ................................................................................................................................................. 80
Individual-Level Variables ................................................................................................................. 80
Neighborhood-Level Variables ........................................................................................................... 84
Analysis .................................................................................................................................................. 88
Hierarchical Linear Modeling (HLM) ................................................................................................ 88
Structural Equation Modeling (SEM) ................................................................................................. 95
CHAPTER 5 : RESULTS ........................................................................................................................... 98
Descriptive Results ................................................................................................................................. 98
Descriptive Results for Individual-Level Variables ............................................................................ 98
Descriptive Results for Neighborhood-Level Variables ................................................................... 104
Multilevel Analysis of Descriptive Norms Regarding Pap Tests ......................................................... 110
Effects of Neighborhood Storytelling Resources on Descriptive Norms ......................................... 112
Contextual Effect on Descriptive Norms .......................................................................................... 116
Multilevel Analysis of Media Recall Regarding Pap Tests .................................................................. 118
Effects of Neighborhood Storytelling Resources on Media Recall .................................................. 119
Contextual Effect on Media Recall Regarding Pap Tests ................................................................. 123
Multilevel Analysis of Connections to Storytelling Resources ............................................................ 125
Contextual Effect on Connections to Storytelling Resources ........................................................... 126
Summary of Multilevel Modeling Analysis Finding ............................................................................ 134
Structural Equation Modeling Analysis of Descriptive Norms ............................................................ 136
Results for Participants Having Lived in Their Neighborhood for 6 Years or More ........................ 137
Results for Participants Having Lived in Their Neighborhood for Less Than 6 Years .................... 144
Post-Hoc Analysis ............................................................................................................................. 150
Difference in Structural Relationships by Years of Residence in Neighborhoods ........................... 152

vii

CHAPTER 6 : DISCUSSION AND CONCLUSION .............................................................................. 155
Key Study Findings ............................................................................................................................... 156
Descriptive Norms and Screening Compliance ................................................................................ 156
Neighborhood Storytelling Resources, Neighborhood Context, and Descriptive Norms ................. 157
Neighborhood Storytelling Resources, Neighborhood Context, and Media Recall.......................... 163
Storytelling Resources and Neighborhood Context .......................................................................... 168
Structural Relationships between Neighborhood Experience, Storytelling Resources, Health
Communication Outcomes, Descriptive Norms and Compliance with Screening Guidelines ......... 170
Contributions of the Current Study ....................................................................................................... 179
Limitations and Suggestions for Future Research ................................................................................ 188
Conclusion ............................................................................................................................................ 193
REFERENCES ......................................................................................................................................... 194
APPENDIX A: DATA COLLECTION OF THE MULTILEVEL PROJECT ......................................... 227
APPENDIX B: PROCEDURE TO DEFINE NEIGHBORHOOD CLUSTERS ...................................... 229
APPENDIX C: 25 NEIGHBORHOOD CLUSTERS DEFINED BY THE MULTILEVEL PROJECT .. 235
APPENDIX D: SPATIAL DISTRIBUTION OF POPULATION CHARACTERISTICS OF THE
STUDY AREA ......................................................................................................................................... 236
APPENDIX E: COMMUNICATION RESOURCES IN THE STUDY AREA ..................................... 246
APPENDIX F: UNREPORTED STATISTICAL TESTS ........................................................................ 247

viii

LIST OF TABLES

Table 5.1 Descriptive Statistics of Socioeconomic Characteristics: Frequencies ........................ 99

Table 5.2 Descriptive Statistics of Health and Healthcare Variables: Frequencies .................... 100

Table 5.3 Descriptive Statistics of English Language Proficiency, Storytelling Resources, Media
Recall, Media Attention, and Descriptive Norms: Means and Standard Deviations .................. 101

Table 5.4 Zero-Order Correlations among Individual-Level Variables Included in Analysis ... 103

Table 5.5 Selection of Neighborhood-level Characteristics for 24 Neighborhood Clusters ...... 105

Table 5.6 Neighborhood-Level Variables Included in Analysis.……………………………… 109

Table 5.7 Zero-Order Correlations among Neighborhood-Level Variables ............................... 110

Table 5.8 Hierarchical Linear Models of Descriptive Norms Regarding Pap Tests .................. 111

Table 5.9 Hierarchical Linear Models of Descriptive Norms Regarding Pap Tests for Non-
compliant Participants ................................................................................................................. 113

Table 5.10 Hierarchical Linear Models of Descriptive Norms Regarding Pap Tests for
Noncompliant Participants: Individual-Level Covariates, ICSN, and Neighborhood-Level
Linguistic Isolation ..................................................................................................................... 117

Table 5.11 Hierarchical Linear Models of Media Recall Regarding Pap Tests ......................... 119

Table 5.12 Hierarchical Linear Models of Media Recall Regarding Pap Tests for Noncompliant
Participants .................................................................................................................................. 121

Table 5.13 Hierarchical Linear Model of Media Recall Regarding Pap Tests for Noncompliant
Participants: Individual-Level Covariates, Neighborhood Storytelling Resources, and
Neighborhood-Level Density of Health Service Providers ........................................................ 124

Table 5.14 Hierarchical Linear Models of Connections to Storytelling Resources ................... 126

Table 5.15 Hierarchical Linear Models of Connections to Neighborhood Storytelling Resources,
English Language Media, and ICSN: Individual-Level Covariates and Neighborhood-Level
Linguistic Isolation ..................................................................................................................... 128

Table 5.16 Hierarchical Linear Models of Connections to Neighborhood Storytelling Resources,
English Language Media, and ICSN: Individual-Level Covariates and Neighborhood-Level
Ethnic Heterogeneity .................................................................................................................. 130

ix

Table 5.17 Hierarchical Linear Models of Connections to Neighborhood Storytelling Resources,
English Language Media, and ICSN: Individual-Level Covariates and Neighborhood-Level
Density of Communication Resources........................................................................................ 132

Table 5.18 Hierarchical Linear Models of Connections to Neighborhood Storytelling Resources,
English Language Media, and ICSN: Individual-Level Covariates and Neighborhood-Level
Density of Health Service Providers ........................................................................................... 133

Table 5.19 Summary of Findings of Hierarchical Linear Models: Descriptive Norms and Media
Recall Regarding Pap tests.......................................................................................................... 135

Table 5.20 Summary of Findings of Hierarchical Linear Models: Neighborhood Storytelling
Resources, English Language Media, and ICSN ........................................................................ 136

Table 5.21 Direct, Indirect, and Total Effects on Descriptive Norms Regarding Pap Tests for
Participants Who Had Lived in Their Neighborhoods for Six Years and More ......................... 142

Table 5.22 Direct, Indirect, and Total Effects on Screening Compliance for Participants Who
Had Lived in Their Neighborhoods for Six Years and More ..................................................... 143

Table 5.23 Direct, Indirect, and Total Effects on Descriptive Norms for Participants Who Had
Lived in Their Neighborhoods for Less Than Six Years ............................................................ 147

Table 5.24 Direct, Indirect, and Total Effects on Screening Compliance for Participants Who
Had Lived in Their Neighborhoods for Less Than Six Years .................................................... 148

Table 5.25 Logistic Regression Models Predicting Compliance with Cervical Screening
Guidelines from Immigration Generation, Age, Health care coverage, Storytelling Resources
Connections, Health Communication Outcomes and Descriptive Norms Regarding Pap Tests for
Longer and Shorter Tenure Groups ............................................................................................ 151

Table 5.26 Summary of Findings of Structural Equation Modeling Analysis by Participants’
Residential Tenure Based on the Revised Models ...................................................................... 153

x

LIST OF FIGURES

Figure 2.1 Integrative Model of Behavioral Prediction.. .............................................................. 24

Figure 3.1 Structural influence model (SIM) of communication.. ............................................... 44

Figure 3.2 Communication infrastructure. .................................................................................... 46

Figure 3.3 Age adjusted cervical cancer incidence rates by race and ethnicity 2000-2012. ........ 59

Figure 3.4 Conceptual model of the structural relations between neighborhood experience,
storytelling resources, health communication outcomes, descriptive norms, and compliance with
cervical cancer screening guidelines ............................................................................................. 72

Figure 5.1 Hypothesized model for participants who had lived in their neighborhoods for six
years or more............................................................................................................................... 138

Figure 5.2 Revised model for participants who had lived in their neighborhoods for six years or
more.. .......................................................................................................................................... 141

Figure 5.3 Hypothesized model for participants who had lived in their neighborhoods for less
than six years............................................................................................................................... 145

Figure 5.4 Revised model for participants who had lived in their neighborhoods for less than six
years. ........................................................................................................................................... 146

xi

ABSTRACT

This dissertation explores the importance of urban ethnic neighborhoods as the context of
everyday life wherein normative influences on health are formed, modified, and maintained.
From the perspective of communication infrastructure theory (Kim & Ball-Rokeach, 2006a,
2006b), this study proposes a multilevel communication infrastructure model and investigates the
development of health-related normative perceptions as a function of people’s connections to
their neighborhood storytelling network consisting of residents, local/ethnic media, and
community organizations, as well as the communication action context. Specifically, people’s
integrated connections to a neighborhood storytelling network (ICSN), where individual
storytellers encourage each other to tell stories about the neighborhood, is considered as a critical
source of health information that could influence people’s perceived norms regarding a particular
health behavior. Moreover, the effect of having integrated connections to a neighborhood
storytelling network is examined not in isolation, but within the context of a specific
neighborhood environment that could strengthen or constrain the function of the neighborhood
storytelling network.
The health domain of interest is cervical cancer screening and detection. Participants for
this study came from the Multilevel Study (R01CA155326 - Murphy/Ball-Rokeach), a large
multilevel and multi-method research project that systematically examines the barriers and
conduits to cervical cancer prevention, detection, and treatment at the individual, interpersonal,
and community level for Hispanic women in Los Angeles. The specific perceived norms
explored in this study are descriptive norms, defined as people’s perceived prevalence of a given
behavior among others “like them”. The three types of descriptive norms examined in the
current study are: 1) perceived prevalence of women never having a Pap test, 2) perceived
xii

prevalence of women having routine Pap tests, and 3) perceived prevalence of women ever
receiving an abnormal Pap test result.
Using hierarchical linear modeling and structural equation modeling, this study examines
two groups of hypotheses and research questions. The first involves the relationship between the
neighborhood storytelling network, neighborhood context, and Latinas’ descriptive norms
regarding Pap tests. The second explores the structural relationship between Latinas’
neighborhood experience, connections to storytelling resources, health communication
outcomes, descriptive norms, and compliance with cervical cancer screening guidelines.
Overall, the present study finds strong empirical evidence indicating the importance of
local communication resources in shaping Latinas’ descriptive norms regarding cervical cancer
screening and detection. Moreover, the communication processes underlying the development of
descriptive norms vary meaningfully across participants in a way that reflects the dynamics
between individuals and their neighborhood environment. This dissertation concludes with the
theoretical, methodological and practical contributions of these findings to the study of
communication, social norms, cancer prevention, and the interdisciplinary field of
neighborhoods and health.
1

CHAPTER 1 : INTRODUCTION

Understanding Normative Influence in the Context of Residential Life
The insight that where we live makes a difference is not novel. As real estate agents
often assert, only three things matter when it comes to buying a house: location, location,
location (Jeffres, 2002). Though globalization and technological advance have fueled the
excitement (and lament) that places no longer matter and that neighborhoods will ultimately
become irrelevant (Giddens, 1991, 2002; Thompson, 1995), a rapidly expanding line of
research across disciplines suggests that the opposite may be true: neighborhoods and local
communities interact with many aspects of our lives, including health (Ball-Rokeach, Kim, &
Matei, 2001; Kawachi & Berkman, 2003; Kim & Ball-Rokeach, 2006b; Kim, Moran, Wilkin, &
Ball-Rokeach, 2011; MacDonald & Sampson, 2012; Sampson, 2012; Sampson, Morenoff, &
Gannon-Rowley, 2002). As addressed in the World Health Organization’s (WHO) Commission
on the Social Determinants of Health, “where people live affects their health and c hances of
leading flourishing lives. Communities and neighborhoods that ensure access and psychological
wellbeing, and that are protective of the natural environment are essential for health equity”
(CSDH, 2008, p. 60).
As is appropriate for such an important topic, for the past two decades, the search for
impacts of neighborhoods on health has been part of the central research agenda in the United
States (Kawachi & Berkman, 2003; Sampson, 2012). The increasing trend of residential
segregation, whether by race and ethnicity or by class, was cited as one of the major motivating
forces driving this multidisciplinary effort (Kawachi & Berkman, 2003). In his 1996 presidential
address to the Population Association of America, Douglas Massey noted that, “urban ization,
2

rising income inequality, and increasing class segregation have produced a geographic
concentration of affluence and poverty throughout the world, creating a radical change in the
geographic basis of human society” (Massey, 1996, p. 395). This challenge is further
complicated by the population diversification that has continued over the past 10 years, largely
due to the growth of ethnic minority populations across the nation
1
. Accompanying this
demographic shift is the proliferation of ethnically and culturally diverse residential communities
nationwide, particularly in immigrant gateway cities such as Los Angeles and New York.
Taking working-class Latino immigrants in Los Angeles alone as an example, the city’s
conglomeration of municipalities, iconic transportation system (e.g., freeways and automobiles),
and highly suburbanized landscape have contributed to their residential pattern that is
substantially different from the concentric configuration in urban city articulated by the Chicago
School. Today, Latino immigrants in Los Angeles are “no longer as place -bound as in previous
generations” (Straughan & Hondagneu-Sotelo, 2001, p. 188). Rather, they settle across the city
instead of only in East Los Angeles. The resultant change in the social fabric in everyday life 
as well as the dynamics between the reshaped social fabric and physical environment  opens a
new, important line of inquiry into how neighborhoods influence health.
In this burgeoning body of work focusing on the impact of neighborhoods,
communication as a discipline is under-represented compared to public health, sociology,
anthropology, geography, economics, and education (Kawachi & Berkman, 2003; Matsaganis,
2015). More significantly, the role of communication as a social process in influencing urban
residents’ health is not adequately studied (Matsaganis, 2015). This omission continues despite
empirical evidence that communication processes at all levels (e.g., individual, group,

1
Using the U.S. Census Bureau’s definition, this dissertation defines minority as any racial and ethnic group other
than non-Hispanic White group alone (Colby & Ortman, 2015).
3

community, media system, culture) play a critical part in shaping people’s knowledge, beliefs,
perceptions, decisions, and behaviors related to health (Frank et al., 2012; Hornik, 2002;
Murphy, Frank, Chatterjee, & Baezconde-Garbanati, 2013; Niederdeppe, Bigman, Gonzales, &
Gollust, 2013; Wilkin, 2013).
One promising line of communication research that may shed light on how
neighborhoods influence health is the study of social norms. Widely recognized as a crucial
social process through which environment impacts human behaviors, social norms comprise the
“rules and standards that are understood by members of a group, and that guide and/or constrain
social behavior without the force of laws” (Cialdini & Trost, 1998, p. 152). Across population
and cultural settings, empirical evidence abounds that our perceptions regarding the prevalence
and acceptability of a given behavior can have a considerable impact on our thoughts and
subsequent actions (Mollen, Rimal, & Lapinski, 2010; Yanovitzky & Rimal, 2006).
In the health arena, perceived social norms have been associated with a broad spectrum
of behaviors, such as alcohol consumption (Gibbons et al., 2010; Ho, Poorisat, Neo, & Detenber,
2013; Rimal, 2008), condom use (Frank, et al., 2012; Rhodes, Stein, Fishbein, Goldstein, &
Rotheram-Borus, 2007), smoking (Echeverría, Gundersen, Manderski, & Delnevo, 2015; Phua,
2012), and cancer prevention (Pasick et al., 2009; Zikmund-Fisher, Windschitl, Exe, & Ubel,
2011). Stemming from this literature, and based on the premise that correcting the misperception
of norms will lead to positive behavioral changes, many health communication campaigns and
programs have incorporated social norms as an integral component (Campo et al., 2003;
Wakefield, Loken, & Hornik, 2010).
Compared to the norm-behavior relationship, sources of normative influence on
behaviors are not well documented (Mead, Rimal, Ferrence, & Cohen, 2014). In health-related
4

literature, previous work exploring this area has extensively focused on interpersonal interaction
(David, Cappella, & Fishbein, 2006; Dunlop, Kashima, & Wakefield, 2010; Frank, et al., 2012)
and mass media (Bleakley, Hennessy, Fishbein, & Jordan, 2011; Gibbons, et al., 2010;
Wakefield, et al., 2010) as sources of normative influence, largely ignoring other significant
sources such as physical environment (Novak, Reardon, Raudenbush, & Buka, 2006; West et al.,
2010). The link between people’s multiple exposures to various sources in their environment
and the development of perceived norms is not sufficiently understood either (Mead, et al.,
2014). Moreover, much of the knowledge regarding sources of normative influence on health is
obtained in the context of health communication campaigns and programs, rather than the
context of everyday life, leading to a lack of generalizability (Seo & Matsaganis, 2013).
This inclusion of the normative influence occurring in daily life on our health beliefs and
behaviors is critical. In the process of everyday life, people come into contact with normative
information through numerous venues. These venues run the gamut from individuals (e.g.,
family, friends, neighbors, or health professionals), media (e.g., “old media” such as television,
radio, or printed-media; and “new” media such as emails, websites, interactive games or videos,
and social media), to local institutional and organizational resources (e.g., church, school,
library, parks, or community organizations) (Kim, et al., 2011; Mead, et al., 2014). Of these,
each source might contribute  either alone or in conjunction with others  to the development
of perceived norms regarding a particular behavior. For example, a woman may learn about Pap
tests from friends and family, local television and newspapers, health fairs organized by local
organizations, and so forth. All these venues may act as sources of normative information on
cervical cancer prevention and detection.
5

It is also reasonable to posit that the roles of different sources in sculpturing perceived
norms may or may not be congruent, depending on the extent to which they are consistent in
terms of the messages that they contain. Furthermore, the centrality of a given information
source to individuals’ lives in general, and to the shaping of normative perceptions in particular,
is hardly a given. Rather, to a great extent, the centrality of a particular information source is the
result of the interaction between individuals and their environment (Wilkin, 2013). From an
ecological perspective, therefore, understanding the influence of information sources on the
development of perceived social norms requires knowledge about both the individuals and their
environment. Ultimately, understanding the formation of perceived norms should be approached
as a multilevel phenomenon both theoretically and analytically (Stokols, 1996).
This dissertation explores the importance of urban residential neighborhood as a crucial
context wherein perceived social norms about health behaviors are formed, shaped, and
maintained in everyday life. The focus on residential neighborhood was made because of two
reasons. First, everyday life is the milieu in which people routinely receive, exchange, and act
upon a myriad of information conveying behavioral norms. Second, residential places are
considered as the environment in which such daily communication activities emerge, unfold, and
evolve over time and space.
Of particular interest are the communication processes that link the information sources
to which people connect for everyday problem-solving with their perceived social norms
regarding health and health behaviors. Central to this research, as described below, is a
neighborhood communication infrastructure approach that articulates the dynamics between
people’s daily communication behaviors and their residential environment (Ball-Rokeach, et al.,
2001; Kim & Ball-Rokeach, 2006a).
6

As urban life continues to diversify on the fronts of ethnic composition, cultural fabrics,
and geographic dimensions (Ball-Rokeach, et al., 2001; Institute of Medicine, 2002; Kim & Ball-
Rokeach, 2006a, 2006b; MacDonald & Sampson, 2012), this dissertation will contribute to social
norm-based health communication research and interventions theoretically, methodologically,
and practically.
A Multilevel Communication Infrastructure Model of Normative Perceptions
This dissertation discusses and tests a communication infrastructure model of normative
perceptions about health behaviors among urban residents. Communication infrastructure is “the
basic communication system of a community” (Wilkin, 2013, p. 2). At the local level,
communication infrastructure consists of two components: the neighborhood storytelling
network and the communication action context (Ball-Rokeach, et al., 2001; Kim & Ball-
Rokeach, 2006a, 2006b). A neighborhood storytelling network is comprised of key local
communication sources, including residents, local/ethnic media, and community organizations.
The communication action context, on the other hand, describes the community-level
characteristics (e.g., psychological, sociocultural, physical, economic, and technological
features) that facilitate or constrain neighborhood storytelling. A viable neighborhood
storytelling network is one that is grounded in a conductive environment, and that creates and
disseminates “stories” about the features of residents’ daily lives (Ball-Rokeach, et al., 2001;
Kim, 2003). The communication infrastructure theory (CIT) model thus represents a socio-
ecological approach to understanding the interplay between community-level characteristics and
residents’ day-to-day life in neighborhood contexts (Kim, 2003; Wilkin, 2013).
At the heart of the CIT approach is the assumption that people’s capacities to solve
everyday problems (e.g., civic engagement, disaster preparedness, and disease prevention) are, at
7

least partially, contingent upon their connections to neighborhood storytelling resources
embedded in the communication action context (Ball-Rokeach, et al., 2001; Kim, et al., 2011).
CIT-based research across diverse urban neighborhoods has unveiled rich variation in both
communication structure and residents’ problem -solving outcomes (Broad, Gonzalez, & Ball-
Rokeach, 2013; N.-T. N. Chen et al., 2013; Kim, et al., 2011; Matsaganis & Wilkin, 2014).
Consistent with CIT’s theoretical predictions, communication infrastructure is f ound to be strong
and integrated in some neighborhoods, but weak or fragmented in others (Ball-Rokeach, et al.,
2001; N.-T. N. Chen, et al., 2013). Specifically, as captured in the concept of “integrated”
connection to neighborhood storytellers, individuals differ in their structural positions in the
everyday communication environment. That is, some individuals have strong connections to all
three aforementioned storytelling agents (i.e., residents, local/ethnic media, community
organizations), whereas others might have a strong connection to one type of storyteller but loose
or nonexistent connections to other storytellers (Kim & Ball-Rokeach, 2006b).
It is theorized that having an integrated connection to local storytelling resources affords
enhanced opportunities for communication, as well as for identifying and solving everyday
problems, including health-related problems (Kim, et al., 2011). For example, a story in local
media that encourages Latinas to get a Pap test might instigate discussions among residents and
bring them in contact with local clinics for the service. Hence, a stronger synergistic effect
occurs than what might have achieved with media reporting alone.
The interplay between the neighborhood storytelling network and the communication
action context receives empirical support too, though mostly in civic engagement research. Kim
and Ball-Rokeach (2006), for instance, observed that residents’ reliance on their local
communication resources for information and resources to become engaged in their
8

neighborhoods or to build a healthy life is heavier in disadvantaged neighborhoods than in
advantaged neighborhoods (Kim & Ball-Rokeach, 2006b).
This dissertation extends the application of CIT to the inquiry into normative influences
on health. Neighborhood storytelling resources grounded in their communication action context
are conceived of as the information sources that people connect to for problem-solving and goal-
attainment in the context of everyday life. Specifically, having an integrated connection to the
storytelling network (ICSN) is explored as a facilitating factor in shaping perceived social norms
regarding health. It is posited that connecting to a strong and integrated storytelling network —
where all storytellers encourage each other to “storytell” the neighborhood issues — makes
residents more aware of a particular issue (e.g., behavioral norms regarding Pap tests) or primes
them to seek additional information which, in turn, contributes to the formation of normative
perceptions. Moreover, this research approaches the development of normative perceptions as a
process that unfolds in the neighborhood environment, here conceptualized as the
communication action context. The effect of the communication action context is examined by
looking at whether and how normative perceptions are shaped by individual- and neighborhood-
level characteristics and the interplay between these levels.
The Research Context
This dissertation draws data from a National Cancer Institute (NCI) funded grant,
“Barriers to Cervical Cancer Prevention in Hispanic Women: A Multilevel Approach” project
(R01CA155326 - Murphy/Ball-Rokeach). Cervical cancer is the third most common type of
cancer of women worldwide, responsible for more than 266,000 deaths in 2012 alone (World
Health Organization, 2014). However, cervical cancer is highly treatable if precancerous lesions
are found through screening (Papanicolaou test, abbreviated as Pap test) and diagnostic tests and
9

removed before they become cancerous (American Cancer Society, 2012). Over the past 70
years, the incidence and mortality rates of cervical cancer have decreased significantly
worldwide and in the United States, largely due to women’s regular use of Pap tests (Centers for
Disease Control and Prevention, 2014; Luciani & Andrus, 2008). Despite the recent advances in
prevention (e.g., Human Papillomavirus vaccination) and detection (e.g., DNA screening), Pap
tests continue to be the most widely available screening tests. However, both globally and
nationally, cervical cancer incidence and mortality rates remain high in certain populations and
geographic areas where access to Pap tests is limited (National Institute of Health, 2013) .
In the United States, women of color endure an unequal burden of cervical cancer, with
Hispanic women continuing to have much higher incidence rates compared to non-Hispanic
White women (10.9 vs. 7.7 per 100,000) (US Cancer Statistics Working Group, 2014). This
disparity is even more striking in Los Angeles County, the location of the current study, where
the incidence of cervical cancer among Hispanic women in 2010 was as high as 14.3 per
100,000, compared to 7.5 per 100,000 among non-Hispanic White women (Los Angeles County
Department of Public Health Office of Women’s Health, 2010) .
Several factors that contribute to cervical cancer disparities among Hispanic women
include underestimation of cervical cancer risk, lack of early and regular screening through Pap
tests, inadequate normative pressure, and cultural beliefs that perpetuate non-compliance to
screenings, follow-up, and treatment (Akers, Newmann, & Smith, 2007; Austin, Ahmad,
McNally, & Stewart, 2002; Byrd, Peterson, Chavez, & Heckert, 2004; Mann, Foley, Tanner,
Sun, & Rhodes, 2014; Murillo et al., 2008). Strategies to decrease such health disparities
represent a critical public health challenge. The magnitude of this challenge is even greater
when considering the exponential increase of Hispanic population in the United States in the past
10

decade and future projections. As projected by the U.S. Census Bureau, more than one quarter
of the U.S. population will be Hispanic or Latino origin by 2060
2
(Colby & Ortman, 2015).
The Multilevel Project is a multi-method, in-depth examination of the barriers and
conduits related to cervical cancer prevention, detection, and treatment for Latinas and Hispanic
women in Los Angeles. There has been a growing awareness in health disparities and cancer
literature that factors playing a part in cancer disparities exist and operate at many different
levels (Holmes et al., 2008). In other words, although disproportionate occurrence of incidence
and death manifests at the level of individuals, the factors that affect women’s capacities and
opportunities to access information, acquire a Pap test, and adhere to treatment exist at the
individual, interpersonal, and community level. Solutions to such complex problems therefore
require an understanding of both individuals and their contexts, such as social and cultural
milieus, neighborhoods, social organizations, and healthcare facilities (Holmes, et al., 2008;
Viswanath & Emmons, 2009).
The Multilevel Project examines the barriers and conduits to cervical cancer prevention,
detection, and treatment at the individual, interpersonal, and community level, as well as
identifies the interactions among these levels. The study population, therefore, consisted of
women who were following the cervical cancer screening guidelines and had received a Pap test
in the past 3 years (hereafter referred to as “compliant” group), and women who were no t
following the screening guideline either because they had never had a Pap test or they had not
received one in over 3 years (hereafter referred to as “noncompliant” group).

2
Hispanics or Latinos refer to people self-classified themselves in one of the specific Spanish, Hispanic, or Latino
categories listed on the Census 2010 questionnaire, and those who indicated that they are "another Hispanic, Latino,
or Spanish origin” (U.S. Census Bureau, 2015). In this dissertation, the terms "Hispanic" and "Latino” are used
interchangeably.
11

Over a two-year period, extensive data were gathered from in-person surveys
administered to 1632 Latinas, as well as 12 focus groups, and field observations of
communication “hotspots” (places where women reported spending time with their friends and
family). A total of 25 neighborhood clusters (see Appendix C) were identified based on the
geographic distribution of survey respondents for the purposes of theorizing and analyzing the
effects of both individual- and neighborhood-level variables on women’s compliance with
cervical cancer screening guidelines
3
.
It is important to note that although the Multilevel Project and the current research focus
exclusively on cervical cancer, they have much broader theoretical and practical implications in
not only cancer prevention, but also health disparities in urban ethnic neighborhoods.
Purpose of the Study
This dissertation theoretically discusses and empirically tests the critical communication
mechanisms underlying the formation of normative perceptions regarding cervical cancer
prevention for Latinas from residential neighborhoods in Los Angeles County. Specifically, a
neighborhood storytelling network grounded in a communication action context is posited as the
sources of normative information regarding Pap tests, as well as the intervening process between
Latinas’ residential environment and their corre sponding normative perceptions.
The normative perception of interest is the descriptive norms: the perception of what is
currently being done by similar others (Lapinski & Rimal, 2005). With hierarchical linear
modeling and structural equation modeling, this dissertation explores two groups of hypotheses
and research questions. The first involves the roles of Latinas’ connections to their local network
of communication resources as an intervening process between neighborhood structural
conditions and descriptive norms regarding Pap tests. The second explores how such

3
See Appendix B for the procedures of defining neighborhood clusters.
12

connections might work in conjunction with connections to major English language media to
impact more proximate health communication outcomes (i.e., media recall about information on
Pap test, attention to information about Pap test in the media, discussion with healthcare
professionals about Pap tests). These, in turn, influence descriptive norms and ultimately
compliance with cervical cancer screening guidelines.
Organization of the Dissertation
Five chapters follow this introduction. Chapter 2 discusses normative influence on health
and prior research on sources of normative influence. First, it specifies the types of norms
examined in the current dissertation. This is followed by a review of major social and behavioral
theories that explore the relations between norms and health behaviors as well as relevant
empirical evidence. Then, it provides an overview of existing work exploring interpersonal
influence, mass media, and physical environment being the sources of normative influence.
Chapter 2 concludes by pointing to the need for communication research that theorizes and tests
the underlying processes linking people’s multiple exposures to normative information in their
social, symbolic, and physical environment in the context of everyday life.
Chapter 3 introduces communication infrastructure theory (CIT). CIT is the multilevel
framework through which this dissertation examines the sources of normative influence on
health behaviors in the context of residential life. This chapter first delineates a general
overview of communication perspective on exploring neighborhood health effect and relevant
ecological models that guide such pursuit. Then, it examines CIT as a multilevel communication
approach to understanding and reducing health disparities, with a focus on geographic
communities. This chapter concludes with the research questions and hypotheses investigated in
this dissertation in the context of cervical cancer screening and detection.
13

Chapter 4 details the methodology for this study, including data collections, measures,
sample characteristics, and analytical approaches. Chapter 5 reports the results of data analysis.
Specifically, this chapter presents each hypothesis and research question, and highlights key
findings in text and tables. Finally, Chapter 6 examines the theoretical, methodological, and
practical implications of the key findings as well as the limitations of this dissertation, followed
by suggestions for future research.

14

CHAPTER 2 : UNDERSTANDING THE SOURCES OF NORMATIVE INFLUENCE
ON HEALTH BEHAVIORS

The study of normative influence has risen in salience in recent health-related scholarship
(Mollen, Rimal, et al., 2010). Empirical evidence for the impact of social norms has been
reported for a wide variety of health domains, including smoking (Echeverría, et al., 2015;
Hoffman, Monge, Chou, & Valente, 2007), alcohol consumption (Halim, Hasking, & Allen,
2012; Rimal, 2008), substance use (Eisenberg, Toumbourou, Catalano, & Hemphill, 2014;
Fujimoto & Valente, 2012), HIV-risk behaviors (Latkin et al., 2013; Young, DiClemente, Halgin,
Sterk, & Havens, 2014), condom use (M. A. Lewis, Patrick, Mittmann, & Kaysen, 2014) and,
most relevant to present purposes, cancer-related behaviors (Moran, Murphy, Frank, &
Baezconde-Garbanati, 2013; Pasick, et al., 2009; Zikmund-Fisher, et al., 2011). Accordingly,
via campaigns and interventions, deliberate efforts to change unhealthy norms and correct
misperceptions of norms have proliferated. The goal of these is to influence individual behaviors,
both domestically and abroad (Frank, et al., 2012; Mullins, Coomber, Broun, & Wakefield, 2013;
M. N. Robinson et al., 2014; Wakefield, et al., 2010). Significant progress notwithstanding,
important questions remain unanswered on both theoretical and empirical grounds, fueling the
debate surrounding the utility of norms in changing health behaviors (Reid, Cialdini, & Aiken,
2010). To address these challenges, one productive strategy is to delineate the mechanisms and
conditions that promote normative influence or inhibit norms from producing enduring effects
(Lapinski & Rimal, 2005; Yanovitzky & Rimal, 2006).
Communication research has much to offer to this inquiry. Social norms wither and
thrive through the process of human communication (Kincaid, 2004). Communication is
15

simultaneously the source of normative perceptions (e.g., people derive their perceptions of the
prevalence of a given behavior based on the depiction of the behavior in the media), and the
vehicle of influence (e.g., people make decisions to act in a given situation based on these
perceived norms) (Lapinski & Rimal, 2005). Yet there remains much to be done to understand
the role of communication in the formation and transmission of normative influence. Findings
from this line of work will not only make important theoretical contributions to norms and
communication scholarship, but also carry significant benefits to related fields such as the health
domain (Reid, et al., 2010; Yanovitzky & Rimal, 2006). Indeed, the central assumption of many
norm-based health campaigns and interventions is one pertaining to the role of communication.
In other words, through a communication intervention often carried out with interpersonal and
mass media channels, people’s misperceptions of true norms can be corrected (Lapinski & Rimal,
2005).
Building on previous literature, this dissertation focuses on the role of communication in
forming normative influence. Specifically, the present study explores the communication
mechanisms that link the sources of normative information in individuals’ environment to their
normative perceptions of health behaviors. Prior research suggests that the sources of normative
information regarding a given behavior, among other factors, affect the magnitude of people’s
perceptions regarding the prevalence and acceptability of the behavior (Borsari & Carey, 2003).
However, compared to the sheer body of evidence regarding the relationship between norms and
intention or behaviors, the sources of normative influence are not adequately studied (Mead, et
al., 2014). Knowledge obtained from this area is considered significant not only theoretically,
but also practically. At the minimum, for example, a holistic understanding of the sources of
perceived norms regarding a certain behavior can shed light into the venues and processes
16

through which erroneous perceptions are formed, spread, and perpetuated, which then serve
points of action for norm-based health campaigns and interventions.
The present chapter is grounded in a discussion regarding the interaction of normative
influence, health behaviors, and communication. The first section discusses the types of norms
examined in this dissertation, followed by an overview of corresponding empirical evidence.
Three major social and behavioral theories that explore normative influence are presented next.
Finally, using cancer screening and detection as an example, the last section discusses the role of
communication in shaping people’s perce ived norms of health behaviors, with a particular focus
on how communication as a social process can link people’s exposure to normative information
emanated from their symbolic, social, and physical environment.
Types of Social Norms
Social norms are common behavioral standards established by and for members of the
same social group (Cialdini & Trost, 1998). Depending on the level at which norms exist and
function, a conceptual line can be drawn between collective norms and perceived norms. From a
social perspective, collective norms operate at the level of the group, community, society, or
culture, providing the “prevailing codes of conduct that either prescribe or proscribe behaviors
that members of a group can enact” (Lapinski & Rimal, 2005, p. 129). From an individual
perspective, perceived norms exist at the level of individual psychology and reflect people’s
interpretation of collective norms. The latter perspective also argues that it is people’s mental
representation of the true norms, not the true norms themselves, that influences behaviors most
strongly (Yanovitzky & Rimal, 2006).
For both theoretical and methodological reasons, it is worth emphasizing that collective
norms should not be treated as the aggregation of perceived norms among members of a social
17

group, or vise versa. Understanding collective norms requires that data be collected at a level
higher than that of individuals. Deriving collective norms from indicators measured at the
individual level will incur individual fallacy (Lapinski & Rimal, 2005). Therefore, studies of
social networks, media environments, or structural characteristics of neighborhoods, cities, or
nations are the preferred sources of information on collective norms. Likewise, because
inferring individuals’ b eliefs and perceptions from collective norms engenders ecological fallacy,
assessing perceived norms requires data be gathered at the individual level (Lapinski & Rimal,
2005).
Perceived norms are the construct of focus in the present study. Broadly referring to
people’s perceptions about other’s beliefs, attitudes, and behaviors, perceived norms have been
commonly and variously coined as “social norms” (Perkins & Berkowitz, 1986), “normative
influence” (Cialdini, Reno, & Kallgren, 1990), “social influences” (Rice, 1993), “subjective
norms” (Ajzen & Fishbein, 1980), or, simply, “norms” (Bendor & Swistak, 2001). Within
perceived norms, a further conceptual distinction exists between descriptive and injunctive
norms. Descriptive norms refer to people’s pe rception of how prevalent a particular behavior is.
Injunctive norms, on the other hand, describe the extent to which people feel pressured into
enacting a certain behavior. Simply put, descriptive norms address the question “What is
currently done?”, whereas injunctive norms address “What is ought to be done” (Lapinski &
Rimal, 2005). Accordingly, each type of perceived norms serves a distinctive function. By
offering a useful heuristic, descriptive norms can be particularly consequential in novel or
ambiguous situations (Cialdini & Trost, 1998). Embedded in an individual’s social milieu,
injunctive norms operate through a threat of incurring social sanction for inappropriate behaviors
or through some sort of reward for appropriate behaviors. It is quite common for descriptive and
18

injunctive norms to mutually support each other. For example, diners might get a clue that
smoking is probably condoned in a restaurant (i.e., injunctive norms), if they notice that plenty
customers are smoking (i.e., descriptive norms). Descriptive and injunctive norms can also
contradict one another. A textbook example of such scenario is the communication of littering in
public places (Cialdini, et al., 1990). Although injunctive norms may be against littering in
places such as parking lots or residential streets, a highly littered environment could well be an
indication of descriptive norms that allow littering.
Relationships of Perceived Norms and Health Behaviors
Both descriptive and injunctive norms have been found to be predictive of a wide range
of health behaviors (Frank, et al., 2012; Rimal, 2008; White, Smith, Terry, Greenslade, &
McKimmie, 2009). Research does, however, indicate a stronger role of descriptive norms in
predicting intentions to perform health-risky behaviors than to perform health-protective
behaviors (Rivis & Sheeran, 2003). In health-related research, a cursory review of relevant
literature shows that the most represented domains are smoking and alcohol consumption
(Mollen, Rimal, et al., 2010; Reid, et al., 2010). This is not surprising, given that the two
behaviors are most frequently engaged in social settings and that the bulk of norm-based
campaign literature focuses on reducing smoking and unsafe drinking on college campuses
(Keyes et al., 2012).
Most relevant to the present study is evidence about cancer screening behaviors. Here,
both classes of perceived norms have demonstrated their influence (Nicholson et al., 2008),
though evidence generally favors the superiority of injunctive over descriptive norms. For
example, the intention to screen for cervical cancer (Bish, Sutton, & Golombok, 2000), breast
cancer, colon cancer, and prostate cancer (Smith-McLallen & Fishbein, 2008) was associated
19

more strongly with injunctive norms than with descriptive norms. Regarding the weaker role of
descriptive norms, some scholars have pointed to the behavioral attributes of screenings. Unlike
more readily visible behaviors such as drinking and smoking, screenings are usually performed
in private and therefore may be less vulnerable to descriptive normative influence, which is by
definition based on people’s observation and perceptions of the world around them (Reid, et al.,
2010; Stiff & Mongeau, 2003).
In many ways, the growing body of empirical evidence has practical implications as well.
Specifically, this line of work has augmented the popularity of campaigns and interventions that
aim to change health behaviors by targeting erroneous normative perceptions. It is important to
note that although the norms literature distinguishes descriptive norms from injunctive norms,
norm-based campaigns typically do not (Rimal & Real, 2005). For example, many antismoking
campaigns usually address injunctive norms, which are often operationalized as peer pressure or
peer influence (Sorensen, Emmons, Stoddard, Linnan, & Avrunin, 2002); campaigns seeking to
reduce excessive drinking among college students, on descriptive norms (Borsari & Carey, 2003);
and those promoting safe sex, on a combination of both (Mizuno, Seals, Kennedy, &
Myllyluoma, 2000). In part, this lack of conceptual consistency might explain why findings
regarding normative influence in these domains are sometimes conflicting (Lapinski & Rimal,
2005).
In practice, failing to recognize the conceptual uniqueness between the two types of
perceived norms can have serious consequence. One prime example is the boomerang effect
produced by norm-based health campaigns targeting risky or unhealthy behaviors of others
(Mollen, Ruiter, & Kok, 2010). In other words, although descriptive norms messages are
designed to inform the public or targeted population the seriousness of those health issues, they
20

might nonetheless reinforce a perception that being unhealthy is normative, and, as a result,
reduce people’s motivation to change. People whose perception of the prevalence of unhealthy
behaviors falls below the actual norms might even increase their behaviors upon learning the true
norms in order not to deviate from the norms. (Mollen, Ruiter, et al., 2010; Reid, et al., 2010).
Social and Behavioral Theories Exploring Normative InfluenceA number of social and
behavioral theories have been used to explicitly link perceived norms with behaviors. In the
health arena, the most popular theoretical frameworks are social norms theory (Perkins &
Berkowitz, 1986), theory of normative social behavior (Lapinski & Rimal, 2005; Rimal, 2008),
theory of reasoned action (Ajzen & Albarracín, 2007; Fishbein, 1979), theory of planned
behavior (Ajzen, 1985, 1991; Ajzen & Fishbein, 2005), and more recently the integrative model
of behavioral prediction (Fishbein & Cappella, 2006). A brief overview of these major theories
is presented below. Because the integrative model of behavioral prediction is an extension of the
theory of reasoned action, as well as the theory of planned behavior, theses theories are reviewed
together. For the purposes of this dissertation, also discussed below is the role of communication
in forming and transmitting normative influence that is articulated in these theories, though to a
varying extent.
Social Norms Theory
Social norms theory primarily focuses on descriptive as opposed to injunctive norms.
The theory states that people are motivated to behave in accordance with the norms in their
social midst. Social norms theory originated when Perkins and Berkowitz (1986) noted a
widespread misperception about alcohol consumption on college campuses, where students
reported a conservative attitude toward heavy drinking themselves, but a liberal attitude of their
peers’. Such discrepancies have been repeatedly found not only in the context of excessive
21

alcohol use and for college students, but also across several health behaviors and for other
populations (Reid, et al., 2010).
Regarding how such misperceptions are formed to begin with, in a later publication after
the seminal 1986 article, Perkins (1997) commented that students spread perceptions about
alcohol use in what he coined as “public co nversation”, even though such information was
frequently inaccurate. It follows that, to the extent that misperceptions of the true norms would
fuel problematic behaviors, providing accurate information about the true norms is a fruitful
strategy to produce behavioral change (Perkins & Berkowitz, 1986; Perkins, 1997). Indeed,
many social norms marketing campaigns have since adopted the social norms theory as the
theoretical backbone to curb unhealthy behaviors by modifying misperceived norms (Campo, et
al., 2003; Mabry & Mackert, 2014). However, what is often neglected in this larger literature is
the question similar to Perkins’s (1997) as for the origins of misperceived norms among targeted
populations (Real & Rimal, 2007).
Theory of Normative Social Behavior
Built on the work of Cialdini and others (Cialdini, et al., 1990; Cialdini & Trost, 1998),
the theory of normative social behaviors (TNSB) theorizes the pathways through which
descriptive norms influence behaviors in conjunction with other factors (Rimal, 2008; Rimal &
Real, 2005). Research following this approach investigates the moderating roles of outcome
expectations, group identity, ego-involvement, and injunctive norms (see Lapinski & Rimal,
2005, for a review). Specifically, descriptive norms were found to be a stronger predictor of a
given behavior when people believed that performing the behavior would result in significant
benefits (Rimal, Lapinski, Cook, & Real, 2005), when they identified with the social group in
22

which the behavior was widespread (Rimal & Real, 2003), and when enacting the behavior
linked to their self-concept in certain ways (Rimal, 2008).
Of particular relevance in TNSB is the interplay between descriptive and injunctive
norms (Rimal & Real, 2003). In line with early empirical research (Asch, 1951; Kitayama &
Burnstein, 1994), where participants did not conform to strong injunctive norms when the
descriptive norms were low, Rimal and Real (2003) found a significant interaction between the
two classes of perceived norms, such that the effect of descriptive norms on behavioral intention
increased with strong injunctive norms and decreased with weak injunctive norms. The authors
went on to argue that strong injunctive norms alone were insufficient to influence behaviors, and
that greater compliance required both norms to be strong.
In more recent TNSB research, the focus shifts to testing whether the inclusion of
communication processes would increase explanatory power for models of normative influence.
For instance, Real and Rimal (2007) found that descriptive norms, peer communication, and their
interactions were all significantly associated with alcohol consumption for college students.
Specifically, the association of descriptive norms and drinking alcohol was stronger when
students talked about alcohol use compared to when they did not. The interaction remained
significant even after accounting for other known predictors of alcohol consumption. Findings
like this support the aforementioned argument that communication, such as peer discussion,
serves as a principle mechanism for forming and transmitting normative influence.
Integrative Model of Behavioral Prediction
Integrative model of behavioral prediction (IMBP) synthesizes constructs from four
psychosocial models of behaviors: theory of reasoned action (TRA), theory of planned behavior
(TPB), social cognitive theory (Bandura, 1977, 1982, 2001), and health belief model (Becker,
23

1974). As shown in Figure 2.1, this model assumes that behavioral intention is the primary
proximal cause of a behavior. The relationship between intention and behavior is moderated by
skills necessary to perform the behavior and the level of environmental constrains. Moreover,
behavioral intention is seen as a function of attitudes, perceived normative pressures, and self-
efficacy, which are then shaped by a corresponding set of underlying beliefs.
Two aspects regarding these proximal variables are worth noting. First, in IMBP, it is
conceptualized that the relative importance of attitudes, perceived norms, and self-efficacy in
determining intention varies across behaviors and populations being studied (Fishbein &
Cappella, 2006). According to this theory, while some behaviors are primarily driven by
attitudes, others might be largely influenced by perceived norms. Similarly, the same behavior
that is predominately normatively driven in one population or culture can be solely attitudinally
driven in another.
Second, in IMBP, level of underlying beliefs manifests the “substantive uniqueness of
each behavior” (Fishbein & Cappella, 2006, p. 53). For instance, the outcome expectations of
getting a Pap test and the barriers to getting a Pap test may differ considerably from those
associated with receiving a flu shot. In order for health promotion programs and interventions to
succeed, according to this theory, one must understand the specific beliefs regarding performing
a certain behavior from the perspective of the population being considered.
Here, it should be noted that because IMBP is an extension of TRA and TPB, in its
original theorizing the perceived norms construct and its underlying beliefs essentially pertain to
subjective norms, a form of injunctive norms (Lapinski & Rimal, 2005). Acknowledging the
conceptual difference between descriptive and injunctive norms, Ajzen and Fishbein (2005)
suggested the inclusion of both norms for future research. Recent IMBP research that followed
24

suit has separately examined the roles of descriptive norms from those of injunctive norms
(Bleakley, et al., 2011; Rhodes, et al., 2007; Robbins & Niederdeppe, 2015).

Figure 2.1 Integrative Model of Behavioral Prediction. Reprinted from “The Role of Theory in
Developing Effective Health Communications”, by M. Fishbein and J. N. Cappella, 2006,
Journal of Health Communication, 56, S1-S17, p.S2. Copyright 2006 by the International
Communication Association.

Finally, IMBP theorizes the roles of “external” or “background” variables, such as past
behaviors, demographic variables, sociocultural factors, and communication variables such as
media exposure. Specifically, the model states that the influence of these background variables
is only mediated by more proximal predictors, based on the premise that such background
variables are unlikely to impact intention or behavior if they are unrelated to the underlying
beliefs regarding attitudes, normative pressures, and self-efficacy (Fishbein & Cappella, 2006).
25

However, whether or not background variables will affect underlying beliefs or more
proximal variables, as theorized in IMBP, is an empirical question. For example, Bleakley and
colleagues (2011) tested the role of exposure to sexual media content as a “background variable”
in influencing adolescent sexual behaviors. Consistent with IMBP, their path analysis
demonstrated that behavioral intention to engage in sexual behavior was determined by attitudes,
perceived norms, and self-efficacy. At the same time, the hypothesized effect of exposure to
sexual media content on proximal variables was only significant for normative beliefs.
Communication and Sources of Normative Influence
As noted in the preceding section, research based on major social and behavioral theories
and designed to examine normative influence has not only documented the effect of perceived
norms on health behaviors, but also increasingly indicated the critical role of communication in
forming and transmitting normative influence. The latter adds empirical support to the recent
development in the larger norm literature, which argues that because social norms emerge,
evolve, and spread through human communication, a better understanding of the communication
mechanisms through which normative influence comes into being represents a promising venue
to enhance this scholarship (Yanovitzky & Rimal, 2006). Specifically, as discussed more fully
below, a communication approach makes important theoretical and practical contributions to the
study of sources of normative influence on health behaviors.
People come into contact with a myriad of information that communicates health-related
norms in the process of everyday life and through a wide variety of venues. Compared to the
relationship between perceived norms and intention or behavior, the sources of normative
influence, as well as the social processes that link information sources to the formation of
perceived norms are not adequately understood (Mead, et al., 2014). Extant literature that
26

pursues this direction largely focuses on interpersonal communication (Cialdini & Trost, 1998;
Khalil & Rintamaki, 2014; Real & Rimal, 2007) and mass-mediated channels (Gibbons, et al.,
2010; Mabry & Mackert, 2014; Wakefield, et al., 2010) as sources of normative influence,
leaving other significant sources less characterized.
This restrictive focus is not surprising, given that a significant portion of relevant work
comes from either traditional mass media research or intervention and campaign contexts, in
which interpersonal discussion and mass media are usually of primary interest. Moreover,
because most research in the large literature on health information seeking and exposure
concerns the strategic and deliberate use of communication channels, such as the use of Internet
or the effect of targeted communication campaigns, much less is known regarding the effects of
casual encounters to normative information (or what is termed as incidental exposure) on the
formation of perceived norms (Hornik et al., 2013).
One recent development in norm literature is particularly relevant to the present
discussion. The concept of social exposure, first introduced in the 2010 report of the Smoke-
Free Ontario Scientific Advisory Committee (Smoke-Free Ontario Scientific Advisory
Committee, 2010), refers to the “composite ways in which people come in contact with or
experience a particular product or behavior in their environment” (Mead, et al., 2014, p. 140).
Using tobacco as an example, Mead et al. (2014) illustrated that individuals are constantly
exposed to tobacco use, as well as cues about tobacco, in their symbolic, social, and physical
environment. This serves as an important type of information that communicates norms about
tobacco use. In other words, individuals learn about how widespread (i.e., descriptive norms)
and socially approved (i.e., injunctive norms) smoking is from depiction of smoking in
entertainment and news media, in tobacco advertising and marketing, and in tobacco control
27

campaigns; from observing others (e.g., family, friends, coworkers) who engage in the behaviors;
and lastly from their physical environment through exposure to tobacco use or products in
venues such as restaurants, schools, workplaces, and tobacco retail outlets in communities. In
this way, the usefulness of the concept of social exposure lies in identifying the aspects of one’s
environment that justify unhealthy norms or promote healthier alternatives. These can be used as
points of intervention to produce change in the environment and subsequently more sustainable
impacts on behaviors.
Taking a social exposure approach, the rest of this section discusses people’s symbolic,
social, and physical environment as sources of normative influence in the case of cancer
screening and detection. Specifically, the discussion explores the roles of communication as a
social process in understanding the impacts of each type of information sources, and the
interactions among them, on the development of people’s perceived norms. In order to account
for the fact that cancer screening is usually conducted in private as opposed to social settings,
hence not as visible as tobacco or alcohol use, the concept of social exposure is loosened to
include not only exposure to actual or cues about a behavior, but also more general types of
information (e.g., statistics on cancer screening, biomedical development).
Symbolic Environment
An individuals’ symbolic environment can be a significant source of information that
conveys norms regarding cancer screening and detection. The symbolic environment in the
present study features a broad array of information sources, such as news and entertainment
media (Chung, 2014; Lee, Long, Slater, & Song, 2014; Murphy, Hether, & Rideout, 2008),
cancer communication campaigns (Marcus & Crane, 1998; Mullins, et al., 2013; Snyder et al.,
2004; Wakefield, et al., 2010), and pharmaceutical advertising (Klein & Stefanek, 2007).
28

Today’s symbolic environment is rife with cancer informati on of all kinds. This
information continues to grow in amount and complexity, both in traditional mass media and on
the Internet (Chou, Prestin, Lyons, & Wen, 2013; Viswanath, 2005; Viswanath et al., 2006). On
the Internet alone, a single search of the word “cancer” may easily generate millions of results.
Nevertheless, prior research has consistently indicated that traditional mass media such as
television news remain a primary source for Americans to obtain health or medical information,
including cancer information (Kelly et al., 2010; Lee, et al., 2014).
As with many things, quantity of cancer information does not assure quality of
information or reflect cancer control priorities. Taking news coverage of cancer as an example,
it is not uncommon to see the same story being covered from different angles or featured with
conflicting information (Smith, Niederdeppe, Blake, & Cappella, 2013). Additionally, the focus
of these news stories is more often on treatment than on prevention, and more often on
individual-level risk factors than on community- or population-level policies (Jensen, Moriarty,
Hurley, & Stryker, 2010)
Several theories from the communication literature offer explanations for why and how
cancer-related stories, images, and statistics presented in news and entertainment media can
affect people’s perceptions of social reality and prevalence of screening behaviors. Cultivation
theory states that the social reality portrayed on television shape individuals’ beliefs about the
general nature of the world (Gerbner, Gross, Morgan, Signorielli, & Shanahan, 2002). This
theory accounts for why heavy television viewers are more likely than light viewers to adopt and
adhere to the viewpoints conveyed on television. Research using cultivation analysis has
consistently reported television’s influence on viewers’ perc eptions and beliefs on topics such as
violence and crime, politics, religion, gender roles and stereotypes, and, more recently, health
29

(Bilandzic & Busselle, 2012; Gerbner, et al., 2002) . One study using a national representative
sample found a positive and significant association between respondents’ fatalistic beliefs about
cancer prevention and both overall amount of television viewing and local television news
viewing (Lee & Niederdeppe, 2011). The latter finding, the authors argued, was consistent with
the observation made in a content analysis (Lee, et al., 2014): cancer coverage in local television
news, compared to that in national television news, was less likely to address cancer prevention
such as screening tests and other preventive behaviors, and less likely to refer audience to
national organizations (e.g., NCI, CDC) for clear recommendations.
Another genre that receives growing attention in cultivation studies is medical drama.
Using a national representative sample, Chung (2014) reported a similar pattern to Lee and
Niederdeppe (2014) in that heavy medical drama viewing was associated with more fatalistic
beliefs about cancer. According to the author, this might result from heavy viewers’ prolonged
exposure to an unrealistically high rate of death from medical conditions portrayed in medical
dramas, which may lead viewers to conclude that fighting cancer is pointless.
More relevant for present purposes, evidence regarding the association between casual
exposure to cancer information and normative beliefs is also emerging. For example, Hornik and
colleagues (2013) suggested that the effect of routine health information exposure, or scanning,
may be more influential in the aggregate than a single episode of deliberate information seeking.
Although cultivation theory was not cited explicitly, measures of such incidental exposure to
some degree capture the dimension of habitual consumption of media content, which is at the
heart of cultivation theory (e.g., “thinking about the past 12 mont hs, did you hear or come across
information about [colonoscopy] … from the media even when you were not actively looking for
it?”) (Hornik, et al., 2013, p. 1425). As for the rationale that routine exposure to health
30

information can serve as a source of normative influence on cancer screening, the authors argued
this: if such information recurrently appeared across a range of major information sources that
people routinely connect to, scanning might result in a belief that the behavior is prevalent
among most other people (i.e., descriptive norms) and/or is socially approved or expected (i.e.,
injunctive norms) (Hornik, et al., 2013).
Social cognitive theory (SCT) is another widely cited theory that attempts to explain the
effects of media portrayal on people’s acquisition of new knowledge, attitudes, and behaviors
(Bandura, 1977, 2001). SCT suggests that learning takes place not only from direct experience,
but also from observing and modeling the behaviors of observed others (Bandura, 2004).
Particularly, such learning occurs more readily when the observed behaviors are demonstrated by
others whom the observers consider similar to themselves (Bandura, 2001, 2004).
Recent years have seen a resurgent interest in SCT in understanding and promoting
health behaviors, due in large part to the growth of entertainment education and narrative
communication scholarship (Hether, Huang, Beck, Murphy, & Valente, 2008; McQueen, Kreuter,
Kalesan, & Alcaraz, 2011; Murphy, et al., 2013; Murphy, Frank, Moran, & Patnoe-Woodley,
2011; Singhal & Rogers, 1999). McQueen et al. (2011), for instance, found that African
American women who watched a narrative video featuring stories of African American women
breast cancer patients reported lower level of perceived barriers and fatalistic beliefs about
cancer. Likewise, Murphy and colleagues (2013) found that Mexican American women
identified more strongly with most of the characters in a cervical cancer-related narrative film
featuring Latinas, and identification with specific characters, in turn, led to their shifts in
knowledge, attitudes, and behaviors regarding Pap tests.
31

Seeing fictional characters adopt Pap tests can also influence viewers’ perceptions about
the prevalence of the behavior. Using the same aforementioned narrative film as the stimulus,
Moran and colleagues (2013) found a significant effect of film type on posttest descriptive norms.
Specifically, among less educated women, those who watched the narrative film thought that
getting a Pap test was much more prevalent than did their counterparts who watched a non-
narrative film on the same topic.
Social Environment
People’s social environment constitutes another significant source of normative influence
that contributes to people’s perceptions of the prevalence and social acceptability of a behavior.
Prior to discussing the role of social environment more fully, a conceptual distinction between
normative and informational social influence made in small group research is worth noting
(Deutsch & Gerard, 1955).
Normative social influence occurs when people are driven by a desire to conform to the
expectations of others, and is the type of social influence exerted in a group setting. Observing
or communicating with others engaging in a certain behavior may lead people to think that the
behavior is widespread, approved, and, therefore, normative. On the other hand, informational
social influence occurs when during the course of discussion people accept the information and
opinions provided by others. The rationale behind informational social influence is that people
learn from the world around them by communicating with others, and, when arguments arise, by
comparing their own opinions with those held by others (Price, Nir, & Cappella, 2006).
Previous research indicates that both types of social influence emanate not only from
proximal sources (e.g., family, close friends) in one’s social network, but also distal sources (e.g.,
coworkers, classmates, co-workers) and non-members (e.g., acquaintances, strangers) (Asch,
32

1951; Borsari & Carey, 2003). Not surprisingly, perhaps, the social distance between a person
and the source is negatively associated with the accuracy of the perceived norms (Baer, Stacy, &
Larimer, 1991). In other words, people are more likely to pay attention to and believe sources in
their inner circle, such as family and friends. However, while the distinction between normative
and informational social influence is clear in concept, separating the normative component from
the informational component of social influence can be difficult in practice (Price, et al., 2006).
The two constructs are also highly correlated, and outcomes frequently result from a combination
of both (Dunlop, et al., 2010; Price, et al., 2006).
In the health arena, the role of social environment in impacting normative perceptions is
often examined in research tackling social influence in one’s social network, which typically
focuses on behavioral domains such as smoking (Valente, Hoffman, Ritt-Olson, Lichtman, &
Johnson, 2003), substance use (Valente et al., 2007), condom use (Frank et al., 2012), and HIV
risk behaviors (Latkin, et al., 2013). In the realm of cancer screening and detection, research
generally points to the importance of social environment in the formation of injunctive norms,
where people stress the influence of family and friends on their screening behaviors (Bazargan,
Bazargan, Farooq, & Baker, 2004; Cogliano et al., 2005; Pasick, et al., 2009; Smith-McLallen &
Fishbein, 2008). More recent research reveals the role of social environment in shaping
descriptive norms as well. Concerning the effect of routine exposure to health information on
cancer screening behaviors (mammography, colonoscopy, prostate-specific antigen [PSA]) as
opposed to the effect of episodic information seeking, Hornik and colleagues (2013) found a
significant and positive association between people’s perceptions of how often peers were
receiving all three types of tests and their routine exposure to health information in prominent
information sources, including family, friends, or coworkers.
33

Social environment can also influence normative perceptions by mediating the effects of
symbolic environment. Empirical research following this direction most often taps into the
question regarding the interplay between interpersonal communication and mass communication
(Southwell & Yzer, 2007; Wakefield, et al., 2010), which has been extensively studied from
perspectives such as two-step flow (E. Katz, 1957; E. Katz & Lazarsfeld, 2006), social cognitive
theory, and diffusion of innovations(Rogers, 1995), just to name a few. Katz and Lazarsfeld
(2006), for example, articulated a two-step flow model to explain voting behaviors, arguing that
information supplied by the media first reach opinion leaders, who then influence their contacts.
Despite being overly simplistic, this model does stress the importance of understanding the effect
of mass media as an information source in shaping people’s perceptions and behaviors in their
social context. Similarly, Bandura (2001) elaborated the direct and indirect effect of media
through peoples’ social networks, o r in his language, “dual paths of influence” (p.285).
Specifically, he argued that “when media influences lead viewers to discuss and negotiate
matters of import with others in their lives, the media set in motion transactional experiences that
further shape the course of change” (Bandura, 2001, p. 286).
In health communication, a variety of studies have examined the influence of
interpersonal communication that is stimulated by media, with much of this scholarship coming
from health campaign research (Chatterjee, Frank, Murphy, & Power, 2009; Frank, et al., 2012;
Valente, Paredes, & Poppe, 1998). Of studies focusing on cancer screening behaviors, including
cervical cancer screening, both interpersonal communication and mass communication were
credited for their impacts on people’s knowledge, attitudes, and behaviors, though nuances were
reported (Marcus & Crane, 1998; Snyder, et al., 2004; Wakefield, et al., 2010). One study, for
example, found that interpersonal communication on breast cancer screening practices were
34

more influential among college-age women, whereas mass media were more influential among
middle-age women (Jones, Denham, & Springston, 2006).
Fewer studies, however, have specifically tested the influence of media-initiated
discussion on normative perceptions (Dunlop, et al., 2010). Among those that have, one study
examined the effect of online group interaction among adolescents on their related attitudes and
behaviors as a result of their exposure to anti-marijuana advertisements (David, et al., 2006).
Group interactions were found to have a negative effect for adolescents who viewed the antidrug
ads, such that those who discussed the ads were more likely to report attitudes and subjective
normative beliefs in favor of marijuana use than those who just viewed the ads. Another study
investigated the impact of exposure to a public communication campaign on behavioral intention
to use condoms among Indian men between the ages of 15 to 49 years (Frank, et al., 2012).
Among high-risk and sexually inactive men, higher level of campaign exposure was associated
with more positive discussions regarding condom use, which enhanced both descriptive and
subjective norms favoring condom use. Findings like these highlight the importance of studying
the effects of information sources on people’s normative beliefs in a particular social and cultural
context, especially when the pre-existing prevailing norms go against the recommended behavior
(Morgan, 2009).
Physical Environment
In their original conceptualization, Mead et al. (2014) argued that people’s physical
environment serves as a significant source of social exposure to normative information in two
fashions. Taking tobacco use as an example, the authors suggested that the use or physical cues
about tobacco use are sufficient to communicate the prevalence and acceptability of smoking.
Following this reasoning, the attributes of the physical environment may also convey normative
35

information about tobacco use, as suggested by previous research that physical availability of
tobacco retail stores in communities was positively associated with adolescents’ tobacco use
(Novak, et al., 2006; West, et al., 2010). Yet, as noted previously, because cancer screening
behaviors are not as visible and observable as smoking, it may be less likely for the physical
environment to communicate normative information by exposing people to related behaviors or
cues about the behaviors.
The physical environment may, however, serve as a venue for an individual’s symbolic
and social environment to convey information about descriptive and injunctive norms regarding
cancer screening. Although empirical research that specifically examines this area is rare,
observations from the larger domain of health communication campaigns and programs support
this contingency. Communication campaigns and programs have used outdoor media such as
billboards, posters, or wall paintings for information dissemination purposes in order to increase
campaign exposure in the targeted communities (Institute of Medicine, 2002; Randolph &
Viswanath, 2004; Wakefield, et al., 2010). Likewise, making health educational materials
available at community places such as clinics, libraries, churches, community organizations, and
trusted business can augment program outreach, especially when the interest is to reach certain
segment of population that traditional mass media campaign may have difficulty in reaching,
such as ethnic minorities or new immigrants (Institute of Medicine, 2002; Mann, et al., 2014).
For example, in one study examining the barriers to cervical cancer screening among a sample of
Hispanic women in the Boston area, the majority of the women, despite their low level of
literacy, stated that they read pamphlets or posters while waiting at the health care providers’
office (Watts et al., 2009). Similarly, the physical environment may interact with the social
environment to communicate normative information related to cancer screening and detection.
36

Outside home, places such as parks, schools, libraries, restaurants, cafes, and supermarkets are
where everyday conversation happens between people with their friends, neighborhoods, or
strangers they randomly encounter (Matsaganis, Gallagher, & Drucker, 2013; Oldenburg, 1999).
Everyday conversation, on the other hand, has been shown to be a significant source of
information, including information that influences health-related normative perceptions (Cline &
Thompson, 2003). Moreover, as shown in a growing body of community-based participatory
research, community places can play a critical role in instigating conversations specifically about
health, and in linking residents with needed social and health services, especially in underserved
areas (Israel et al., 2010; Kreuter, Kegler, Joseph, Redwood, & Hooker, 2012; Matsaganis,
Golden, & Scott, 2014). For instance, prior research recruiting Latinas from churches and
training them to deliver evidence-based screening interventions to peers (e.g., one-on-one
outreach, small group education, health fair) has reported success in increasing conversations and
awareness about Pap tests, and promoting cervical cancer screenings (Allen et al., 2014; Mann,
et al., 2014). Taken together, to the extent that community places can enable or constrain
communication activities of individuals, the availability of those places, as well as the attributes
that make them conducive places for social interactions in general and conversations about
health in specific, can serve as another explanation for the critical role of the physical
environment in forming perceived norms regarding health actions, including cancer screenings.
Conclusion
Over the course of decades, the importance of perceived norms on health behaviors has
been extensively studied in a multitude of behavioral domains. Despite significant progress,
there remain questions awaiting further theoretical and empirical inquiry. One of these questions
pertains to the mechanisms and processes through which normative influence exerts its impact,
37

an area to which the field of communication has much to contribute. Against this backdrop, the
present chapter centers its discussion around the intersection of literature on normative influence,
health behaviors, and communication, with a particular focus on the sources of normative
influence that people encounter in their environment. Using cancer screening and detection as an
example, three types of environmental sources of normative information are explored, namely
symbolic, social, and physical environment. In short, people can learn about the prevalence and
social acceptability of cancer screenings from their social environment by talking with others
(e.g., family, friends, neighbors, coworkers, or acquaintances); from various forms of portrayals
of cancer screenings in their symbolic environment; and, lastly, from seeing or hearing
information about cancer screening in their physical environment. Communication is thus the
critical process that underlies the influence of each type of information sources on the formation
of perceived norms, and of the interactive effects between these sources. Knowledge obtained
from this front has both theoretical and practical significance. In terms of theoretical
contributions, studying the communication mechanisms that account for the influence of
information sources on the formation of perceived norms can help identify the processes and
conditions that promote or constrain normative influence, a question that has intrigued both
norms and communication scholars. In terms of practical contributions, such knowledge can
help researchers and practitioners to diagnose the aspects of people’s environments that uphold
unhealthy norms or perpetuate misperceived norms, which can serve as points of intervention.
One observation of this literature is worth noting. A significant portion of current
knowledge regarding the influence of information sources on people’s health -related normative
perceptions, among other cognitive and behavioral outcomes, is derived from the context of
health communication campaigns and programs. This may explain why the bulk of evidence on
38

information sources focuses on interpersonal communication and mass media, as both are
heavily drawn upon in health communication efforts (Wakefield, et al., 2010). However, in the
context of day-to-day life, people encounter normative information through numerous types of
sources that go beyond the range of channels that a campaign typically targets. Casual exposure
to health information (including cancer information) in nonmedical sources has also been
suggested to be more influential than episodic and deliberate information seeking or formal
contact with medical providers (Hornik, et al., 2013). Further, different sources of normative
information may not carry equal weight. Depending on the interaction between individuals and
their environment, information sources differ from each other in terms of relative importance in
shaping people’s perceptions about the world around them in general, and health -related
behavioral norms in particular (Wilkin, 2013). For example, individuals might discuss with
friends and family about cancer screening related issues more often if they have limited access to
medical providers in their neighborhoods (Wilkin & Ball-Rokeach, 2011). Those who watch
local television news more frequently might be more likely to have a distorted perception
regarding cancer screenings compared to those who mainly watch national television news. This
distortion might result from the systematic difference in cancer coverage between local and
national news (Lee, et al., 2014) or the media market in which people live (Preacher, Rucker, &
Hayes, 2007; Snyder, Fleming-Milici, Slater, Sun, & Strizhakova, 2006). Lastly, individuals
who are considered linguistically isolated in their neighborhood may have reduced
communication opportunities with others in the neighborhood due to their difficulty in
navigating the environment. To conclude, communication of perceived norms is a function of
the characteristics of both individuals and their environment, and the formation of perceived
norms is a multilevel phenomenon.
39

CHAPTER 3 : A MULTILEVEL COMMUNICATION INFRASTRUCTURE MODEL OF
NORMATIVE PERCEPTIONS

Residential neighborhoods are the environment where we experience the conditions of
everyday life most deeply. In the health arena, the interdependent relationship that we have with
our neighborhoods is embodied in a research question that has become increasingly salient: How
do the places we live influence our health?
A rough idea of the rapid growth of this research field can be gleaned from looking at the
number of publications returned by online citation search engines. During the decade between
2003 and 2013 alone, over one hundred thousand studies were published on the neighborhoods
health effects, which nearly doubles the number for the decade prior (Matsaganis, 2015).
Accompanying this growth is the research agenda that has progressively shifted from early work.
Until the 1990s, neighborhood health effects research almost exclusively focused on drawing
associations between area-level characteristics (e.g., poverty or instability at the level of census
tracts) and people’s physical and mental health. Since the outset of 1990s, there has been a
resurgence of interest in the social processes and mechanisms through which neighborhoods
effects on many aspects of people’s lives, including health, are produced (Sampson, et al., 2002).
The vast majority of these studies stem from the fields of sociology and public health. Some of
the proposed processes and mechanisms include social disorganization, social ties or interaction
and social capital, norms of reciprocity and collective efficacy, institutional resources, and
routine activities (see Sampson et al, 2002, and Matsaganis, 2015, for a review).
Intriguingly, outside the discipline of communication, the role of communication as a
social process through which neighborhoods influence people’s well -being is largely ignored. In
40

sociological work on neighborhood effects, for instance, communication is often treated as
equivalent to social ties or social networks (Sampson, et al., 2002), and the role of media is
noticeably absent from discussion. This is despite the observation made in the early days of the
Chicago School, on which much of the extant literature on neighborhood effects has theoretically
based, that communication is one of the key mechanisms that organize and transform urban life.
In his seminal work, Park noted that “transportation and communication…are primary factors in
the ecological organization of the city” (R. E. Park, 1925, p. 2). In relevant public health
literature, communication is typically employed in public health and social marketing
campaigns, with the assumption that place-based campaigns can extend the reach to otherwise
neglected areas or populations (Daniel, Bernhardt, & Eroğlu, 2009) . However, with most
campaign research and programs focusing on finding the optimal combination of communication
channels (e.g., mass media and interpersonal discussion) to maximize campaign outreach, the
role of places in shaping people’s access to available co mmunication resources and their
exposure to health information conveyed by those resources is less understood (Matsaganis,
2015).
The goal of this chapter is to present a socio-ecological, multilevel communication
framework to understand the sources of normative perceptions regarding Pap tests in the context
of everyday life among Latinas living in Los Angeles’ diverse neighborhoods. The premise of
the framework is that communication is an elementary social process that not only helps explain
the relationship between where people live and their health-related knowledge, beliefs, and
behaviors, but also, over time, may “give birth to and sustain more complex social mechanisms”
underlying the impacts of neighborhoods on health (Matsaganis, 2015, p. 44). To this end, the
first section below provides a general overview of communication research that examines
41

people’s health behaviors and health -related outcomes, either theoretically or analytically, from
an ecological perspective, a perspective that has a long history in the multidisciplinary field of
neighborhood effects (Kreps & Maibach, 2008). The strengths and limitations of two ecological
communication frameworks that have been fruitfully applied to addressing the larger question
regarding the relationship between place (including neighborhoods) and people are discussed.
The second section elaborates communication infrastructure theory (CIT), a socio-ecological
theory that is developed specifically in the context of diverse urban neighborhoods. Relevant
CIT research that investigates residents’ civic engagement and health -related outcomes in urban
racial and ethnic communities is discussed in terms of the utility and potential of CIT for the
present pursuit. The last section applies CIT to the study of normative perceptions in the
behavioral context of cervical cancer screening and detection for Latinas in Los Angeles,
followed by hypotheses and research questions.
Communication Research Examining Health-Related Outcomes in Ecological Context
The role of communication process in shaping people’s health -related beliefs, attitudes,
and behaviors has been extensively documented in the field of communication. However, this
line of literature is dominated by behavioral change theories that primarily focus on cognitive or
behavioral changes at the individual level. Much more work needs to be done to account for the
dynamics between health behaviors and the ecological context in which those behaviors are
performed, modified, and sustained. More relevant to present purposes, as noted by Kreps and
Maibach (2008), ecological perspective is one of the promising areas of transdisciplinary
common ground that can bridge communication and public health research, especially for efforts
exploring the relationship between people and places. Further, one key characteristic of
ecological models is, to put it simply, that “they incorporate two or more analytic levels (e.g.,
42

personal, organizational, community)” (Stokols, 1996, p. 287), which subsequently implies the
use of analytic techniques suitable for studying multilevel phenomena. However, despite a
growing advocacy over the past decade (Hampton, 2010; Hayes, 2006; H. S. Park, Jr, & Cudeck,
2008; Slater, Snyder, & Hayes, 2006), communication research that spans levels to theorize and
test communication as a multilevel social process remains relatively few.
Of health communication studies that embrace an ecological perspective resembling
Stokols’ (1996) definition, the majority are established on the followin g theoretical frameworks:
the knowledge gap hypothesis, the structural influence model of communication (SIM), and
communication infrastructure theory (CIT) (see Matsaganis, 2015, for a review). Of these,
knowledge gap hypothesis and SIM contribute to the present discussion by showing how
individual- and neighborhood-level characteristics influence residents’ lives and well -beings by
interacting with communication inequalities, defined as “differences among social groups in the
generation, manipulation, and distribution of information at the group level, and differences in
access, use, and holding of relevant information, as well as the capacity to take advantage of it at
the individual level” (Viswanath & Emmons, 2009, p. 284). In this context, communication
inequalities are broadly conceived of as a social determinant of health disparities, and, with right
strategies, a readily modifiable and addressable one compared to other social determinants (e.g.,
policies, social class).
Knowledge Gap Hypothesis
The central thesis of the knowledge gap hypothesis is that the distribution of knowledge
is shaped by the structural positions of individuals, groups, and communities (Viswanath &
Finnegan, 1996). Accordingly, when new information is disseminated to the society,
individuals, groups, or communities that are in more desirable structural positions will access the
43

information faster, thus increasing the gap in knowledge. In application, however, studies
adopting this framework have mostly focused on individual-level socioeconomic characteristics
(e.g., education) as determinants of knowledge gap and health outcomes, rather than exploring
community-level characteristics or accounting for the ecological context in which the population
of interest are situated. Of those that have, one study found that the association between
education and cancer-prevention knowledge was moderated by the difference in newspaper
cancer-prevention coverage across U.S. regions, or areas of comparable media market size
(Slater, Hayes, Reineke, Long, & Bettinghaus, 2009).
Structural Influence Model (SIM)
By comparison, SIM elaborates and tests the pathways that link structural characteristics
to individual-level communication and health outcomes, thereby more explicitly bridging
communication and public health literature to explore communication inequalities as a social
determinant of health disparities (Bekalu & Eggermont, 2014; Viswanath & Emmons, 2009). As
shown in Figure 3.1, SIM posits that people’s socioeconomic positions and structural factors
related to where they live interact with their sociodemographic characteristics and social
resources, which in turn give rise to differential health communication outcomes (e.g., media
exposure, attention, information seeking), and, ultimately, health outcomes (e.g., health beliefs,
screening). In a recent study, Bekalu and Eggermont (2014) inquired about the effects of
education and area of residence (urban vs. rural), characterized as a socio-ecological factor, on
people’s HIV/AIDS knowledge and perceived risk. Analysis indicated that both education and
urbanity versus rurality had a direct effect on the outcome variables of interest. Consistent with
the prediction of SIM, such effect was also partially mediated by communication inequalities,
44

such as people’s difference in their information need on HIV/AIDS, related media use, and
perceived salience of relevant information.

Figure 3.1 Structural influence model (SIM) of communication. Reprinted from Health
Communication and Communication Inequalities in Addressing Cancer Disparities (p.285), by
K. Viswanath and K. Emmons, in H. K. Koh (Ed.), Toward the Elimination of Cancer
Disparities, 2009, New York, NY: Springer. Copyright 2009 by the Springer Science +
Business Media.

Two limitations inherent to the extant SIM literature are worth noting. First, current SIM
work attending to the contextual influence on disparities in communication and health-related
outcomes does not adequately elaborate how specific physical characteristics of a neighborhood
(e.g., availability of parks and recreational areas) strengthen or constrain the role of residents’
social environment in impacting health-related beliefs and behaviors (Matsaganis, 2015).
Second, as with many conceptual frameworks in public health literature, SIM assumes a linear
pathway that travels from distal predictors, such as socioeconomic and demographic variables, to
communication processes proposed as intervening variables, and to health behaviors and related
outcomes. This linear mechanism potentially masks the complexity between people and their
ecological context (including neighborhoods) in relation to a given health-related outcome. For
45

example, area-level characteristics such as percentage of ethnic minorities may impact the media
landscape in a given area (e.g., ethnic media are more dominant in ethnic minority
neighborhoods). This, in turn, can lead to differential access to media content among residents
of different areas. On the other hand, residents’ differential access to media content may also
result in divergent perceptions regarding their neighborhoods (Gerbner, et al., 2002; Matei, Ball-
Rokeach, & Qiu, 2001), which can have substantial implications for their health as well. For
example, depending on their respective cultural or linguistic orientations, residents living in the
same ethnically heterogeneous neighborhood might connect to different media outlets to learn
about their neighborhoods. If some media do not provide sufficient information about certain
local resources (e.g., quality groceries, places to exercise), residents who connect to those media
outlets might have a skewed perception that they have to go elsewhere for their corresponding
needs, even though relevant resource in their own area could be plenty.
Communication Infrastructure Theory
Putting communication front and center, communication infrastructure theory (CIT) is an
socio-ecological framework focusing on the importance of communication infrastructure to local
residents’ everyday information seeking and problem solving (Ball-Rokeach et al, 2001; Kim &
Ball-Rokeach, 2006a, 2006b). CIT has a theoretical root in media system dependency theory
(MSD) (Ball-Rokeach, 1985, 1998; Ball-Rokeach & DeFleur, 1976). MSD makes an argument
in favor of the necessity of interdependence among stakeholders due to ambiguity and threat that
features modern society. To this end, mass media is conceived of as a central institution that
most people must connect to for goal attainment. A focus on interdependence means that,
however, MSD assumes neither a powerful media nor an active or powerless audience. Rather,
the centrality of mass media in individuals’ lives hinges upon the various goals of individuals,
46

who, in turn, negotiate with mass media that control necessary resources for them to achieve the
goals (Ball-Rokeach, 1998).
Extending the scope of MSD, CIT shifts the almost exclusive focus on the power
dependency relation between mass media system and individuals to the dynamic interplay
between a more inclusive communication environment and individuals or community (Ball-
Rokeach & Jung, 2003). A communication infrastructure is defined as a storytelling system
grounded in a communication action context (Figure 3.2).

Figure 3.2 Communication infrastructure. Adapted from Metamorphosis Project, Retrieved
June 22, 2015, from http://www.metamorph.org/research/theory/. Copyright Metamorphosis,
Annenberg School for Communication, USC.

47

The multilevel storytelling system comprises agents who have “the capacity to tell stori es
about a community, whole city, the nation, and the world for large audiences (i.e., macro level),
specific communities (i.e., meso level), and individuals (i.e., micro level)” (Wilkin, 2013, p.2).
In health communication, examples of macro-level storytellers include, but not limited to,
government organizations and programs, large-scale institutions, advertising companies, and
mainstream media (e.g., national television news, entertainment television programming,
newspapers, and radios). Meso-level storytelling agents run the gamut from community
organizations, schools, churches, libraries, to local and ethnic media (see Wilkin, 2013 for a
review). Micro-level storytellers are individuals and their interpersonal network, such as family,
friends, neighbors, colleagues, members of social groups, health practitioners, and church
leaders, and so forth.
At the community level, neighborhood storytelling network (STN) symbolizes a
triangular network of residents (micro-level), community organizations (meso-level), and
local/ethnic media (meso-level) that target a particular geographic area or, in some cases, certain
residents in that area (e.g., a particular ethnic group). According to CIT, a fully integrated STN
is one with storytellers encouraging each other to tell stories about the community. These
stories, ideally, are sustained and enlarged by community organizations and local/ethnic media
that address the shared concern or interest in the community. Meanwhile, although they are not
conceptualized as a component of STN at the community level, macro-level agents that tell
stories about the city, state, nation, or the world can shape the communicative activities within
STN and the content of its stories. For this reason, a community’s STN and macro -level
storytellers, such as mainstream media, need to be considered together as a community’s broader
storytelling system (Ball-Rokeach, et al., 2001; Matsaganis, 2015). Furthermore, although
48

residents many times discuss things that happen far from their neighborhoods (Wyatt, Katz, &
Kim, 2000), they nevertheless “carry the most burden of storytelling neighborhood” (Ball-
Rokeach, et al., 2001, p. 397), meaning that the most prevalent and pressing issues in a
community are, ideally, being discussed among residents’ social network.
Community action context (CAC), on the other hand, refers to a wide range of elements
of residential environment that facilitate or constrain the function of neighborhood storytelling
system. Physical elements pertain to natural and built environment, such as an area’s layout of
transportation grid, or the presence of communication incipient places that bring residents
together (e.g., community organizations, churches, libraries, movie theaters, farmer’s market,
grocery stores, parks, cafes). Psychological elements concern people’s perceptions regarding
how safe or comfortable it is for them to engage one another in the community (Matei, et al.,
2001). Sociocultural elements refer to the degree of ethnic and cultural similarity among
residents in a community (e.g., the level of dominance of individualism versus collectivism).
Ecological elements concern the resources available in one’s neighborhoods for people to engage
in everyday conversation and problem-solving. Lastly, technological elements can include
residents’ access to communication technologies such as broadband connections (Matei, et al.,
2001; Matsaganis, 2015).
Prior CIT research has predominantly focused on geo-ethnic communities in Los
Angeles, based on the premise that both geography and ethnicity matter (Wilkin, 2013). That is,
the same ethnic group from different neighborhoods might connect to different storytelling
systems. Similarly, different ethnic groups sharing the same geographic space might connect to
different storytelling agents. Early research mainly explores the relationship between the STN
and civic engagement. Consistent with CIT’s theoretical predictions, rich variation has been
49

found across diverse urban neighborhoods in the relationships not only between residents and
individual neighborhood storytellers, but also between individual neighborhood storytellers and
civic engagement outcomes (Ball-Rokeach, et al., 2001; Broad, et al., 2013; N.-T. N. Chen, et
al., 2013; Kim & Ball-Rokeach, 2006b).
Along this line of work, researchers developed a measure to capture the synergistic
effects of connecting to multiple neighborhood storytellers on people’s capability of everyday
problem-solving. Coined as the integrated connections to a neighborhood storytelling network
(ICSN), the measure evaluates the extent to which connecting to one storyteller increases the
likelihood of connecting to the other two storytellers. Analytically, ICSN is a summation of
three interaction terms between residents’ intensity of discussion with others about things
happening in the neighborhood, their scope of connections to local/ethnic media, and their scope
of connections to community organizations (Kim & Ball-Rokeach, 2006b).
Research employing ICSN generally reports its association with civic engagement across
racial and ethnic communities. For example, in their study of residents in Glendale, a suburb of
Los Angeles, Kim and Ball-Rokeach (2006b) found that an integrated connection to the STN
was positively associated with all three dimensions of civic engagement, including neighborhood
belonging (subjective and objective attachment to their neighborhood), collective efficacy
(perception about how willing neighbors are to solve problems shared by the neighborhood), and
civic participation (actual participation in civic activities that influence local community
decision-making and policy-making process, such as signing and circulating petitions, writing
letters to the councilmen, meeting elected officials, attending city council meetings, or
participating in a protest and so on).
50

Neighborhood Storytelling Resources and Health Outcomes
More recently, CIT research has seen a growing interest in the relationship between
connections to neighborhood storytelling resources and health outcomes in urban communities.
One rationale that explains the health effects of STN connections concerns the association
between STN and civic engagement. When residents of a neighborhood come together to solve a
shared problem, positive change can take place in the neighborhood, which, in turn, contributes
to better health outcomes (Hyyppä & Mäki, 2001; Rudd, Kirsch, & Yamamoto, 2004; Wilkin,
2013). For example, communities with high rate of crime and violence can mobilize to
transform public spaces such as parks into safe and welcoming places for residents to take part in
recreational activities and physical exercises, and for community organizations to hold health
and resource fairs (Fischer & Teutsch, 2014; Los Angeles County Department of Public Health,
2014). Likewise, communities that deal with food deserts or inadequate access to healthy food
can work together with residents and community groups to build community gardens (Feenstra,
McGrew, & Campbell, 1999; Teig et al., 2009).
The empirical base for the health effects of STN connections through enhanced level of
civic engagement, however, is mixed. For example, one study in South Los Angeles found that
residents’ connections to the STN were po sitively associated with their level of physical
exercise, but not with their consumption of fruits and vegetables (Wilkin, Katz, Hether, & Ball-
Rokeach, 2012, October). This finding echoes earlier observations from the same area, where
the limited nutritional resource available made it challenging for residents from poorer
neighborhoods to eat healthy (L. B. Lewis et al., 2005). Examined from the lens of CIT, results
like these might suggest that STN connections may be instrumental to addressing some health
issues or disparities by bringing people together, but insufficient to solving others, especially
51

when it involves overcoming structural barriers such as dearth of quality groceries or healthy
food options at the community level.
In a more recent study in South Los Angeles, Matsaganis and Wilkin (2014) directly
tested the mediating role of civic engagement (in their case collective efficacy) between STN
connections and residents’ perceived access to local resources that enhance health (e.g., public
recreation spaces, healthier food options, and healthcare services). As expected, residents who
had a stronger and integrated connection to the neighborhood storytelling network reported a
higher level of collective efficacy, which was positively related to their perceived access to
health-enhancing sources. According to the authors, the positive effect of collective efficacy
might be because residents were confident that they could count on each other to locate the
resources they need, or act on behalf of each other to raise demands (e.g., sign and circulate
petitions or hold town hall meetings). However, more conclusive explanation for the underlying
relationship between STN connections, civic engagement, and health outcomes is pending for
future research.
More relevant to present purposes, another line of reasoning to explicate the relationship
between STN connections and health effects points to the increased communication opportunity
and better access to health information that STN connections afford. As noted above, integrating
into a STN enhances individuals’ engagement with a network of community -based information
resources consisting of local interpersonal network, community organizations, and local or ethnic
media outlets. This, in turn, likely allows individuals to become more aware of local issues,
including health issues. This assertion has received empirical support. For example, Viswanath
and colleagues (2006) found that the number of community ties, a proxy measure of community
52

integration, was positively associated with people’s recall of health messages related to
cardiovascular disease.
Several CIT studies are based on this premise, though, here, the findings are mixed as
well. Combining the knowledge gap hypothesis and CIT, Kim and colleagues (2011) shed light
into how educational disparities shape disparities in obtaining health knowledge through
community-based communication resources. For a sample of Latino and African American
residents from South Los Angeles, higher level of education was associated with higher level of
integration into the neighborhood storytelling network, which was associated with higher level of
chronic disease knowledge. The authors explained that people “with more education also have
the advantage of a network of information (or storytelling) resources to help them access,
understand, and apply health information in their daily lives” (Kim, et al., 2011, pp. 407-408). In
the end, however, their findings were mixed. The association between accessing neighborhood
storytelling resources and health knowledge was established for diabetes and breast cancer
knowledge, but not for hypertension and prostate cancer knowledge. The selective impact of
STN connections could be due to the actual amount of storytelling on these health issues in one’s
neighborhood (Wilkin, 2013). For example, prior research indicates that local television and
newspapers tend to overemphasize breast cancer, but underestimate lung cancer and colon cancer
(Cohen et al., 2008; Lee, et al., 2014). Thus, one can expect an association between STN
connections and a particular health issue only when said issue is being discussed in the
neighborhood storytelling environment.
Mixed findings have been reported for the relationship between STN connections and
perceived access to healthcare and health-enhancing resources too. Drawing data from a 2002-
2003 community survey, one study found that having an integrated connection to the STN
53

predicted to greater ease to finding healthcare for new immigrant Latinos in two Los Angeles
communities (Wilkin & Ball-Rokeach, 2011). The opposite was reported in a later study of both
Latinos and African American residents in Crenshaw from South Los Angeles (Matsaganis,
2008). One plausible explanation is, again, the amount and characteristics of neighborhood
storytelling on the issue. A few months before the survey, a series of stories appeared in Los
Angeles mainstream media regarding the impending closure of a local medical center not far
from the study area (Wilkin, 2013). The reporting and the frames of the stories might have
colored the perceptions of participants. In other words, when the content of neighborhood
storytelling is shaped by that of macro-level agents (i.e., mass media), participants who connect
to neighborhood storytelling network more strongly might perceive more difficulty in accessing
healthcare (Nicholson, et al., 2008; Wilkin, 2013).
Furthermore, as shown in a more recent study of the same area, how stories about health
issues and health disparities are being told in one’s storytelling system might impact
disadvantaged residents more strongly (Matsaganis & Wilkin, 2014). Using survey data
collected from Latinos and African Americans living in both Crenshaw and Figueroa Corridor,
the authors initially found a non-significant and positive association between STN connections
and residents’ perception to accessing health-enhancing resources. This association became
significant but negative after the analysis accounted for health insurance and self-reported health
status, suggesting a potential inconsistent mediation or a suppressor effect resulting from the two
control variables (Matsaganis & Wilkin, 2014). It follows, then, that an integrated connection to
neighborhood storytelling resources might amplify the magnitude of perceived difficulty in
accessing health-enhancing resources in the community among residents lacking health
insurance or in poorer health.
54

To conclude, connecting to the neighborhood storytelling resource can have both direct
and indirect, and either positive or negative, effect on people’s health -related outcomes. A
complete understanding of such effect, therefore, must be obtained within the specific behavioral
domains, and in the particular geographic and ethnic contexts. To this end, identifying the
conditions at the community level that promote or constrain the influence of health storytelling is
important, which pertains to the critical role of communication action context (CAC).
Communication Action Context and Health Outcomes
At the community level, CAC refers to the features of residential environment that
constrain or facilitate the function and vitality of STN. Compared to STN, CAC is relatively less
explored in CIT literature, and much of the quantitative evidence is regarding civic engagement.
Focusing on two CAC factors, namely residential stability and ethnic heterogeneity, Kim and
Ball-Rokeach (2006b) demonstrated a direct effect of both CAC factors on civic engagement and
their cross-level interactive effects with connecting to an integrated neighborhood storytelling
network, or ICSN. Respondents from stable and ethnically homogenous communities displayed
a significantly higher level of neighborhood belonging and collective efficacy than their
counterparts from unstable and ethnically heterogeneous communities. Additionally, ICSN had
a significant and positive direct effect on all three dimensions of civic engagement, and, more
relevant to present purposes, a conditional effect on civic participation such that the effect of
ICSN was stronger in unstable and ethnically diverse communities. According to the authors,
the relative importance of ICSN for civic participation in disadvantaged communities manifests
the dynamic relationship between neighborhood storytelling resources and its communication
action context. In their case, stable and ethnically homogeneous communities could be better
positioned to create and share neighborhood stories, which might explain the somewhat limited,
55

though still important, effect of ICSN on civic participation in these communities. By
comparison, information about one’s neighborhood might be m ore difficult to come by and
accumulate in unstable and ethnically heterogeneous areas, which could be the reason that
residents had to rely on local communication resources more strongly for such information to
become civically engaged (Kim & Ball-Rokeach, 2006b).
In the health arena, research exploring CAC emerged only recently. Unlike prior CIT
literature, which is largely based on surveys of residents or communities with known health
issues, this burgeoning line of research has mainly been conducted with qualitative techniques
and in the context of community-based interventions to identify and reduce health disparities.
For example, in the Accountable Communities Healthy Together (ACHT) project in Atlanta,
Georgia, researchers and community members collaborated to address mental health and
depression, a shared health concern that presented low diagnosis and treatment rates due to the
stigma attached to the issue (Kreuter, et al., 2012; Wilkin, 2013). Using PhotoVoice techniques,
the project found that the larger number of vacant properties was among the top features of built
environment (e.g., a CAC factor) that adversely affected residents’ mental health and perceived
safety. Through local STN, the ACHT project mobilized the community to engage in
community clean-up and petition signing activities. These efforts persuaded the city to agree to
help non-profit organizations acquire and reuse vacant housings, beginning with areas closest to
parks and schools. As one would imagine, a more traditional route of health communication
campaign would be encouraging changes in related health beliefs and behaviors. But by
focusing on ecological factors, such as built environment, projects like ACHT have the potential
to not only address targeted health disparities in a more sustainable fashion, but also other
possible disparities related to built environment (Matsaganis, 2015; Wilkin, 2013).
56

Another two CAC features examined in this emerging work are the presence of
communication hotspots, defined as places where residents naturally engage each other and talk,
and comfort zones, defined as community institutions and business which residents feel closely
connected (Ball-Rokeach, Moran, Hether, & Frank, 2010). In the aforementioned ACHT
project, a CAC-based strategy was used to recruit residents into a healthcare assistance program
at the local health hospital (Kreuter, et al., 2012). Mini-health fairs were held in locations that
residents frequent often, such as convenience and grocery stores and church (communication
hotspots). Posters and flyers were made available at trusted community business and
organizations (comfort zones). The strategies were proven to be effective. Over two weeks, the
number of residents enrolled to the program nearly doubled. (Kreuter, et al., 2012).
In another community-based intervention that aimed to promote reproductive health
among African American women, the potential of communication hotspots and comfort zones
were explored to promote residents’ connections with community -based organizations (CBOs).
As part of the efforts to bridge the communicative disjuncture between residents and CBOs,
health events were held in a local community hotspot (the community room of an apartment
complex) for residents to meet CBO representatives. The positive feedback toward the events
indicated that, in CIT terms, the engagement between residents and CBOs increased residents’
knowledge of their CAC (e.g., availability of CBOs that offer the social service they needed). To
quote one resident who attended the event, “now I know that I can — there’s other places I can
go to get the [health] insurance like the [organization I met at the event]” (Matsaganis, et al.,
2014, p. 1503). Interactions with CBOs also seemed to extend residents’ range of c omfort zones
(e.g., CBOs they felt connected to). One resident described how the event made her feel
comfortable and willing to seek service from the CBOs in the future, “They was talking to me,
57

they was real good to me . . . And they was listening and that was making me feel good ’cause
they was listening and answering my questions the way that I wanted them answered” (p.1503).
In summary, although quantitative evidence on the interplay among CAC, STN
connections, and health outcomes is still to be seen, the important part that CAC plays in shaping
health outcomes has been documented in a small but growing number of community-based
interventions addressing health disparities in urban communities. Evidence from these studies is
informative to future research that employs a quantitative design and an ecological framework to
investigate the multilevel effects of neighborhoods on health outcomes. Taken together, research
that explores both CAC and STN demonstrates the potential of CIT in theorizing and testing the
communication process that involves multiple actors in a residential neighborhood (e.g.,
residents, community organizations, local media outlets), that unfolds at and across multiple
levels of analysis within a neighborhood, and that explains why and how where people live
influences their health (Matsaganis, 2015).
A Multilevel Communication Infrastructural Model of Normative Perceptions
Situating itself in the context of residential lives, this dissertation extends the application
of CIT to understand the sources of normative influence on health behaviors. As noted in
Chapter 2, people encounter a myriad of information that conveys norms in the process of
everyday life; and from not only their social environment, but also their symbolic and physical
environment. The communication and formation of perceived norms should be approached as a
multilevel phenomenon, both theoretically and analytically, for two reasons. First, information
sources at different levels interact with each other to shape people’s perceived norms regarding a
particular health behavior. Second, the roles of information sources are shaped by the larger
58

context within which the sources are situated. Based on a review of prior literature, the present
study proposes that CIT as a socio-ecological framework can be fruitfully applied to this inquiry.
As articulated in the preceding sections, CIT provides a theoretical and analytic
framework to conceptualize and test the communication processes and mechanisms that mediate
the impacts of neighborhoods on residents’ well -beings and health. Consisting of locally situated
interpersonal, media, and organizational communication resources, neighborhood storytelling
network (STN) has demonstrated its critical importance as an information source that affords
residents necessary information and resources for solving everyday problems, including health-
related problems. Furthermore, both quantitative and qualitative research evidence is growing
that residents’ varying degrees of STN connection s, as well as the cognitive and behavioral
outcomes resulting from those connections, are shaped by their neighborhood context, which is
characteristic of a range of CAC factors that span social, cultural, and physical dimensions.
The chosen behavioral domain of the present study is cervical cancer screening and
detection. Despite decades of deliberate efforts in research and intervention, cervical cancer
remains a major public health problem among women globally, nationally, and locally. As
shown in Figure 3.3, in the United States, women of color have disproportionately endured the
burden of cervical cancer, with Latinas and Hispanic women continuing to be the most affected
group (Centers for Disease Control and Prevention, 2014). In Los Angeles alone, Hispanic
women are nearly twice as likely to develop cervical cancer as non-Hispanic White women (Los
Angeles County Department of Public Health Office of Women’s Health, 2010) . Key strategies
identified to eliminate such disparities include improving both access to care as well as
information and communication (Green, Davis, Rivers, Buchanan, & Rivers, 2014).

59

Figure 3.3 Age adjusted cervical cancer incidence rates by race and ethnicity 2000-2012.
Reprinted from Fast Stats, In Surveillance, Epidemiology, and End Results Program (SEER),
n.d., Retrieved June 22, 2015 from http://seer.cancer.gov/faststats/selections.php?series=race.
Copyright 2015 by the Surveillance Research Program, NCI.
60

The type of perceived norms of interest is descriptive norms, or the perception regarding
the prevalence of a behavior among similar others (Lapinski & Rimal, 2005). This construct was
chosen for two reasons. First, with TPB and TRA continuing to be the most widely used
frameworks to explore normative influence on health behaviors, where subjective norms (a type
of injunctive norm) are the perceived norms construct, descriptive norms remain understudied
both in health communication research in general (Mollen, Rimal, et al., 2010) and in cervical
cancer communication in specific (Jennings-Dozier, 1999; Roncancio et al., 2015). Second,
from an intervention point of view, because injunctive norms entail a sense of social sanction
(e.g., motivation to comply to important others’ expectations or approvals), they are by definition
most responsive to the influence from social environment, and subsequently not as modifiable as
descriptive norms through communication campaign messages via media or other
communication channels (Dillard, 2011).
This dissertation explores two groups of hypotheses and research questions. The first
group addresses the relationship between the neighborhood storytelling network, neighborhood
context (operationalized as CAC), and descriptive norms regarding Pap tests. Within this line of
exploration, the primary goal is to examine the roles of Latinas’ connections to neighborhood
storytelling resources in mediating the impact of neighborhood context on the development of
descriptive norms regarding Pap tests, and on one specific health communication outcome,
namely, recall of having seen or heard information about Pap tests in the media. For a better
understanding of the interaction between the neighborhood storytelling network and its
neighborhood context, the secondary goal of this group of hypothesis testing and research
question inquiry is to examine the influence of neighborhood context on Latinas’ connections to
neighborhood storytelling resources. Next, the second group of hypotheses and research
61

questions addresses the structural relationships between Latinas’ neighborhood experience
(operationalized as residential tenure and immigration generations), connections to storytelling
resources, health communication outcomes (i.e., media recall, media attention, and discussion
about Pap tests with healthcare professionals), descriptive norms, and, ultimately, their
compliance with cervical cancer screening guidelines. The rationales for these two lines of
investigations and respective research questions and hypotheses are articulated below.
Hypotheses and Research Questions Group 1
Neighborhood Storytelling Resources and Descriptive Norms
The present study hypothesizes the roles of neighborhood storytelling resources in
shaping Latinas’ descriptive norms regarding P ap tests. Prior CIT research across diverse urban
neighborhoods demonstrates that connections to neighborhood storytelling network (STN),
which consists of residents, local/ethnic media, and community organizations, provide
information crucial for residents to solve everyday problems, including health related problems.
For example, residents who reported a higher level of STN connections were found to be more
knowledgeable about preventing and detecting diabetes and breast cancer (Kim, et al., 2011) and
to seek health information more actively (Wilkin & Ball-Rokeach, 2011).
Although the link between STN connections and perceived norms has not been examined
to date, such association is not unreasonable to posit. As noted earlier, STN represents a local
network of communication resources that provides critical information for residents to address
everyday problem and shared concerns. To the extent that STN creates and circulates stories
about certain local health issues, connecting to STN will increase residents’ opportunities to be
aware of those issues or particular health behaviors, or prime them to seek relevant information
more actively (Kim, et al., 2011; Viswanath, Randolph Steele, & Finnegan, 2006), which, in
62

turn, may shape residents’ descri ptive norms. Additionally, from an information scanning
perspective, if the same information regarding a given health behavior appears repeatedly across
a range of local communication resources that residents routinely connect to, scanning may
reinforce residents’ normative perception that the performance (or lack of it) of the behavior is
widespread (Hornik, et al., 2013).
It should be recognized that most quantitative CIT-based health research to date examines
the role of ICSN, a measure that captures the synergistic effects of connecting to multiple
neighborhood storytellers. Supportive evidence from larger norms and health communication
literature is available for proposing the roles of individual storytelling agents as information
sources that shape residents’ health-related descriptive norms as well. First, the association
between interpersonal discussion and descriptive norms has been reported both in the context of
active, deliberate information seeking, such as health communication campaigns (Chatterjee, et
al., 2009; David, et al., 2006; Frank, et al., 2012; Schuster et al., 2006), and in the context of
information scanning, where the encounter to health information is more incidental (Hornik, et
al., 2013). The latter is particularly relevant to present purposes, as it highlights the importance
of everyday conversation in constructing social reality and shaping perceived norms (Cline &
Thompson, 2003). Relevant research evidence on descriptive norms specific to Latina
population and in the field of cervical cancer screening was reported too, though to a less extent
than was the evidence on injunctive norms. For example, one study observed that only 61
percent of a sample of young Latinas living along the United States – Mexico border reported
that most young women they knew had Pap tests, speaking to a suboptimal level of descriptive
norms (Byrd, et al., 2004).
63

Second, in terms of symbolic environment (typically media environment) as a source of
normative information, a substantial body of literature documents how exposure to health
information in news and entertainment media leads to people’s inflated beliefs regarding a range
of health and health behaviors (Bleakley, et al., 2011; Chatterjee, et al., 2009; Gibbons, et al.,
2010), including cancer screening and detection (Chung, 2014; Hornik, et al., 2013; Jones, et al.,
2006). More relevant to present discussion, cumulative exposure to local television news was
more likely to produce exaggerated beliefs about cancer prevention (e.g., fatalistic beliefs) than
exposure to national television news (Lee & Niederdeppe, 2011; Niederdeppe, Fowler,
Goldstein, & Pribble, 2010). However, observations from most of these studies are based on
national representative samples, where local variance in people’s communication pattern at the
neighborhood level is consequently obscured. For example, Ball-Rokeach and Wilkin (2009)
examined the difference in Hispanic respondents’ health information seeking behaviors between
those from a national representative sample and those from a community telephone survey in Los
Angeles. Although media, such as television, were listed as one of the top health information
sources by Hispanic respondents from both samples, over 80% of the Hispanic respondents in
the community survey sample specified that their choice of television was locally or ethnically
oriented, typically in Spanish language. But this distinction between mainstream and
local/ethnic media was not made in the national sample.
In the context of cervical cancer prevention, media have been repeatedly cited as a major
source of information or campaign venue to promote cervical cancer screening and detection
among Latinas (Mann, et al., 2014). Although quantitative research that explicitly tests the
association between media connections and descriptive norms regarding Pap tests in this
population remains limited, the possible role of media connections in shaping Latinas’ beliefs
64

and perceptions has been noted. For example, one study of Latinas in Boston found that over
70% of participants believed that a cervical cancer diagnosis is fatal; irrespective of age and
years of residents in the United States, the vast majority of the sample preferred radio and
television for learning new information and reported regular use of those venues, preferably in
Spanish language (Watts, et al., 2009)
Last, the role of community organizations in promoting descriptive norms regarding
cervical cancer screening and detection is also posited. Prior research indicates that community
organizations and groups play a crucial part in positively shifting residents’ health -related beliefs
and behaviors by providing both informational and instrumental support and resources
(Campbell et al., 2007; Matsaganis, et al., 2014; Stephens, Rimal, & Flora, 2004). For example,
health communication campaigns and programs that recruit trusted local organizations for
material placement and distribution can have a broader reach to otherwise neglected populations
in a community (Kreuter, et al., 2012; Miller & Shinn, 2005; Randolph & Viswanath, 2004).
Community organizations have also proven to be an indispensable player in community
mobilizing efforts that bring together community leaders and residents to address the shared
health concerns in the community (Wilkin, 2013). Relative few studies, however, empirically
examine the association between community organization membership and health-related
perceptions and behaviors (Matsaganis, 2008; Viswanath, Randolph Steele, et al., 2006), though
models incorporating community organization membership in addition to media use and
demographic predictors was found to explain significantly greater variance in health outcomes
(Stephens, et al., 2004).
65

Neighborhood Storytelling Resources and Media Recall
Following a similar line of reasoning, the present study hypothesizes the association
between connections to neighborhood storytelling resources and Latinas’ recall of having seen or
heard information about Pap tests in the media. As discussed earlier, an effect of STN
connections on people’s awareness, knowledge, or perceptions with respect to a particular health
condition or behavior can only be expected when related stories appear in people’s storytelling
environment (Wilkin, 2013). Meanwhile, mass media have been cited as an important source of
health information in general (Ball-Rokeach & Wilkin, 2009; Hornik, et al., 2013) and cancer
information in specific (Lee, et al., 2014; Niederdeppe, et al., 2010). Thus, when such
neighborhood “stories” about cancer s creening and detection are present in media channels
within one’s broad storytelling system (including both neighborhood storytelling network and
mainstream media), those with a strong and integrated STN are expected to be more aware of
those stories, thereby higher level of recall of having ever seen or heard of such information.
This reasoning receives theoretical support from SIM as well, which posits that health
communication variables, such as exposure to health information in the media, are antecedents to
corresponding beliefs, perceptions, or behaviors (Viswanath & Emmons, 2009).
Contextual Influence of Neighborhood Environment
In the present study, the contextual influence of neighborhood environment on
neighborhood storytelling resources, media recall, and descriptive norms are explored on four
dimensions: neighborhood level linguistic isolation, ethnic heterogeneity, density of
communication resources, and density of health service providers. In line with CIT, it is posited
that neighborhood context conditions Latinas’ connections to neighborhood storytelling
resources as well as the effects of such connections on descriptive norms and media recall
66

regarding Pap tests. In other words, to the extent that the four neighborhood-level characteristics
will affect the degree of Latinas’ integration into their neighborhood storytelling network, it is
possible that these neighborhood-level characteristics are responsible for the observed difference
in descriptive norms and media recall in the study area.
Linguistic isolation. In neighborhood effect and public health literature, linguistic
isolation is usually referred to individuals’ skills to speak, read, and understand the dominant
language in a larger societal environment (Y. Park, Neckerman, Quinn, Weiss, & Rundle, 2008).
At the individual level, linguistic barriers have been found to be inversely associated with
residents’ participation in civic and social activities (N.-T. N. Chen, et al., 2013) as well as with
healthcare access and quality (De Jesus & Xiao, 2014) in diverse urban neighborhoods. At the
neighborhood or aggregate level, linguistic isolation is typically operationalized as the
percentage of population who do not speak the dominant language (e.g., English) very well
(Wang & Luo, 2005). Residents from linguistically isolated neighborhoods may experience
substantial difficulty in navigating their everyday social, symbolic, and physical environment
due to the prevalent linguistic barrier. This, in CIT terms, indicates residents’ reduc ed likelihood
of talking with neighbors and other residents, becoming a member of community organizations,
or receiving information distributed from local media that are not in their preferred language.
However, the opposite might also be true. For example, when residents from linguistically
isolated neighborhoods have an unmet information need for everyday problem-solving due to
their linguistic barriers, they might resort to linguistically friendly communication resources such
as local media or organizations that are ethnically oriented to their ethnic own group.
Ethnic heterogeneity. Neighborhood-level ethnic heterogeneity describes the extent to
which diverse ethnic groups share the same geographic area. In ethnically heterogeneous
67

neighborhoods, it is possible that the widespread difference in languages, cultural practices, or
religions hinder residents from engaging in everyday conversations with one another (Wickrama
& Bryant, 2003). From a CIT perspective, different ethnic groups from the same geographic
area might simply connect to different local media and organizational resources that target their
own ethnic group, giving rise to an ethnically-bounded and fragmented storytelling network
within each ethnic group in the area (Ball-Rokeach, et al., 2001; N.-T. N. Chen, et al., 2013;
Kim, 2003). Furthermore, empirical evidence from diverse urban neighborhoods indicates that
ethnic heterogeneity can influence the strengths of STN effects on civic engagement outcomes,
such that the positive effect of connection to an integrated STN on civic participation was
stronger for residents in ethnically heterogeneous neighborhoods than for residents in ethnically
homogenous neighborhoods (Kim & Ball-Rokeach, 2006b). According to the authors, this might
be because in ethnically diverse areas, residents have more difficulty in finding and accumulating
information and resource required to become engaged in their community, thereby relying on
local storytelling resources more heavily for that purpose.
Density of communication resources. Density of communication resources provides an
estimate of the availability of physical communication recipient places in a local community, i.e.
churches, schools, libraries, parks, and community organizations offering social and cultural
services. Basic to everyday life, such local institutional and organizational resources are critical
to the vitality of neighborhoods (Sampson, 2012; Small, Jacobs, & Massengill, 2008; Small &
McDermott, 2006). From the lens of CIT, some of those places can be classified as
communication hotspots, where residents congregate and engage each other in everyday
conversations. Others, such as community organizations, belong to the category of community
comfort zones, which residents feel connected to and trust (Ball-Rokeach, et al., 2010).
68

Neighborhoods with high density of communication resources, therefore, may be endowed with
increased communication opportunities for their residents, which, in turn, likely strengthens
residents’ connections to other neighborhood storytelling resources.
Density of healthcare providers. The presence and density of healthcare providers can,
too, affect residents’ connections to neighborhood storytell ing network. Medical professionals
and health service providers (e.g., hospitals and clinics) are credible sources for residents seeking
health information. Community clinics may also be places for patients to engage in
conversations with each other about their health or the health of their family. From this
perspective, neighborhoods characterized with low density of healthcare providers may see
reduced communication opportunities among residents when residents have to seek healthcare
outside their neighborhoods. Area-level low density of healthcare providers might also result in
residents’ stronger reliance on other available communication resources for health information,
such as their social network and local or mainstream media.
Hypotheses and Research Questions Group 1
Neighborhood storytelling resources, neighborhood context, and descriptive norms
After controlling for individual level covariates and neighborhood level factors,
H 1a: Frequency of interpersonal conversations about one’s neighborhood will have a
direct effect on descriptive norms regarding Pap tests.
H 1b: Scope of connections to local/ethnic media will have a direct effect on descriptive
norms regarding Pap tests.
H 1c: Scope of connections to community organizations will have a direct effect on
descriptive norms regarding Pap tests.
H 1d: ICSN will have a direct effect on descriptive norms regarding Pap tests.
69

RQ1a: Is there a direct effect of values of (1) neighborhood-level linguistic isolation, (2)
ethnic heterogeneity, (3) density of communication resources, and (4) density of health
service providers on descriptive norms regarding Pap tests, after accounting for
individual-level variables?
RQ1b: Will the marginal effect of ICSN on descriptive norms vary depending on the
values of (1) neighborhood-level linguistic isolation, (2) ethnic heterogeneity, (3) density
of communication resources, and (4) density of health service providers on descriptive
norms regarding Pap tests?
Neighborhood storytelling resources, neighborhood context, and media recall
After controlling for individual level covariates and neighborhood level factors,
H 1e: Frequency of interpersonal conversations about one’s neighborhood will have a
direct effect on media recall regarding Pap tests.
H 1f: Scope of connections to local/ethnic media will have a direct effect on media recall
regarding Pap tests.
H 1g: Scope of connections to community organizations will have a direct effect on
media recall regarding Pap tests.
H 1h: ICSN will have a direct effect on recall of having seen or heard information about
Pap tests in the media in the past 30 days, after controlling for individual level covariates
and neighborhood level factors.
RQ1c: Is there a direct effect of values of (1) neighborhood-level linguistic isolation, (2)
ethnic heterogeneity, (3) density of communication resources, and (4) density of health
service providers on recall of having seen or heard information about Pap tests in the
media?
70

RQ1d: Will the marginal effect of ICSN on media recall vary depending on the values of
values of (1) neighborhood-level linguistic isolation, (2) ethnic heterogeneity, (3) density
of communication resources, and (4) density of health service providers on descriptive
norms regarding Pap tests?
Neighborhood context and storytelling resources
H1i: Neighborhood-level linguistic isolation will have a direct effect on connections to
(1) individual neighborhood storytelling resources, (2) English media, and (3) ICSN.
H1j: Neighborhood-level ethnic heterogeneity will have a direct effect on connections to
(1) individual neighborhood storytelling resources, (2) English media, and (3) ICSN.
H1k: Neighborhood-level density of communication resources will have a direct effect
on connections to (1) individual neighborhood storytelling resources, (2) English media,
and (3) ICSN.
H1l: Neighborhood-level density of health service providers will have a direct effect on
connections to (1) individual neighborhood storytelling resources, (2) English media, and
(3) ICSN.
Hypotheses and Research Questions Group 2
The second group of hypotheses and research questions examine the structural
relationship between Latinas’ neighborhood experience, connections to storytel ling resources,
health communication outcomes, descriptive norms, and, lastly, their compliance with cervical
cancer screening guidelines. The purpose is to obtain a more in-depth understanding of how
connecting to a broad storytelling system (including macro-, meso-, and micro-level storytellers)
impacts Latina’s perceived norms regarding Pap tests and, subsequently, their screening
behaviors. Figure 3.4 illustrates the storytelling paths to descriptive norms and screening
71

compliance. In addition to the hypothesized paths that are described in more details below, the
paths from health insurance status as a control variable to storytelling variables, descriptive
norms and screening compliance are also illustrated. Within this model, connections to macro-
level storytellers are operationalized as Latinas’ connections to mainstream English language
media, whereas connections to neighborhood storytellers are explored via Latinas’ integrated
connections to their neighborhood storytelling network (ICSN). It must be noted that the
linearity of the model is determined by the requirement of model testing, rather than by
theoretical underpinnings. For example, bidirectional paths between macro-level and
neighborhood-level storytelling resources are plausible within CIT, though rarely the case in
reality (Ball-Rokeach, et al., 2001).
Neighborhood experience as structural condition
Two neighborhood experience variables considered in the current discussion are
immigration generation and residential tenure. According to CIT, residents who are old
immigrants or who have lived in their neighborhoods for a longer period of time are more
inclined to engage in storytelling their neighborhoods, and to have more opportunity to do so
(e.g., established social network in the area, knowledgeable about available local media or
community organizations). Empirical evidence gleaned from diverse urban neighborhoods
generally supports this anticipation (Ball-Rokeach, et al., 2001), though the association between
immigration generation and neighborhood storytelling resources is less conclusive (N.-T. N.
Chen, et al., 2013). Additionally, it has been found that Latinas who are recent immigrants
experience disproportionately more barriers to cervical cancer screening (Behbakht, Lynch, Teal,
Degeest, & Massad, 2004; Mann, et al., 2014).
72

Figure 3.4 Conceptual model of the structural relations between neighborhood experience, storytelling resources, health
communication outcomes, descriptive norms, and compliance with cervical cancer screening guidelines
73

Based on prior research, this study posits the following. First, it hypothesizes a direct
effect of immigration generation (first generation vs. second generation and higher) on
compliance with cervical cancer screening guidelines. Second, it hypothesizes that immigration
generation will condition the communication mechanisms of interest by influencing Latinas’
connections to both ICSN and English language media. Last, it inquires whether there exists a
difference in the structural relations among the variables of interest between Latinas who have
lived in their area for a longer period of time and Latinas whose years of residence in their areas
are shorter.
Indirect effects of storytelling resources
The present study hypothesizes that the effects of ICSN and English media on
descriptive norms will be mediated by three health communication outcomes specifically related
to Pap tests: 1) recall of having seen or heard information about Pap tests in the media in the past
30 days, 2) attention to information about Pap tests in the media, and 3) discussion about Pap
tests with a doctor, nurse, or other health service providers. These proposed paths are consistent
with SIM, which conceives health communication outcomes as more proximate predictors to
people’s beliefs and behaviors (Viswanath & Emmons, 2009). The hypothesized association
between connections to neighborhood storytelling resources and media recall was discussed
above. In short, one can only expect an effect of connections to storytelling resources on
Latinas’ perceived norms related to Pap tests when relevant information is generated and
disseminated by the storytellers, such as local/ethnic media. When such information is present,
from the perspective of information screening, incidental encounters with the information in the
media through regular television viewing, newspaper reading, or radio listening may generate
74

some level of recall later, provided sufficient attention being paid to the information (Hornik, et
al., 2013; Niederdeppe et al., 2007).
On the other hand, from the perspective of active information seeking, such recall might
be more likely to happen among individuals with higher level of ICSN, which describes a
person’s structural position in her communication environment where three neighborhood
storytellers encourage each other to “storytell” the shared concerns in the neighborhood. In this
case, for example, it is possible that Latinas first hear about information or services about Pap
tests from routine interactions with people in their neighborhoods or from local organizations,
which in turn leads them to seek more information in local media such as health programming on
televisions or health columns in newspapers. The hypothesized relationship between storytelling
resources and media attention follows a similar line of reasoning. Further, a strong and
integrated connection to a neighborhood storytelling network can not only enable Latinas to be
more aware of local issues, including health-related issues such as Pap tests, but also prime them
to seek additional information more actively. The latter scenario is captured by hypothesizing
the effect of STN connection on discussion about Pap tests with healthcare professionals.
Moreover, because mainstream English media has also been cited as a major source of
cancer information (Viswanath, 2005; Viswanath, Breen, et al., 2006), the effect of English
media connection on media recall, media attention, and discussion with healthcare professionals
is similarly proposed. It is also hypothesized that the effect of English language media will also
be, at least partially, indirect through ICSN, based on the theoretical assumption that people are
more likely to attend to or engage in local storytelling when mainstream media stories touch
upon local concerns (Ball-Rokeach, et al., 2001), such as those related to the barriers to getting a
Pap test for a certain ethnic group on in a particular area.
75

Direct and indirect effects of health communication outcomes
Scholars in media effects suggest that both exposure (e.g., the ability to recall media
message) and attention to the messages are necessary for a message to have an effect on beliefs
and behaviors (McGuire, Rice, & Atkin, 2001; Slater & Rasinski, 2005; Stevens & Hornik,
2014). To model the relationship between exposure and attention, a linear relationship between
the two constructs is suggested for an information-rich stimulus, such as cancer information,
where attention at least partially mediates the effect of exposure (McGuire, et al., 2001).
Following this line of literature, the current study hypothesizes an indirect effect of media recall,
such that the effect of media recall will be mediated by media attention, which in turn will
influence descriptive norms.
Additionally, it is posited that media attention will have an indirect effect through
discussion with healthcare professionals, in order to account for the potential pathway from
information exposure in a more or less incidental manner to a more active, deliberate effort of
information seeking. Discussion with healthcare professionals is included here due to the
established importance of patient-provider communication in cancer screening, both as
interpreters of media messages and as credible sources of information and advice that patients
look to (Viswanath, 2005).
Lastly, research among underserved ethnic minority women, including Latinas, reports
that advice about Pap tests from healthcare providers is particularly critical for women to stay
up-to-date with cervical cancer screening (Bazargan, et al., 2004; Mann, et al., 2014). For these
reasons, the current study hypothesizes the direct effects of discussion with healthcare
professionals on descriptive norms and on compliance with cervical cancer screening guidelines.
76

Direct effects of descriptive norms
Following the theoretical predictions of IMBP (Fishbein & Cappella, 2006; Frank, et al.,
2012; Moran, et al., 2013), this study posits that descriptive norms regarding Pap tests will have
a direct effect on screening compliance.
Hypotheses and Research Questions Group 2
Effects of immigration generation
H2a: Immigration generation will have a direct effect on screening compliance.
H2b: Immigration generation will have a direct effect on (1) ICSN and (2) English
language media connections.
Effects of storytelling resources
H2c: ICSN will have a direct effect on (1) media recall regarding Pap tests, (2) media
attention to content about Pap tests in the media, and (3) discussion about Pap tests with
healthcare professionals.
H2d: Connections to English language media will have a direct effect on (1) media
recall, (2) media attention, and (3) discussion with healthcare professionals.
H2e: Connections to English language media will have an indirect effect on ICSN.
Effects of health communication outcomes
H2f: Media recall will have a positive effect on attention such that the effect of media
recall will be mediated by media attention.
H2g: Attention will have a positive effect on discussion such that the effect of media
attention will be partially mediated by discussion.
H2h: Attention will have a direct effect on descriptive norms regarding (1) never having
a Pap test, (2) having had regular Pap tests, and (3) having had abnormal test results.
77

H2i: Discussion will have a direct effect on descriptive norms regarding (1) never having
a Pap test, (2) having had regular Pap tests, and (3) having had abnormal test results.
H2j: Discussion will have a direct effect on compliance with screening guidelines.
Effects of descriptive norms
H2k: Descriptive norms regarding (1) never having a Pap test, (2) having had regular Pap
tests, and (3) having had abnormal test results will have a direct effect on compliance
with cervical cancer screening guidelines.
Difference in structural relations by participants’ year of residence
RQ2: Will there be different structural relations among storytelling resources, health
communication outcomes, descriptive norms, and screening compliance between participants
who have lived in their neighborhoods for a longer period of time and participants who have
lived in their neighborhoods for a shorter period of time?

78

CHAPTER 4 : METHODOLOGY

Data Collection
Individual-Level Data Collection
Data used in this dissertation were part of the Multilevel Study (R01CA155326 -
Murphy/Ball-Rokeach), a large-scale multi-component research grant that seeks to understand
the barriers and conduits to cervical cancer prevention, detection, and treatment at individual,
interpersonal, and community level among Hispanic women in Los Angeles. The main
component of the Multilevel Study involved a survey of knowledge, beliefs, attitudes, and
behaviors related to cervical cancer as well as everyday communication behaviors. Eligible
participants for the survey were Hispanic or Latina females, between the ages of 21 to 50, not
currently pregnant, not having undergone a hysterectomy, and with no pre-existing
cervical/ovarian/uterine/vaginal/vulvar cancer. The study population consisted of two subgroups
of women: women who were following the cervical cancer screening guidelines and had
received a Pap test in the past 3 years (hereafter referred to as “compliant” women), and women
who were not following the screening guideline either because they had never had a Pap test or
they had not received one in over 3 years (hereafter referred to as “noncompliant” women).
Recruitment occurred at clinics (including LAC+USC Women’s Clinic and two
participating community clinics) and local public spaces from April 2012 to December 2013.
The compliant group was recruited by having research staff make a phone call to potentially
eligible patients with appointments at the clinics the night before the appointments, or approach
and/or hand out a flyer to potentially qualified patients who were waiting to receive a Pap test in
a clinic’s lobby. The flyer invited potential participants to a study on Latina women’s health. It
79

briefly described the procedure of the survey, where an interviewer would read questions to the
participant while the participant waits for her appointment. It also stated that participants of the
study will receive a gift card worth $20, and that women’s decision to participate will not affect
the care they will receive at any future appointments. Likewise, the noncompliant group was
recruited from potentially eligible women who were at the clinics, but were not there to receive
care for themselves, or from local public spaces (such as parks, bus stops, laundromats, or
community events). In both recruitment scenarios, research staff determined the potential
participants’ eligibility, and introduced the purpose of the study to those who satisfied the
recruitment criteria. Women who were interested in participation proceeded with informed
consent, where a staff member fully informed women the details including the length of the
survey, the benefits and risks involved in participation, and the rights of participants. Upon
obtaining consent, a staff member administered the Multilevel survey in participants’ preferred
language (i.e., English or Spanish). The survey lasted about 45 to 60 minutes. The final sample
consisted of 1632 participants, including 1121 compliant participants and 511 noncompliant
participants (see more details in Appendix A).
Neighborhood-level data collection
During the survey, participants were asked to provide their residential address
information specific to street level. The residential addresses were imported into an online
geographic information processing platform
4
for batch geocoding, followed by manual geocode
correction to ensure the desired quality of address information (Goldberg, 2013). Defined by the
U.S. Census Bureau, ZIP Code Tabulation Areas (ZCTA) are generalized areal representations
of United States Postal Service (USPS) ZIP Code service areas (www.census.gov). A centroid is
the geometric center of a spatial feature, here a ZCTA unit. Eventually, addresses for 1616

4
http://geoservices.tamu.edu/Services/Geocode/
80

participants were geocoded to a resolution more accurate than ZCTA centroid. Of those, 1150
(71%) were within the 25 neighborhood clusters defined by the Multilevel Project
5
. Participants
whose addresses could not be geocoded to exact parcel and street segment centroids during
manual correction were excluded (N = 16). Moreover, because the multilevel modeling analysis
of the present study involved using population-adjusted density measures, where low populations
produce unreliable estimates of the measures, one cluster that contains a census tract with
population below 200 was excluded from subsequent analysis. This resulted in 24 clusters that
are composed of 414 census tracts, which represented the homes of 1116 Multilevel survey
participants. These 414 tracts were used as the geographic units for which census data were
obtained from the 2010 Census and American Community Survey 5-Year Estimate 2007-2011.
Measures
Individual-Level Variables
Perceived Norms
Descriptive norms. Descriptive norms regarding three aspects of cervical cancer
screening with a Pap test were investigated. Participants were prompted to imagine 100 Latinas
like them, and then asked how many of these 100 women they think 1) have never had a Pap test,
2) have Pap tests at least every 3 years, and 3) have ever had abnormal Pap test results.
Health communication Outcomes Variables
Media recall. Participants were asked on a 4-point scale, “ In the past 30 days, how often
have you seen or heard information about the Pap test in the media” (1 = not at all, 2 = about
once, 3 = about once per week, 4 = more than once per week).

5
See Appendix B for details on the procedure of defining neighborhood clusters.
81

Media attention. On a 10-point scale, participants were asked to indicate the amount of
attention they had paid to information about Pap test in the media (from 1 = no attention at all to
10 = a great deal of attention).
Discussion with healthcare providers. Participants were asked if they had discussed
Pap tests with a doctor, nurse or other health care providers (1 = yes, 0 = no).
Neighborhood storytelling variables
Intensity of interpersonal neighborhood storytelling (INS). On a 10-point scale,
where 1 represents “never” and 10 “all the time”, participants’ frequency of neighborhood
discussion was measured by their response to the question, “How often do you have discussions
with other people about things happening in your neighborhood?”
Scope of connections to community organizations (OC). To measure the scope of
connections to local community organizations, this study first asked participants to indicate if
anyone in their family participated in any of the following categories of community
organizations: 1) sport or recreational groups; 2) cultural, ethnic or religious groups; 3)
neighborhood group or homeowner’s association; 4) political or educational groups; and 5) any
other organizations or groups. An affirmative answer for each type of organization was coded as
1 and negative answer as 0. Next, following the lead of Kim (2003), this study assigned “1” to
participants who were not a member in a religious organization, but later during the survey
reported going to a religious service regularly. These scores were summed to create a synthetic
variable that reflects participants’ scope of connections to local organizations.
Scope of connections to local/ethnic media (LC). Participants were asked to assess
approximately how many hours they spent last week on 1) reading newspapers that are produced
for their area or for Latinos; 2) listening to radio stations that target their area or are produced for
82

Latinos; and 3) watching television and cable channels that target their area or are produced for
Latinos (from 0 = none to 6 = 5 or more hours). Response to each type of local/ethnic media was
dichotomized such that “1” represents at least a couple minutes being spent on a given type of
local/ethnic media. These values were then summed across all three types of media to create a
scope variable that captures the breadth of a woman’s connections to local/ethnic media.
Scope of connections to major English language media. A parallel procedure was
used to gauge participants’ sco pe of connections to major English language media by asking
about the approximate number of hours they spent last week on 1) reading major English
language newspapers (e.g., Los Angeles Times); 2) listening to major English language radio
stations (e.g., KPCC, KNX or KIIS FM); and 3) watching major English language cable and
television channels, such as ABC, NBC or CBS (0 = none, 6 = 5 or more hours). Responses to
each type of media were dichotomized such that “1” represents any time spent on a given type of
major English language media. These values were then summed across all three types of media
to generate a composite score.
Integrated Connections to Storytelling Network (ICSN). ICSN was calculated as a
weighted summation of three interaction terms between the following variables: frequency of
interpersonal discussion about things happening in one’s neighborhood; the scope of connections
to local/ethnic media; and the scope of connections to community organizations. The equation
with which to create the ICSN measure has been reported elsewhere (Kim, 2003; Kim & Ball-
Rokeach, 2006b), and is shown as below,

where INS is the intensity of interpersonal neighborhood storytelling, LC is the scope of
connections to local/ethnic media, and OC is the scope of connections to local organizations.
83

Consistent with the conceptualization of ICSN, this measure reflects a person’s structural
position in her communication environment, rather than the connections to individual storytelling
agents (Kim & Ball-Rokeach, 2006b). Consequently, a person will receive a high score of ICSN
only when she has strong connections with all three neighborhood storytelling resources.
Individual-level covariates
Age was calculated by subtracting participants’ ye ar of birth from the year of survey.
Education was gauged by five categories: “8th grade or less”; “some high school
(including 9th through 11th grades)”; “high school degree (including those with vocational/trade
school)”; “some college (including junior and community college)” and “college degree or
professional/graduate school”.
Residential tenure was assessed by having participants indicate the number of years that
they had lived in their neighborhood.
Immigration generation. Immigration generation was assessed in three steps.
Participants were asked if they were born in the United States. Those who were born in the
United States were then asked if their parents and any of their grandparents were born in the
United States. Following the definition of the U.S. Census Bureau (U.S. Census Bureau, 2013a),
participants who were not born in the United States were coded as 1 (i.e., first generation); those
who were born in the United States, and had at least one foreign-born parent, were coded as 2
(i.e., second generation); and the rest of the participants who had two U.S. native parents were
coded as 3 (i.e., third generation or more).
English linguistic proficiency. Participants’ English linguistic proficiency was
measured with a previously validated 12-item scale that seeks to assess Hispanics’ ability to
speak, read, and write in Spanish and English, with each linguistic domain being assessed by six
84

questions (Marin & Gamba, 1996). Example items included, “How well do you speak
English?”, “How well do you understand television programs in English?”, and “How well do
you understand radio programs in English”. All questions were on a 4 -point scale (1 = very well,
2 = poor, 3 = poorly, 4 = very poorly). The Cronbach’s alphas for the subscales of Engli sh and
Spanish linguistic proficiency were at .99 and .96 for the study sample, respectively. For both
English and Spanish domains, the mean of the respective six items served as the measure of
linguistic proficiency. The composite scores were then reverse coded such that higher values
indicate better linguistic fluency. A closer look at the Spanish subscale, however, revealed a
considerable and negative skew, where over 90% of the study sample’s Spanish linguistic
proficiency was anchored at either “ver y well” or “well”. Therefore, this dissertation included
only the subscale on English linguistic proficiency for subsequent analysis.
Insurance was measured by asking if participants had any kind of healthcare coverage (0
= none, 1 = yes).
Compliance with Pap test screening guidelines was evaluated in two steps. First,
participants were asked, “Have you ever had a Pap smear, Pap test or Papanicolaou where a
nurse or doctor takes a sample of cells from your cervix?” For those who responded “yes” to the
question, they were then asked to indicate in months when they had their most recent Pap test.
Participants were coded as “compliant” if they had their last Pap test within 36 months (i.e., 3
years) of the survey. Participants were coded as “noncompliant ” if they had never had a Pap test,
or had a Pap test in over 36 months.
Neighborhood-Level Variables
Linguistic isolation was operationalized as the neighborhood-level percentage of
population with limited English ability. The measure was derived from the census tract-level
85

estimate for population aged 5 years and older, available in the American Community (ACS)
Survey 2007-2011 5-Year Estimate. ACS assesses people’s self -reported ability to speak
English on a four-point scale: “very well”, “well”, “not well” and “not at all”. By the definition
of the U.S. Census Bureau, individuals whose ability to speak English is less than “very well”
can be considered as having some difficulty with English, or “linguistically isolated” (U.S.
Census Bureau, 2013b). To calculate the neighborhood-level linguistic isolation variable for the
current study, census tract-level “linguistically isolated” population were first summed up to the
cluster level for each neighborhood cluster, and then divided by the total population aged 5 years
and older for that cluster.
1

Ethnic heterogeneity was calculated for each neighborhood cluster using the formula
below (Alesina & Ferrara, 2000), where S
i
denotes the proportion of ethnic group i in an area,
Ethnic Heterogeneity=1

Following this formula, a neighborhood cluster gets a score of 0 if only one ethnic group
presents in the area. By comparison, the ethnic heterogeneity score approaches 1, if a growing
number of ethnic groups share the same geographic area evenly in terms of percentage of
population. The current study first obtained population estimates from the 2010 Census at the
level of census tracts for Latinos, Whites, Blacks or African Americans, Asians, and others
(including American Indians and Alaska Native, Native Hawaiian and Other Pacific Islander,
other races alone, and two or more races). For each racial and ethnic group (e.g., Latinos),
census tract-level population estimates were summed for each neighborhood cluster, and divided
by the total population of the cluster to obtain the proportion of the ethnic group for the area.
These proportions were then used to calculate ethnic heterogeneity for the neighborhood cluster.
86

Density of communication resources. To determine the type and number of
communication resources, this study began with a full inventory available at the Los Angeles
County GIS Data Portal
6
. The inventory contains information for over 73,000 locations, which
are broken into more than 270 unique categories (e.g., arts and recreation, education, community
groups, health and mental health, municipal services, social services). For the purpose of this
study, churches, schools, libraries, community organizations and social services, and parks and
recreation centers were identified as categories of interest. Also, in the inventory, a single
geographic location may get to be grouped into more than one category, depending on the
number of services offered at the site. For example, a community organization that provides
both job training and food assistance would have two separate data entries in the database.
Because the interest of this study lies in the density of actual communication resources, duplicate
location information was consolidated to ensure that only unique geographic point data were
entered into the subsequent analysis. Last, in order to account for the spatial influence of
resources that are beyond the study area’s boundaries, but reachable within a reasonable walking
distance, this study selected locations that fall within 0.5 mile buffer of the analysis area (Smiley
et al., 2010). This resulted in 2150 unique geographic locations for subsequent analysis.
The density of communication resources was estimated with a fixed kernel smoothing
method using ArcGIS 10.2. Compared to non-spatial measures of density, which typically
divide raw counts of locations by area or population, kernel estimates generate a continuous
density layer over the entire study area. With this approach, all resources within a search radius
are allowed to exert some influence, but resources near the center of a catchment area are given
more weight than those in the periphery (Guagliardo, 2004). By doing so, kernel estimates take

6
Available at: http://egis3.lacounty.gov/dataportal/2014/07/07/locationspoints-of-interest-lms-data/
87

into account the spatial pattern of locations, thereby shedding light on spatial accessibility rather
than mere availability.
Computational details about kernel estimates are beyond the scope of this chapter (see
Guagliardo, 2004, for more details on kernel density method). To put it succinctly, a kernel
density layer consists of small grid cells (100 square meters in this case). The communication
resource density was estimated for all cells using a 0.5 mile radius. Next, because population
density affects communication resource density, a population density layer on the basis of census
block group centroid was generated in a similar fashion, using the same cell size and extent, but
with a one-mile radius as suggested in previous literature (Guagliardo, Ronzio, Cheung, Chacko,
& Joseph, 2004). Then, the raw density layer of resources was divided by the population density
layer and multiplied by 10,000 to achieve a layer of resources-to-population ratio in the units of
communication resources per 10,000 population members. As noted earlier, because low
populations produce unreliable estimates of ratios, one cluster was removed from subsequent
analysis as it contains a census tract with population below 200. The resulting area for analysis
consisted of 24 neighborhood clusters
2
, which were then overlaid with the ratio layer. Finally,
population adjusted density of communication resources was calculated for each neighborhood
cluster in the form of mean resources-to-population ratio.
Density of health service providers. Using a similar procedure, this study selected from
the geographic database 301 unique locations for health service providers within a 0.5 mile
buffer of the analysis area. The kernel density for health service providers was estimated using
same cell size and extent. The searching radius was set as 3 miles, a distance beyond which
providers’ influence was found to be negligible (Guagliardo, 2004). The resultant provider
88

density layer was then divided by the population density layer, and multiplied by 10,000
population members.
Analysis
Hierarchical Linear Modeling (HLM)
General HLM Modeling
In order to account for the possible interdependence of observations in the dataset,
hypotheses and research questions regarding the relationships among neighborhood
characteristics, storytelling resources, and the dependent variables were examined using
hierarchical linear modeling (HLM). HLM is appropriate for data with nested structure (e.g.,
students in specific schools, patients in specific hospitals, or residents in particular
neighborhoods), where the assumption of independence of observations needed in traditional
statistical techniques is frequently violated (Hayes, 2006; H. S. Park, et al., 2008; Raudenbush &
Bryk, 2002). Conventional statistical models such as ordinary least squares regression (OLS)
and analysis of variance (ANOVA) require that each observation is randomly obtained from
some larger population, a procedure that produces statistical independence between observations.
This assumption is open to question when the structure of data involves lower-level units nested
within higher-level clusters (e.g., a student in a particular classroom within a particular school),
because units within the same cluster are typically more like each other than they are like units in
other clusters (Slater, et al., 2006).
When clustering is present, it is informative for a researcher to assess the degree of
nonindependence across lower-level units by calculating intraclass correlation coefficient
(Raudenbush & Bryk, 2002). Intraclass correlation coefficient measures the proportion of total
variance in a lower-level dependent variable (e.g., perceived social norms of residents) that exists
89

between higher-level units (e.g., neighborhood clusters). If intraclass correlation coefficient is
statistically significant or substantially meaningful
3
, analysis that fails to address data clustering
has multiple disadvantages. For instance, analyzing health outcomes for 3000 residents recruited
from 10 neighborhoods only at the individual level discounts the commonalities among residents
who live in the same neighborhood. Analysis conducted only at the neighborhood level won’t be
a correct course of action either. With all individual variables aggregated to the neighborhood
level, models performed only at the neighborhood level will result in considerable reduction in
statistical power, and probably obscure important effects of individual variables on residents’
health outcomes (Slater, et al., 2006).
Largely based on single-level analysis, traditional statistical techniques have proven
limited in accommodating nested data structure. That is, when variables at both lower and
higher levels are of interest, including variables in one single-level analytical model, such as
OLS, implies the use of either data disaggregation (e.g., assigning neighborhood attributes to
individual residents) or aggregation (e.g., combining information about individual residents to
represent attributes of a neighborhood). Generalizing results obtained at a lower level to the
higher level incurs what is termed as individual fallacy (Hox, 2010). Drawing inference at a
lower level solely based on data aggregated to the higher level leads to ecological fallacy (W. S.
Robinson, 2009).
In contrast, the HLM technique allows researchers to examine variables at two or more
levels simultaneously, rather than to restricting themselves to one level of analysis without
theoretical justification. Compared to a conventional regression model, the intercept and the
slope in a HLM model can both vary randomly. The Level 1 and Level 2 equations for a two-
level model are presented below (Raudenbush & Bryk, 2002),
90

Level 1: ij ij j j ij r X Y    1 0  
Level 2:
j
j j u W
0 01 00
0      

j
j j u W
1 11 10
1      
where,
i indicates Level-1unit;
j indicates Level 2 units;
Y
ij
is the value of a dependent variable for Level 1 unit i within Level 2 unit j;
X
ij
is the value of a Level 1 independent variable for unit i in Level 2 unit j;
W
j
is the value of a Level 2 independent variable for Level 2 unit j;
β
0j
is the intercept of the Level 1 model;
β
1j
is the slope of the Level 1 model;
γ
00
is the mean of the dependent variable, controlling for W;
γ
01
is the Level 2 mean of the effects on the Level 1 intercept;
γ
10
is the average slope, controlling for W;
γ
11
is the Level 2 mean of the effects on the Level 1 slope;
r
ij
is the variance of the dependent variable for Level 1 unit i within Level 2 unit j;
u
0j
is the unique effect of Level 2 unit j on the mean of the dependent variable, holding W
j

constant; and
u
1j
is the unique effect of Level 2 unit j on the average slope, holding W
j
constant.
Combining the Level 1 and Level 2 models leads to the following equation,
ij
j
ij
j
ij j ij j ij r u X u X W X W Y       
0 1 11 10 01 00
   
The combined model shows why single-level analysis with traditional statistical
techniques will be inappropriate when nested data structure is involved. The primary assumption
91

of a conventional regression model, for example, is that the random errors need to be normally
distributed, independent, and with constant variance (Raudenbush & Bryk, 2002). With data
clustering, violation of all these assumptions can be inevitable. In the combined model above,
the random error is ( i
j
ij
j
r u X u  
0 1
). This random error is no longer independent, but contingent
upon the values of u
0j
and u
1j
, both of which are common to every Level 1 unit i within Level 2
unit j. Also, because the random error additionally depends upon the value of X
ij
, which varies
across all Level 1 units, it has unequal variance too. For example, for woman i living in
neighborhood j, assume that Y
ij
stands for her attention to information about Pap tests in the
media, and X
ij
for her connections local/ethnic media. Consequently, u
0j
denotes the unique
effect of neighborhood j on the mean of attention for all women in neighborhood j, and u
1j
denotes the unique effect of neighborhood j on the effect (i.e., slope) of local/ethnic media
connections (X) on attention (Y) in the neighborhood. As a result, the random error (i.e.,
i
j
ij
j
r u X u  
0 1
) is now interdependent, because u
0j
and u
1j
are shared by all women in
neighborhood j. The random error term does not have equal variance either, because the level of
connections to local/ethnic media (X
ij
)

varies across all women.
Depending on the research questions of interest, a HLM model can have a randomly
varying intercept and/or slopes. A cross-level model is in order if a researcher is interested in
testing the interaction between a lower-level predictor and a higher-level predictor. The
following text introduces two HLM models used in this dissertation: random intercept model and
cross-level model.
Random Intercept Model
A random intercept model is a HLM model where only the intercept coefficient is set as
random. The simplest random intercept model, which is also termed as “fully unconditional
92

model”, is an intercept -only model without individual predictors at either lower or higher level.
Equivalent to a one-way analysis of variance model with random effects, the model tests whether
there is a significant difference among Level 2 units (e.g., neighborhoods) for a Level 1 outcome
variable (e.g., descriptive norms). Because of this, a fully unconditional model is recommended
as the preliminary step in a multilevel modeling analysis (Raudenbush & Bryk, 2002). The
Level 1 and Level 2 equations are shown as below,
Level 1: ij j ij r Y   0 
Level 2:
j
j u
0 00
0    
Combined model: ij
j
ij r u Y   
0 00

In the context of this dissertation, i and j indicate participant i and neighborhood j respectively.
In the Level 1 model, β
0j
serves the intercept of the Level 1 model, and r
ij
represents within-
neighborhood variance of the dependent variable (e.g., descriptive norms) with a mean of zero
and a variance of σ
2
. In the Level 2 model, γ
00
denotes the grand mean of the dependent variable
in the study population, whereas u
0j
is the unique effect of neighborhood cluster j on the
dependent variable with a mean of zero and a variance τ
00
. Taken together, as shown in the
combined model, u
0j
is the random effect at the neighborhood level, whereas r
ij
is the random
effect at the individual level. The decomposition of variance within- and between-neighborhood
for the dependent variable is written as follows,
Variance (Y
ij
) = Variance (γ
00
+u
0j
+ r
ij
) = Variance (u
0j
+ r
ij
)
which is equivalent to the one-way ANOVA, as the total variance is composed of between-group
variance and within-group variance. Based on the decomposition of variance, intraclass
correlation coefficient can be calculated using the following formula to provide information on
93

the proportion of variance in the dependent variable that exists at the neighborhood level as
opposed to at the individual level,
 
2
00
00
 




where τ
00
is the neighborhood-level variance and σ
2
is the individual-level variance.
Often, a fully unconditional model itself is not of great research interest. One step further
is to include Level 1 and/or Level 2 predictors into the fully unconditional model, while fixing
the Level 2 random effects of the slope parameters (u
1j
) to zero. This results in a model
equivalent to a one-way ANCOVA model with random effects (Raudenbush & Bryk, 2002). To
illustrate, the Level 1 and Level 2 models are shown below that include one Level 1 predictor (X)
only, with
j
u
1
set to zero,
Level 1: ij ij j j ij r X Y    1 0  
Level 2:
j
j u
0 00
0    

10
1    j
Combined model: ij
j
ij ij r u X Y    
0 10 00
 
In these models, β
1j
is the Level 1 slope parameter, whose value denotes the change in mean
score for the dependent variable (e.g., descriptive norms) with a one unit increase in the Level
1predictor (e.g., ICSN) in neighborhood cluster j. γ
10
represents the average regression slope
across neighborhood clusters.
In the same fashion, one can proceed with including Level 2 predictors (W),
Level 1: ij ij j j ij r X Y    1 0  
Level 2:
j
j j u W
0 01 00
0      
94

10
1    j
Combined model: ij
j
ij j ij r u X W Y     
0 10 01 00
  
Note that u
0j
is still the unique effect of neighborhood cluster j on the dependent variable, but
now conditional on the Level 2 predictor.
Cross-Level Model
A cross-level model allows for the investigation of whether the effects of Level 1
predictors on the dependent variable differ depending on the values of Level 2 predictors. Built
upon the previous random intercept model, a cross-level model with one interaction term is
presented as follows,
Level 1: ij ij j j ij r X Y    1 0  
Level 2:
j
j j u W
0 01 00
0      
j j W
11 10
1     
Combined model: ij
j
ij j ij j ij r u X W X W Y      
0 11 10 01 00
   
where γ
11
is the mean difference in regression slope depending on the values of the Level 2
predictor (e.g., mean difference in the effect of ICSN on descriptive norms between
neighborhood clusters with high level of linguistic isolation and those with low level of linguistic
isolation).
Analysis Strategy
All HLM models were estimated with maximum likelihood procedures using Stata 13
(Rabe-Hesketh & Skrondal, 2012; Raudenbush & Bryk, 2002). For all dependent variables, an
unconditional model was first analyzed in order to compute intraclass correlation coefficient.
For dependent variables that exhibited meaningful difference among neighborhood clusters,
analysis proceeded with two-level random intercept models and cross-level models. For the
95

random intercept models predicting descriptive norms and media recall, Level 1 demographic
and socioeconomic variables were entered first, followed by storytelling variables (such as
ICSN) and Level 2 neighborhood characteristics (such as linguistic isolation). Neighborhood-
level variables and their interaction terms with Level 1 ICSN were entered one at a time.
Additionally, all Level 1 predictors measured on a ratio or interval scale were centered to
their respective cluster means when added to the models. All Level 2 variables were centered to
their respective grand means. Centering is recommended for meaningful interpretation of the
intercept when the values of predictors do not have a true zero (Raudenbush & Bryk, 2002). For
Level 1 predictors, centering to the cluster mean is preferred over centering to the grand mean
when a substantive interest of hypothesis testing involves Level 1 predictors and cross-level
interactions, because the former generates less biased estimates (Enders & Tofighi, 2007).
Structural Equation Modeling (SEM)
The structural relationship among neighborhood experience, storytelling resources, media
recall, media attention, discussion with healthcare professionals, descriptive norms, and
compliance with cervical screening guidelines were investigated using Structural Equation
Modeling (SEM). SEM allows for simultaneously evaluating a system of hypothesized
relationships to determine the consistency between a theoretical process and empirical data. For
this reason, SEM is particularly useful for communication research, where the study of
communication process is fundamental to theory building and testing (Cappella, 1980; Holbert &
Stephenson, 2008).
The hypotheses and research questions involved in the aforementioned structural
relationship were examined in LISREL 8.7 (Joreskog & Sorbom, 1996). As some of the
variables were dichotomous, the data were processed first in PRELIS to produce polychoric and
96

polyserial correlation and asymptotic covariance matrices before estimating the model. Cases
that contained missing values were excluded from subsequent analysis.
4
Weighted least squares
(WLS) estimation was used to accommodate categorical variables and the presence of skewness
and kurtosis in some of the endogenous variables (Joreskog & Sorbom, 1996). The model fits
were determined by the following goodness-of-fit measures: 1) Chi-square statistic (non-
significant); 2) the goodness of fit index (GFI; greater than .95); 3) the comparative fit index
(CFI; greater than .95); 4) the normed fit index (NFI; greater than .95); and 5) the root mean
square error of approximation was also satisfactory (RMSEA; less than .05) (Hooper, Coughlan,
& Mullen, 2008; Schumacker & Lomax, 2004). The significance of each path within a model
was assessed individually, with the alpha level set at .05.

97

Footnotes
1
Neighborhood-level homeownership was included in preliminary analysis, but was dropped
later because of its strong correlation with linguistic isolation (r=-.763, p<.001).
2
The excluded cluster is Downey, which is composed of 25 census tracts and 34 participants.
The tract with a population below 200 within Downey is used for railroads a large medical center.
3
Statistically speaking, when intraclass correlation coefficient approaches zero sufficiently, data
observations at the individual-level can be seen as independent and multilevel modeling is
therefore unnecessary. Some argue, however, that what quantifies “sufficiently” remains
debatable, and that hypothesis testing can be invalid when the values of intraclass correlation
coefficient are as small as 0.05 (Hayes, 2006).
4
Removing cases with missing values was achieved with listwise deletion using PRELIS.
Listwise deletion was preferred over imputation for two reasons. First, a primary assumption of
this study is that individuals differ substantially in terms of knowledge, beliefs, and behaviors
across neighborhoods. This difference might be masked when imputation is used to replace
missing values, especially grand-mean imputation. Second, because data collection of the
Multilevel Study was through in-person survey, the number of cases with missing values for
most variables is sufficiently low (less than 5%).

98

CHAPTER 5 : RESULTS

Descriptive Results
Descriptive Results for Individual-Level Variables
This research examines the effects of both individual-level and neighborhood-level
variables on the formation of descriptive norms regarding Pap tests among Latinas from Los
Angeles. Table 5.1 and Table 5.2 present summary statistics of the individual-level
characteristics related to socioeconomic status, health, and healthcare for the survey participants
from the 24 neighborhood clusters. Participants of the full sample (N = 1116) had a median age
of 38 years (Range = 21-50, M = 37.18, SD = 8.63). The median combined annual household
income was between $10,000 and $20,000, and over 80% of the sample made less than $50,000.
Thirty two point six percent of the sample reported 8
th
grade or less of schooling, 22.8% had
some high school, 27.6% had a high school diploma or equivalent credential, and about 16.1%
had some college or higher level of education. In terms of neighborhood experience, the average
years of residence for the participants were 11.50 years (SD = 9.62), and approximately 68% of
the participants had lived in their neighborhoods for 6 years or more. The vast majority of the
sample rented rather than owned their homes (89.4%). Additionally, 81.9% of the participants
were first generation immigrants, 15.9% were second generation, and 2.2% were third and higher
generation. In terms of health-related measures, over half of the sample had some kind of health
care coverage. The majority of the participants rated their general health status as fair (46.8%) or
good (30.2%). Nearly 70% of the sample reported having discussed Pap tests with a doctor,
nurse, or other health care provider.

99

Table 5.1 Descriptive Statistics of Socioeconomic Characteristics: Frequencies
Variables
Total Compliant Noncompliant
N = 1116 N = 720 N = 396
N (%)
Age (years)

21-24 125(11.2) 63(8.8) 62(15.7)
25-29 133(11.9) 80(11.1) 53(13.4)
30-34 151(13.5) 94(13.1) 57(14.4)
35-39 209(18.7) 142(19.7) 67(16.9)
40-45 215(19.3) 153(21.3) 62(15.7)
46-50 283(25.4) 188(26.1) 95(24.0)
Income ($)

Less than 10,000 373(33.4) 244(33.9) 129(32.6)
10,000 to less than 50,000 579(51.9) 370(51.4) 209(52.8)
50,000 to less than 100,000 27(2.4) 14(1.9) 13(3.3)
100,000 and more 3(.3) 1(.1) 2(.5)
Missing 134(12.0) 91(12.6) 43(10.9)
Education

8
th
grade or less 364(32.6) 267(37.1) 97(24.5)
Some high school (9
th
to 11
th
grades) 255(22.8) 166(23.1) 89(22.5)
High school (including vocation/trade
school)
308(27.6) 190(26.4) 118(29.8)
Some college (including junior/community
college)
129(11.6) 61(8.5) 68(17.2)
College degree or more 50(4.5) 29(4.0) 21(5.3)
Missing 10(.9) 7(1.0) 3(.8)
Tenure

< 6 years 352(31.5) 218(30.3) 134(33.8)
6 years and more 757(67.8) 498(69.2) 259(65.4)
Missing 7(.6) 4(.6) 3(.8)
Homeowner

Yes 68(6.1) 46(6.4) 22(5.6)
No 998(89.4) 648(90.0) 350(88.4)
Missing 1(.1) 1(.1) 0(0)
Immigration

First generation 914(81.9) 626(86.9)
b
288(72.7)
a

Second and higher generation 202(18.1) 94(13.1)
a
108(23.3)
b

Notes: Percentages within rows that do not share superscripts differ at p <.05.

100

When the data were broken down by status of compliance with cervical cancer screening
guidelines, a significant difference was found between noncompliant participants (N = 396) and
compliant participants (N = 720) on immigration generation, health care coverage, self-reported
general health, and prior discussion about Pap tests with healthcare providers. Compared to
compliant group, noncompliant group had significantly lower percentages of women who were
first generation immigrants (86.9% vs. 72.7%; χ
2
= 34.84, p <.001), who had health care
coverage (64.4% vs. 37.1%; χ
2
= 82.16, p <.001), and who had discussed Pap tests with health
care providers (74.2% vs. 58.3%; χ
2
= 31.23, p <.001); but significantly higher percentages of
women who felt they were in good health (27.2% vs. 35.6% ; χ
2
= 8.84, p <.01) and very good
or excellent health (11.1% vs. 17.9%; χ
2
= 10.36, p <.01).

Table 5.2 Descriptive Statistics of Health and Healthcare Variables: Frequencies
Variables
Total Compliant Noncompliant
N = 1116 N = 720 N = 396
N (%)
Health care coverage

Yes 611(54.7) 464(64.4)
b
147(37.1)
a

No 490(43.9) 243(33.8)
a
247(62.4)
b

Missing 15(1.3) 13(1.8) 2(.5)
General Health Status

Very Poor or Poor 98(8.8) 73(10.1) 25(6.3)
Fair 522(46.8) 367(51.0)
b
155(39.1)
a

Good 337(30.2) 196(27.2)
a
141(35.6)
b

Very Good or Excellent 151(13.5) 80(11.1)
a
71(17.9)
b

Missing 8(.7) 4(.6) 4(1.0)
Had Discussed Pap tests with a Doctor, Nurse,
or Other Healthcare Professional

Yes 765(68.5) 534(74.2)
b
231(58.3)
a

No 321(28.8) 167(23.2)
a
154(38.9)
b

Missing 30(2.7) 19(2.6) 11(2.8)
Notes: Percentages within rows that do not share superscripts differ at p <.05.

101

Table 5.3 Descriptive Statistics of English Language Proficiency, Storytelling Resources, Media
Recall, Media Attention, and Descriptive Norms: Means and Standard Deviations
Variables
Total Compliant Noncompliant
N = 1116 N = 720 N = 396
M (SD)
English linguistic proficiency (1-4) 2.25 (1.11) 2.07
a
(1.04) 2.58
b
(1.16)
Storytelling

ICSN (3-15) 8.25(3.08) 8.29(3.05) 8.16(3.14)
Interpersonal discussion (1-10) 4.15(2.96) 4.15(2.96) 4.13(2.98)
Community Org (0-5) .95(.80) .99(.80) .90(.78)
Local/Ethnic Media (0-3) 1.95(.87) 1.94(.84) 1.96(.91)
English Language Media (0-3) 1.17(1.10) 1.06
a
(1.05) 1.37
b
(1.16)
Media recall (1-4) 1.71(1.04) 1.77
b
(1.09) 1.16
a
(.95)
Media attention (1-10) 6.02(3.36) 6.39
b
(3.38) 5.37
a
(3.25)
Descriptive norms (1-100)

Have never had a Pap 46.71(25.63) 45.25
a
(25.44) 49.33
b
(25.80)
Have routine Pap tests 50.14(23.10) 52.70
b
(23.33) 45.58
a
(22.00)
Have ever had abnormal Pap test results 39.79(23.30) 41.55
b
(23.91) 36.76
a
(21.91)
Notes: Means within rows that do not share superscripts differ at p <.05.
Table 5.3 reports descriptive statistics of the individual-level characteristics related to
participants’ English linguistic proficiency, connections to storytelling resources, as well as
media recall, media attention, and descriptive norms regarding Pap tests. As shown in the table,
noncompliant women (M = 2.58, SD = 1.16) scored significantly higher on English linguistic
proficiency than did compliant women (M = 2.07, SD = 1.04), t (1112) = 7.47, p <.001. In other
words, compared to compliant women, noncompliant women in general were more fluent in
speaking, reading, and writing English, as well as in understanding English language television
and radio programs. In terms of storytelling variables, noncompliant women had a broader scope
of connections to English language media (M = 1.37, SD = 1.16) compared to compliant women
102

(M = 1.06, SD = 1.05), t (1112) = 4.46, p <.001. Nevertheless, the two groups were comparable
in their frequency of interpersonal discussion about their neighborhoods, scopes of connections
to local/ethnic media and community organizations. One of the key variables of interest in this
dissertation, integrated connection to a storytelling network (ICSN), did not differ significantly
between noncompliant and compliant women either.
In summary, results of descriptive analysis indicate substantial similarities between
noncompliant and compliant participants on most of the demographic and socioeconomic
characteristics. But significant differences existed between the two groups of women in English
linguistic proficiency, immigration generation, healthcare coverage, and self-reported general
health status. Specifically, noncompliant women in the study sample were significantly more
proficient in English, more likely to be second or higher generation immigrants, more likely to
be uninsured, and more likely to feel that they were in good health. In terms of connections to
storytelling resources, although noncompliant women did not differ significantly from compliant
women in their connections to individual neighborhood storytelling agents or ICSN, they had a
significantly broader scope of connections to English language media. In terms of health
communication outcomes, noncompliant women reported a significantly lower level of recall of
having seen or heard information about Pap tests in the media in the past month, attention to Pap
tests-related information in the media, and discussion about Pap tests with healthcare
professionals. Finally, the two groups of participants differed significantly on all three measures
of descriptive norms. Compared to compliant participants, noncompliant participants reported
significantly higher estimates of women (out of 100) who had never had a Pap test, who had
received routine Pap tests, and who had received an abnormal Pap test result.
103

Table 5.4 Zero-Order Correlations among Individual-Level Variables Included in Analysis

Variables
(2)
EDU
(3)
RET
(4)
GEN
(5)
ENG
(6)
HCC
(7)
EC
(8)
INS
(9)
LC
(10)
OC
(11)
ICSN
(12)
MC
(13)
MA
(14)
DIS
(15)
DNN
(16)
DNR
(17)
DNA
(1) Age -.341
***
.127
***
-.350
***
-.450
***
.005 -.308
***
.088
**
.143
***
.171
***
.184
***
.161
***
.130
***
-.015 -.076
*
.138
***
-.003
(2) Education (EDU) .075
*
.393
***
.569
***
-.016 .436
***
-.028 -.104
**
-.019 -.060
*
-.137
**
-.118
**
.026

.059 -.123
**
-.033
(3) Residential Tenure (RET) .178
***
.201
***
.014 .123
***
.139
***
.024 .050 .113
***
-.057 -.036 .061
*
-.035 .031 -.023
(4) Immigration Generation
(GEN)
.620
***
.020 .403
***
-.001 -.224
***
-.098
**
-.133
***
-.140
***
-.192
***
.059 .012 -.120
***
.007
(5) English Linguistic Proficiency
(ENG)
.017 .668
***
.017 -.164
***
-.055 -.085
**
-.140
***
-.151
***
.088
**
.043 -.176
***
.001
(6) Health Care Coverage
(HCC)
.049 .019 .027 .015 .031 .048 .066
*
.048 -.032 .112
***
.053
(7) English Language Media
(EC)
.065
*
.069
*
.038 .092
**
-.046 -.028 .058 -.001 -.138
***
-.027
(8) Interpersonal Discussion
(INS)
.190
***
.203
***
.673
***
.113
***
.121
***
.022 .096
**
.075
*
.076
*
(9) Local/Ethnic Media
(LC)
.143
***
.629
***
.184
***
.142
***
-.023 -.030 .021 -.039
(10) Community Organization
(OC)
.632
***
.130
***
.153
***
.014 .008 .061
*
.025
(11) ICSN .203
***
.194
***
.005 .028 .065
*
.021
(12) Media Recall
(MC)
.494
***
.012 .025 .050 .040
(13) Media Attention
(MA)
.060 .043 .029 .073
*
(14) Discussion Pap tests with
health care providers (DIS)
-.062
*
-.060

-.061

(15) Descriptive Norms (Never)
(DNN)
-.155
***
.246
***
(16) Descriptive Norms (Routine)
(DNR)
.196
***
(17) Descriptive Norms (Abnormal
Results)(DNA)

--
*p <.05, **p <.01, ***p <.001
104

Table 5.4 presents the zero-order correlation among the individual-level variables
included in analysis. It is of note that the correlation coefficient between descriptive norms
regarding never having a Pap test and descriptive norms regarding regular Pap tests was -.155, p
<.001. The negative direction of the correlation is not unexpected, given that the two types of
descriptive norms can be seen as conceptually opposite to each another. The magnitude of the
correlation coefficient is small, however. To further explore the correlation between the two
types of descriptive norms, the data were broken down by status of compliance with cervical
cancer screening guidelines. For compliant participants, the correlation between descriptive
norms regarding never having a Pap test and descriptive norms regarding regular Pap test
remained significant, with a slight increase in magnitude of the correlation coefficient, r = -.208,
p <.001. This correlation was no significant for noncompliant participants, r = -.031, p = .555.
Implications of these findings are discussed in Chapter 6.
Descriptive Results for Neighborhood-Level Variables
This study focus on four neighborhood-level variables, which are linguistic isolation,
ethnic heterogeneity, density of communication resources, and density of health service
providers. Table 5.5 presents a selection of population characteristics for the study area based on
the American Community Survey 2007-2011 5-Year Estimate. The four neighborhood-level
variables included in analysis are shown in Table 5.6. Both tables are organized by the larger
regions into which a particular neighborhood cluster falls
7
. Appendix D illustrates the spatial
distribution of several population characteristics of the study area.

7
Regions are defined according to the neighborhood boundaries delineated by the Mapping Project maintained by
the Los Angeles Times, available at http://maps.latimes.com/neighborhoods.
105

Table 5.5 Selection of Neighborhood-level Characteristics for the 24 Neighborhood Clusters (By Regions of Los Angeles)
a

Neighborhood Clusters
Area
(Miles
2
)
Total
Pop.
Pop. speaking
language other
than English at
home
b
(%)
Foreign
Born
Pop.
(%)
Latino
(%)
White
(%)
Black
(%)
Asian
(%)
Country of Origins for
Hispanic Population
c

Median
Income
Home
owner
(%)
Mexican
American
(%)
Central
American
(%)

East Los Angeles

Boyle Heights
4.02
69909 91.40 49.12 95.30 1.77 0.72 1.79 91.67 6.19 34,743 27.15

East LA
5.82
104730 88.63 45.63 97.81 1.21 0.20 0.53 91.89 6.09 38,373 35.12

El Sereno
3.62
33542 77.68 41.97 76.78 5.42 1.20 15.50 86.02 8.05 51,432 54.79

Lincoln Heights
1.67
22534 87.20 47.30 70.75 4.22 0.71 24.08 86.21 11.41 31,722 26.88

Central Los Angeles

East Hollywood
2.42
67613 81.68 61.26 58.36 19.28 3.25 17.27 38.53 54.89 32,654 9.53

Koreatown
2.81
113688 86.24 65.51 57.33 5.68 4.13 31.19 47.98 46.06 32,424 5.39

Pico Union
1.00
21326 91.03 59.70 88.61 3.54 2.65 4.03 53.99 42.13 27,017 10.43

Westlake
2.94
104806 87.91 63.07 71.56 5.05 4.02 18.36 50.31 45.32 28,306 5.10

South Los Angeles (East of 110 )

Central Alameda
1.54
28685 88.77 47.53 90.05 0.65 8.01 0.76 81.56 17.40 31,367 29.56

East Adams
1.09
20294 88.28 50.18 90.34 0.15 8.91 0.13 82.54 16.45 31,181 28.08

Florence
4.01
74454 82.46 44.05 84.56 0.74 14.25 0.13 85.50 13.41 34,919 36.15

Historic South Central
2.19
28984 90.81 52.64 91.75 1.58 4.43 1.69 84.16 14.14 28,322 20.72

Vernon Main
2.89
107098 57.67 49.48 86.05 1.34 11.63 0.03 77.31 20.97 32,202 27.36

Watts
6.79
109702 69.44 34.92 71.34 0.86 26.98 0.14 81.06 17.28 31,494 36.99

South Los Angeles (West of 110)

Crenshaw
6.54
114359 61.86 40.50 58.91 3.17 32.08 3.93 55.60 40.08 32,229 28.97

Vermont/Slauson
6.42
60916 85.80 35.49 57.13 1.27 39.55 0.81 56.07 41.07 29,457 38.90
106

Neighborhood Clusters
Area
(Miles
2
)
Total
Pop.
Pop. speaking
language other
than English at
home
a
(%)
Foreign
Born
Pop.
(%)
Latino
(%)
White
(%)
Black
(%)
Asian
(%)
Country of Origins for
Hispanic Population
b

Median
Income
Home
owner
(%)
Mexican
American
(%)
Central
American
(%)

Southeast Los Angeles

Huntington Park
2.07
41012 94.85 51.00 97.76 1.00 0.20 0.79 83.31 13.22 34,661 22.80

Maywood
4.07
82441 91.98 48.76 96.30 2.77 0.37 0.23 83.90 12.21 39,843 24.90

South Gate
6.57
120277 90.95 48.08 96.90 1.89 0.22 0.55 83.50 12.42 40,069 40.79

Other regions

El Monte
11.90
129561 85.03 53.46 67.84 4.36 0.48 26.43 90.29 6.79 43,581 40.44

Highland Park
1.30
24780 77.08 45.41 79.08 10.22 1.45 8.52 70.51 23.38 44,630 31.69

Montebello
7.52
57928 77.83 37.57 90.44 5.74 0.72 2.00 87.52 9.06 46,003 46.72

Pasadena
1.99
27792 60.02 33.29 55.79 15.88 17.03 8.53 78.14 15.12 43,037 27.22

Pomona
16.73
123294 67.75 37.59 72.09 9.80 6.80 10.10 87.33 8.70 52,245 53.74
a
Regions are defined according to the neighborhood boundaries delineated by the Mapping Project maintained by the Los Angeles
Times, available at http://maps.latimes.com/neighborhoods.
b
In the American Community Survey, the question regarding language
spoken at home was asked to population aged 5 years or older.
c
Breakdown by country of origin is in reference to the total Latino
population in a given neighborhood cluster.

107

Table 5.5 and the maps in Appendix D reveal a considerable concentration of certain
demographic and socioeconomic characteristics at the neighborhood level. Overall, Koreatown,
Vernon Main, and Westlake are among the smallest neighborhood clusters by land area, but the
largest clusters by population, suggesting high population density. In terms of demographic
characteristics, East, Central, South, and Southeast Los Angeles all have neighborhoods where
over 90% of the population (aged 5 years and older) report speaking language other than English
at home. Neighborhood clusters in Central Los Angeles have the highest percentage of foreign-
born population compared to other neighborhoods. Moreover, East, Southeast and South Los
Angeles are home to neighborhood clusters with the highest percentages of Latino population.
Of these, South Los Angeles is notably and spatially divided into two sub-areas (west and east)
by ethnic composition. In particular, the vast majority of the population in the neighborhoods
east of the 110 Freeway are Latino, whereas on average approximately one third of the
population in the neighborhoods west of the 110 Freeways are African Americans.
Table 5.5 also shows the breakdown by country of origin for the Latino and Hispanic
population. It appears that clusters in East Los Angeles, Southeast Los Angeles, and South Los
Angeles east of 110 Freeway have the highest percentage of Latinos of Mexican origin. By
comparison, neighborhoods in Central Los Angeles and South Los Angeles west of 110 Freeway
have higher proportion of Latinos of Central American descent. Moreover, most neighborhood
clusters in the study area have a relatively low representation of Asian population, except some
clusters in Central Los Angeles (e.g., Koreatown) and East Los Angeles (e.g., Lincoln Heights).
Further, with a median annual household income above $50,000, and over half of the
housing units occupied by homeowners as opposed to renters, El Sereno and Pomona are the
most economically better-off neighborhoods in the study area. By comparison, neighborhood
108

clusters in Central Los Angeles have the lowest level of homeownership, where less than 10
percent of the housing units are owner-occupied on average.
Table 5.6 summarizes the level of linguistic isolation and ethnic heterogeneity.
Consistent with what is observed in Table 5.5, East, Central, Southeast Los Angeles and South
Los Angeles east of 110 Freeway have the most linguistically isolated neighborhood clusters.
For example, in all neighborhood clusters in Central Los Angeles, the percentage of population
who speak English less than “very well” is over 50%. In terms of ethnic heterogeneity, the most
diverse neighborhood cluster is Pasadena, followed by East Hollywood and Koreatown from
Central Los Angeles. It is also here that the aforementioned spatial divide in South Los Angeles
emerges again. While the two neighborhoods west of 110 Freeway are markedly diverse in
ethnic composition, most neighborhoods east of 110 Freeway are not.
The rest of Table 5.6 provides information about local communication resources (e.g.,
parks, churches, and community organizations)
8
and health service providers. This information
is presented in both raw counts and population-adjusted density. In terms of raw counts, Watts,
El Monte, Koreatown, Pomona, Crenshaw and East LA have the highest number of
communication resources; Westlake and Koreatown have the highest number of health service
providers. After accounting for population density, Pasadena, Highland Park, Historic South
Central and Lincoln Heights all have a density of communication resources above 15 per 10000
residents. Historic South Central, Pico Union, Lincoln Heights, and Westlake top the list of the
neighborhood clusters with the highest density of health service providers. Table 5.7 presents
the zero-order correlation parameters among the neighborhood-level characteristics discussed
above.

8
See Table F1 in Appendix F for types of communication resources included for each neighborhood cluster.
109

Table 5.6 Neighborhood-Level Variables Included in Analysis (By Regions of Los Angeles)
a

Neighborhood Clusters
Linguistic
Isolation
(%)
Ethnic
Heterogeneity
Counts
b

(N)
Density
c

(per 10000 population)
Comm.
Resources
Health
Services
Comm.
Resources
Health
Services

East Los Angeles
1 Boyle Heights
51.30 0.09 76 16 13.03 1.43
6 East LA 47.99 0.04 100 13 11.80 1.55
8 El Sereno
32.17 0.38 26 5 11.53 1.84
14 Lincoln Heights
48.22 0.44 30 3 16.40 2.22

Central Los Angeles
5 East Hollywood
51.31 0.59 49 14 10.63 1.79
13 Koreatown
59.06 0.57 134 22 12.83 1.44
18 Pico Union
57.90 0.21 26 3 10.31 2.38
24 Westlake
56.98 0.45 99 27 12.63 2.09

South Los Angeles (East of 110)
2 Central Alameda
48.36 0.18 13 1 5.02 0.81
4 East Adams
47.44 0.18 22 0 12.51 1.16
9 Florence
44.24 0.26 40 3 5.30 0.62
11 Historic South Central
50.35 0.16 48 10 17.66 2.92
22 Vernon Main
50.63 0.25 29 12 4.86 1.07
23 Watts
34.33 0.42 142 4 13.50 0.51

South Los Angeles (West of 110)
3 Crenshaw
35.81 0.55 106 12 13.39 1.47
21 Vermont/Slauson
33.01 0.52 96 4 9.77 0.68

Southeast Los Angeles
15 Maywood
54.66 0.07 46 6 6.77 0.65
12 Huntington Park
52.10 0.04 40 6 9.27 0.61
20 South Gate
46.84 0.06 55 3 5.12 0.67

Other regions
7 El Monte
48.56 0.47 136 17 12.98 1.18
10 Highland Park
37.37 0.36 25 2 19.47 0.79
16 Montebello
33.45 0.18 30 6 7.68 0.99
17 Pasadena
31.22 0.63 36 3 20.04 0.56
19 Pomona
30.98 0.46 120 7 9.61 0.64
a
Regions are defined according to the neighborhood boundaries delineated by the Mapping Project maintained by
the Los Angeles Times, available at http://maps.latimes.com/neighborhoods.
b
Raw counts of actual geographic
locations that strictly fall within the boundary of neighborhood clusters, regardless of numbers of services offered.
For the raw counts of communication resources for each specific category included, see Appendix E.
c
Mean
cluster-level density of communication resources is reported per 10000 population. Because the analysis is based on
locations within 0.5 mile buffer of the analysis area, for both communication resources and health service providers,
the total count of locations used to calculate the density is larger than the total count reported in the table.
110

Table 5.7 Zero-Order Correlations among Neighborhood-Level Variables

Variables
(2)
Whites
(3)
Blacks
(4)
Asians
(5)
FBP
(6)
MI
(7)
HO
(8)
LI
(9)
EH
(10)
DCR
(11)
DHS
(1) Latinos (%) -.560
**
-.528
**
-.602
**
.079 .011 .125 .334 -.988
***
-.410
*
-.009
(2) Whites (%) -.124 .446
*
.037 .366 -.143 -.230 .610
**
.386 .021
(3) Blacks (%) -.284 -.523
**
-.303 .128 -.504
*
.460
*
.040 -.303
(4) Asians (%) .459
*
.133 -.253 .198 .636
**
.368 .347
(5) Foreign Born Population (%)
(FBP)
-.448
*
-.772
***
.899
***
-.067 -.055 .546
**
(6) Median Income ($)
(MI)
.721
***
-.569
**
.044 .021 -.378
(7) Homeownership (%)
(HO)
-.763
***
-.109 -.189 -.447
*
(8) Linguistic Isolation (%)
(LI)
-.345 -.192 .429
*
(9) Ethnic Heterogeneity
(EH)
.431
*
.039
(10) Density of Comm. Resources
(DCR)
.346
(11) Density of Health Service
(DHS)
--
*p <.05, **p <.01, ***p <.001

Multilevel Analysis of Descriptive Norms Regarding Pap Tests
The current section reports results from multilevel modeling analysis that was conducted
to address H1a through H1d and RQ1a through RQ1b. Before testing the proposed models, fully
unconditional models were run first to determine the necessity of multilevel modeling analysis
(Raudenbush & Bryk, 2002). Also termed as “ intercept-only models”, these models are
equivalent to a one-way ANOVA model with random effects. They tested whether there was a
significant difference in descriptive norms across 24 neighborhood clusters.
Table 5.8 presents the results from testing the between-cluster differences in three types
of descriptive norms: 1) perceived prevalence of women never having a Pap test, 2) perceived
prevalence of women having routine Pap tests, and 3) perceived prevalence of women ever
receiving an abnormal Pap test result. For each model, the intercept equaled the grand mean of
the perceived norm tested across all clusters. Because noncompliant and compliant participants
111

differed substantially on several socioeconomic characteristics and, especially, on Pap test-
related variables, they were analyzed separately.

Table 5.8 Hierarchical Linear Models of Descriptive Norms Regarding Pap Tests (Fully Un-
conditional Models)

Noncompliant Women

Compliant Women
Never Routine Results

Never Routine Results

M(SD)

M(SD)

Neighborhood Mean
Descriptive Norms
49.521
(1.692)***
45.580
(1.126)***
36.756
(1.145)***

45.253
(.973)***
52.699
(.895)***
41.535
(.958)***

Level 2 Variance 23.995 1.04E-19 9.18E-12

2.26E-07 1.32E-16 3.31E-16
Level 1 Variance 641.037 482.857 478.940

646.492 543.438 570.732
Chi-Square 2.57* 0 0

0 0 0
ICC .036 2.15E-22 1.92E-14 3.49E-10 2.43E-19 5.80E-19
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

Of these models, only the model predicting descriptive norms regarding never having a
Pap test for noncompliant women revealed a meaningful between-cluster difference. The
coefficient of the intercept was 49.521, with a standard error as 1.692 and a p value smaller than
.001. The Chi-square score afforded by a likelihood ratio test comparing the current model with
a conventional regression model was significant (χ
2
= 2.57, p = .054), suggesting that the
random intercept model was a meaningfully better fit to the data than a regression model. The
intraclass correlation coefficient (ICC) was .03, indicating that 3% of the variance in descriptive
norms was explained by neighborhood clusters. Although the intercepts were statistically
significant for the remaining models, the Chi-square scores produced by likelihood ratio tests for
those models were all non-significant. In addition, the remaining models all invariably returned
an ICC that was extremely low. Thus, for each of those models, it can be concluded that a
conventional regression model would be a better alternative to the data compared to a multilevel
112

model, and that the difference in the intercept was best explained by individual-level
characteristics alone for the study sample (Rabe-Hesketh & Skrondal, 2012).
Given these findings, the rest of the hypotheses regarding the relationship between
descriptive norms and individual- and neighborhood-level variables were examined only for the
descriptive norms regarding women’s never having a Pap test among noncompliant women.
Only major significant results from hypotheses testing are discussed in detail in this chapter,
whereas non-significant findings are reported in Appendix F.
Effects of Neighborhood Storytelling Resources on Descriptive Norms
The dependent variable in this section is descriptive norms regarding the prevalence of
never having a Pap test among Latinas. The first three hypotheses respectively concerned the
roles of frequency of interpersonal discussion about one’s neighborhood (H 1a), scope of
connections to local/ethnic media (H1b), and scope of connections to community organizations
(H1c) on descriptive norms, controlling for both individual-level and neighborhood-level
variables. Because significant between-neighborhood variance was found only for the
descriptive norms regarding noncompliant participants never having had a Pap test, these
hypotheses were examined for this particular descriptive norms construct and for noncompliant
participants only.
Table 5.9 presents the results of six random intercept models of descriptive norms,
accounting for individual-level covariates. Because these models do not yet include
neighborhood-level variables, they are also called level-2 unconditional models (Raudenbush &
Bryk, 2002). For the full intercept-models that included each of the four neighborhood-level
variables one at a time, see Table F1 in Appendix F.

113

Table 5.9 Hierarchical Linear Models of Descriptive Norms Regarding Pap Tests for Non-
compliant Participants (Level-2 Unconditional Models)
Predictors
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Baseline SES NGH
Experience
English
Media
NGH
Storytelling
ICSN

Coefficient (Standard Error)

Neighborhood Mean
Descriptive Norms
49.521
(1.692)***
43.547
(3.193)***
42.514
(4.497)***
42.187
(4.513)***
41.118
(4.486)***
42.209
(4.517)***

Age

-.283
(.162)
¶

-.306
(.176)
¶

-.310
(.176)
¶

-.404
(.178)*
-.276
(.180)*

8th Grade and above
(Dummy)

5.640
(3.552)
5.571
(3.596)
5.948
(3.629)
6.166
(3.598)
¶

5.806
(3.635)

Health Care Coverage

4.682
(2.731)
¶

4.832
(2.765)
¶

4.757
(2.765)
¶

4.699
(2.750)
¶

4.828
(2.775)
¶

English Proficiency

-.913
(1.463)
-.600
(1.790)
.166
(2.062)
.572
(2.141)
.911
(2.136)

Residential Tenure

.

.012
(.154)
.017
(.154)
-.067
(.155)
-.030
(.156)

First Generation
Immigrants (Dummy)

1.332
(4.165)
1.396
(4.163)
2.532
(4.205)
1.331
(4.215)

English Language
Media

-1.207
(1.619)
-1.816
(1.701)

Interpersonal
Discussion

1.367
(.478)**

Local/Ethnic Media

-1.627
(1.580)

Community
Organizations

1.629
(1.734)

ICSN

.833
(.469)
¶

Level 2 Variance 23.995 31.400 31.925 31.273 27.939 29.420
Level 1 Variance 641.037 616.168 619.762 619.165 605.334 617.771
Chi-Square 2.57
¶
4.09* 4.07* 3.92* 3.17* 3.45*
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001
114

Model 1 in Table 5.9 was a baseline model and was identical with the significant
intercept-only model in Table 5.8. Because the Chi-square test result was significant, (χ
2
= 2.57,
p = .054), a multilevel modeling analysis to further explore the sources of variance in descriptive
norms was warranted. To this end, this study began with building a SES model (Model2) by
including socioeconomic variables (i.e., age, education, health care coverage, and English
linguistic proficiency). Results from testing this model showed that both age (B = -.283, SE =
.162, p = .081) and health insurance status (B = 4.682, SE = 2.731, p = .086) were marginally
significant for the sample of noncompliant participants. In other words, the estimates of the
number of women (out of 100 Latinas like them) who had never had a Pap test tended to be
higher for younger women than for older women, and tended to be higher for insured women
than for uninsured women. Built on Model 2, Model 3 included residential tenure and
immigration history, but neither of these two neighborhood experience variables bore a
significant relationship with descriptive norms. The same was true for the newly added variable
in Model 4, where the association between connections to English language media and
descriptive norms was also non-significant.
The first hypothesis related to neighborhood storytelling resources predicted that the
frequency of interpersonal discussion about one’s neighborhood would have a direct effect on
descriptive norms regarding Pap tests (H1a). This hypothesis was confirmed by testing whether
the coefficient of interpersonal discussion would remain significant after accounting for
individual-level covariates and neighborhood-level covariate (s). Model 5 in Table 5.9 added
interpersonal discussion about one’s neighborhoods in conjunction with other two neighborhood
storytelling variables (i.e., scope of connections to local/ethnic media, scope of connections to
community organizations). Holding other individual-level covariates constant, interpersonal
115

discussion about one’s neighborhood had a positive and significant effect on descriptive norms
regarding Pap test prevalence (B = 1.367, SE = .478, p < 0.01). This indicates that, for the
noncompliant sample, women who talked with others about their neighborhoods more frequently
were more likely to think it was common for Latinas to never have had a Pap test than women
who discussed their neighborhoods less often. Interpersonal discussion about one’s
neighborhood remained significant as a predictor of descriptive norms even after further
controlling for neighborhood-level variables (see Table F1 in Appendix F). Thus, H1a, which
proposed a significant effect of interpersonal discussion about one’s neighborhood on descriptive
norms regarding Pap tests, was supported for the noncompliant sample.
The second and third hypothesis respectively predicted that scope of connections to
local/ethnic media (H1b) and scope of connections to community organizations (H1c) would
have a direct effect on descriptive norms regarding Pap tests, accounting for individual- and
neighborhood-level covariate(s). As shown in Model 5 in Table 5.9, neither of these two
variables had a significant association with descriptive norms, when other individual-level
covariates were held constant. This pattern remained the same across all models after
neighborhood-level variables were entered (see Table F1 in Appendix F). Therefore, the
hypothesized effects of connections to local/ethnic media (H1b) and connections to community
organizations (H1c) on descriptive norms were not supported.
The fourth hypotheses (H1d) regarding the relationship between neighborhood
storytelling resources and descriptive norms concerned the role of ICSN, a measure that captures
the synergetic effect of connecting to all three neighborhood storytelling resources. As shown in
Model 6, ICSN was a marginally significant predictor of descriptive norms (B = .833, SE = .469,
p = .076), controlling for other individual-level covariates. In other words, women with a higher
116

level of ICSN perceived it as more common for Latinas to never have had a Pap test than women
with a lower level of ICSN. The effect of ICSN remained marginally significant after
neighborhood-level variables were added (see Table 5.10 below and Table F2 to F5 in Appendix
F). Thus, though suggestive, the statistical evidence did not confirm the proposed effect of ICSN
on descriptive norms (H1d).
Contextual Effect on Descriptive Norms
The analysis thus far revealed significant cross-neighborhood differences in descriptive
norms regarding never having a Pap test among noncompliant women, implying the existence of
contextual effects. RQ1a inquired about whether the cross-neighborhood variation in descriptive
norms can be explained by neighborhood-level linguistic isolation (RQ1a-1), ethnic
heterogeneity (RQ1a-2), density of communication resources (RQ1a-3), and density of health
service providers (RQ1a-4). The statistics exploring these questions with random intercept
models showed that none of these neighborhood-level variables presented a significant direct
effect on descriptive norms (see Table F1 in Appendix F). In other words, no significant
difference was found in descriptive norms between more linguistically isolated and less
linguistically isolated neighborhoods, between ethnically heterogeneous and homogenous
neighborhoods, between areas with higher density of communication resources and areas with
lower density of such resources, or between areas with higher density of health service providers
and areas with lower density of such services.
Besides direct effects, it was posited that contextual influences might also occur with
neighborhood-level variables indirectly affecting descriptive norms by interacting with
individual-level predictors. Following this rationale, RQ1b asked whether the marginal effect of
ICSN on descriptive norms would vary as a function of neighborhood-level variables. To test
117

this hypothesis, four cross-level models were constructed with the interactions of ICSN with
each of the four neighborhood-level variables added one at a time. Of those, only the interaction
between ICSN and linguistic isolation was significant, B = .096, SE = .050, p = .052 (Table
5.10). Meanwhile, the main effect of ICSN no longer retained its marginal significance in the
cross-level model once the interaction was included.
Table 5.10 Hierarchical Linear Models of Descriptive Norms Regarding Pap Tests for
Noncompliant Participants: Individual-Level Covariates, ICSN, and Neighborhood-Level
Linguistic Isolation (Full Model and Cross-Level Model)
Predictors
Model 1 Model 2
Coefficient (Standard Error)

Neighborhood Mean Descriptive Norms

42.215 (4.511)***

42.347 (4.487)***

Individual-Level Variables
Age -.373 (.180)* -.365 (.179)*

8th Grade and above (Dummy) 6.103 (3.678)
¶
5.758 (3.678)

Health Care Coverage (Dummy) 4.792 (2.775)
¶
5.115 (2.767)
¶

English Proficiency .979 (2.148) .621 (2.139)

Residential Tenure -.033 (.156) -.043 (.155)

First Generation Immigrants (Dummy) 1.031 (4.252) 1.012 (4.231)

English Language Media -2.147 (1.694) -1.962 (1.688)

ICSN .831 (.469)
¶
.687 (.473)

Neighborhood-Level Variable
Linguistic Isolation .108 (.202) .111 (.201)

Cross-Level Interaction
Linguistic Isolation ×ICSN

.096 (.050)

*

Level 2 Variance 28.313 27.630
Level 1 Variance 617.910 611.936
Chi-Square 3.36* 3.24*
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001
118

In summary, for the study sample of noncompliant participants, there was a contextual
effect on the strength of ICSN effect, such that the effect of ICSN on descriptive norms
regarding never having a Pap test was stronger for women living in more linguistically isolated
neighborhoods than for those from less linguistically isolated areas. It was thus concluded that
the interaction of ICSN with linguistic isolation (RQ1b-1) was significant, but the interactions of
ICSN with ethnic heterogeneity (RQ1b-2), density of communication resources (RQ1b-3), and
density of health service providers (RQ1b-4) were not. For models that included the other three
neighborhood-level variables, see Table F2, Table F3, and Table F4 in Appendix F.
Multilevel Analysis of Media Recall Regarding Pap Tests
Another interest of multilevel modeling analysis of this study concerned media recall, or
participants’ recall of having seen or heard information about Pap tests in the media in the past
30 days. This inquiry involves examining H1e through H1h and RQ1c through RQ1d. With a
parallel approach, an intercept-only model was performed predicting media recall for both
noncompliant and compliant participants. As shown in Table 5.11, for both models the intercept
(i.e., the grand mean of media recall across 24 clusters) was significant with a p value smaller
than .001. However, only the model for noncompliant women was a significantly better fit to the
data compared to single-level regression models, χ
2
= 6.89, p = .004. In other words, there was
a significant cross-neighborhood difference in women’s recall of having heard or seen
information about Pap tests in the media in the past month for the noncompliant group, but not
for the compliant group.
Furthermore, suggested by intraclass correlation coefficient (ICC), about 6.5% of the
variance in media recall was explained by neighborhood clusters for the noncompliant group.
119

Therefore, the remaining hypotheses concerning the relationship between media recall and
individual- and neighborhood-level variables were examined only for noncompliant women.

Table 5.11 Hierarchical Linear Models of Media Recall Regarding Pap Tests (Fully
Unconditional Models)

Noncompliant Women Compliant Women

Coefficient (Standard Error)

Neighborhood mean media recall 1.621 (.070)*** 1.765 (.046)***
Level 2 variance 0.059 0.010
Level 1 variance 0.845 1.173
Chi-square 6.89** 0.91
ICC 0.065 0.009
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

Effects of Neighborhood Storytelling Resources on Media Recall
Media recall regarding Pap tests is the dependent variable in this section. As for the
relationship between connections to individual neighborhood storytelling resources with media
recall, it was hypothesized that frequency of interpersonal discussion about one’s neighborhood
(H1e), scope of connections to local/ethnic media (H1f), and scope of connections to community
organizations (H1g) would each have a direct effect on media recall, holding constant both
individual-level and neighborhood-level variables. These hypotheses were tested for
noncompliant women only, because only they showed a significant cross-neighborhood
difference in media recall.
Table 5.12 presents the results of six random intercept models of media recall that only
included individual-level covariates. The full intercept-models that further added neighborhood-
level variables can be found in Table F5 in Appendix F. Model 1 in Table 5.12 was a baseline
model providing identical statistics as with those seen in Table 5.11. A significant Chi-square
score (χ
2
= 6.89, p = .004) produced by a likelihood ratio verified the existence of variance in
120

media recall at the neighborhood level, therefore paving the way for subsequent multilevel
modeling analysis.
From Model 2 to Model 6, individual-level variables were entered sequentially to assess
the extent to which they would account for the variance in media recall. Results from testing
Model 2, a SES model, revealed significant effects of age (B = .012, SE = .005, p = .031), health
care coverage (B = .192, SE = .097, p = .047), and English proficiency (B = -.099, SE = .051, p
= .054) on media recall. That is, women in this noncompliant sample were more likely to recall
that they had seen or heard information about Pap tests in the media in the past 30 days, if they
were older, had some kind of health care coverage, or were more fluent in English.
Model 3 added two neighborhood experience variables to the equation, namely
residential tenure and immigration history. However, neither of these two variables was found to
associate with media recall. Connection to English language media as introduced by Model 4
was also not a significant predictor.
H1e, H1f, and H1g proposed that interpersonal discussion about one’s neighborhood,
scope of connections to local/ethnic media, and scope of community organizations would have
a direct effect on media recall, after taking into consideration both individual- and neighborhood-
level variables. As shown in Model 5 in Table 5.12, there was a marginally significant effect of
interpersonal discussion (B = .028, SE = .017, p = .092) and a significant effect of local/ethnic
media connections (B = .147, SE = .055, p = .008) on media recall. Yet the effect of community
organization connections on media recall was not significant. In other words, holding other
individual-level covariates constant, women who talked with others about things happening in
their neighborhoods more often tended to recall having seen or heard information about Pap tests
in the media more frequently. Women with a broader scope of connections to local/ethnic media
121

Table 5.12 Hierarchical Linear Models of Media Recall Regarding Pap Tests for Noncompliant
Participants (Level-2 Unconditional Models)
Predictors
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Baseline SES NGH
Experience
English
Media
NGH
Storytelling
ICSN

Coefficient (Standard Error)

Neighborhood Mean
Media Recall
1.621
(.070)***
1.599
(.118)***
1.578
(.162)***
1.602
(.162)***
1.632
(.161)***
1.613
(.160)***

Age

.012
(.006)*
.012
(.006)*
.012
(.006)*
.010
(.006)
.009
(.006)

8th Grade and above
(Dummy)

-.005
(.127)
-.020
(.127)
-.051
(.128)
-.060
(.127)
-.059
(.127)

Health Care Coverage

.192
(.097)*
.181
(.097)
¶

.185
(.097)
¶

.192
(.096)

*
.197
(.096)

*

English Proficiency

-.099
(.051)

*
-.083
(.063)
-.141
(.072)

*
-.082
(.075)
-.101
(.074)

Residential Tenure

.

-.002
(.005)
-.003
(.005)
-.005
(.005)
-.005
(.005)

First Generation
Immigrants (Dummy)

.037
(.147)
.035
(.146)
-.021
(.147)
.011
(.147)

English Language
Media

.091
(.057)
.021
(.060)

Interpersonal
Discussion

.028
(.017)
¶

Local/Ethnic Media

.147
(.055)**

Community
Organizations

.020
(.061)

ICSN

.047
(.160)

**

Level 2 Variance .059 .058 .055 .055 .055 .052
Level 1 Variance .845 .800 .794 .788 .763 .770
Chi-Square 6.89** 7.17** 6.41** 6.56** 6.70** 6.07**
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001
122

(e.g., connecting to local/ethnic television, radio, and newspapers) reported a significantly
higher level of media recall regarding Pap tests than those with a narrower scope of connections
(e.g., connecting to local/ethnic television only).
On the other hand, media recall did not differ between women who had a broad scope of
connections to community organizations (e.g., participating in a number of community
organizations such as churches, homeowner’s associations, or sports groups) and women whose
scope of connections was narrow (e.g., participating in one community organization only). A
similar pattern of these findings persisted even after further controlling for neighborhood-level
variables (see Table F5 in Appendix F). Taken together, though suggestive, the predicted effect
of neighborhood discussion on media recall was not confirmed (H1e). The proposed effect of
local/ethnic media on media recall was supported (H1f). The proposed effect of community
organization connection on media recall was rejected (H1g).
The next hypothesis posited the effect of having an integrated connection to
neighborhood storytelling resources, or ICSN, on media recall (H1h). Model 6 showed that
ICSN was a significant predictor of media recall for the present sample of noncompliant women
(B = .047, SE = .016, p = .004), controlling for other individual-level covariates. In other words,
women connecting to a strong and integrated network of neighborhood storytelling resources
also reported higher exposure to information about Pap tests in the media than women
connecting to a weak or fragmented local network of information sources. This direct effect of
ICSN on media recall continued to be significant even after accounting for the possible influence
of neighborhood-level variables (see Table F6 through Table F9 in Appendix F). Thus, the
hypothesis that ICNS would have a direct effect on media recall controlling for both individual-
and neighborhood-level variables was confirmed (H1h).
123

Contextual Effect on Media Recall Regarding Pap Tests
The intercept-model in Table 5.11 showed that the sources of variance in media recall
regarding Pap tests were present in both individual and neighborhood-level factors for
noncompliant participants. RQ1c asked that whether neighborhood-level linguistic isolation
(RQ1c-1), ethnic heterogeneity (RQ1c-2), density of communication resources (RQ1c-3), and
density of health service providers (RQ1c-4) would be able to explain the contextual effects on
media recall. Of those variables tested, only density of health service providers was found to be
a predictor of individual-level media recall, B = -.205, SE = .102, p = .044, controlling for
individual-level covariates (Table 5.13). In other words, women from areas with lower density
of health service providers were more likely to report having seen or heard information about
Pap tests in the media than women from areas with higher density of such resources. Yet there
was no significant difference in media recall found between areas that differed by the level of
linguistic isolation, ethnicity heterogeneity, or density of communication resources (see statistics
in Table F5 in Appendix F).
Meanwhile, contextual effects might also occur when neighborhood-level variables
interact with individual-level variables to affect media recall. From this standpoint, RQ1d asked
whether the significant effect of ICSN on media recall would vary as a function of
neighborhood-level variables such as linguistic isolation, ethnic heterogeneity, density of
communication resources, and density of health service providers.
Four cross-level models were built to include the interactions of ICSN with each of these
four neighborhood-level variables one at a time. In each model, the previously reported main
effect of ICSN retained its significance after a particular neighborhood-level variable was added.
However, none of the interaction terms between ICSN and the neighborhood-level variables was
124

Table 5.13 Hierarchical Linear Model of Media Recall Regarding Pap Tests for Noncompliant
Participants: Individual-Level Covariates, Neighborhood Storytelling Resources, and
Neighborhood-Level Density of Health Service Providers (Full Model)
Predictors Coefficient (Standard Error)

Neighborhood Mean Media Recall 1.634 (.159)***

Individual-Level Variables

Age .010 (.006)

8th Grade and above (Dummy) -.079(.126)

Health Care Coverage (Dummy) .197(.096)*

English Proficiency -.074(.075)

Residential Tenure -.005(.005)

First Generation Immigrants (Dummy) -.007(.147)

English Language Media .026(.060)

Interpersonal Discussion .029(.017) ¶

Local/Ethnic Media .149(.055)**

Community Org. .023(.061)

Neighborhood-Level Variable

Density of Health Service Providers

-.205(.102)*

Level 2 Variance .040
Level 1 Variance .762
Chi-Square
4.12*
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

125

significant. In other words, for noncompliant women, there was no evidence to suggest that the
strength of ICSN effect on media recall would differ between more linguistically isolated and
less linguistically isolated areas, between ethnically heterogeneous and homogenous areas,
between areas with high density of communication resources and areas with low density of such
resources, or between areas where density of health service providers is high and areas where
such density is low. For detailed statistics, see Table F6 through Table F9 in Appendix F.
Multilevel Analysis of Connections to Storytelling Resources
The present study additionally explored the sources of variation in individuals’
connections to storytelling resources in everyday communication environment. Participants’
connections to each of the three types of neighborhood storytellers, to English language media,
and their level of ICSN are the dependent variables in this section. Because the dependent
variables this time are about people’s communicatio n behaviors in general, rather than their
perceptions regarding Pap tests more specifically, the research questions and hypotheses were
examined using the combined sample of both noncompliant and compliant participants while
controlling for compliance status.
Following the same approach shown above, a series of intercept-only models were
conducted to determine if there was a neighborhood-level difference in people’s connections to
neighborhood storytelling resources, English language media, and ICSN. As shown in Table
5.14, all but the model predicting connections to community organizations represented a valid
and better solution to the data compared to conventional single-level techniques, as judged by the
Chi-square statistics. The ICCs for the intercept-only models of interpersonal discussion,
connections to local/ethnic media, connections to English language media, and ICSN were all
around .02, indicating that, for each dependent variable, about 2% of its variance was explained
126

by neighborhood clusters. By comparison, the portion of variance in connections to community
organizations at the neighborhood level was negligibly small.

Table 5.14 Hierarchical Linear Models of Connections to Storytelling Resources (Fully
Unconditional Models)

Interpersonal
Discussion
Local/Ethnic
Media
Community
Orgs.
English
Media ICSN

Coefficient (Standard Error)
Intercept
4.123
(.133)***
1.935
(.039)***
.955
(.027)***
1.165
(.047)***
8.225
(.139)***
Level 2 Variance 0.222 0.019 0.003 0.026 0.244
Level 1 Variance 8.564 0.737 0.633 1.182 9.257
Chi-Square 8.12** 8.89** 0.52 6.63** 8.77**
ICC 0.025 0.025 0.005 0.021 0.026
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

Contextual Effect on Connections to Storytelling Resources
This study proposed that neighborhood-level linguistic isolation (H1i), ethnic
heterogeneity (H1j), density of communication resources (H1k), and density of health service
providers (H1l) would each have a direct effect on people’s connections to neighborhood
storytelling resources, to English language media, and level of ICSN. For neighborhood
storytelling resources, because previous intercept-only models did not find neighborhood-level
difference in people’s scope of connections to community organizations, only interpersonal
discussion about one’s neighborhood and scope of connections to local/ethnic media were
included in hypothesis testing. Results from testing these hypotheses are presented respectively
in Table 5.15 to Table 5.18. Individual-level variables that were found to be predictive of
people’s connections to storytelling resources in previous research were entered as covariates for
each model (Ball-Rokeach et al, 2001; Kim & Ball-Rokeach, 2006b). These variables included
age, education, English proficiency, residential tenure, and immigration generation.
127

All multilevel models were a meaningful fit to the data given the significant Chi-square
statistics produced by likelihood ratio tests. Across all four tables, frequency of interpersonal
discussion about one’s neighborhood consistently and significantly increased with age and years
of residence for the present sample, controlling for other individual-level covariates and
neighborhood-level variables. Residential tenure was also a marginally significant predictor of
local/ethnic media connection across models, indicating that participants tended to connect to a
broader range of local/ethnic media with increased years of residence in their neighborhoods.
Moreover, participants who were first generation immigrants had a significantly broader scope of
connections to local/ethnic media than did participants who were second and higher generation
immigrants.
On the other hand, participants who had at least 8
th
grade of schooling and who were
more fluent in English had a significantly broader scope of connections to English language
media, compared to those who were less educated or less proficient in English. Additionally,
when controlling for ethnic heterogeneity (Table 5.16), density of communication resources
(Table 5.17), and density of health service providers (Table 5.18), both education and English
proficiency had a marginally significant association with connections to local/ethnic media,
suggesting that participants tended to connect to local/ethnic media more if they were more
educated or less fluent in English. When controlling for neighborhood-level linguistic isolation
(Table 5.15), while education remained a marginally significant predictor of local/ethnic media
connections, English proficiency became a statistically significant one. The latter means that,
within neighborhoods, participants less proficient in English were more likely to report a broader
scope of local/ethnic media connection than their more proficient counterparts were, even after
taking into account the neighborhood-level difference in linguistic isolation. Last, across all four
128

Table 5.15 Hierarchical Linear Models of Connections to Neighborhood Storytelling Resources,
English Language Media, and ICSN: Individual-Level Covariates and Neighborhood-Level
Linguistic Isolation (Full Models)
Predictors
Interpersonal
Discussion
Local/Ethnic
Media
English
Media
ICSN
Coefficient (Standard Error)

Intercept

3.781
(.339)***
1.600
(.097)***
1.028
(.097)***
6.054
(.345)***

Individual-Level Variables

Age .033
(.012)**
.004
(.004)
-.002
(.003)
.053
(.013)***

8th Grade and Above .284
(.217)
.118
(.063)
¶

.180
(.060)**
.656
(.222)**

English Proficiency .111
(.119)
-.067
(.034) *
.625
(.033)***
.017
(.122)

Residential Tenure .039
(.011)***
.005
(.003)
¶

-.001
(.003)
.036
(.011)**

First Generation
Immigrants
.170
(.313)
.378
(.090)***
-.006
(.087)
1.012
(.320)**

Compliance .013
(.189)
-.088
(.055)
.032
(.053)
-.015
(.193)

Neighborhood-Level Variable

Linguistic Isolation -.013
(.016)

.008
(.004)
¶

-.009
(.005)

*
.024
(.015)

Level 2 Variance
.227 .013 .028 .197
Level 1 Variance
8.346 .703 .647 8.757
Chi-Square
8.50** 3.90* 18.16*** 6.02**
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

129

tables, the level of ICSN was higher for older, more educated participants who had lived in their
neighborhoods for a longer period of time, or who were first generation immigrants.
H1i proposed a direct effect of neighborhood-level linguistic isolation on participants’
connections to neighborhood storytelling resources (H1i-1; here, interpersonal discussion and
local/ethnic media connections), connections to English language media (H1i-2), and ICSN
(H1i-3). As shown in Table 5.15, controlling for individual-level covariates, linguistic isolation
was a marginally significant neighborhood-level predictor of participants’ connections to
local/ethnic media (B = .008, SE = .004, p = .061), and a significant predictor of participants’
connections to English language media (B = -.009, SE = .005, p = .054). In other words, holding
individual-level covariates constant, participants living in more linguistically isolated areas
tended to connect to local/ethnic media more, and were significantly less likely to connect to
English-language media, compared to those from less linguistically isolated areas.
On the other hand, the analysis did not reveal significant differences in participants’
intensity of interpersonal discussion about their neighborhoods and in the level of ICSN. Thus,
the proposed influence of neighborhood-level linguistic isolation received marginal support for
neighborhood storytelling resources (H1i-1; for local/ethnic media connections), sufficient
support for connections to English language media (H1i-2), but no support for ICSN (H1i-3).
In the same fashion, H1j posited the direct effects of neighborhood-level ethnic
heterogeneity. Table 5.16 shows that ethnic heterogeneity had a negative and marginally
significant effect on interpersonal discussion (B = -1.140, SE = .691, p = .099), a negative and
significant effect on local/ethnic media connection (B = -.395, SE = .184, p = .033), and a
negative and significant effect on ICSN (B = -1.342, SE = .692, p = .052). That is, compared to
participants in ethnically heterogeneous neighborhoods, participants living in more ethnically
130

Table 5.16 Hierarchical Linear Models of Connections to Neighborhood Storytelling Resources,
English Language Media, and ICSN: Individual-Level Covariates and Neighborhood-Level
Ethnic Heterogeneity (Full Models)
Predictors
Interpersonal
Discussion
Local/Ethnic
Media
English
Media
ICSN
Coefficient (Standard Error)

Intercept

3.768
(.337)***
1.594
(.096)***
1.027
(.098)***
6.934
(.345)***

Individual-Level Variables

Age .033
(.012)**
.004
(.004)
-.001
(.003)
.052
(.013)***

8th Grade and Above .266
(.215)
.105
(.062)
¶

.188
(.060)**
.622
(.220)**

English Proficiency .119
(.119)
-.062
(.034)
¶

.621
(.033)***
.034
(.121)

Residential Tenure .039
(.011)***
.005
(.003)
¶

-.001
(.003)
.036
(.011)**

First Generation
Immigrants
.192
(.311)
.393
(.089)***
-.018
(.087)
1.060
(.318)**

Compliance .018
(.189)
-.085
(.055)
.031
(.053)
-.006
(.193)

Neighborhood-Level Variable

Ethnic Heterogeneity -1.140
(.691)
¶

-.395
(.184)

*
-.199
(.241)

-1.342
(.692)

*

Level 2 Variance .202 .012 .033 .195
Level 1 Variance 8.342 .703 .647 8.748
Chi-square 7.78** 4.38* 20.29*** 7.25**
¶
p <.10,*p <.05, ** p <.01, ***p <.001

131

homogeneous neighborhoods tended to talk about things happening in their neighborhoods less
often, were significantly more likely to connect to local/ethnic media, and were significantly
more likely to have an integrated neighborhood storytelling network. Thus, H1j-1 received
sufficient support for local/ethnic media connection, and marginal support for interpersonal
discussion; H1j-3 too had sufficient support. By comparison, no significant difference was
found in participants’ English lang uage media connection between ethnically homogenous and
heterogeneous areas. H1j-2 was thus rejected.
Table 5.17 presents the results from testing H1k, which hypothesized the direct effect of
neighborhood-level density of communication resources. As shown in the first column, there
was a negative and marginally significant effect of the density of communication resources on
interpersonal discussion, B = -.059, SE = .032, p = .068. This indicates a tendency that
participants from areas with fewer communication resources per 10,000 residents discussed their
neighborhoods more often. Meanwhile, no significant or marginally significant difference
existed in local/ethnic media connections, English language media connection (H1k-2), or ICSN
(H1k-3) between areas differed by density of communication resources. Thus, the proposed
effect of neighborhood-level density of communication resources received marginal support for
individual neighborhood storytelling resources (H1k-1) when interpersonal discussion was
concerned, but not for connections to English language media (H1k-2) or ICSN (H1k-3).
Finally, Table 5.18 provides the results from testing H1l, which proposed the direct effect
of neighborhood-level density of health service providers on connections to individual
storytelling resources and ICSN. As it turned out, none of the models tested revealed a
significant effect of density of health service providers. Thus, it can be concluded that, for the
study sample, Latina’s frequency of interpersonal discussio n about their neighborhoods,
132

Table 5.17 Hierarchical Linear Models of Connections to Neighborhood Storytelling Resources,
English Language Media, and ICSN: Individual-Level Covariates and Neighborhood-Level
Density of Communication Resources (Full Models)
Predictors
Interpersonal
Discussion
Local/Ethnic
Media
English
Media
ICSN
Coefficient (Standard Error)

Intercept

3.764
(.336)***
1.596
(.098)***
1.028
(.098)***
6.935
(.346)***

Individual-Level Variables
Age .033
(.012)**
.004
(.004)
-.001
(.003)
.052
(.013)***

8th Grade and Above .251
(.215)
.107
(.062)
¶

.188
(.060)**
.616
(.221)**

English Proficiency .123
(.119)
-.063
(.034)
¶

.622
(.033)***
.035
(.121)

Residential Tenure .039
(.011)***
.005
(.003)
¶

-.001
(.003)
.036
(.011)**

First Generation
Immigrants
.202
(.311)
.392
(.090)***
-.016
(.087)
1.063
(.319)**

Compliance .011
(.189)
-.089
(.055)
.031
(.053)
-.014
(.193)

Neighborhood-Level Variable
Density of Comm.
Resources
-.059
(.032)
¶

-.008
(.010)
.005
(.011)

-.044
(.034)

Level 2 Variance .181 .017 .034 .218
Level 1 Variance 8.348 .703 .647 8.753
Chi-Square 6.03** 8.46** 22.58*** 8.14**
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

133

Table 5.18 Hierarchical Linear Models of Connections to Neighborhood Storytelling Resources,
English Language Media, and ICSN: Individual-Level Covariates and Neighborhood-Level
Density of Health Service Providers (Full Models)
Predictors
Interpersonal
Discussion
Local/Ethnic
Media
English
Media
ICSN
Coefficient (Standard Error)

Intercept

3.786
(.340)***
1.602
(.098)***
1.029
(.098)***
6.958
(.348)***

Individual-Level Variables
Age

.033
(.012)**
.004
(.004)
-.001
(.003)
.052
(.013)***

8th Grade and Above .257
(.216)
.105
(.062)
¶

.186
(.060)**
.616
(.221)**

English Proficiency .120
(.119)
-.063
(.034)
¶

.622
(.033)***
.032
(.121)

Residential Tenure .039
(.011)***
.006
(.003)
¶

-.001
(.003)
.036
(.011)**

First Generation
Immigrants
.197
(.311)
.392
(.090)***
-.016
(.087)
1.058
(.319)**

Compliance .010
(.189)
-.090
(.055)
.030
(.053)
-.018
(.193)

Neighborhood-Level Variable
Density of Comm.
Resources
-.189
(.215)
-.076
(.061)
-.063
(.072)

-.224
(.220)

Level 2 Variance 0.225 0.018 0.033 0.237
Level 1 Variance 8.346 0.702 0.647 8.749
Chi-Square 8.87** 9.57** 22.52*** 9.79***
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

134

local/ethnic media connections, English language media connections, or ICSN did not differ by
density of health service providers at the neighborhood level. Taken together, the proposed
effect of neighborhood-level density of health service providers did not receive support for
connections to neighborhood storytelling resources (H1l-1), connections to English language
media (H1l-2), or ICSN (H1l-3).
Summary of Multilevel Modeling Analysis Finding
The previous sections report the findings about the effects of connecting to neighborhood
storytelling resources as well as neighborhood context on descriptive norms and media recall
regarding Pap tests, and the effects of neighborhood context on connections to storytelling
resources. Table 5.19 summarizes the results pertaining to descriptive norms (perceived
prevalence of Latinas who have never had a Pap test) and media recall for the noncompliant
sample. Interpersonal discussion with others about one’s neighborhoods had a positive and
significant effect on descriptive norms, and a positive and marginally significant effect on media
recall. Connections to local/ethnic media had a significant and positive effect on media recall.
ICSN had a positive yet marginally significant effect on descriptive norms, and a positive and
significant effect on media recall. Neighborhood-level linguistic isolation had a significant
effect on the strength of ICSN as a predictor of descriptive norms. Neighborhood-level density
of health service providers had a negative and significant effect on media recall.
Table 5.20 provides a summary of findings regarding the contextual effects on
connections to storytelling resources for the full sample. At the level of .05, the hypothesized
effects of neighborhood characteristics were confirmed for linguistic isolation on connections to
English language media, and for ethnic heterogeneity on connections to local/ethnic media and
ICSN. Moreover, linguistic isolation was a marginally significant and positive predictor of
135

connections to local/ethnic media. Ethnic heterogeneity was a marginally significant and
negative predictor of interpersonal discussion. Last, density of communication resources served
as a marginally significant and negative predictor of interpersonal discussion. In the next
section, findings from the structural equation modeling analysis of the structural relationships
between neighborhood experience variables, connections to storytelling resources, health
communication outcomes, descriptive norms, and compliance with cervical cancer screening
guidelines are presented. The theoretical, methodological, and practical implications of these
findings are discussed in Chapter 6.

Table 5.19 Summary of Findings of Hierarchical Linear Models: Descriptive Norms and Media
Recall Regarding Pap tests (Noncompliant Sample)
Hypotheses and Research Questions
Re:
Descriptive Norms
(never had a Pap test)
Media Recall

Neighborhood Storytelling Resources
Interpersonal Discussion H1a: Yes (positive) H1e: Marginal (positive)
Local/Ethnic Media H1b: NS H1f: Yes (positive)
Community Organizations H1c: NS H1g: NS
ICSN H1d: Marginal (positive) H1h: Yes (positive)

Neighborhood-Level Characteristics
Linguistic Isolation RQ1a-1: NS RQ1c-1: NS
Ethnic Heterogeneity RQ1a-2: NS RQ1c-2: NS
Density of Comm. Resources RQ1a-3: NS RQ1c-3: NS
Density of Health Service Providers RQ1a-4: NS RQ1c-4: Yes (negative)

Cross-Level Interactions
ICSN × Linguistic Isolation RQ1b-1: Yes (positive) RQ1d-1: NS
ICSN × Ethnic Heterogeneity RQ1b-2: NS RQ1d-2: NS
ICSN × Density of Comm. Resources RQ1b-3: NS RQ1d-3: NS
ICSN × Density of Health Services RQ1b-4: NS RQ1d-4: NS
Note: “Yes (positive)” = positive and significant effect, “Marginal ( positive)” = positive and
marginally significant effect, “ Yes (negative)” = negative and significant effect, “NS”= non -
significant effect.

136

Table 5.20 Summary of Findings of Hierarchical Linear Models: Neighborhood Storytelling
Resources, English Language Media, and ICSN (Full Sample)
Hypotheses Re:
Interpersonal
Discussion
Local/Ethnic
Media
English
Media
ICSN
H1i: Linguistic Isolation NS Marginal
(positive)
Yes
(negative)

NS
H1j: Ethnic Heterogeneity Marginal
(negative)
Yes
(negative)

NS Yes
(negative)
H1k: Density of Comm. Resources Marginal
(negative)

NS NS NS
H1l: Density of Health Service
Providers
NS NS NS NS
Note: “Yes (positive)” = positive and significant effect, “Marginal ( positive)” = positive and
marginally significant effect, “ Yes (negative)” = negative and significant effect, “Marginal
(negative)” = negative and marginally significant effect, “NS”= non -significant effect.

Structural Equation Modeling Analysis of Descriptive Norms
The previous sections present the results of multilevel analysis of descriptive norms
regarding Pap tests, recall of having seen or heard information about Pap tests in the media, and
neighborhood storytelling resources. The purpose of the current section is to examine the
structural relation among neighborhood experience, storytelling resources, media recall, attention
to information about Pat tests in the media, discussion about Pap tests with healthcare
professionals, descriptive norms regarding Pap tests, and screening compliance for the whole
study sample (N = 1116). The hypotheses put forth to address the structural relation among these
variables were explored using structural equation modeling via LISREL 8.7. All the
hypothesized paths were estimated separately for participants who had lived in their
neighborhoods for 6 years or more (N = 757; hereafter referred to as the “longer tenure group”)
and for those who have lived in their neighborhoods for less than 6 years (N = 352; hereafter
137

referred to as the “shorter tenure group”). The 6-year cutoff point was based on data distribution,
where about 68% of the study sample had lived in their neighborhoods for six years or more.
Results for Participants Having Lived in Their Neighborhood for 6 Years or More
Estimation of the hypothesized model yielded a satisfactory fit for the longer tenure
group. The Chi-square was non-significant (χ
2
= 38.73, p = .067, df = 27), suggesting a fair
amount of similarity between the reconstructed and observed matrices. The ratio of Chi-square
to degrees of freedom was acceptable at χ
2
/df = 1.43 (Wheaton, Muthen, Alwin, & Summers,
1977). The goodness of fit index (GFI), comparative fit index (CFI), and normed fit index (NFI)
all approached 1 (GFI = .99, CFI = .98, NFI = .95). The root mean square error of
approximation was also satisfactory (RMSEA = .026, 90% CI = 0 to.044).
Figure 5.1 presents the coefficients for the structural model that tested H2a to H2k. The
model estimation provided empirical support for all 22 but 3 hypothesized paths. The three
unsupported hypotheses posited a direct association between ICSN and attention to information
about Pap tests in the media (H2c-2; γ = .17, SE = .22, p > .05), between attention and
descriptive norms regarding never having a Pap test (H2h-1; β = -.09, SE = .08, p > .05), and
between attention and descriptive norms regarding having ever received abnormal Pap test
results (H2h-3; β = -.02, SE = .07, p > .05).
Consistent with H2a, immigration generation had a significant direct effect on screening
compliance for the current sample of participants whose years of residence in their
neighborhoods were six years or more. Compared to second or higher generation immigrants,
first generation immigrants were more likely to be in compliance with cervical cancer screening
guidelines (γ = .30, SE = .04, p <.001). In line with H2b-1 and H2b-2, immigration generation
was also significantly associated with two types of storytelling resources: ICSN and English
138

Figure 5.1 Hypothesized model for participants who had lived in their neighborhoods for six years or more. Standardized solution.
Non-significant paths are presented with dash lines. *p ≤ .05, ** p ≤ .01, *** p ≤ .001

139

language media. That is, in the longer tenure group, first generation immigrants had a greater
level of connections to an integrated storytelling network (γ = 1.00, SE = .29, p <.001), but a
narrower scope of connections to English language media (γ = -.76, SE = .06, p <.001),
compared to second or higher generation immigrants.
In terms of the effects of connecting to storytelling resources, participants’ recall of
having seen or heard information about Pap tests in the media in the past 30 days was
significantly and positively influenced by ICSN (H2c-1; β = .45, SE = .08, p <.001), but
negatively influenced by connections to English language media (H2d-1; β = -.26, SE = .03, p
<.001). ICSN was negatively associated with discussion about Pap tests with healthcare
professionals (H2c-3; β = -.11, SE = .06, p <.05). Connection to English media was related to
decreased attention to information about Pap tests in the media (H2d-2; β = -.19, SE = .06, p
<.001), but increased discussion about Pap tests with healthcare professionals (H2d-3; β = .19,
SE = .04, p <.001). The indirect effect of English media connections through ICSN was also
confirmed (H2e; β = 1.08, SE = .32, p <.001). That is, the more participants were connected to
mainstream English language media, the more likely they were to be connected to an integrated
storytelling network consisting of local communication resources as well.
Furthermore, media recall had a significant and positive association with attention to Pap
tests-related information in the media (β = .37, SE = .13, p <.01), which, in turn, positively
predicted discussion with healthcare professionals (β = .18, SE = .04, p <.001), respectively in
accordance with H2f and H2g. Attention had a significant and positive association with
descriptive norms regarding other women having had regular Pap tests, supporting H2h-2 (β =
.65, SE = .11, p <.001). All paths proposed from discussion about Pap tests with healthcare
140

professionals to the three types of descriptive norms were significant, supporting H2i-1 (β = -.06,
SE = .03, p <.05), H2i-2 (β = -.15, SE = .04, p <.001), and H2i-3 (β = -.11, SE = .03, p <.001).
In terms of the direct effects on screening compliance, discussion with healthcare
providers positively predicted screening compliance (H2j; β = .27, SE = .06, p <.001). The
proposed paths from descriptive norms regarding never having a Pap test (H2k-1; β = -.08, SE =
.03, p <.01), having regular Pap tests (H2k-2; β = .07, SE = .03, p <.05), and having received
abnormal Pap test results to cancer screening compliance were also all confirmed (H2k-3; β =
.07, SE = .03, p <.05). Finally, the control variable, healthcare coverage, had a significant effect
on descriptive norms regarding having regular Pap tests (γ = .21, SE = .03, p <.001) and having
received abnormal Pap test results (γ = .07, SE = .03, p <.05). In other words, participants with
healthcare coverage were more likely to perceive that it was common for Latinas to receive
regular Pap tests and to have ever received abnormal results. Having healthcare coverage was
also positively associated with screening guidelines (γ = .42, SE = .05, p <.001).
Using LISREL 8.7, a revised model (Figure 5.2) was produced by sequentially deleting
three non-significant paths mentioned above. Non-significant covariances between exogenous
variables were also restricted to zero. Table 5.21 and Table 5.22 respectively present the
corresponding direct, indirect, and total effects on descriptive norms and screening compliance
for the hypothesized and revised models. The revised model did not differ significantly from the
hypothesized model in global model fit (χ
2
= 43.04, p = .11, df = 33, χ
2
/df = 1.30; GFI = .99,
CFI = .99, NFI =. 95; RMSEA = .022, 90% CI = 0 to.039). As shown in the tables, all included
paths changed very little from the hypothesized model to the revised model.
141

Figure 5.2 Revised model for participants who had lived in their neighborhoods for six years or more. Standardized solution.
*p ≤ .05, ** p ≤ .01, *** p ≤ .001

142

Table 5.21 Direct, Indirect, and Total Effects on Descriptive Norms Regarding Pap Tests for Participants Who Had Lived in Their
Neighborhoods for Six Years and More (N=630
a
)
Variables
Descriptive Norms
(Never had a Pap test)

Descriptive Norms
(Have had regular Pap tests)
Descriptive Norms
(Have had abnormal test results)
Direct
Effect
Indirect
Effect
Total
Effect
Direct
Effect
Indirect
Effect
Total
Effect
Direct
Effect
Indirect
Effect
Total
Effect

Hypothesized model
First gen. immigrants -- -.02 -.02 -- .20*** .20*** -- .01 .01
Healthcare coverage -.04 .00 -.04 .21*** -.01 .20*** .07* .00 .07*
ICSN -- -.03 -.03 -- .23 .23 -- .00 .00
English media -- -.01 -.01 -- -.04 -.04 -- -.01 -.01
Media Recall -- -.04 -.04 -- .23** .23** -- -.01 -.01
Media attention -.09 -.01 -.10 .65*** -.03** .62*** -.02 -.02** -.04
Discussion -.06* -- -.06* -.15*** -- -.15*** -.11*** -- -.11***
Revised model
First gen. immigrants -- .01* .01* -- .19*** .19*** -- .01* .01*
Healthcare coverage -- -- -- .20*** -- .20*** .08** -- .08**
ICSN -- .00 .00 -- .11** .11** -- .00 .00
English media -- -.01 -.01 -- -.10 -.10 -- -.01 -.01
Media recall -- .00 .00 -- .25** .25** -- -.01* -.01*
Media attention -- -.01* -.01* .63*** -.03** .61*** -- -.02** -.02**
Discussion -.07* -- -.07* -.15*** -- -.15*** -.11*** -- -.11***
a
N represents the number of participants who provided valid responses to all variables included in the estimation.
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001
143

Table 5.22 Direct, Indirect, and Total Effects on Screening Compliance for Participants Who Had Lived in Their Neighborhoods for
Six Years and More (N = 630
a
)
Variables
Hypothesized Model Revised Model
Direct
Effect
Indirect
Effect
Total
Effect

Direct
Effect
Indirect
Effect
Total
Effect

First generation immigrants .30*** -.02 .29*** .30***
-.01 .28***
Healthcare coverage .42*** .02*** . 45*** .44***
.02*** . 46***
ICSN -- .00 .00 --
.00 .00
English media -- .03 .03 --
.01 .01
Media recall -- .04* .04* --
.04** .04**
Media attention -- .10*** .10*** --
.10*** .10***
Discussion .27*** -.01 .26*** .26***
-.01 .25**
Descriptive norms
(Never had a Pap test)
-.08** -- -.08** -.10*** -- -.10***
Descriptive norms
(Have had regular Pap tests)
.07* -- .07* .09** -- .09***
Descriptive norms
(Have received abnormal results)
.07* -- .07* .06* -- .06*
a
N represents the number of participants who provided valid responses to all variables included in the estimation.
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

144

Results for Participants Having Lived in Their Neighborhood for Less Than 6 Years
Model estimation and revision were conducted in a similar fashion for the shorter tenure
group. Estimation of the hypothesized model indicates an acceptable fit to the data. The Chi-
square was non-significant (χ
2
= 15.67, p = .96, df = 27), and the ratio of Chi-square to degrees
of freedom was satisfactory at χ
2
/df =.58 (Wheaton, et al., 1977). The values of goodness of fit
index (GFI), comparative fit index (CFI) and normed fit index (NFI) were either close or at 1
(GFI = 1.00, CFI = 1.00, NFI = .96). The root mean square error of approximation took the
value of zero. This is not surprising. RMSEA is estimated as zero when the Chi-square value is
smaller than the degrees of freedom (Kenny, Kaniskan, & McCoach, 2014).
Coefficients produced from testing the hypothesized model are illustrated in Figure 5.3.
Only half of the 22 hypothesized paths were significant from testing the conceptual model,
suggesting room for improvement. To enhance the mode, non-significant paths were
sequentially deleted, and non-significant covariances between exogenous variables were
constrained to zero. During this iterative process, some of the non-significant paths became
significant, and vice versa.
The final model for participants having lived in their neighborhoods for less than 6 years
contained paths that were both theoretically plausible and statically significant (Figure 5.4).
Direct, indirect, and total effects on descriptive norms and screening compliance are respectively
presented in Table 5.23 and Table 5.24 for hypothesized and revised model. Assessment of the
revised model pointed to a satisfactory model fit (χ
2
= 44.95, p = .27, df = 40, χ
2
/df = 1.12; GFI
= .99, CFI = .98, NFI =. 88; RMSEA = .021, 90% CI = 0 to.048).
145

Figure 5.3 Hypothesized model for participants who had lived in their neighborhoods for less than six years. Standardized solution.
Non-significant paths are presented with dash lines.
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

146

Figure 5.4 Revised model for participants who had lived in their neighborhoods for less than six years. Standardized solution.
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001
147

Table 5.23 Direct, Indirect, and Total Effects on Descriptive Norms for Participants Who Had Lived in Their Neighborhoods for Less
Than Six Years (N = 284
a
)
Variables
Descriptive Norms
(Never had a Pap test)

Descriptive Norms
(Have had regular Pap tests)
Descriptive Norms
(Have had abnormal test results)
Direct
Effect
Indirect
Effect
Total
Effect
Direct
Effect
Indirect
Effect
Total
Effect
Direct
Effect
Indirect
Effect
Total
Effect

Hypothesized model
First gen. immigrants -- .00 .00 -- .09** .09** -- .05 .05
Healthcare coverage .02 .00 .02 -.04 .03 -.01 .10* .01 .12**
ICSN -- -.03 -.03 -- .29* .29* -- .16 .16
English media -- -.02 -.02 -- .10 .10 -- .05 .05
Media Recall -- -.02 -.02 -- .24 .24 -- .13 .13
Media attention -.04 -.01 -.05 .49** -.02 .47** .27 -.01 .26
Discussion -.10 -- -.10 -.16** -- -.16** -.10* -- -.10*
Revised model
First gen. immigrants -- -- -- -- .09*** .09*** -- .00 .00
Healthcare coverage -- -- -- -- .03* .03* .11** -- .11**
ICSN -- -- -- -- .29*** .29*** -- -- --
English media -- -- -- -- -.01 -.01 -- -.01 -.01
Media recall -- -- -- -- .55*** .55*** -- -- --
Media attention -- -- -- .65*** -- .65*** -- -- --
Discussion -- -- -- -.11** -- -.11** -.08* -- -.08*
a
N represents the number of participants who provided valid responses to all variables included in the estimation.
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001
148

Table 5.24 Direct, Indirect, and Total Effects on Screening Compliance for Participants Who Had Lived in Their Neighborhoods for
Less Than Six Years (N =284
a
)
Variables
Hypothesized Model Revised Model
Direct
Effect
Indirect
Effect
Total
Effect

Direct
Effect
Indirect
Effect
Total
Effect

Immigration generation .17** .01 .19** .18** .00 .18**
Healthcare coverage .45*** .03 .48*** ¤ .49*** .03** .53***
ICSN -- .11** .11** -- .05*** .05***
English media -- .08* .08* -- .02 .02
Media recall -- .09* .09* -- .10*** .10***
Media attention -- .17*** .17*** -- .12*** .12***
Discussion .34*** -.05** .29** .28*** -.04*** .24**
Descriptive norms
(Never had a Pap test)
-.01 -- -.01 -- -- --
Descriptive norms
(Have had regular Pap tests)
.16** -- .16** .18*** -- .18***
Descriptive norms
(Have received abnormal results)
.24*** -- .24*** .26*** -- .26***
a
N represents the number of participants who provided valid responses to all variables included in the estimation.
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

149

As for individual paths, Latinas who immigrated from other countries were significantly
more likely to be in compliance with cervical cancer screening guidelines than Latinas who were
born in the United States (γ = .18, SE = .06, p <.01). Latinas who were first generation
immigrants also reported significantly higher level of ICSN (γ = .30, SE = .04, p <.001), but
fewer connections to English language media (γ = -.48, SE = .06, p <.001). ICSN was a
significant predictor of participants’ recall of having seen or heard information about Pap tests in
the media (β = .52, SE = .08, p <.001), but was not associated with either attention to
corresponding information in the media or discussion about Pap tests with healthcare
professionals.
Connections to English language media had a positive association with discussion with
healthcare providers (β = .10, SE = .04, p <.05), but not with media recall or attention. Notably,
the proposed indirect effect of English language media connections through ICSN was not
confirmed for the current sample of participants who had lived in their neighborhoods for less
than 6 years. This indicates that, in this subsample, there was a disconnect between participants’
connections to English language media and local storytelling resources, such that connections to
a broader scope of English language media did not increase (or decrease) participants’
connections to the integrated neighborhood storytelling network.
Next, recall of having seen or heard information about Pap tests in the media was
significantly and positively associated with attention to such information in the media (β = .85,
SE = .13, p <.001). Contrary to what was anticipated, attention was not associated with
discussion with healthcare providers about Pap tests. Furthermore, attention to information
about Pap tests in the media had a direct effect on descriptive norms regarding having regular
Pap tests only (β = .65, SE = .18, p <.001), whereas discussion with healthcare professionals
150

about Pap tests had a direct effect on descriptive norms with respect to both having regular Pap
tests (β = -.11, SE = .04, p <.01) and having received abnormal Pap tests results (β = -.08, SE =
.04, p <.05), as well as a positive effect on screening compliance (β = .28, SE = .08, p <.001).
Of the three descriptive norms constructs, descriptive norms regarding having had regular Pap
tests (β = .18, SE = .04, p <.001) and having received abnormal tests results (β = .26, SE = .04, p
<.001) both had a significant effect on screening compliance, but descriptive norms with respect
to never having a Pap test did not.
Finally, the control variable, healthcare coverage, was significantly and positively
associated with ICSN (γ = .10, SE = .04, p <.05), descriptive norms regarding abnormal Pap test
results (γ = .11, SE = .04, p <.01), and screening compliance (γ = .49, SE = .08, p <.001). In
other words, compared to uninsured women, insured women were more likely to have an
integrated neighborhood storytelling network, to perceive that it was common for Latinas to have
received an abnormal Pap test result, and to be in compliance with cervical cancer screening
guidelines for participants who had lived in their neighborhoods for less than 6 years.
Post-Hoc Analysis
In both longer and shorter tenure groups, path analyses showed that first generation
immigrants were more likely to have been in compliance with cervical cancer screening than
second or high generation immigrants. Because this finding was contrary to expectation, post-
hoc analyses were conducted to further investigate the association between immigration
generation and screening compliance. A closer look of the data revealed that first generation
immigrants were significantly older than second and higher generation immigrants in both longer
tenure group (M
native
= 40.00, SD
native
= 7.20, M
foreign-born
= 29.01, SD
foreign-born
= 7.49, t = -16.32,
151

p <.001) and shorter tenure groups (M
native
= 36.65, SD
native
= 8.62, M
foreign-born
= 30.16, SD
foreign-born
= 8.07, t = -5.30, p <.001).
Table 5.25 Logistic Regression Models Predicting Compliance with Cervical Screening
Guidelines from Immigration Generation, Age, Health care coverage, Storytelling Resources
Connections, Health Communication Outcomes and Descriptive Norms Regarding Pap Tests for
Longer and Shorter Tenure Groups
Predictors
Longer Tenure Group
OR (95%CI)
Shorter Tenure Group
OR (95%CI)
Age

.991 (.967 – 1.017) 1.008 (.974 – 1.044)
First generation immigrants

1.833 (1.042 – 3.225) 2.274 (1.025 – 5.044)
Health care coverage

3.819 (2.633 – 5.539) 4.189 (2.405 – 7.296)
ICSN

.971 (.914 – 1.032) 1.052 (.948 – 1.166)
English Language Media

.735 (.605 – .891) 1.054 (.796 – 1.395)
Media Recall

.984 (.800 – 1.211) 1.081(.783 – 1.494)
Media Attention

1.061 (.998 – 1.129) 1.072(.976 – 1.177)
Pap Discussions with Healthcare Providers

2.427 (1.628 – 3.619) 2.417(1.324 – 4.413)
Descriptive norms (Never had a Pap test)

.991 (.984 – .999) .994(.982 – 1.006)
Descriptive norms (Have regular Pap test)

1.007 (.999 – 1.016) 1.007 (.994 – 1.020)
Descriptive norms (Have abnormal results)

1.005 (.996 – 1.014) 1.021 (1.008 – 1.034)

A logistic regression was then run for both groups in order to test if the association
between immigration generation and screening compliance would retain its significance after
accounting for the potential effect of age. As shown in Table 5.25, being first generation
immigrations was significantly and positively associated with compliance status for both longer
and shorter tenure groups, controlling for age and other variables examined in the structural
equation models above. In the longer tenure group, first generation immigrations were about
152

83% (OR=1.833, 95% CI=1.042 – 3.225) more likely than were second or higher generation
immigrants to have been in compliance with cervical cancer screening recommendations.
Likewise, in the shorter tenure group, first generation immigrants were approximately 127%
more likely to have been in compliance than second or higher generation immigrants (OR=2.274,
95% CI=1.025 – 5.044). By comparison, age was not significantly associated with compliance
status for either group.
Difference in Structural Relationships by Years of Residence in Neighborhoods
RQ2 inquired if there would be substantial difference in the structural relationships
among neighborhood experience, storytelling resources, health communication outcomes,
descriptive norms, and screening compliance between participants in the longer and shorter
tenure groups. Table 5.26 summarizes the results from testing H2a to H2k by participants’ years
of residence. Together with aforementioned tables and figures produced from examining the
structural relationships of interest, these back-to-back comparisons revealed considerable
differences between the two groups of participants. For both groups, the influence of
immigration generation, which was considered an individual-level structural position variable in
the analysis, was partially mediated by participants’ connections to storytelling resources.
Specifically, being first generation immigrants increased participants’ level of integration into
their neighborhood storytelling network, but decreased their scope of connections to English
language media. A meaningful difference emerged with a closer look at the size of the
coefficients as shown in Figure 5.2 and Figure 5.4. For example, in the longer tenure group, first
generation immigrants saw an increase in their level of ICSN that was nearly three times that for
their counterparts in the shorter tenure group. Moreover, with a direct path from English media
connections to ICSN, participants in the longer tenure group also had a notably more integrated
153

Table 5.26 Summary of Findings of Structural Equation Modeling Analysis by Participants’
Residential Tenure Based on the Revised Models
Hypotheses and Research Questions Re:
Residential Tenure
≥ 6 years
Residential Tenure
< 6 years
Effects of immigration generation
H2a (First generation immigration  compliance)
Yes (Positive)
Yes (Positive)
H2b-1 (Immigration generation  ICSN)
Yes (Positive)
Yes (Positive)
H2b-2 (Immigration generation  English media) Yes (Negative) Yes (Negative)
Effects of storytelling resources
H2c-1 (ICSN  media recall)
Yes (Positive) Yes (Positive)
H2c-2 (ICSN  media attention)
NS NS
H2c-3 (ICSN  discussion with providers) Yes (Negative) NS
H2d-1 (English media  media recall) Yes (Negative) NS
H2d-2 (English media  media attention) Yes (Negative) NS
H2d-3 (English media  discussion with providers)
Yes (Positive) Yes (Positive)
H2e (English media  ICSN)
Yes (Positive) NS
Effects of health communication outcomes
H2f (media recall  attention) Yes (Positive)
Yes (Positive)
H2g (media attention  discussion) Yes (Positive)
NS
H2h-1 (attention  descriptive norms [never])
NS NS
H2h-2 (attention  descriptive norms [regular])
Yes (Positive) Yes (Positive)
H2h-3 (attention  descriptive norms [results])
NS NS
H2i-1 (discussion  descriptive norms [never]) Yes (Negative)
NS
H2i-2 (discussion  descriptive norms [regular]) Yes (Negative) Yes (Negative)
H2i-3 (discussion  descriptive norms [results])
Yes (Negative)
Yes (Negative)
H2j (discussion  compliance)
Yes (Positive) Yes (Positive)
Direct effects of descriptive norms
H2k-1 (descriptive norms [never] compliance)
Yes (Negative) NS
H2k-2 (descriptive norms [regular] compliance) Yes (Positive) Yes (Positive)
H2k-3 (descriptive norms [results] compliance) Yes (Positive) Yes (Positive)
Direct effects of healthcare coverage (control variable)
Healthcare coverage  ICSN NS Yes (Positive)
Healthcare coverage  English media
NS NS
Healthcare coverage  descriptive norms [never]
NS NS
Healthcare coverage  descriptive norms [regular] Yes (Positive)
NS
Healthcare coverage  descriptive norms [results] Yes (Positive) Yes (Positive)
Healthcare coverage compliance Yes (Positive) Yes (Positive)
Note: “Yes (positive)” = positive and significant effect, “ Yes (negative)” = negative significant
effect, “NS”= non-significant effect.

154

storytelling system compared to those in the shorter tenure group. The difference in the paths
from storytelling resources to health communication outcomes between the two groups of
participants was particularly noteworthy. Although in both groups ICSN had an indirect effect
through media recall (i.e., recall of having seen or heard information about Pap tests in the media
in the past 30 days) and on media attention (i.e., attention to information about Pap tests in the
media), only media attention had a direct effect on discussion about Pap tests with healthcare
professionals only for participants in the longer tenure group.
Among participants with more years of residence, discussion with healthcare
professionals had a direct effect on all three types of descriptive norms, which each predicted
screening compliance. This, however, was not replicated among participants in the shorter
tenure group, where descriptive norms regarding never having a Pap test were not associated
with any of the variables in the model. Together with the findings of the multilevel modeling
analysis, results obtained from the structural equation modeling analysis and their implications
will be discussed in Chapter 6.

155

CHAPTER 6 : DISCUSSION AND CONCLUSION

This dissertation examines the communication mechanisms that underlie the formation of
perceived norms regarding Pap tests for Latinas living in urban ethnic neighborhoods in Los
Angeles. Communication in this study is conceptualized as the elementary social process that
links people’s multiple exposure to normative information in their day -to-day environment with
the development of perceived norms. The specific perceived norms explored in this study
pertain to descriptive norms, defined as people’s perceived p revalence of a given behavior
among similar others. Central to this investigation are the roles of a neighborhood storytelling
network (STN) — a local information network made up of residents, local/ethnic media, and
community organizations — in shaping descriptive norms in a neighborhood environment,
wherein people’s everyday life unfolds.
Participants for this study came from the Multilevel Study (R01CA155326 -
Murphy/Ball-Rokeach), a large multilevel and multi-method research project that systematically
examines the barriers and conduits to cervical cancer prevention, detection, and treatment at the
individual, interpersonal, and community level for Hispanic women in Los Angeles. These
participants fall into two subgroups, depending on their compliance status with cervical cancer
screening guidelines: Latinas who had received a Pap test in the past 3 years (the “compliant”
group); and Latinas who had never had a Pap test or had not received one in over 3 years (the
“noncompliant” group).
The preceding chapter reports the findings from exploring two groups of hypotheses and
research question. The first group examined the relationship between the neighborhood
storytelling network, neighborhood context, and Latinas’ descriptive norms regarding Pap tests .
156

The second investigated the structural relationship between Latinas’ neighborhood experience,
connections to storytelling resources, health communication outcomes, descriptive norms, and
the primary dependent variable of interest.
Overall, strong empirical evidence was found supporting the important role of local
communication resources in shaping Latinas’ perceived prevalence of cervical cancer screening
and detection among their ethnic group. Moreover, as expected, the communication processes
underlying the development of descriptive norms varied meaningfully across participants in a
way that reflected the dynamic interaction between participants and their neighborhood
environment. The key findings of this study are discussed below, and followed by their
theoretical, methodological, and practical contributions.
Key Study Findings
Descriptive Norms and Screening Compliance
Results of descriptive analysis revealed significant differences in descriptive norms
regarding Pap tests between compliant and noncompliant participants. The three descriptive
norms constructs examined in this study are: 1) perceived prevalence of never having a Pap test;
2) perceived prevalence of having a Pap test at least every 3 years; and 3) perceived prevalence
of having ever had an abnormal Pap test result. Compared to compliant participants,
noncompliant participants perceived it as significantly more common for Latinas like them to
never have had a Pap test, and perceived it as significantly less common to have had a Pap test in
at least every three years. As discussed in more details in next sections, patterns of these
findings are largely consistent with the theoretical predictions of normative influence (e.g.,
integrative model of behavioral prediction and theory of normative social behavior). Meanwhile,
although research on cervical cancer-related descriptive norms specific to Latina population is
157

relatively limited, similar observations have been reported. For example, Byrd and colleagues
(2004) found that only 61 percent of their sample of young Latinas reported that most young
women they knew had received a Pap test.
In addition, although descriptive norms regarding never having a Pap test and descriptive
norms regarding regular Pap tests were conceived of as two opposite ends on a continuum of
descriptive norms related to cervical cancer screening and detection behaviors, a negative
correlation between the two constructs was found for compliant participants, but not for
noncompliant participants. One plausible reason for this finding may be the composition of the
current noncompliant sample. As mentioned above, noncompliant participants in this study
include women who had never received a Pap test and those who had not had a Pap test in the
past 3 years. The majority of the present noncompliant sample (83%) belonged to the latter
category. Therefore, it is possible that for the current noncompliant sample, what they perceived
as the opposite to having regular Pap tests was not never having a Pap test, but having not had a
Pap test in the past 3 years (“overdue” Pap test), therefore the lack of significant correlation
between descriptive norms regarding never having a Pap test and descriptive norms regarding
regular Pap tests.
Neighborhood Storytelling Resources, Neighborhood Context, and Descriptive Norms
This study was based on the premise that the formation of descriptive norms is a
multilevel phenomenon under the influence of people’s specific symbolic, social, and
neighborhood environment in the context of everyday life. Analytically, this translates into a
question concerning the existence of significant between-neighborhood variance in descriptive
norms, an indication that factors related to both people and their neighborhood contexts are at
158

work. As noted in the preliminary steps of multilevel modeling analysis in Chapter 5, the
significant between-neighborhood variance was present only for noncompliant participants.
The pattern of this finding might have to do with the sites of survey participant
recruitment in the Multilevel Study. Noncompliant women were recruited from clinic waiting
rooms, where the women awaited not for their own care but for friends or family, and from
public spaces in the 25 neighborhoods defined by the Multilevel Project. Compliant participants
were recruited from women who were waiting to receive a Pap test in the clinics. As discussed
in more details below, having discussed Pap tests with healthcare professionals may help correct
patients’ misperceived norms and inflated beliefs regarding Pap tests (Moldovan-Johnson, Tan,
& Hornik, 2014; Viswanath, 2005). This may be one of the reasons for the lack of geographical
variations in descriptive norms among compliant participants.
In terms of specific types of descriptive norms, there was a significant between-
neighborhood variance in descriptive norms regarding never having a Pap test for noncompliant
participants. In other words, in the noncompliant sample, participants’ perception that it was
common for women like them to never have had a Pap test varied significantly across
neighborhoods in the study area. Thus, for a fuller understanding of the development of this
particular normative perception, one should adopt a socio-ecologically oriented approach, and
expand the scope of investigation to include not only individual-level factors, such as
participants’ connections to communication resources t hat convey normative information, but
also participants’ neighborhood context that promotes or constrains such connections.
Of many types of communication resources, this dissertation was particularly interested
in neighborhood storytelling resources. Both the roles of connecting to individual neighborhood
storytelling resources (i.e., residents, local/ethnic media, and community organizations) and to an
159

integrated neighborhood storytelling network (ICSN) in influencing descriptive norms regarding
never having a Pap test were tested for the noncompliant sample, accounting for individual- and
neighborhood-level covariates.
First, among noncompliant participants, women who talked about their neighborhoods
with others more frequently on average reported a higher level of perception that it was common
for Latinas to have never been screened for cervical cancer. This finding is consistent with the
conceptualization of communication infrastructure theory (CIT) regarding the relative
importance of neighborhood discussion in generating and disseminating neighborhood “stories”
that subsequently influence residents’ perceived reality and their capacity to solve related
problems. In their observation of diverse urban neighborhoods in Los Angeles, Ball-Rokeach
and her colleagues (2001) commented that although residents many times talk about things that
happen far beyond the geographical boundaries of their neighborhoods, they carry the most
weight in telling stories about the neighborhoods and their own ethnic groups. Consequently,
one may expect that the most pressing issues affecting a neighborhood or residents’ own ethnic
groups are being discussed within residents’ interpersonal network. Although empirical
evidence in the health domain remains to be seen, this reasoning has received extensive support
from prior research on civic engagement. For example, neighborhood discussion among
residents has been consistently found to have a positive association with residents’ sense of
neighborhood belonging and civic participation in urban ethnic communities, in which civic
engagement has been identified as a pressing challenge due to residents’ diversity on multiple
dimensions such as ethnicity, culture, language, religion, political orientation, and so on (Ball-
Rokeach, et al., 2001; N.-T. N. Chen, et al., 2013). In the context of this study, to the extent that
there is substantial amount of neighborhood storytelling about Latinas’ low compliance with
160

cervical cancer screening, one can expect that stories as such would also be prevalent in the
interpersonal network for the current sample of noncompliant participants. These stories, in turn,
could help shape the corresponding descriptive norms among participants.
Second, though only marginally significant, the positive effect of ICSN on descriptive
norms regarding never having a Pap test provided meaningful support for the proposed role of
connecting to an integrated neighborhood storytelling network in shaping descriptive norms. By
definition, ICSN captures an individual’s structural position in her communication environment,
where her connection to one neighborhood storyteller leads her to other storytellers. A person
receives a high ICSN score only when she is strongly connected to all three storytellers (i.e.,
residents, local/ethnic media, community organizations), rather than strongly connected to one
but weakly to others. According to CIT, a strong, integrated neighborhood storytelling network
not only affords residents increased opportunity to the information they need for solving
everyday problems, but also primes them on pressing issues that concern their neighborhood or
their ethnic groups, including health-related problems and issues (Wilkin, 2013). For example,
one study found that ICSN was positively associated with breast cancer knowledge and diabetes
knowledge for Latino and African American residents living in South Los Angeles (Kim, et al.,
2011). Another study suggested that residents reporting higher level of ICSN were more likely
to be aware of the closure of several local healthcare facilities, which may have influenced their
perceived ease of accessing those resources in their area (Matsaganis & Wilkin, 2014). In the
present study, the direction of the marginally significant effect of ICSN on descriptive norms
corroborates this line of evidence. In other words, to the extent that there are stories regarding
Latinas’ low compliance with cerv ical cancer screening being told through a neighborhood
storytelling network, noncompliant Latinas who are integrated into their neighborhood
161

storytelling network may perceive that it is common for women like them to never have had a
Pap test.
The size of the marginally significant effect of ICSN, on the other hand, might be
accounted for by the possibility that neighborhood stories specifically about Latinas’ high
prevalence of never having been screened, if any, were told infrequently and not told by all
storytellers, thereby limiting the synergistic effect of ICSN. This is likely, considering that the
proposed effect of individual storytellers on descriptive norms was established only for
neighborhood discussion, but not for local/ethnic media connections or community organization
connections.
Yet another possibility may also hold true. As suggested in the cancer communication
literature, the astonishing growth of cancer information of all kinds in a multitude of channels is,
for better or worse, accompanied by a rapid increase of information complexity as well
(Viswanath, 2005). In addition, largely owing to the advanced communication technologies that
have tremendously changed how people access and digest health information, it has become
increasingly common for the public to see the same issue being addressed from different angles,
and to encounter conflicting messages about the same issue in different information sources
(Kelly, et al., 2010; Viswanath, Breen, et al., 2006). From this perspective, it is reasonable to
suspect that the local/ethnic media and community organizations in the study area, too, might
have created and disseminated “stories” about Pap tests. However, their individual messages
may be considerably different from what was most widespread within participants’ interpersonal
network (i.e., low uptake of Pap tests among Latinas). Therefore, being more integrated into
one’s neighborhood storytelling network could increase participants’ access to multiple
162

narratives about Pap tests, hence the diminished magnitude of the effect of ICSN on descriptive
norms regarding never having a Pap test compared to that of neighborhood discussion.
Third, the present study found a contextual influence on the strength of ICSN on
descriptive norms. For the noncompliant sample, the effect of ICSN on descriptive norms was
stronger for women living in more linguistically isolated neighborhoods than for women living
in less linguistically isolated neighborhoods. This finding illustrates the dynamic relationship
between connecting to an integrated neighborhood storytelling network and the communication
action context. From a CIT perspective, a high percentage of population who do not speak
English very well at the neighborhood level may be an indication of widespread differences in
ethnic, cultural, and linguistic backgrounds among residents in the neighborhood, which can
limit a neighborhood’s capacity of shared storytelling. Prior CIT research from urban ethnic
communities, where reduced communication opportunities across ethnic groups that share the
same urban space manifest in ethnically bounded and fragmented storytelling network within
each ethnic group, supports this reasoning (N.-T. N. Chen, et al., 2013). For each ethnic group,
their ethnically bounded storytelling network could nevertheless serve as an important
information source, when the widespread ethnic, cultural, or linguistic barriers in their
neighborhoods make it hard for them to access and accumulate necessary information for
everyday problem-solving. For example, residents living in ethnically heterogeneous
neighborhoods were found to rely more strongly on their neighborhood storytelling network for
information to become engaged in their neighborhoods than their counterparts living in
ethnically homogenous neighborhoods (Kim & Ball-Rokeach, 2006b). The present study
extends this line of inquiry into the health domain. Specifically, the interaction between ICSN
and linguistic isolation on descriptive norms underscores the increased importance of locally and
163

ethnically-oriented communication resources in providing critical health information for Latinas
in linguistically isolated areas.
Neighborhood Storytelling Resources, Neighborhood Context, and Media Recall
Operationalized as the frequency of having seen or heard information about Pap tests in
the media in the past 30 days, media recall in this study is conceived of as an intermediate
outcome that lies on the pathways between communication resources and descriptive norms.
Participants’ level of media recall therefore served as a proxy measure of the actual
neighborhood storytelling regarding Pap tests in the study area at the time of the Multilevel
Survey. After all, as commented in prior research adopting communication infrastructure theory
(CIT), one can expect an association between connections to neighborhood storytelling resources
and a given health behavior only when related health issues are being discussed in the
storytelling network (Wilkin, 2013).
Results of descriptive analysis indicates that, on average, participants of the Multilevel
Survey reported having seen or heard information about Pap tests less than once in the past
month, with noncompliant participants reporting significantly lower level of media recall than
compliant participants. Moreover, media recall displayed a significant geographical variation for
noncompliant participants, indicating the influence of both individual- and neighborhood-level
factors. Although similar observations in the neighborhood health effect literature are rare,
which is probably due to the relative lack of discussion on the role of media in that literature,
communication research exploring the larger relationship between place and health knowledge
echoes this finding. For example, Slater and colleagues (2009) reported that people’s cancer
prevention knowledge differed significantly across U.S. regions, defined as geographical areas of
similar newspaper market size. In the present study, the next question, then, became what
164

communication resources at the individual level and characteristics at the neighborhood level
would explain the spatially-varied media recall for noncompliant sample.
First, as expected, noncompliant women with a broader scope of connections to
local/ethnic media (e.g., connecting to more than one type of local/ethnic media, such as
television, newspaper, and radio) reported having seen or heard information about Pap tests in
the media significantly more often than those with a narrower scope of local/ethnic media
connections (e.g., connecting to only one type of local/ethnic media, such as radio). It is worth
noting that this relationship was not found between participants’ connections to mainstream
English language media and media recall. The pattern of these findings seems to imply that
local/ethnic media, not mainstream English language media, served as a critical source of cancer
prevention information such as Pap tests for noncompliant participants in the study area. This is
consistent with prior research comparing the difference in cancer news reporting between ethnic
and mainstream media. For example, one content analysis found that ethnic newspapers were
more likely to print cancer awareness stories, to cover cancers that exhibited a larger ethnic
disparity (e.g., breast, prostate, colorectal, and feminine reproductive cancer), and to provide
cancer screening and detection information than mainstream newspaper (Stryker, Emmons, &
Viswanath, 2006). The importance of ethnic media as information source was also noted in
previous CIT research on health information seeking behaviors. For instance, Ball-Rokeach and
Wilkin (2009) observed that Latino residents from Southeast L.A. and Pico Union  two
neighborhoods within the present study area  specified local/ethnic television (usually Spanish
language television) as their primary source for health and medical information.
Second, noncompliant women who discussed things happening in their neighborhoods
with others also tended to report a higher level of media recall. Though only marginally
165

significant, this finding may suggest that frequent neighborhood discussion could render
information about Pap tests in the media environment more salient (Viswanath, Randolph Steele,
et al., 2006). As a result, women discussing their neighborhoods more often with others tend to
be well-informed about Pap test-related information. Additional support for this reasoning
comes from recent work on health information scanning. This line of research argues that health
information acquisition occurs not only through active seeking, but also within routine exposure
to interpersonal and mediated sources, sometimes in a purely incidental fashion, where
information receives some amount of attention at first, sufficient to produce certain level of
recall at a later time (Hornik, et al., 2013; Niederdeppe, et al., 2007). From this perspective, it is
plausible that women in the noncompliant sample might have been exposed to information about
Pap tests in their media environment, and later recalled the information during the course of
talking with others about their neighborhoods when relevant topics were brought up.
Third, contrary to expectations, connections to community organizations did not bear any
significant effect on media recall for the noncompliant sample. As noted earlier, the rationale for
the hypothesized effect of local organizational ties was that greater participation in community
groups, institutions, and organizations should not only promote interpersonal interactions among
residents, but also prime them to attend to information about local resources and events, thereby
enhancing the salience of media messages in residents’ environment. Using data from the
Minnesota Heart Health Program, for example, Viswanath et al. (2006) tested this hypothesis and
found that the number of cardiovascular disease messages recalled by survey respondents
increased as a function of community group memberships.
One possible explanation to why the anticipated relationship between community
organization connections and media recall regarding Pap tests was not established for the present
166

noncompliant sample may lie in the level and type of organizational membership reported by
participants. This study measured connections to community organizations by counting the
number of self-reported membership in 5 different categories of local groups, which ranged from
sports or recreational groups, neighborhood group or homeowner’s association, political or
education groups, to cultural, ethnic or religious groups. As presented in Chapter 5, the average
organizational membership for noncompliant participants was less than one. More significantly,
about 33% of them reported no membership in any of the aforementioned groups, 47% reported
having membership in one category of organization, and 18% reported having membership in
two categories of organizations. Further, a close look at the types of organizational membership
revealed that participants most often reported affiliations in ethnic, cultural, or religious groups
(58.1%) and sports and recreational groups (23.1%), whereas affiliations to other types of
organizations were all below 6%.
This profile of organizational connections was not surprising, given the socioeconomic
characteristics of the noncompliant sample (N.-T. N. Chen, et al., 2013). As presented in the
preceding chapter, approximately 47% of the participants had an educational level of less than
high school; about one third had lived in their neighborhoods for less than 6 years; and close to
90% rented rather than owned their homes. Furthermore, although prior research has
demonstrated the important part that churches play in health promotion and prevention efforts
targeting ethnic minorities, including cervical cancer screening and detection (Allen, et al., 2014;
Campbell, et al., 2007), much less is known regarding the same capacity for sports and
recreational organizations.
Fourth, the synergistic effect of connecting to all individual neighborhood storytelling
resources on media recall was observed for the noncompliant sample. Compared to women with
167

low level of ICSN, women with high level of ICSN reported having seen or heard information
about Pap tests in their media environment in the past month significantly more often. As
discussed above, ICSN privileges individuals whose connections to one neighborhood storyteller
strengthen their connections to other storytellers. To illustrate, a woman might connect to all
three types of neighborhood storytellers (i.e., residents, local/ethnic media, and community
organizations) to receive information about things happening in her neighborhood, but spent only
ten minutes on each information source a week. In contrast, for various reasons (e.g., linguistic
proficiency, limited resources, or lack of alternative options), another woman might connect only
to local/ethnic radio to get to know her neighborhood, and spent 10 hours a day listening to a
local radio channel. ICSN privileges the former scenario because it portrays a better structural
position of individuals in their communication environment that features “networked” rather than
isolated information sources (Kim, 2003). In the present study, it may be that women’s greater
integration into their neighborhood storytelling network could amplify their opportunities to
access or to be primed to pay more attention to information about Pap tests in their media
environment, hence the heightened level of recall.
Finally, a significant contextual effect was found on media recall resulting from
neighborhood-level density of health service providers. Among noncompliant participants,
women from neighborhoods with a lower density of health service providers reported having
seen or heard information about Pap tests more often than those from neighborhoods with a
higher density of health service providers. The importance of health service providers as a
credible source of cancer information has been extensively established (Moldovan-Johnson, et
al., 2014; Viswanath, 2005). Previous research also suggests the association between
geographical access to healthcare providers and regional level of health service utilization,
168

including cancer screening and detection (H.-Y. Chen, Kessler, Mori, & Chauhan, 2012; Frieden,
2012). On the other hand, recent research on health information seeking has increasingly
acknowledged the indispensable role played by nonmedical sources of health information, such
as media, organizations, and friends or family (Hornik, et al., 2013; Kelly, et al., 2010; Ramírez
et al., 2013). These information sources may have particular implications for populations with
restricted access to formal healthcare resources, such as ethnic minorities and new immigrants
(Ball-Rokeach & Wilkin, 2009; V. S. Katz, Ang, & Suro, 2012). That is, population who
experience difficulty in receiving health and medical information from formal information
sources, such as health service providers, might engage in alternative health communication
channels to compensate their information need. The present study provided empirical support
for this explanation, as the higher level of media recall among participants from areas with fewer
health service providers per capita may indicate their heavy reliance on local/ethnic media for
health information, including information on cervical cancer prevention.
Storytelling Resources and Neighborhood Context
According to communication infrastructure theory (CIT), communication action context
promotes or constrains people’s scope of connections to local storytelling resources as well as
the interconnectedness among those resources (Kim & Ball-Rokeach, 2006a). Prior research
also suggested that the importance of having a broad and integrated network of storytelling
resources became greater for residents from disadvantaged neighborhoods, where the challenge
to access, store, and apply necessary information for everyday problem-solving was harder (Kim
& Ball-Rokeach, 2006b). The ecological interrelationship between neighborhood contexts and
storytelling resources received further empirical evidence in this study.
169

First, regardless of their screening compliance status, participants from linguistically
isolated neighborhoods were significantly less connected to major English language media and
tended to be more connected to local/ethnic media, compared to participants living in
neighborhoods where linguistic barriers were less widespread. In other words, in linguistically
isolated neighborhoods, local/ethnic media and major English language media seem to compete
with each other, and the former tend to be the resource that participants connect to for local news
and information.
Second, compared to participants from ethnically homogenous neighborhoods,
participants from ethnically diverse neighborhoods tended to discuss their neighborhoods less
often, were significantly less connected to local/ethnic media, and had a significantly lower level
of ICSN. It is worth noting that ethnically heterogeneous neighborhoods may also be diverse on
a linguistic dimension, or, in the case of new immigrant neighborhoods, have considerable
percentage of population with limited English ability (e.g., Koreatown in the study area).
However, ethnic heterogeneity and linguistic isolation should not be interpreted as overlapping
or redundant. Conceptually, the measure of ethnic heterogeneity used in this study captures
neighborhood-level fragmentation not only linguistically, but also along cultural, religious, and
perhaps political lines as well (Alesina & Ferrara, 2000). This conceptual difference between
ethnic heterogeneity and linguistic isolation may help explain why the two constructs were not
significantly correlated in the present study (see Table 5.7 in Chapter 5). Furthermore, the fact
that ethnic heterogeneity reflects the differences among residents on multiple fronts may help
elucidate why ethnic heterogeneity had an overall broader and greater influence on participants’
connections to neighborhood storytelling resources than did linguistic isolation. As mentioned
above, in ethnically diverse neighborhoods, participants not only had constrained connections to
170

their locally situated interpersonal and mediated sources of information, but also had a less
integrated neighborhood storytelling network. The magnitudes of the coefficients of ethnic
heterogeneity were also considerably larger than those of linguistic isolation.
Third, contrary to expectation, participants living in neighborhoods with more
communication resources per capita tended to discuss their neighborhoods less often than their
counterparts from areas with fewer communication resources per capita. For analytical purposes,
this study regarded parks, schools, churches, libraries, community organizations, and social
services as potential communication recipient places in a neighborhood (e.g., communication
hotspot). According to CIT, neighborhoods with high density of communication resources may
produce increased communication opportunities among residents, and likely strengthen
residents’ connections with other storytellers too. One reason for the marginally significant,
negative effect of density of communication resources in the current study may lie in the fact that
the construct was positively associated with ethnic heterogeneity, r = .431, p <.05 (see Table 5.7
in Chapter 5), and that ethnic heterogeneity had a marginally significant and negative effect on
neighborhood discussion. Alternatively, it may also be because that, as the population-adjusted
number of communication “hotspot” increases in a neighborhood, the density of residents at each
place becomes lower, hence the reduced communication opportunities among residents.
Structural Relationships between Neighborhood Experience, Storytelling Resources, Health
Communication Outcomes, Descriptive Norms and Compliance with Screening Guidelines
Results from testing the structural equation models provided a nuanced portrayal of the
communication pathways underlying the formation of descriptive norms regarding Pap tests in a
residential context. A side-by-side comparison of the two revised models indicates both
similarities and differences between participants who had lived in their neighborhoods for six
171

years or more (the “longer tenure group”) and those whose residential tenure was shorter (the
“shorter tenure group”).
In terms of similarities, participants in both groups were significantly more likely to have
been in compliance with cervical cancer screening guidelines, if they 1) were first generation
immigrants; 2) had any kind of healthcare coverage; 3) perceived that it was common for Latinas
like them to have had regular Pap test; and 4) perceived that it was common for Latinas like them
to have received abnormal Pap test results. Perhaps the most counterintuitive of these findings,
first generation immigrants were more likely to have been in compliance with cervical cancer
screening guidelines, given that extant research generally suggests lower uptake of Pap screening
among foreign-born women than those born in the United States (Downs, Smith, Scarinci,
Flowers, & Parham, 2008; Hewitt, Devesa, & Breen, 2004; Mann, et al., 2014; Siegel,
Naishadham, & Jemal, 2012). At the first sight, one plausible explanation for this comes from
the fact that native-born participants were significantly younger than foreign-born participants in
both longer and shorter tenure group for this study (see Table 5.25 in Chapter 5). Evidence from
recent research supports this speculation. For example, one study based on a large-scale mail
survey showed that the lowest understanding of the importance of routine Pap tests appeared
among women in the youngest age range, which in their case was 18-34 years (Hawkins, Cooper,
Saraiya, Gelb, & Polonec., 2011). Another study tracking the national trend of Pap screening
from 1993 to 2010 found that women aged 30-59 years were consistently less likely to report
never having a Pap test (below 3%) than women aged 21-29 (8%) (H.-Y. Chen, et al., 2012).
However, as indicated in Table 5.25, a post-hoc logistic regression analysis showed that the
positive association between being first generation immigrants and screening compliance
172

retained its significance after accounting for the influence of age, whereas age was not
significantly associated with screening compliance.
The pattern of this finding may indicate an epidemiological paradox that has attracted a
great deal of attention in public health literature. Despite the fact that Latino population in
general has less income, less education, and is less likely to have adequate access to medical
care, Latino population as a whole, especially foreign-born Latinos, have better health outcomes
than Whites and African American populations in the United States (Markides & Coreil, 1986;
Pérez-Escamilla, Garcia, & Song, 2010). Therein lies the Hispanic health paradox, also known
as the immigrant paradox (Peréa, 2015). One explanation that has received extensive research
scrutiny is acculturation. From this perspective, foreign-born Latinos’ health outcomes may
deteriorate with their years’ of residence in the United States, depending on the degree to which
they learn about and adopt their recipient country’s (e.g., Untied States) cultural norms and
behavioral practices, while maintain their heritage culture (Schwartz, Unger, Zamboanga, &
Szapocznik, 2010). Empirical evidence for this hypothesis, however, is mixed (Abraído-Lanza,
Chao, & Flórez, 2005; Schwartz et al., 2011). In the current study, first generation immigrants in
both longer and shorter tenure groups were more likely to have been in compliance with cervical
cancer screening recommendations than second or higher generation immigrants, even after
controlling for the potential effect of age. Although the SEM models did not explicitly test
mechanisms of the immigrant paradox, such as acculturation, the findings on communication
pathways may offer an explanation for the paradoxical association between immigration
generation and screening compliance. This is discussed in more details in next section.
The positive relationship between descriptive norms regarding regular Pap tests and
compliance status is consistent with major social behavioral theories that articulate normative
173

influence on health behaviors (e.g., integrative model of behavioral prediction and theory of
normative social behavior). The positive relationship between descriptive norms regarding
abnormal Pap test results and compliance status is less straightforward to assess, however. For
example, previous research on Latinas’ adherence to follow -up after an abnormal Pap test result
attributes the low adherence to psychological barriers resulting from an abnormal result, such as
fatalistic beliefs, fear of cancer, and embarrassment of examinations (Eggleston, Coker, Das,
Cordray, & Luchok, 2007; Mann, et al., 2014). Given these barriers, one might expect that
Latinas who believe that a high percentage of women like them have received abnormal results
might avoid getting a Pap test. Yet the opposite was found for the current sample, where
participants who thought it was common for Latinas like them to have had an abnormal Pap test
result were more likely to be in compliance with cervical cancer screening guidelines. One
potential reason for this counterintuitive finding might be that, in both longer and shorter tenure
groups, compliant participants were also more likely to have discussed Pap tests with a doctor,
nurse, or other healthcare professional. Discussion with healthcare professionals, in turn, was
negatively related to descriptive norms regarding abnormal Pap test results. That is, participants
who had discussed Pap tests with healthcare professionals perceived it as less common for
women like them to have received an abnormal Pap test result. It is possible that during such
interactions healthcare professionals were able to provide accurate medical information to
correct patients’ misperceptions related to cancer prevention, detection, and treatment
(Moldovan-Johnson, et al., 2014; Viswanath, 2005), including inflated perceptions regarding
abnormal Pap test results. Upon obtaining an accurate understanding of the incidence and causes
of abnormal results, women may bring their actions in line with cervical cancer screening
recommendations.
174

The critical importance of healthcare providers as a source of credible medical
information becomes even more salient when considering the overall pattern of the associations
between Pap discussions with healthcare providers and descriptive norms. In both longer and
shorter tenure groups, in addition to having a negative association with descriptive norms
regarding abnormal Pap test results, Pap discussions with healthcare professionals also had a
negative association with descriptive norms regarding regular Pap tests. In other words,
participants who had discussed Pap tests with a healthcare professional perceived it as less
common for Latinas like them to have had Pap tests in at least every 3 years. One possible
reason could be that healthcare professionals may have explained to participants the generally
low compliance with cervical cancer screening recommendations among Latinas, hence lowering
participants’ perception that it is common for Latinas to have had routine Pap tests.
Moreover, in the longer tenure group, Pap discussions had a negative association with
descriptive norms regarding never having a Pap test, such that participants having discussed Pap
tests with a healthcare professional perceived it as less widespread for Latinas like them to never
have had a Pap test. This suggests that during patient-provider interactions, healthcare
professionals might have addressed participants’ exaggerated beliefs related to Latina women’s
never having a Pap tests, thereby lowering their corresponding perceived norms. This is
particularly important when considering the role played by descriptive norms regarding never
having a Pap test in shaping screening compliance for this group. While the paths leading to and
from descriptive norms regarding never having a Pap test were all non-significant for the shorter
tenure group, the construct had a negative association with screening compliance for the longer
tenure group. In other words, in the longer tenure group, participants who believed that a high
number of women like them had never had a Pap test were less likely to have been in compliance
175

with cervical cancer screening guidelines. This finding is consistent with previous norms
literature noting the unintended adverse effects of social norms, especially in health
communication campaigns. When people believe that unhealthy behaviors are prevalent, or
“normative”, they are less likely to adopt the healthy alternative (Mollen, Ruiter, et al., 2010;
Reid, et al., 2010). From this perspective, a perception that it is common for Latinas to never
have had a Pap test may reinforce women’s belief that it is “normal” to not participate in cervical
cancer screening at all or to postpone the screening to a later time. Information able to correct
this perception therefore has significant implications for increasing cervical cancer screening and
detection. As shown in this study, in the longer tenure group, such information came from
healthcare professionals rather than media, as Pap discussions with healthcare professionals was
the only health communication outcome variables of interest that was significantly associated
with descriptive norms regarding never having a Pap test.
The most notable difference in the structural models between the longer and shorter
tenure groups appeared in the communication pathways involving storytelling resources and
health communication outcomes. First, in both groups, first generation immigrants were more
likely to report having an integrated neighborhood storytelling network (ICSN), and less likely to
report connecting to major English language media. However, the reliance on ICSN seemed to
be much stronger for those in the longer tenure group, given that the magnitude of the coefficient
for the path from immigration generation to ICSN for the longer tenure group (γ = .99) was more
than three times that for the shorter tenure group (γ = .30). This pattern is not unexpected,
considering that participants who had lived in their neighborhoods for a shorter period of time
may be less likely to establish interpersonal network with other residents, or to learn about
176

available local information resources such as media outlets and community organizations in their
areas (Ball-Rokeach, et al., 2001).
Second, the longer tenure group had a relatively integrated storytelling model with most
of the proposed pathways supported by the analysis. Specifically, there was a significant and
positive path from major English language media connections to ICSN, indicating that, to some
degree, connections to the two types of storytelling resources supplemented each other (Ball-
Rokeach, et al., 2001). This has particular implications, given the differential influences that
ICSN and English language media had on health communication outcomes. For example,
English language media connections, not ICSN, had a positive effect on Pap discussions with
healthcare professionals. On the other hand, ICSN, but not English language media connections,
displayed an indirect effect through media recall on attention to information about Pap tests in
the media, with attention being positively associated with Pap discussions with healthcare
professionals. Taken together, despite the divergent influence of ICSN and English language
media on health communication outcomes, the relatively integrated storytelling model enabled
participants in the longer tenure group to receive normative information regarding Pap tests from
both media and healthcare professionals.
In contrast, the storytelling model for the shorter tenure group was a fragmented one.
The path from English language media connections to ICSN was missing, suggesting that the
two types of storytelling resources were independent of each other as source of local news and
information for participants in the shorter tenure group. The direct path from English language
media connections to Pap discussions with healthcare professionals, as well as the indirect path
from ICSN to media attention through media recall, all remained significant. However, the
direct path from media attention to discussion with healthcare professionals did not. In addition,
177

in the revised models, the magnitude of the coefficient for the path from media recall to media
attention (β = .83) for the shorter tenure group was twice as high as that for the longer tenure
group (β = .41). This suggests that, compared to participants in the longer tenure group, although
participants in the shorter tenure group might be less reliant on ICSN for local news and
information, ICSN nevertheless play a relatively larger role as a source of Pap test information
for this group. As noted above, individuals with constrained access to formal source of health
information may find it particularly important to have informal health communication
connections, such as connections to family and friends, television, newspapers, churches, and
community organizations (V. S. Katz, et al., 2012). It has also been found that local/ethnic
media as well as friends and family both received higher rankings than healthcare professionals
as primary sources of health and medical information for Latino residents from new immigrant
communities in South Los Angeles (Ball-Rokeach & Wilkin, 2009). In the current study, it may
be that participants in the longer tenure group are better positioned in their communication
environment to access, store, and act on health information. The smaller coefficient for the path
from media recall to media attention for this group therefore suggests that the role of ICSN as a
source of health information (including information on cervical cancer prevention) may be
somewhat limited, though still significant. In contrast, participants in the shorter tenure group,
especially first generation immigrants, might find it harder to find and accumulate health
information in their neighborhoods, partially due to their lack of experience in their
neighborhoods and their fragmented communication connections. Thus, it is possible that they
would pay particularly more attention to health information (including information on Pap tests)
in the media channels that they connect to for local news and information, hence the larger
coefficient for the path from media recall to media attention for this group.
178

Taken together, findings summarized above show that ICSN partially mediates the
influence of individual-level structural characteristics (i.e., immigration generation) on
descriptive norms in the study sample. Consistent with the results of multilevel modeling
analysis, the significant paths from ICSN to media attention through media recall indicate that
ICSN, not major English language media, serve as the source of Pap test information for
participants. However, the relative weight of ICSN in influencing health communication
outcomes directly, as well as descriptive norms and screening compliance indirectly, can vary
considerably as a result of participants’ years of residence in their neighborhoods. Compared to
participants in the shorter tenure group, participants in the longer tenure group had increased
opportunities to access, store, and apply information about Pap tests, which is, in part, due to
their relatively integrated storytelling system in which neighborhood storytelling resources and
major English language media supplemented each other as information sources. In addition, a
higher level of ICSN may have primed participants to not only attend to information about Pap
tests in the media, but also to actively seek relevant information from medical sources (Wilkin,
2013). As a result, although it seems that ICSN mainly shaped descriptive norms regarding
regular Pap test, as indicated by its corresponding significant indirect effect, the formation of
descriptive norms for the longer tenure group was shaped by both nonmedical sources (i.e.,
storytelling resources) and medical sources (i.e., healthcare professionals).
It appears that the longer tenure group reaps the advantage of connecting to an integrated
neighborhood storytelling network, as opposed to connecting to isolated neighborhood
storytellers (e.g., local television). As noted earlier, a recent research indicates that cancer
coverage in local television news was less likely to address screening behaviors, or to refer
audience to credible organizations for clear recommendations, compared to national television
179

news (Lee, et al., 2014). In Los Angeles area, a content analysis found that the health stories on
Spanish-language television did not provide adequate information regarding relevant local health
resources (Wilkin & González, 2006). Yet this problem might be ameliorated when individuals
are connected to not just one information source, such as local television, but an integrated
network of information sources that affords increased opportunity to not only access information,
but also retain and act upon the information (e.g., seeking advice from healthcare providers).
Contributions of the Current Study
The present study contributes to the study of social norms, communication, cancer
prevention, and the larger interdisciplinary field of neighborhoods and health theoretically,
methodologically, and practically. First, this study theorized and tested the communication
pathways underlying the formation of perceived norms in the context of residential environment.
As acknowledged in recent norms literature, people come into contact with information that
conveys norms from a wide array of venues situated in their social, symbolic, and physical
environment. However, compared to the extensively established norm-behavior relationship, the
influence of information sources on the formation of normative perceptions is not well
documented (Mead, et al., 2014). Much less is known regarding the processes and mechanisms
through which people’s environment shapes their perceived norms regarding a given behavior.
From a theoretical perspective, identifying such mechanisms could shed light on the conditions
where normative influence is strong versus weak.
Second, communication scholarship has increasingly advocated for a multilevel approach
to the study of communication, both theoretically and analytically (Slater, et al., 2006). This
advocacy is based on the premise that  considering that communication always involves
human interaction within various contexts (e.g., schools, organizations, neighborhoods, media
180

system)  communication is inherently a multilevel phenomenon. In health communication,
this has led to research examining the relationship between health and place from a socio-
ecological perspective. Situating itself in the context of urban ethnic neighborhoods, this study
demonstrated that the communication and formation of perceived norms regarding Pap tests
indeed was a multilevel phenomenon shaped by not only individual-level factors, such as
people’s connections to local/ethnic media, but also the characteristics of their neighborhood
environment, such as the percentage of population with limited English abilities at the
neighborhood level.
Third, this study examined the potential of communication infrastructure theory (CIT) to
explain the dynamic relationship between people’s symbolic, social, and physical environ ment,
and how such relationship shaped their perceived norms regarding cancer screening and
detection. CIT elaborates the roles of a neighborhood storytelling network consisting of
residents, local/ethnic media, and community organizations. An integrated neighborhood
storytelling network (ICSN) provides residents increased opportunities to access, store, and
apply information necessary for everyday problem-solving. Prior CIT studies of diverse urban
neighborhoods exploring the application of ICSN in the health domains generally support this
conceptualization (Kim, et al., 2011; Matsaganis & Wilkin, 2014). However, researchers have
warned against drawing overly simplistic conclusions regarding the benefits (or disadvantages)
of connecting to neighborhood storytelling resources, arguing that such outcomes must be
understood in their specific ethnic, cultural, and geographical contexts (Wilkin, 2013). In
addition, an effect of ICSN on people’s perceptions or behaviors in a particular health domain
can be expected only when related stories are being created and discussed in their neighborhood
communication environment. In other words, both connections to an integrated neighborhood
181

storytelling network as well as the types of health stories circulated within the network could
make a difference (Kim, et al., 2011; Wilkin & Ball-Rokeach, 2011).
The current study extends this body of work by applying CIT to cervical cancer screening
and detection. Results from testing multilevel models confirm the potential of ICSN as a source
of cervical cancer information for participants who were not in compliance with cervical cancer
screening guidelines. Results from testing structural equation models shows that, depending on
participants’ neighborhood experience, ICSN was not only the source from which participants
learned about Pap tests, but also may have primed them to actively seek additional information
from medical sources such as healthcare professionals. Moreover, despite a lack of knowledge
regarding the actual neighborhood storytelling on Pap tests in the study area, it seems that the
normative information about Pap tests in participants’ local media environment mainly related to
routine Pap tests, whereas such information circulated in participants’ interpersona l network
mostly had to do with low uptake of Pap screening. Consistent with prior research, this finding
indicates that a neighborhood storytelling network could promote both healthy and unhealthy
health beliefs (Kim, et al., 2011; Wilkin & Ball-Rokeach, 2011). As discussed in more details
below, efforts exploring the utility of CIT in reducing health disparities among ethnic minority
and immigrant neighborhoods require a comprehensive diagnosis of not only residents’ level of
connections to their neighborhood storytelling network, but also the types of health stories being
told through the network.
Fourth, the current study explored the effect of a communication action context (CAC) on
the formation of descriptive norms. Although CAC has been articulated as an important
component of CIT, quantitative evidence on the role of CAC in the health domain is limited.
The four CAC factors tested in this study are linguistic isolation, ethnic heterogeneity, density of
182

communication resources and density of health service providers. By including these factors in
the analysis, the present study was able to elaborate how participants’ interpersonal, media, and
neighborhood environment together shaped participants’ perceived norms regarding Pap tests.
For example, among noncompliant participants, there was a significant interaction between
ICSN and linguistic isolation on descriptive norms regarding never having a Pap test. That is,
the importance of ICSN as information source in shaping noncompliant women’s perc eived
number of Latinas like them who had never had a Pap test was greater for those living in
linguistically isolated neighborhoods. There was also a significant contextual effect of
neighborhood-level density of health service providers on noncompliant participants’ level of
recall of having seen or heard information about Pap tests in the media. Specifically, self-
reported recall of encountering Pap tests-related information in the media was higher for
noncompliant participants in areas where health care resources were less plenty. Consistent with
prior research, these findings indicate that, in disadvantaged neighborhoods (e.g., neighborhoods
characterized by linguistic isolation and low density of health service providers), local
communication resources could play a greater role as health information source for residents
who might find it difficult to access and accumulate health information in their day-to-day
environment. (Kim & Ball-Rokeach, 2006b).
Fifth, by testing the utility of a communication framework in understanding the formation
of descriptive norms in a neighborhood context, this study provides empirical evidence on the
crucial role of communication in understanding the influence of neighborhoods on health, an
area that the larger neighborhood health effect literature has not adequately addressed. As noted
earlier, despite the prominent role that communication plays in organizing community life,
research on the role of communication in neighborhood health effect is often limited to social
183

ties or social interactions, especially in sociological literature (Matsaganis, 2015). In relevant
public health research, the focus is more on identifying the optimal combinations of
communication channels, whereas how places affect the availability of those channels in
people’s communication environment is largely ignored. With a communication framework that
explicitly elaborates the dynamic relationships between people’s connections to communication
resources and their neighborhood environment, this study was able to examine the roles of not
only micro-level communication resources (i.e., residents), but also meso-level and (i.e.,
local/ethnic media and community organizations) and macro-level resources (i.e., major English
language media), as well as whether and how these resources functioned in relation to the
neighborhood environment in which they were situated. Additionally, as shown in the analysis,
although not every outcome variables (i.e., descriptive norms and media recall) examined in this
study varied across neighborhoods, connections to most of the storytelling resources did. This
suggests that a discussion of the influence of information sources on people’s health -related
outcomes, such as descriptive norms and Pap screening compliance, must not be isolated from
people’s specific neighborhood contexts.
Sixth, the current study offers a potential explanation for the immigrant paradox from a
communication perspective. Although Latino population, particularly first generation
immigrants, usually are in the lower socioeconomic group with less access to medical care than
Whites and African Americans in the United States, they as a whole have been found to have
better health outcomes (Peréa, 2015). Of a number of explanations proposed to elucidate this
paradox, acculturation hypothesis argues that foreign-born Latinos’ health advantages would
diminish, as they learn about and adopt the cultural beliefs and behaviors in the United States
(Schwartz, et al., 2010). Yet empirical support for this hypothesis is mixed, underscoring the
184

importance of more research to better understand the relationship between immigration status
and health outcomes (Abraído-Lanza, Armbrister, Flórez, & Aguirre, 2006; Abraído-Lanza, et
al., 2005; Pérez-Escamilla, et al., 2010). In the current study, in addition to a direct effect on
Latinas’ compliance with cervical cancer screening recommendations, immigration generation
also had a direct effect on their connections to storytelling resources. Specifically, in both longer
and shorter tenure groups, first generation immigrants were more likely to connect to
neighborhood storytelling resources (measured by ICSN), and less likely to connect to major
English language media. ICSN, in turn, served as an important source of Pap tests-related
information for Latinas in both groups, and, more significantly, prime Latinas in the longer
tenure group to actively seek information from healthcare professionals. Moreover, in both
longer and shorter tenure groups, being first generation immigrants had a significant, positive,
and indirect association with descriptive norms regarding Pap tests, which were positively
associated with screening compliance. Taken together, these communication pathways shed
light into the social process that, at least partially, explain why compliance with cervical cancer
screening recommendations was higher among first generation immigrants than among second
and higher generation immigrants for the current sample.
Seventh, previous CIT studies found that local storytelling networks can be ethnically
bounded, where different ethnic groups living in the same area each connect to a neighborhood
storytelling network oriented to their own ethnic group (N.-T. N. Chen, et al., 2013). The timing
of immigration has also been noted as an additional important factor that shapes residents’
connections to neighborhood storytelling resources. For example, Ball-Rokeach and her
colleagues (2001) found their storytelling model of neighborhood belonging to perform very well
among participants considered as old generation immigrants (i.e., Caucasians and African
185

Americans), but less well among new immigrant participants (i.e., Latinos and Asians).
Specifically, while the model was a more integrated one for old immigrant participants, it
appeared to be fragmented for new immigrant participants given the missing linkages between
macro- and meso-level storytellers. Findings of the present study corroborate this line of
evidence. For both longer and shorter tenure groups, being first generation immigrants were
more likely to connect to neighborhood storytelling network (as captured by ICSN), and less
likely to connect to English language media. These patterns indicate the bounding role of
immigration generation in shaping people’s connection to a neighborhood storytelling network.
Eighth, this study makes methodological contributions to the aforementioned literature
with the innovative ways of defining neighborhood boundaries developed by the Multilevel
Study (R01CA155326 - Murphy/Ball-Rokeach). Previous research using multilevel modeling to
study neighborhood health effect usually delineates neighborhoods based on arbitrarily drawn
boundaries, such as census tracts and ZIP code (Schaefer-McDaniel, O'Brien Caughy, O'Campo,
& Gearey, 2010; Weiss, Ompad, Galea, & Vlahov, 2007). However, this approach has long been
criticized for not being able to fully capture the cultural, institutional, and physical characteristics
of neighborhoods, which may partially explain the inconsistency of results regarding the effects
of neighborhoods on health (Matsaganis, 2015). As the boundaries of a neighborhood represent
the geographical scope within which a proposed process underlying a specific health-related
outcome operates, solutions to this methodological issue have theoretical implications. As
discussed in more details in Appendix B, the Multilevel Project responded to this challenge and
delineated 25 distinctive neighborhood clusters with unique socioeconomic, ethnic, cultural, and
physical characteristics. More important, these neighborhoods attempted to maximally reflect
186

residents’ day-to-day communication environment where people receive, process, and act on
health information, including information related to cancer screening and detection.
Ninth, this research makes practical contributions by identifying several points of action
to promote cervical cancer prevention among Latinas from urban ethnic neighborhoods. As
summarized above, this study demonstrates that neighborhood storytelling resources can serve as
antecedents for health communication outcomes (e.g., media recall, media attention, and
discussion with healthcare providers about Pap tests) that subsequently influence descriptive
norms and Pap screening compliance among a large sample of Latinas. To the extent that
individuals’ connections to neighborhood storytelling resources vary across geographical
communities, researchers and health professionals should assess these connections  both in
terms of breadth and interconnectedness  in each target population’s specific cultural, social,
and geographical context.
Researchers and practitioners should be cognizant of the vast complexity of people’s
communication environment, where facts and myths surrounding a particular health topic could
vary greatly across neighborhoods and population. Findings of this study provide only a
snapshot of this complexity. For example, the prevalent storytelling about Pap tests within
residents’ interpersonal network involved Latinas’ low compliance with Pap screening, whereas
stories covered by local/ethnic media seemed to be related to routine Pap tests. Thus, for health
messages to effectively travel through the noise and reach targeted audience, efforts should be
directed to investigate not only people’s connections to their neighborhood storytelling network ,
but also the amount, type, and framing of health stories being told through the network. This
knowledge could help identify venues in which positive health storytelling could be created. For
instance, in neighborhoods where residents discuss Latina’s l ow compliance with Pap tests that
187

reinforces perceived norms justifying noncompliance, disseminating stories regarding the
importance of routine Pap tests to residents’ social network may help.
Further, it is important that interventions geared toward encouraging positive health
stories involve local/ethnic media and community organizations, and, more important, strengthen
linkages between neighborhood storytellers (Wilkin, 2013). As shown in prior research, the
disjuncture between neighborhood storytellers, especially between meso-level storytellers such
as local media and community organizations, is a critical challenge to overcome in order to
produce sustainable change in health and well-being in a community (Matsaganis, et al., 2014).
In this study, the findings that different Pap tests-related stories were told in participants’
interpersonal network and in local/ethnic media suggest missing linkages between the two
neighborhood storytellers. The fact that connections to community organizations did not bear a
significant relationship with any outcome variables provides further evidence of a fragmented
health storytelling network. These findings indicate the need for interventions that not only
increase positive health storytelling, but also forge missing links among neighborhood
storytellers. For example, in the current study area, one might increase media coverage referring
residents to local organizational resources that provide relevant information and services
(including healthcare resources in the area), interview community organizations to understand
their outreach to media outlets and the community, or hold community events to create more
communication opportunities between residents and local organizations. Strategies addressing
the missing linkages between neighborhood storytellers would not only address disparities in
cervical cancer screenings among Latinas, but also health disparities in general that affect diverse
population and communities.
188

Limitations and Suggestions for Future Research
This research has several limitations that invite future investigations. First, this
dissertation focuses on three types of descriptive norms related to Pap test, which are descriptive
norms regarding never having a Pap test, having had a Pap test at least in past 3 years, and
having received an abnormal Pap test result. There remain unexplored descriptive norms related
to other points of action on the continuum of cervical cancer prevention, detection, and
treatment. As noted above, the weak correlation between descriptive norms regarding never
having a Pap test and descriptive norms regarding regular Pap tests for the present sample of
complaint participants, as well as the non-significant correlation between the two constructs for
noncompliant participants, both suggest that descriptive norms regarding having not received a
Pap test regularly could be a construct for future research. Researchers could explore the level of
Latinas’ perceived prevalence of women like them who have not received a Pap t est in the past 3
years, as well as the communication factors that lead to this normative perception that justifies an
“overdue” Pap screening test. Similarly, future research could examine normative perceptions
regarding Latinas’ adherence to follow -up treatment after receiving an abnormal Pap test result.
Findings obtained from these venues could help identify points of intervention for health
education and promotion efforts to enhance Pap test screening compliance in this population.
Second, following the lead of Kim (2003), this research suggests more systematic work
to refine the ICSN measure. The ICSN measure was conceptualized to capture the synergistic
effect of connecting to all neighborhood storytellers. Although it makes theoretical sense that
each storyteller is weighted differently in the process of everyday life and depending on specific
behavioral, cultural, or even geographical context, CIT research has been allocating equal weight
to each storyteller in calculating the measure. The present study followed this practice.
189

However, future research should invest in finding appropriate weights given the specific
outcomes of interest.
Third, future research exploring the potential of local information resources on health-
related outcomes should consider expanding the scope of investigation from neighborhood
storytelling resources to communication ecologies. This is to capture the complexity of
individuals’ communication environment that consists of not only neighborhood storytellers, but
also more formal information sources known to have critical influence on health and health
behaviors (e.g., health service providers), social media, and macro-level media and
organizational resources (e.g., mainstream media, CDC, NCI) (Wilkin, 2013). This ecology of
communication resources could be expanded if one includes not only “traditional print, broadcast
and telecommunication ecology”, but also “social networking applications for peer to peer
modes of communication, transport infrastructure that enables face to face interactions, as well
as public and private places where people meet, chat, gossip” (Hearn & Foth, 2007; n.p.). The
importance of this shift is that individuals affected most by health disparities might not have an
integrated neighborhood storytelling network to benefit from (Wilkin, 2013; Wilkin & Ball-
Rokeach, 2011). In this study, the fragmented storytelling model for participants with shorter
residential tenure provides further empirical evidence to this line of advocacy.
Fourth, on a related note, it is important to also examine the roles of alternative practice
and medicine as well as transnational resources in shaping Latino immigrants’ health decision
making and healthcare utilization (Ransford, Carrillo, & Rivera, 2010). For example, about 61%
of the noncompliant participants and 50% of the compliant participants in this study reported
having used products from a botanica or a Latino market. In addition, about 38% of the
noncompliant participants and 27% of the compliant participants reported having used
190

prescription medicine from their home country. These findings suggest that, for research that
exploring the role of local healthcare resources in health issues such as Latino immigrants’
general health, chronic disease prevention and management, or healthcare utilization, neglecting
such information and only focusing on local healthcare resources may lead to a skewed or
fragmented picture of the outcomes of interest.
Fifth, this study did not have a qualitative component to its data collection geared
towards understanding the actual health storytelling taking place within a given study area. As
mentioned earlier, both facts and myths about a particular health issue, such as Pap test, can
circulate within a neighborhood storytelling network and in people’s wider communication
ecologies. In-depth knowledge as such could supplement quantitative findings obtained from
survey research to better understand the information environment where normative perceptions
regarding the health issue are shaped and maintained. In this regard, future research should
include qualitative research methods such as media monitoring, content analysis, and focus
groups with residents, in order to best capture the nature and flow of health storytelling among
the targeted populations and areas.
Sixth, this study did not detect a significant effect of neighborhood-level density of
communication resources on either descriptive norms or media recall. One plausible reason
might be the way the measure was created. By lumping five categories of communication
resources together, the current measure might end up losing its ability to capture the relationship
between density of each specific type of communication resource and the outcomes of interest.
This is possible especially given that prior research has documented the roles played by some of
the resources  such as churches, parks, and community organizations  in health promotion
191

efforts in urban ethnic communities (Allen, et al., 2014; Los Angeles County Department of
Public Health, 2014; Matsaganis, et al., 2014).
Seventh, the measure of density of healthcare providers used in this study had the same
weakness. For analytic purposes, this study selected relevant health centers located in the study
area from an existing geographical database maintained by the Los Angeles County. However,
these clinics may still differ vastly from each other in terms of service provided and population
served. It is also possible that these healthcare resources might not be the usual places that the
present Latina population would go for medical care. Thus, when resources permit, future
research should identify an inventory of usual places for medical and health care services (e.g.,
community clinics) that targeted population can and do access. This inventory can then be used
to create density measures in order to understand how area-level accessibility of healthcare
resources influence residents’ well -being. Additionally, research can also calculate the distance
from residents’ homes to their usual places of health care, and use the distance as an individual -
level measure of healthcare providers.
Eighth, it should be mentioned that the four neighborhood-level characteristics explored
in this study (i.e., linguistic isolation, ethnic heterogeneity, density of communication resources,
density of health service providers) may not capture the uniqueness of the neighborhood clusters
in the study area. Future research should explore other dimensions of neighborhoods that have
theoretical implications for a given research question. For example, neighborhood-level
percentage of foreign-born population may be particularly relevant for studies that focus on
immigrant health.
Ninth, the utility of the current neighborhood cluster boundaries should be explored
further in other health or behavioral domains, and in a more in-depth fashion. One primary
192

research aim of the Multilevel Study was to track research participants’ screening compliance
status based on their medical records (see more in Appendix A). This led to a purposive sample
based on certain clinics where medical records could be accessed. Thus, the sample was not
envisioned as random or geographically representative. As a result, the current study wasn’t able
to generalize to specific geographic areas or to the larger population. If resources permitted,
future research should consider conducting research with random samples recruited from these
ethnic neighborhoods to allow for a fuller picture of the interaction between geography and
ethnicity, and how such interaction influence the well-beings of individuals and communities.
Last, future research that contemplates a similar study design and population of interest
with the Multilevel Study should be cognizant of the daunting challenge that recruitment process
might impose. Location-constrained recruitment can be both time and resource consuming.
Compared to Caucasians, racial and ethnic minority populations have also seen a low
participation rate in health research projects, due to barriers such as mistrust of medical research
(George, Duran, & Norris, 2013). In the Multilevel Study, the recruitment process spanned
approximately a year and a half. In light of recruitment challenges, the research team proposed
alternative strategies half way through the recruitment, including recruiting women from two
additional community clinics and from community public spaces to supplement the sample
recruited at the main clinic site. At the minimum, solutions to these challenges will require
interdisciplinary collaborations and community connections (e.g., community clinics and
organizations). Well-thought-out strategies that more fully engage individuals and communities
in research design and implementation are also critical to the success of research projects.
193

Conclusion
This study found that the formation of descriptive norms was a multilevel phenomenon
shaped by both individual-level and neighborhood-level factors for the current sample of Latinas
from diverse urban neighborhoods in Los Angeles. Specifically, with a communication
infrastructure model, analysis of this study revealed that neighborhood storytelling resources
could serve as a critical source of normative information that affected Latinas’ descriptive norms
regarding cervical cancer screening and detection in the context of neighborhood environment,
where their everyday life unfolded. Findings of this study underscore the importance of local
information sources in mediating the influence of structural conditions at both individual level
(e.g., residential tenure and immigration generation) and neighborhood level (e.g., linguistic
isolation, ethnic heterogeneity, density of communication resources and density of health service
providers) on residents’ h ealth-related perceptions and behaviors. Additionally, these findings
have significant theoretical, methodological, and practical implications for research that explores
the relationship between people and place, particularly for research that focuses on urban ethnic
neighborhoods. These implications can apply to not only disparities in cervical cancer screening
and prevention for Latinas, but also health disparities in other domains that affect diverse
population and communities.

194

REFERENCES
Abraído-Lanza, A. F., Armbrister, A. N., Flórez, K. R., & Aguirre, A. N. (2006). Toward a
theory-driven model of acculturation in public health research. American Journal of
Public Health, 96(8), 1342-1346. doi: 10.2105/ajph.2005.064980
Abraído-Lanza, A. F., Chao, M. T., & Flórez, K. R. (2005). Do healthy behaviors decline with
greater acculturation? Implications for the Latino mortality paradox. Social Science
& Medicine, 61(6), 1243-1255. doi: 10.1016/j.socscimed.2005.01.016
Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In J. Kuhl & J.
Beckman (Eds.), Action-control: From cognition to behavior (pp. 11-39). Heidelberg,
Germany: Springer.
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision
Processes, 50(2), 179-211. doi: 10.1016/0749-5978(91)90020-t
Ajzen, I., & Albarracín, D. (2007). Predicting and changing behavior: A reasoned action
approach. In I. Ajzen, D. Albarracín & R. C. Hornik (Eds.), Prediction and change of
health behavior (pp. 1-22). Mahwah, NJ: Lawrence Erlbaum.
Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior.
Englewood Cliffs, NJ: Prentice-Hall, Inc.
Ajzen, I., & Fishbein, M. (2005). The influence of attitudes on behavior. In D. Albarracín, B. T.
Johnson & M. P. Zanna (Eds.), The handbook of attitudes (pp. 173-221). Mahwah, NJ:
Lawrence Erlbaum Associates.
Akers, A. Y., Newmann, S. J., & Smith, J. S. (2007). Factors underlying disparities in cervical
cancer incidence, screening, and treatment in the United States. Current problems in
cancer, 31(3), 157-181.
195

Alesina, A., & Ferrara, E. L. (2000). Participation in heterogeneous communities. The Quarterly
Journal of Economics, 115(3), 847-904.
Allen, J. D., Pérez, J. E., Tom, L., Leyva, B., Diaz, D., & Torres, M. I. (2014). A pilot test of a
church-based intervention to promote multiple cancer-screening behaviors among
Latinas. Journal of Cancer Education, 29(1), 136-143.
American Cancer Society. (2012). Cancer facts & figures for Hispanics/Latinos 2012-2014, from
http://www.cancer.org/acs/groups/content/@epidemiologysurveilance/documents/docum
ent/acspc-034778.pdf
Asch, S. E. (1951). Effects of group pressure upon the modification and distortion of judgments.
In H. Guetzkow (Ed.), Groups, leardship, and men (pp. 177-190). Pittsburgh, PA:
Carnegie Press.
Austin, L. T., Ahmad, F., McNally, M.-J., & Stewart, D. E. (2002). Breast and cervical cancer
screening in Hispanic women: a literature review using the health belief model. Women's
Health Issues, 12(3), 122-128. doi: http://dx.doi.org/10.1016/S1049-3867(02)00132-9
Baer, J. S., Stacy, A., & Larimer, M. (1991). Biases in the perception of drinking norms among
college students. Journal of Studies on Alcohol and Drugs, 52(6), 580-586.
Ball-Rokeach, S. J. (1985). The origins of individual media-system dependency a sociological
framework. Communication Research, 12(4), 485-510.
Ball-Rokeach, S. J. (1998). A theory of media power and a theory of media use: Different
stories, questions, and ways of thinking. Mass Communication and Society, 1(1-2), 5-40.
Ball-Rokeach, S. J., & DeFleur, M. L. (1976). A dependency model of mass-media effects.
Communication Research, 3(1), 3-21.
196

Ball-Rokeach, S. J., & Jung, J. (2003). Media system dependency to communication
infrastructure: A review of the evolution of MSD theory and a proposal for a new
concept. Paper presented at the International Conference on Mass Media in the Era of
Globalization, Marketing, and Hi-Technology, Beijing, China.
Ball-Rokeach, S. J., Kim, Y.-C., & Matei, S. (2001). Storytelling neighborhood. Communication
Research, 28(4), 392-428. doi: 10.1177/009365001028004003
Ball-Rokeach, S. J., Moran, M. B., Hether, H. J., & Frank, L. B. (2010). Communication
hotspots and comfort zones: A report presented to the California Endowment. Los
Angeles, CA: University of Southern California Metamorphosis Project.
Ball-Rokeach, S. J., & Wilkin, H. A. (2009). Ethnic differences in health information-seeking
behavior: Methodological and applied Issues. [Article]. Communication Research
Reports, 26(1), 22-29. doi: 10.1080/08824090802636983
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change.
Psychological Review, 84, 191-215.
Bandura, A. (1982). Self-efficacy mechanism in human agency. American Psychologist, 37(2),
122-147. doi: 10.1037/0003-066x.37.2.122
Bandura, A. (2001). Social cognitive theory of mass communication. [Article]. Media
Psychology, 3(3), 265-299.
Bandura, A. (2004). Health promotion by social cognitive means. Health Education & Behavior,
31(2), 143-164. doi: 10.1177/1090198104263660
Bazargan, M., Bazargan, S. H., Farooq, M., & Baker, R. S. (2004). Correlates of cervical cancer
screening among underserved Hispanic and African-American women. Preventive
Medicine, 39(3), 465-473. doi: http://dx.doi.org/10.1016/j.ypmed.2004.05.003
197

Becker, M. H. (1974). The health belief model and personal health behavior. Health
Education Monographs, 2(4).
Behbakht, K., Lynch, A., Teal, S., Degeest, K., & Massad, S. (2004). Social and cultural barriers
to Papanicolaou test screening in an urban population. Obstetrics & Gynecology, 104(6),
1355-1361.
Bekalu, M. A., & Eggermont, S. (2014). Exposure to HIV/AIDS-related media content and HIV
testing intention: Applying the integrative model of behavioral prediction. Mass
Communication and Society, null-null. doi: 10.1080/15205436.2013.878362
Bendor, J., & Swistak, P. (2001). The evolution of Norms. American Journal of Sociology,
106(6), 1493-1545.
Bilandzic, H., & Busselle, R. (2012). A narrative perspective on genre-specific cultivation.
Living with television now: Advances in cultivation theory and research, 261-285.
Bish, A., Sutton, S., & Golombok, S. (2000). Predicting uptake of a routine cervical smear test:
A comparison of the health belief model and the theory of planned behaviour. Psychology
and Health, 15(1), 35-50.
Bleakley, A., Hennessy, M., Fishbein, M., & Jordan, A. (2011). Using the integrative model to
explain how exposure to sexual media content influences adolescent sexual behavior.
Health Education & Behavior, 38(5), 530-540. doi: 10.1177/1090198110385775
Borsari, B., & Carey, K. B. (2003). Descriptive and injunctive norms in college drinking: A
meta-analytic integration. Journal of studies on alcohol, 64(3), 331.
Broad, G. M., Gonzalez, C., & Ball-Rokeach, S. J. (2013). Intergroup relations in South Los
Angeles – Combining communication infrastructure and contact hypothesis approaches.
198

International Journal of Intercultural Relations(0). doi:
http://dx.doi.org/10.1016/j.ijintrel.2013.06.001
Byrd, T. L., Peterson, S. K., Chavez, R., & Heckert, A. (2004). Cervical cancer screening beliefs
among young Hispanic women. Preventive Medicine, 38(2), 192-197. doi:
http://dx.doi.org/10.1016/j.ypmed.2003.09.017
Campbell, M. K., Hudson, M. A., Resnicow, K., Blakeney, N., Paxton, A., & Baskin, M. (2007).
Church-Based health promotion interventions: Evidence and lessons learned. Annual
Review of Public Health, 28(1), 213-234. doi:
10.1146/annurev.publhealth.28.021406.144016
Campo, S., Brossard, D., Frazer, M. S., Marchell, T., Lewis, D., & Talbot, J. (2003). Are social
norms campaigns really magic bullets? Assessing the effects of students' misperceptions
on drinking behavior. Health Communication, 15(4), 481.
Cappella, J. N. (1980). Structural equation modeling: An introduction. In J. N. Cappella & P. R.
Monge (Eds.), Multivariate techniques in human communication research (pp. 57-110).
New York: Academic Press.
Centers for Disease Control and Prevention. (2014). Cervical cancer screening rates, 2015, from
http://www.cdc.gov/cancer/cervical/statistics/screening.htm
Chatterjee, J. S., Frank, L. B., Murphy, S. T., & Power, G. (2009). The importance of
interpersonal discussion and self-efficacy in knowledge, attitude, and practice models.
International Journal of Communication, 3, 607-634.
Chen, H.-Y., Kessler, C. L., Mori, N., & Chauhan, S. P. (2012). Cervical cancer screening in the
United States, 1993–2010: Characteristics of women who are never screened. Journal of
Women's Health, 21(11), 1132-1138.
199

Chen, N.-T. N., Ognyanova, K., Zhao, N., Liu, W., Gerson, D., Ball-Rokeach, S., & Parks, M.
(2013). Communication and socio-demographic forces shaping civic engagement patterns
in a multiethnic city. In P. Moy (Ed.), Communication and Community. New York, NY:
Hampton Press.
Chou, W.-y. S., Prestin, A., Lyons, C., & Wen, K.-y. (2013). Web 2.0 for health promotion:
Reviewing the current evidence. American Journal of Public Health, 103(1), e9-e18. doi:
10.2105/ajph.2012.301071
Chung, J. E. (2014). Medical dramas and viewer perception of health: Testing cultivation effects.
Human Communication Research, 40(3), 333-349. doi: 10.1111/hcre.12026
Cialdini, R. B., Reno, R. R., & Kallgren, C. A. (1990). A focus theory of normative conduct:
Recycling the concept of norms to reduce littering in public places. Journal of
Personality and Social Psychology, 58(6), 1015-1026.
Cialdini, R. B., & Trost, M. R. (1998). Social influence: Social norms, conformity and
compliance. In D. T. Gilbert, S. Fiske & G. Lindzey (Eds.), The Handbook of Social
Psychology (4 ed., pp. 151-192). New York: McGraw-Hill.
Cline, R. J. W., & Thompson, T. (2003). Everyday interpersonal communication and health. In
T. Thompson, A. Dorsey, K. Miller & R. Parrott (Eds.), Handbook of health
communication (pp. 285-313). Mahwah, NJ: Erlbaum.
Cogliano, V., Baan, R., Straif, K., Grosse, Y., Secretan, B., & Ghissassi, F. E. (2005).
Carcinogenicity of human papillomaviruses. The Lancet Oncology, 6(4), 204. doi:
http://dx.doi.org/10.1016/S1470-2045(05)70086-3
200

Cohen, E. L., Caburnay, C. A., Luke, D. A., Rodgers, S., Cameron, G. T., & Kreuter, M. W.
(2008). Cancer coverage in general-audience and black newspapers. Health
Communication, 23(5), 427-435.
Colby, S. L., & Ortman, J. M. (2015). Projections of the size and composition of the U . S .
population: 2014 to 2060 current population reports: United States Census Bureau.
CSDH. (2008). Closing the gap in a generation. Health equity through action on the social
determinants of health. Geneva: World Health Organization.
Daniel, K. L., Bernhardt, J. M., & Eroğlu, D. (2009). Social marketing and health
communication: from people to places. American Journal of Public Health, 99(12), 2120-
2122.
David, C., Cappella, J. N., & Fishbein, M. (2006). The social diffusion of influence among
adolescents: Group interaction in a chat room environment about antidrug
advertisements. Communication Theory, 16(1), 118-140. doi: 10.1111/j.1468-
2885.2006.00008.x
De Jesus, M., & Xiao, C. (2014). Predicting health care utilization among Latinos: Health locus
of control beliefs or access factors? Health Education & Behavior. doi:
10.1177/1090198114529130
Deutsch, M., & Gerard, H. B. (1955). A study of normative and informatinoal social influences
upon individual judgment. Journal of Abnormal and Social Psychology, 51(3), 629-636.
Dillard, J. P. (2011). An application of the integrative model to women's intention to be
vaccinated against HPV: Implications for message design. Health Communication, 26(5),
479-486. doi: 10.1080/10410236.2011.554170
201

Downs, L. S., Smith, J. S., Scarinci, I., Flowers, L., & Parham, G. (2008). The disparity of
cervical cancer in diverse populations. Gynecologic Oncology, 109(2, Supplement 1),
S22-S30. doi: 10.1016/j.ygyno.2008.01.003
Dunlop, S. M., Kashima, Y., & Wakefield, M. (2010). Predictors and consequences of
conversations about health promoting media messages. Communication Monographs,
77(4), 518-539. doi: 10.1080/03637751.2010.502537
Echeverría, S. E., Gundersen, D. A., Manderski, M. T. B., & Delnevo, C. D. (2015). Social
norms and its correlates as a pathway to smoking among young Latino adults. Social
Science & Medicine, 124(0), 187-195. doi:
http://dx.doi.org/10.1016/j.socscimed.2014.11.034
Eggleston, K. S., Coker, A. L., Das, I. P., Cordray, S. T., & Luchok, K. J. (2007). Understanding
barriers for adherence to follow-up care for abnormal pap tests. Journal of Women's
Health, 16(3), 311-330.
Eisenberg, M. E., Toumbourou, J. W., Catalano, R. F., & Hemphill, S. A. (2014). Social norms
in the development of adolescent substance use: A longitudinal analysis of the
international youth development study. Journal of youth and adolescence, 43(9), 1486-
1497.
Enders, C. K., & Tofighi, D. (2007). Centering predictor variables in cross-sectional multilevel
models: A new look at an old issue. Psychological Methods, 12(2), 121-138. doi:
http://dx.doi.org/10.1037/1082-989X.12.2.121
Feenstra, G. W., McGrew, S., & Campbell, D. (1999). Entrepreneurial community gardens:
Growing food, skills, jobs and communities (Vol. 21587): UCANR Publications.
202

Fischer, K. N., & Teutsch, S. M. (2014). Safe summer parks programs reduce violence and
improve health in Los Angeles County: Institute of Medicine of the National Academies.
Fishbein, M. (1979). A theory of reasoned action: Some applications and implications.
Fishbein, M., & Cappella, J. N. (2006). The role of theory in developing effective health
communications. Journal of Communication, 56, S1-S17. doi: 10.1111/j.1460-
2466.2006.00280.x
Frank, L. B., Chatterjee, J. S., Chaudhuri, S. T., Lapsansky, C., Bhanot, A., & Murphy, S. T.
(2012). Conversation and compliance: Role of interpersonal discussion and social norms
in public communication campaigns. Journal of Health Communication, 17(9), 1050-
1067. doi: 10.1080/10810730.2012.665426
Frieden, T. (2012). Use of selected clinical preventive services among adults--United States,
2007–2010. MMWR. Morbidity and mortality weekly report, 61, 1-2.
Fujimoto, K., & Valente, T. W. (2012). Social network influences on adolescent substance use:
Disentangling structural equivalence from cohesion. Social Science & Medicine, 74(12),
1952-1960. doi: http://dx.doi.org/10.1016/j.socscimed.2012.02.009
George, S., Duran, N., & Norris, K. (2013). A systematic review of barriers and facilitators to
minority research participation among African Americans, Latinos, Asian Americans,
and Pacific Islanders. American Journal of Public Health, 104(2), e16-e31. doi:
10.2105/ajph.2013.301706
Gerbner, G., Gross, L., Morgan, M., Signorielli, N., & Shanahan, J. (2002). Growing up with
television: Cultivation processes. In J. Bryant & D. Zillmann (Eds.), Media effects:
Advances in theory and research (2 ed., pp. 43-67).
203

Gibbons, F. X., Pomery, E. A., Gerrard, M., Sargent, J. D., Weng, C.-Y., Wills, T. A., . . . Yeh,
H.-C. (2010). Media as social influence: Racial differences in the effects of peers and
media on adolescent alcohol cognitions and consumption. Psychology of Addictive
Behaviors, 24(4), 649-659. doi: 10.1097/01.psy.0000221275.75056.d8
Giddens, A. (1991). Modernity and self-identity: Self and society in the late modern age:
Stanford University Press.
Giddens, A. (2002). Runaway world: How globalisation is reshaping our lives: Profile books.
Goldberg, D. W. (2013). Texas A&M University Geoservices. Available online at
http://geoservices.tamu.edu. , 2013
Green, B. L., Davis, J., Rivers, D., Buchanan, K., & Rivers, B. (2014). Cancer health disparities.
In D. Alberts & L. M. Hess (Eds.), Fundamentals of Cancer Prevention (pp. 151-193):
Springer Berlin Heidelberg.
Guagliardo, M. F. (2004). Spatial accessibility of primary care: Concepts, methods and
challenges. International Journal of Health Geographics, 3, 3-3. doi: 10.1186/1476-
072x-3-3
Guagliardo, M. F., Ronzio, C. R., Cheung, I., Chacko, E., & Joseph, J. G. (2004). Physician
accessibility: An urban case study of pediatric providers. Health & Place, 10(3), 273-283.
Halim, A., Hasking, P., & Allen, F. (2012). The role of social drinking motives in the
relationship between social norms and alcohol consumption. Addictive behaviors, 37(12),
1335-1341.
Hampton, K. N. (2010). Internet use and the concentration of disadvantage: Glocalization and
the urban underclass. American Behavioral Scientist, 53(8), 1111-1132. doi:
10.1177/0002764209356244
204

Hawkins, N. A., Cooper, C. P., Saraiya, M., Gelb, C. A., & Polonec., L. (2011). Why the Pap
test? Awareness and use of the Pap test among women in the United States. Journal of
Women's Health., 20(4), 511-515. doi: 10.1089/jwh.2011.2730.
Hayes, A. F. (2006). A primer on multilevel modeling. Human Communication Research, 32(4),
385-410.
Hearn, G. N., & Foth, M. (2007). Communicative ecologies: Editorial preface. Electronic
Journal of Communication, 17(1-2).
Hether, H. J., Huang, G. C., Beck, V., Murphy, S. T., & Valente, T. W. (2008). Entertainment-
education in a media-saturated environment: Examining the impact of single and multiple
exposures to breast cancer storylines on two popular medical dramas. [Article]. Journal
of Health Communication, 13(8), 808-823. doi: 10.1080/10810730802487471
Hewitt, M., Devesa, S. S., & Breen, N. (2004). Cervical cancer screening among US women:
Analyses of the 2000 National Health Interview Survey. Preventive Medicine, 39(2), 270-
278.
Ho, S. S., Poorisat, T., Neo, R. L., & Detenber, B. H. (2013). Examining how presumed media
influence affects social norms and adolescents' attitudes and drinking behavior intentions
in rural Thailand. Journal of Health Communication, 19(3), 282-302. doi:
10.1080/10810730.2013.811329
Hoffman, B. R., Monge, P. R., Chou, C.-P., & Valente, T. W. (2007). Perceived peer influence
and peer selection on adolescent smoking. Addictive Behaviors, 32(8), 1546-1554. doi:
10.1016/j.addbeh.2006.11.016
Holbert, R. L., & Stephenson, M. T. (2008). Commentary on the Uses and Misuses of Structural
Equation Modeling in Communication Research. In A. F. Hayes, M. D. Slater & L. B.
205

Snyder (Eds.), The SAGE Sourcebook of Advanced Data Analysis Methods for
Communication Research (pp. 185-219). Thousand Oaks, CA: Sage Publications, Inc.
Holmes, J. H., Lehman, A., Hade, E., Ferketich, A. K., Gehlert, S., Rauscher, G. H., . . . Bird, C.
E. (2008). Challenges for multilevel health disparities research in a transdisciplinary
environment. American Journal of Preventive Medicine, 35(2, Supplement), S182-S192.
doi: http://dx.doi.org/10.1016/j.amepre.2008.05.019
Hooper, D., Coughlan, J., & Mullen, M. R. (2008). Structural equation modelling: Guidelines for
determining model fit. [Article]. Electronic Journal of Business Research Methods, 6(1),
53-59.
Hornik, R. C. (2002). Public health communication: Evidence for behavior change. Mahwah,
NJ: Lawrence Erlbaum.
Hornik, R. C., Parvanta, S., Mello, S., Freres, D., Kelly, B., & Schwartz, J. S. (2013). Effects of
scanning (routine health information exposure) on cancer screening and prevention
behaviors in the general population. Journal of Health Communication, 18(12), 1422-
1435. doi: 10.1080/10810730.2013.798381
Hox, J. (2010). Multilevel analysis: Techniques and applications (2 ed.). New York, NY:
Routledge.
Hyyppä, M. T., & Mäki, J. (2001). Individual-level relationships between social capital and self-
rated health in a bilingual community. Preventive Medicine, 32(2), 148-155.
Institute of Medicine. (2002). Speaking of health: Assessing health communication strategies for
diverse populations: Joseph Henry Press.
Israel, B. A., Coombe, C. M., Cheezum, R. R., Schulz, A. J., McGranaghan, R. J., Lichtenstein,
R., . . . Burris, A. (2010). Community-based participatory research: A capacity-building
206

approach for policy advocacy aimed at eliminating health disparities. American Journal
of Public Health, 100(11), 2094-2102. doi: 10.2105/ajph.2009.170506
Jeffres, L. W. (2002). Urban communication systems: Neighborhoods and the search for
community. Cresskill, NJ: Hampton Press, Inc.
Jennings-Dozier, K. (1999). Predicting intentions to obtain a Pap Smear among African
American and Latina women: Testing the theory of planned behavior. Nursing Research,
48(4), 198-205.
Jensen, J. D., Moriarty, C. M., Hurley, R. J., & Stryker, J. E. (2010). Making sense of cancer
news coverage trends: A comparison of three comprehensive content analyses. Journal of
Health Communication, 15(2), 136-151. doi: 10.1080/10810730903528025
Jones, K. O., Denham, B. E., & Springston, J. K. (2006). Effects of mass and interpersonal
communication on breast cancer screening: Advancing agenda-setting theory in health
contexts. Journal of Applied Communication Research, 34(1), 94-113.
Joreskog, K. G., & Sorbom, D. (1996). LISREL 8: User's Reference Guide. Chicago, IL:
Scientific Software International.
Katz, E. (1957). The two-step flow of communication: An up-to-date report on an hypothesis.
Public Opinion Quarterly, 21(1), 61-78.
Katz, E., & Lazarsfeld, P. F. (2006). Personal Influence:The part played by people in the flow of
mass communications: Free Press. (Original work published 1955).
Katz, V. S., Ang, A., & Suro, R. (2012). An ecological perspective on U.S. Latinos’ health
communication behaviors, access, and outcomes. Hispanic Journal of Behavioral
Sciences, 34(3), 437-456. doi: 10.1177/0739986312445566
207

Kawachi, I., & Berkman, L. F. (2003). Neighborhoods and health. Oxford ;New York: Oxford
University Press.
Kelly, B., Hornik, R. C., Romantan, A., Schwartz, J. S., Armstrong, K., DeMichele, A., . . .
Wong, N. (2010). Cancer information scanning and seeking in the general population.
Journal of Health Communication, 15(7), 734-753. doi: 10.1080/10810730.2010.514029
Kenny, D. A., Kaniskan, B., & McCoach, D. B. (2014). The performance of RMSEA in models
with small degrees of freedom. Sociological Methods & Research, 0049124114543236.
Keyes, K. M., Schulenberg, J. E., O'Malley, P. M., Johnston, L. D., Bachman, J. G., Li, G., &
Hasin, D. (2012). Birth cohort effects on adolescent alcohol use: The influence of social
norms from 1976 to 2007. Archives of General Psychiatry, 69(12), 1304-1313. doi:
10.1001/archgenpsychiatry.2012.787
Khalil, G. E., & Rintamaki, L. S. (2014). A televised entertainment-education drama to promote
positive discussion about organ donation. Health Education Research, 29(2), 284-296.
doi: 10.1093/her/cyt106
Kim, Y. C. (2003). Storytelling community: Communication infrastructure and civic
engagements in urban spaces. University of Southern California.
Kim, Y. C., & Ball-Rokeach, S. J. (2006a). Civic engagement from a communication
infrastructure perspective. Communication Theory, 16(2), 173-197. doi: 10.1111/j.1468-
2885.2006.00267.x
Kim, Y. C., & Ball-Rokeach, S. J. (2006b). Community storytelling network, neighborhood
context, and civic engagement: A multilevel approach. Human Communication Research,
32(4), 411-439. doi: 10.1111/j.1468-2958.2006.00282.x
208

Kim, Y. C., Moran, M. B., Wilkin, H. A., & Ball-Rokeach, S. J. (2011). Integrated connection to
neighborhood storytelling network, education, and chronic disease knowledge among
African Americans and Latinos in Los Angeles. Journal of Health Communication,
16(4), 393-415. doi: 10.1080/10810730.2010.546483
Kincaid, D. L. (2004). From innovation to social norm: Bounded normative influence. Journal of
Health Communication, 9(sup1), 37-57. doi: 10.1080/10810730490271511
Kitayama, S., & Burnstein, E. (1994). Social influences, persuasion, and group decision making.
Persuasion: Psychological Insights and Perspectives. Allyn and Beacon, 175-194.
Klein, W. M. P., & Stefanek, M. E. (2007). Cancer risk elicitation and communication: Lessons
from the psychology of risk perception. CA Cancer J Clin, 57(3), 147-167. doi:
10.3322/canjclin.57.3.147
Kreps, G. L., & Maibach, E. W. (2008). Transdisciplinary science: The nexus between
communication and public health. Journal of Communication, 58(4), 732-748.
Kreuter, M. W., Kegler, M. C., Joseph, K. T., Redwood, Y. A., & Hooker, M. (2012). The
impact of implementing selected CBPR strategies to address disparities in urban Atlanta:
a retrospective case study. Health Education Research, 27(4), 729-741.
Lapinski, M. K., & Rimal, R. N. (2005). An explication of social norms. Communication Theory,
15(2), 127-147. doi: 10.1111/j.1468-2885.2005.tb00329.x
Latkin, C., Donnell, D., Liu, T.-Y., Davey-Rothwell, M., Celentano, D., & Metzger, D. (2013).
The dynamic relationship between social norms and behaviors: the results of an HIV
prevention network intervention for injection drug users. Addiction, 114-135. doi:
10.1111/add.12095
209

Lee, C.-J., Long, M., Slater, M. D., & Song, W. (2014). Comparing local TV news with national
TV news in cancer coverage: An exploratory content analysis. Journal of Health
Communication, 19(12), 1330-1342. doi: 10.1080/10810730.2014.894598
Lee, C.-j., & Niederdeppe, J. (2011). Genre-specific cultivation effects: Lagged associations
between overall TV viewing, local TV News viewing, and fatalistic beliefs about cancer
prevention. Communication Research, 38(6), 731-753. doi: 10.1177/0093650210384990
Lewis, L. B., Sloane, D. C., Nascimento, L. M., Diamant, A. L., Guinyard, J. J., Yancey, A. K.,
& Flynn, G. (2005). African Americans’ access to healthy food options in South Los
Angeles restaurants. Journal Information, 95(4), 668-673.
Lewis, M. A., Patrick, M. E., Mittmann, A., & Kaysen, D. L. (2014). Sex on the beach: the
influence of social norms and trip companion on spring break sexual behavior.
Prevention Science, 15(3), 408-418.
Los Angeles County Department of Public Health. (2014). Parks after dark turns parks into safe
havens that promote community cohesion and healthy physical activity. Los Angeles
County Department of Public Health Division of Chronic Disease and Injury Prevention,
from http://publichealth.lacounty.gov/chronic/docs/CTG_Parks_After_Dark.pdf
Los Angeles County Department of Public Health Office of Women’s Health. (201 0). Health
indicators for women in Los Angeles County: Highlighting disparities by ethnicity and
poverty Level, February 2010. Retrieved June 10, 2014, from
http://publichealth.lacounty.gov/owh/docs/Health-Indicators- 2010.pdf
Luciani, S., & Andrus, J. K. (2008). A Pan American health organization strategy for cervical
cancer prevention and control in Latin America and the Caribbean. Reproductive Health
Matters, 16(32), 59-66. doi: 10.2307/25475430
210

Mabry, A., & Mackert, M. (2014). Advancing use of norms for social marketing: extending the
theory of normative social behavior. International Review on Public and Nonprofit
Marketing, 11(2), 129-143. doi: 10.1007/s12208-013-0109-5
MacDonald, J., & Sampson, R. J. (2012). The world in a city: Immigration and America’s
changing social fabric. The ANNALS of the American Academy of Political and Social
Science, 641(1), 6-15.
Mann, L., Foley, K., Tanner, A., Sun, C., & Rhodes, S. (2014). Increasing cervical cancer
screening among US Hispanics/Latinas: A qualitative systematic review. Journal of
Cancer Education, 1-14. doi: 10.1007/s13187-014-0716-9
Marcus, A. C., & Crane, L. A. (1998). A Review of cervical cancer screening intervention
research: Implications for public health programs and future research. Preventive
Medicine, 27(1), 13-31. doi: http://dx.doi.org/10.1006/pmed.1997.0251
Marin, G., & Gamba, R. J. (1996). A new measurement of acculturation for Hispanics: The
bidimensional acculturation scale for Hispanics (BAS). Hispanic Journal of Behavioral
Sciences, 18(3), 297-316. doi: 10.1177/07399863960183002
Markides, K. S., & Coreil, J. (1986). The health of Hispanics in the southwestern United States:
An epidemiologic paradox. Public health reports, 101(3), 253.
Massey, D. S. (1996). The age of extremes: Concentrated affluence and poverty in the twenty-
first century. Demography, 33(4), 395-412.
Matei, S., Ball-Rokeach, S. J., & Qiu, J. L. (2001). Fear and misperception of Los Angeles urban
space: A spatial-statistical study of communication-shaped mental maps. Communication
Research, 28(4), 429-463. doi: 10.1177/009365001028004004
211

Matsaganis, M. D. (2008). Rediscovering the communication engine of neighborhood effects:
How the interaction of residents and community institutions impacts health literacy and
how it can be leveraged to improve health care access. Communication Ph.D.,
Communication, United States -- California. Retrieved from
https://libproxy.usc.edu/login?url=http://search.proquest.com/docview/304462621?accou
ntid=14749 ProQuest database.
Matsaganis, M. D. (2015). How do the places we live in impact our health? Challenges for, and
insights from, communication research. In E. L. Cohen (Ed.), Communication Yearbook
(Vol. 39, pp. 33–65). New York: Routledge.
Matsaganis, M. D., Gallagher, V. J., & Drucker, S. J. (2013). Communicative cities in the 21st
century: The urban communication reader III (Vol. 3). New York: Peter Lang.
Matsaganis, M. D., Golden, A. G., & Scott, M. E. (2014). Communication infrastructure theory
and reproductive health disparities: Enhancing storytelling network integration by
developing interstitial actors. International Journal of Communication, 8, 21.
Matsaganis, M. D., & Wilkin, H. A. (2014). Communicative social capital and collective
efficacy as determinants of access to health-enhancing resources in residential
communities. Journal of Health Communication, 1-10. doi:
10.1080/10810730.2014.927037
McGuire, W. J., Rice, R., & Atkin, C. (2001). Input and output variables currently promising for
constructing persuasive communications. In R. E. Rice & C. K. Atkin (Eds.), Public
communication campaigns (pp. 22-48). Thousand Oaks, CA: Sage.
McQueen, A., Kreuter, M. W., Kalesan, B., & Alcaraz, K. I. (2011). Understanding narrative
effects: The impact of breast cancer survivor stories on message processing, attitudes, and
212

beliefs among African American women. Health Psychology, 30(6), 674-682. doi:
10.1080/1070551990954011810.1080/107055199095401181998-03102-001
Mead, E. L., Rimal, R. N., Ferrence, R., & Cohen, J. E. (2014). Understanding the sources of
normative influence on behavior: The example of tobacco. Social Science & Medicine,
115(0), 139-143. doi: http://dx.doi.org/10.1016/j.socscimed.2014.05.030
Miller, R., & Shinn, M. (2005). Learning from communities: Overcoming difficulties in
dissemination of prevention and promotion efforts. American Journal of Community
Psychology, 35(3-4), 169-183. doi: 10.1007/s10464-005-3395-1
Mizuno, Y., Seals, B., Kennedy, M., & Myllyluoma, J. (2000). Predictors of teens' attitudes
toward condoms: Gender differences in the effects of norms. Journal of Applied Social
Psychology, 30(7), 1381-1395.
Moldovan-Johnson, M., Tan, A. S. L., & Hornik, R. C. (2014). Navigating the cancer
information environment: The reciprocal relationship between patient-clinician
information engagement and information seeking from nonmedical sources. Health
Communication, 29(10), 974-983. doi: 10.1080/10410236.2013.822770
Mollen, S., Rimal, R. N., & Lapinski, M. K. (2010). What is normative in health communication
research on norms? A review and recommendations for future scholarship. Health
Communication, 25(6-7), 544-547. doi: 10.1080/10410236.2010.496704
Mollen, S., Ruiter, R. A. C., & Kok, G. (2010). Current issues and new directions in psychology
and health: What are the oughts? The adverse effects of using social norms in health
communication. Psychology & Health, 25(3), 265-270.
213

Moran, M. B., Murphy, S. T., Frank, L., & Baezconde-Garbanati, L. (2013). The ability of
narrative communication to address health-related social norms. International Review of
Social Research, 3(2), 131.
Morgan, S. E. (2009). The intersection of conversation, cognitions, and campaigns: The social
representation of organ donation. Communication Theory, 19(1), 29-48.
Mullins, R., Coomber, K., Broun, K., & Wakefield, M. (2013). Promoting cervical screening
after introduction of the human papillomavirus vaccine: the effect of repeated mass media
campaigns. Journal of Medical Screening, 20(1), 27-32. doi:
10.1177/0969141313478588
Murillo, R., Almonte, M., Pereira, A., Ferrer, E., Gamboa, O. A., Jerónimo, J., & Lazcano-
Ponce, E. (2008). Cervical cancer screening programs in Latin America and the
Caribbean. Vaccine, 26, Supplement 11(0), L37-L48. doi:
http://dx.doi.org/10.1016/j.vaccine.2008.06.013
Murphy, S. T., Frank, L. B., Chatterjee, J. S., & Baezconde-Garbanati, L. (2013). Narrative
versus nonnarrative: The role of identification, transportation, and emotion in reducing
health disparities. Journal of Communication, 63(1), 116-137.
Murphy, S. T., Frank, L. B., Moran, M. B., & Patnoe-Woodley, P. (2011). Involved, transported,
or emotional? Exploring the determinants of change in knowledge, attitudes, and
behavior in entertainment-education. Journal of Communication, 61(3), 407-431. doi:
10.1111/j.1460-2466.2011.01554.x
Murphy, S. T., Hether, H., & Rideout, V. (2008). How healthy is prime time? An analysis of
health content in popular primetime television programs: The Kaiser Family Foundation.
214

National Institute of Health. (2013). NIH fact sheets: Cervical cancer, from
http://report.nih.gov/nihfactsheets/viewfactsheet.aspx?csid=76
Nicholson, R. A., Kreuter, M. W., Lapka, C., Wellborn, R., Clark, E. M., Sanders-Thompson, V.,
. . . Casey, C. (2008). Unintended effects of emphasizing disparities in cancer
communication to African-Americans. Cancer Epidemiology Biomarkers & Prevention,
17(11), 2946-2953. doi: 10.1158/1055-9965.epi-08-0101
Niederdeppe, J., Bigman, C. A., Gonzales, A. L., & Gollust, S. E. (2013). Communication about
health disparities in the mass media. Journal of Communication, 63(1), 8-30. doi:
10.1111/jcom.12003
Niederdeppe, J., Fowler, E. F., Goldstein, K., & Pribble, J. (2010). Does local television news
coverage cultivate fatalistic beliefs about cancer prevention? Journal of Communication,
60(2), 230-253. doi: 10.1111/j.1460-2466.2009.01474.x
Niederdeppe, J., Hornik, R. C., Kelly, B. J., Frosch, D. L., Romantan, A., Stevens, R. S., . . .
Schwartz, J. S. (2007). Examining the dimensions of cancer-related information seeking
and scanning Behavior. Health Communication, 22(2), 153-167. doi:
10.1080/10410230701454189
Novak, S. P., Reardon, S. F., Raudenbush, S. W., & Buka, S. L. (2006). Retail tobacco outlet
density and youth cigarette smoking: a propensity-modeling approach. American Journal
of Public Health, 96(4), 670-676.
Oldenburg, R. (1999). The great good place: Cafes, coffee shops, bookstores, bars, hair salons,
and other hangouts at the heart of a community: Marlowe New York.
Park, H. S., Jr, W. P. E., & Cudeck, R. (2008). Multilevel modeling: Studying people in contexts.
In A. F. Hayes, M. D. Slater & L. B. Snyder (Eds.), The SAGE Sourcebook of Advanced
215

Data Analysis Methods for Communication Research (pp. 219-247). Thousand Oaks,
CA: Sage Publications, Inc.
Park, R. E. (1925). The city: Suggestions for the investigation of human behavior in the urban
environment. In R. E. Park & E. W. Burgess (Eds.), The city (pp. 1-46). Chicago:
University of Chicago Press.
Park, Y., Neckerman, K. M., Quinn, J., Weiss, C., & Rundle, A. (2008). Place of birth, duration
of residence, neighborhood immigrant composition and body mass index in New York
City. The International Journal of Behavioral Nutrition and Physical Activity, 5, 19-19.
doi: 10.1186/1479-5868-5-19
Pasick, R. J., Barker, J. C., Otero-Sabogal, R., Burke, N. J., Joseph, G., & Guerra, C. (2009).
Intention, subjective norms, and cancer screening in the context of relational culture.
Health Education & Behavior, 36(5 suppl), 91S-110S. doi: 10.1177/1090198109338919
Peréa, F. (2015). Hispanic health paradox. In S. Loue & M. Sajatovic (Eds.), Encyclopedia of
Immigrant Health (pp. 828-830): Springer New York.
Pérez-Escamilla, R., Garcia, J., & Song, D. (2010). Health care access among Hispanic
immigrants. NAPA Bulletin, 34(1), 47-67. doi: 10.1111/j.1556-4797.2010.01051.x
Perkins, H. W., & Berkowitz, A. D. (1986). Perceiving the community norms of alcohol use
among students: some research implications for campus alcohol education
programming*. Substance use & misuse, 21(9-10), 961-976.
Phua, J. J. (2012). The reference group perspective for smoking cessation: An examination of the
influence of social norms and social identification with reference groups on smoking
cessation self-efficacy. Psychology of Addictive Behaviors.
216

Preacher, K. J., Rucker, D. D., & Hayes, A. F. (2007). Addressing moderated mediation
hypotheses: Theory, methods, and prescriptions. Multivariate Behavioral Research,
42(1), 185-227. doi: 10.1080/00273170701341316
Price, V., Nir, L., & Cappella, J. N. (2006). Normative and informational influences in online
political discussions. Communication Theory, 16(1), 47-74.
Rabe-Hesketh, S., & Skrondal, A. (2012). Multilevel and longitudinal modeling using Stata (3
ed.). College Station, TX: STATA press.
Ramírez, A. S., Freres, D., Martinez, L. S., Lewis, N., Bourgoin, A., Kelly, B. J., . . . Hornik, R.
C. (2013). Information seeking from media and family/friends increases the likelihood of
engaging in healthy lifestyle behaviors. Journal of Health Communication, 18(5), 527-
542. doi: 10.1080/10810730.2012.743632
Randolph, W., & Viswanath, K. (2004). Lessons learned from public health mass media
campaigns: Marketing health in a crowded media world. Annual Review of Public Health,
25(1), 419-437. doi: doi:10.1146/annurev.publhealth.25.101802.123046
Ransford, H. E., Carrillo, F. R., & Rivera, Y. (2010). Health care-seeking among Latino
immigrants: Blocked access, use of traditional medicine, and the role of religion. Journal
of health care for the poor and underserved, 21(3), 862-878. doi: 10.1353/hpu.0.0348
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data
analysis methods (Vol. 1): Sage.
Real, K., & Rimal, R. N. (2007). Friends talk to friends about drinking: Exploring the role of
peer communication in the theory of normative social behavior. Health Communication,
22(2), 169-180. doi: 10.1080/10410230701454254
217

Reid, A., Cialdini, R., & Aiken, L. (2010). Social norms and health behavior. In A. Steptoe (Ed.),
Handbook of Behavioral Medicine (pp. 263-274): Springer New York.
Rhodes, F., Stein, J., Fishbein, M., Goldstein, R., & Rotheram-Borus, M. (2007). Using theory to
understand how interventions work: Project RESPECT, condom use, and the integrative
model. AIDS and Behavior, 11(3), 393-407. doi: 10.1007/s10461-007-9208-9
Rice, R. E. (1993). Using network concepts to clarify sources and mechanisms of social
influence. Progress in communication sciences, 12, 43-62.
Rimal, R. N. (2008). Modeling the relationship between descriptive norms and behaviors: A test
and extension of the theory of normative social behavior (TNSB). [Article]. Health
Communication, 23(2), 103-116. doi: 10.1080/10410230801967791
Rimal, R. N., Lapinski, M. K., Cook, R. J., & Real, K. (2005). Moving toward a theory of
normative influences: How perceived benefits and similarity moderate the impact of
descriptive norms on behaviors. [Article]. Journal of Health Communication, 10(5), 433-
450. doi: 10.1080/10810730591009880
Rimal, R. N., & Real, K. (2003). Understanding the influence of perceived norms on behaviors.
Communication Theory, 13(2), 184-203. doi: 10.1111/j.1468-2885.2003.tb00288.x
Rimal, R. N., & Real, K. (2005). How behaviors are influenced by perceived norms.
Communication Research, 32(3), 389-414. doi: 10.1177/0093650205275385
Rivis, A., & Sheeran, P. (2003). Descriptive norms as an additional predictor in the theory of
planned behaviour: A meta-analysis. Current Psychology, 22(3), 218-233.
Robbins, R., & Niederdeppe, J. (2015). Using the integrative model of behavioral prediction to
identify promising message strategies to promote healthy sleep behavior among college
students. Health Communication, 30(1), 26-38.
218

Robinson, M. N., Tansil, K. A., Elder, R. W., Soler, R. E., Labre, M. P., Mercer, S. L., . . .
Fridinger, F. (2014). Mass media health communication campaigns combined with
health-related product distribution: a Community Guide systematic review. American
Journal of Preventive Medicine, 47(3), 360-371.
Robinson, W. S. (2009). Ecological correlations and the behavior of individuals. International
Journal of Epidemiology, 38(2), 337-341.
Rogers, E. M. (1995). Diffusion of innovations (5 ed.): Free Press.
Roncancio, A. M., Ward, K. K., Sanchez, I. A., Cano, M. A., Byrd, T. L., Vernon, S. W., . . .
Fernandez, M. E. (2015). Using the theory of planned behavior to understand cervical
cancer screening among Latinas. Health Education & Behavior, 1090198115571364.
Rudd, R., Kirsch, I., & Yamamoto, K. (2004). Literacy and health in America. Policy
information Report. Educational Testing Service.
Sampson, R. J. (2012). Great American city: Chicago and the enduring neighborhood effect:
University of Chicago Press.
Sampson, R. J., Morenoff, J. D., & Gannon-Rowley, T. (2002). Assessing" neighborhood
effects": Social processes and new directions in research. Annual Review of Sociology,
443-478.
Schaefer-McDaniel, N., O'Brien Caughy, M., O'Campo, P., & Gearey, W. (2010). Examining
methodological details of neighbourhood observations and the relationship to health: A
literature review. Social Science & Medicine, 70(2), 277-292. doi:
http://dx.doi.org/10.1016/j.socscimed.2009.10.018
Schumacker, R. E., & Lomax, R. G. (2004). A beginner's guide to structural equation modeling
(3 ed.). New York, NY: Psychology Press.
219

Schuster, D. V., Valente, T. W., Skara, S. N., Wenten, M. R., Unger, J. B., Cruz, T. B., &
Rohrbach, L. A. (2006). Intermedia processes in the adoption of tobacco control activities
among opinion leaders in California. Communication Theory, 16(1), 91-117.
Schwartz, S. J., Unger, J. B., Zamboanga, B. L., & Szapocznik, J. (2010). Rethinking the concept
of acculturation: Implications for theory and research. American Psychologist, 65(4),
237-251.
Schwartz, S. J., Weisskirch, R. S., Zamboanga, B. L., Castillo, L. G., Ham, L. S., Huynh, Q.-L., .
. . Cano, M. A. (2011). Dimensions of acculturation: Associations with health risk
behaviors among college students from immigrant families Journal of Counseling
Psychology (Vol. 58, pp. 27-41). United States: American Psychological Association.
Seo, M., & Matsaganis, M. D. (2013). How interpersonal communication mediates the
relationship of multichannel communication connections to health-enhancing and health-
threatening behaviors. Journal of Health Communication, 18(8), 1002-1020. doi:
10.1080/10810730.2013.768726
Siegel, R., Naishadham, D., & Jemal, A. (2012). Cancer statistics, 2012. CA: A Cancer Journal
for Clinicians, 62(1), 10-29. doi: 10.3322/caac.20138
Singhal, A., & Rogers, E. M. (1999). Entertainment education: A communication strategy for
social change. Mahwah, NJ: Lawrence Erlbaum Associates.
Slater, M. D., Hayes, A. F., Reineke, J. B., Long, M., & Bettinghaus, E. P. (2009). Newspaper
coverage of cancer prevention: Multilevel evidence for knowledge-gap effects. Journal of
Communication, 59(3), 514-533. doi: 10.1111/j.1460-2466.2009.01433.x
220

Slater, M. D., & Rasinski, K. A. (2005). Media exposure and attention as mediating variables
influencing social risk judgments. Journal of Communication, 55(4), 810-827. doi:
10.1111/j.1460-2466.2005.tb03024.x
Slater, M. D., Snyder, L. B., & Hayes, A. F. (2006). Thinking and modeling at multiple levels:
The potential contribution of multilevel modeling to communication theory and research.
Human Communication Research, 32(4), 375-384. doi: 10.1111/j.1468-
2958.2006.00292.x
Small, M. L., Jacobs, E. M., & Massengill, R. P. (2008). Why organizational ties matter for
neighborhood effects: Resource access through childcare centers. Social Forces, 87(1),
387-414. doi: 10.1353/sof.0.0079
Small, M. L., & McDermott, M. (2006). The presence of organizational resources in poor urban
neighborhoods: An analysis of average and contextual effects. Social Forces, 84(3),
1697-1724. doi: 10.1353/sof.2006.0067
Smiley, M. J., Diez Roux, A. V., Brines, S. J., Brown, D. G., Evenson, K. R., & Rodriguez, D.
A. (2010). A spatial analysis of health-related resources in three diverse metropolitan
areas. Health & Place, 16(5), 885-892.
Smith-McLallen, A., & Fishbein, M. (2008). Predictors of intentions to perform six cancer-
related behaviours: Roles for injunctive and descriptive norms. Psychology, Health &
Medicine, 13(4), 389-401.
Smith, K. C., Niederdeppe, J., Blake, K. D., & Cappella, J. N. (2013). Advancing cancer control
research in an emerging news media environment. JNCI Monographs, 2013(47), 175-
181. doi: 10.1093/jncimonographs/lgt023
221

Smoke-Free Ontario Scientific Advisory Committee. (2010). Evidence to guide action:
Comprehensive tobacco control in Ontario. Ontario Agency for Health Protection and
Promotion Toronto, ON.
Snyder, L. B., Fleming-Milici, F., Slater, M. D., Sun, H., & Strizhakova, Y. (2006). Effects of
alcohol advertising exposure on drinking among youth. Archives of pediatrics &
adolescent medicine, 160(1), 18-24. doi: 10.1001/archpedi.160.1.18
Snyder, L. B., Hamilton, M. A., Mitchell, E. W., Kiwanuka-Tondo, J., Fleming-Milici, F., &
Proctor, D. (2004). A meta-analysis of the effect of mediated health communication
campaigns on behavior change in the United States. Journal of Health Communication,
9(sup1), 71-96. doi: 10.1080/10810730490271548
Sorensen, G., Emmons, K., Stoddard, A. M., Linnan, L., & Avrunin, J. (2002). Do social
influences contribute to occupational differences in quitting smoking and attitudes toward
quitting? American journal of health promotion, 16(3), 135-141.
Southwell, B. G., & Yzer, M. C. (2007). The roles of interpersonal communication in mass
media campaigns. In C. Beck (Ed.), Communication Yearbook 31 (pp. 420-462). New
York: Lawrence Erlbaum.
Stephens, K. K., Rimal, R. N., & Flora, J. A. (2004). Expanding the reach of health campaigns:
Community organizations as meta-channels for the dissemination of health information.
Journal of Health Communication, 9(sup1), 97-111. doi: 10.1080/10810730490271557
Stevens, R., & Hornik, R. C. (2014). AIDS in black and white: The influence of newspaper
coverage of HIV/AIDS on HIV/AIDS testing among African Americans and White
Americans, 1993–2007. Journal of Health Communication, 19(8), 893-906.
Stiff, J. B., & Mongeau, P. A. (2003). Persuasive communication: Guilford press.
222

Stokols, D. (1996). Translating social ecological theory into guidelines for community health
promotion. American journal of health promotion, 10(4), 282-298.
Straughan, J., & Hondagneu-Sotelo, P. (2001). From immigrants in the city, to immigrant city. In
M. J. Dear & J. D. Dishman (Eds.), From Chicago to LA: Making sense of urban theory
(pp. 187-211). Thousand Oaks, CA: Sage Publications.
Stryker, J. E., Emmons, K. M., & Viswanath, K. (2006). Uncovering differences across the
cancer control continuum: A comparison of ethnic and mainstream cancer newspaper
stories. Preventive Medicine, 44(1), 20-25.
Teig, E., Amulya, J., Bardwell, L., Buchenau, M., Marshall, J. A., & Litt, J. S. (2009). Collective
efficacy in Denver, Colorado: Strengthening neighborhoods and health through
community gardens. Health & Place, 15(4), 1115-1122.
Thompson, J. B. (1995). The media and modernity: A social theory of the media: Stanford
University Press.
U.S. Census Bureau. (2013a). Foreign born, from
https://www.census.gov/topics/population/foreign-born/about.html
U.S. Census Bureau. (2013b). Language use, from
http://www.census.gov/hhes/socdemo/language/about/faqs.html
U.S. Census Bureau. (2015). Population estimates Retrieved April 29, 2015, from
http://www.census.gov/popest/estimates.html
U.S. Cancer Statistics Working Group. United States Cancer Statistics: 1999–2011 Incidence and
Mortality Web-based Report (2014).
Valente, T. W., Hoffman, B. R., Ritt-Olson, A., Lichtman, K., & Johnson, C. A. (2003). Effects
of a social-network method for group assignment strategies on peer-led tobacco
223

prevention programs in schools. Am J Public Health, 93(11), 1837-1843. doi:
10.2105/ajph.93.11.1837
Valente, T. W., Paredes, P., & Poppe, P. R. (1998). Matching the message to the process The
telative ordering of knowledge, attitudes, and practices in behavior change research.
Human Communication Research, 24(3), 366-385. doi: 10.1111/j.1468-
2958.1998.tb00421.x
Valente, T. W., Ritt-Olson, A., Stacy, A., Unger, J. B., Okamoto, J., & Sussman, S. (2007). Peer
acceleration: Effects of a social network tailored substance abuse prevention program
among high-risk adolescents. Addiction, 102(11), 1804-1815. doi: 10.1111/j.1360-
0443.2007.01992.x
Viswanath, K. (2005). The communications revolution and cancer control. [10.1038/nrc1718].
Nat Rev Cancer, 5(10), 828-835.
Viswanath, K., Breen, N., Meissner, H., Moser, R. P., Hesse, B., Steele, W. R., & Rakowski, W.
(2006). Cancer knowledge and disparities in the information age. Journal of Health
Communication, 11(sup001), 1-17. doi: 10.1080/10810730600637426
Viswanath, K., & Emmons, K. (2009). Health communication and communication inequalities in
addressing cancer disparities. In H. K. Koh (Ed.), Toward the Elimination of Cancer
Disparities (pp. 277-298). New York, NY: Springer New York.
Viswanath, K., & Finnegan, J. R. (1996). The knowledge gap hypothesis: Twenty-five years
later. Communication Yearbook, 19, 187-228.
Viswanath, K., Randolph Steele, W., & Finnegan, J. R. (2006). Social capital and health: Civic
engagement, community size, and recall of health messages. American Journal of Public
Health, 96(8), 1456-1461. doi: 10.2105/ajph.2003.029793
224

Wakefield, M. A., Loken, B., & Hornik, R. C. (2010). Use of mass media campaigns to change
health behaviour. The Lancet, 376(9748), 1261-1271. doi:
http://dx.doi.org/10.1016/S0140-6736(10)60809-4
Wang, F., & Luo, W. (2005). Assessing spatial and nonspatial factors for healthcare access:
towards an integrated approach to defining health professional shortage areas. Health &
Place, 11(2), 131-146.
Watts, L., Joseph, N., Velazquez, A., Gonzalez, M., Munro, E., Muzikansky, A., . . . del Carmen,
M. G. (2009). Understanding barriers to cervical cancer screening among Hispanic
women. American Journal of Obstetrics and Gynecology, 201(2), 199.e191-199.e198.
doi: 10.1016/j.ajog.2009.05.014
Weiss, L., Ompad, D., Galea, S., & Vlahov, D. (2007). Defining neighborhood boundaries for
urban health research. American Journal of Preventive Medicine, 32(6, Supplement),
S154-S159. doi: http://dx.doi.org/10.1016/j.amepre.2007.02.034
West, J. H., Blumberg, E. J., Kelley, N. J., Hill, L., Sipan, C. L., Schmitz, K. E., . . . Hovell, M.
F. (2010). Does proximity to retailers influence alcohol and tobacco use among Latino
adolescents? Journal of Immigrant and Minority Health, 12(5), 626-633.
Wheaton, B., Muthen, B., Alwin, D. F., & Summers, G. F. (1977). Assessing reliability and
stability in panel models. . Sociological Methodology, 8, 84-136.
White, K. M., Smith, J. R., Terry, D. J., Greenslade, J. H., & McKimmie, B. M. (2009). Social
influence in the theory of planned behaviour: The role of descriptive, injunctive, and in-
group norms. British Journal of Social Psychology, 48(1), 135-158. doi:
10.1348/014466608x295207
225

Wickrama, K. A. S., & Bryant, C. M. (2003). Community context of social resources and
adolescent mental health. Journal of Marriage and Family, 65(4), 850-866. doi:
10.1111/j.1741-3737.2003.00850.x
Wilkin, H. A. (2013). Exploring the potential of communication infrastructure theory for
informing efforts to reduce health disparities. Journal of Communication, 63(1), 181-200.
doi: 10.1111/jcom.12006
Wilkin, H. A., & Ball-Rokeach, S. J. (2011). Hard-to-reach? Using health access status as a way
to more effectively target segments of the Latino audience. Health Education Research,
26(2), 239-253. doi: 10.1093/her/cyq090
Wilkin, H. A., & González, C. (2006). Are Spanish-language television shows connecting
Latino residents in Los Angeles to their health storytelling networks? . Paper presented at
the International Communication Association Annual Conference, Dresden, Germany.
Wilkin, H. A., Katz, V. S., Hether, H. J., & Ball-Rokeach, S. (2012, October). Community and
family support for obesity prevention behaviors among Latinos and African Americans.
Paper presented at the 140th American PublicHealth Association Annual Meeting, San
Francisco, CA.
World Health Organization. (2014). Comprehensive cervical cancer control: A guide to essential
practice (2 ed.): World Health Organization.
Wyatt, R. O., Katz, E., & Kim, J. (2000). Bridging the spheres: Political and personal
conversation in public and private spaces. Journal of Communication, 50(1), 71-92. doi:
10.1111/j.1460-2466.2000.tb02834.x
Yanovitzky, I., & Rimal, R. N. (2006). Communication and normative influence: An
introduction to the special issue. Communication Theory, 16(1), 1-6.
226

Young, A. M., DiClemente, R. J., Halgin, D. S., Sterk, C. E., & Havens, J. R. (2014). Drug
users’ willingness to encourage social, sexual, and drug network members to receive an
HIV vaccine: A social network analysis. AIDS and Behavior, 18(9), 1753-1763. doi:
10.1007/s10461-014-0797-9
Zikmund-Fisher, B. J., Windschitl, P. D., Exe, N., & Ubel, P. A. (2011). ‘I’ll do what they did”:
Social norm information and cancer treatment decisions. Patient Education and
Counseling, 85(2), 225-229. doi: http://dx.doi.org/10.1016/j.pec.2011.01.031

227

APPENDIX A: DATA COLLECTION OF THE MULTILEVEL PROJECT

The Multilevel Project (R01CA155326 - Murphy/Ball-Rokeach) seeks to understand the
barriers and conduits that exist and operate at individual, interpersonal, and community level to
affect knowledge, beliefs, attitudes, and behaviors related to cervical cancer prevention,
detection, and treatment for Hispanic women in Los Angeles. The multi-method, multiple-
component project consisted of an in-person survey, 12 focus groups, and field observations of
communication “hotspots” (places where wo men spend time with their friends and family) in 6
neighborhoods in the study area. For the survey component, the study population included a)
women who were following the cervical cancer screening guidelines and had received a Pap test
in the past 3 years (hereafter referred to as “compliant” group); and b) women who were not
following the screening guideline either because they had never had a Pap test or they had not
received one in over 3 years (hereafter referred to as “noncompliant” group).
Data collections took place from April 2012 to December 2013 at the LAC+USC
Women’s Clinic, South Central Family Health Center (SCFHC), Mosaic Family Care clinic, and
public spaces (such as parks, bus stops, laundromats, or community events). Compliant
participants were recruited by having trained research staffs approach and/or hand out a flyer to
potentially eligible women waiting to get a Pap test at the clinics. With the same fashion,
noncompliant participants were recruited from women who were waiting in the clinics’ lobby but
were not there to receive care themselves, and from women at public spaces.
Specifically, this project adopted an organic, interactive, and neighborhood-centered
recruitment approach that was determined by its research aims. One of the main objectives of
the project was to identify critical factors cross levels that differentiate Hispanic women who
228

engage in different cervical cancer detection activities (e.g., factors that differentiate those who
complied with the screening guidelines and those who did not, those who returned for a
colposcopy following an abnormal Pap test result from those who did not, and those who
followed through treatment from those who did not). In order to detect sufficient statistical
difference at the neighborhood level in multilevel analysis, this research aim required that each
neighborhood cluster had a minimum of 5 participants, and that the ratio of compliant to
noncompliant participants within each cluster was approximately 2 to 1. Since August 2012, the
research team closely monitored the spatial concentration of participants to ensure the
satisfactory size and composition of participants for each neighborhood cluster
9
.
Moreover, for the behavioral outcomes related to cervical cancer prevention, detection,
and treatment, the research team collaborated with the participating clinics and retrieved the
numbers of compliant participants who acquired a Pap test, who subsequently received an
abnormal result, who returned for colposcopy for an abnormal result, and finally who did and did
not follow up with treatment. The data were then merged with the survey data. The final dataset
consists of 1632 participants who completed the survey, and contains the records of behavioral
outcomes for the compliant participants.

9
See more details on defining neighborhood clusters in Appendix B.
229

APPENDIX B: PROCEDURE TO DEFINE NEIGHBORHOOD CLUSTERS

An essential step of the Multilevel Project is to delineate boundaries of neighborhood
clusters for the study area in Los Angeles (LA). Studies using multilevel modeling to analyze
neighborhood health effect typically define neighborhoods based on contiguous units with
arbitrarily drawn boundaries such as census tracts (Matsaganis, 2015; Sampson, et al., 2002).
This approach may not necessarily reflect the cultural, institutional, and physical characteristics
of neighborhoods, however (Weiss, et al., 2007). Based on previous work of the Metamorphosis
Project (http://www.metamorph.org/), the Multilevel Project developed a novel approach to
define neighborhoods in a way that responds to this challenge and captures residents’ everyday
communication environment where decisions related to health  such as cervical cancer
prevention, detection, and treatment  are shaped and acted upon.
Furthermore, the majority of research on the dynamics between neighborhoods and
aggregate or individual outcomes define neighborhoods either in advance (e.g., for recruitment
purposes) or post hoc (e.g., using secondary data collected at various geographic units) (Sampson
et al, 2002). By comparison, the Multilevel Project’s approach features an organic and iterative
procedure that considered the geographic catchment areas in which the survey participants
concentrate. The geographic distribution of participants guided the initial development of
neighborhood clusters and informed the place-based, targeted recruitment strategy, which
informed the process of refining the pre-identified clusters. The text below presents in order of
priority some of the guidelines that served the cluster development, followed by the multistage
procedure to determine the clusters.

230

Guidelines
Potential for efficient recruitment
In order to properly perform multilevel modeling analysis, it is required that each cluster
is small enough to capture unique neighborhood characteristics, but large enough to ensure
feasible survey recruitment (i.e., 1:2 ratio of noncompliant versus compliant participants per
cluster). This criterion was taken into consideration during the initial stage of drawing
neighborhood clusters based on the spatial concentration of participants.
Census tract boundaries
Though widely used in neighborhood effects literature, census tracts are an imperfect
proxy of neighborhoods because they are frequently out of sync with the historic, cultural,
socioeconomic, and geographical characteristics that define neighborhoods(Matsaganis, 2015).
Using census tracts, however, allows for determining the census-derived neighborhood
characteristics for multilevel modeling analysis. Because of this, tracts became the smallest units
of defining clusters in this project
10
.
Evidence of homogeneity within clusters and heterogeneity across clusters
Relative homogeneity within a cluster and heterogeneity across clusters is important not
only for multilevel analysis, but also for appropriate attribution of census-derived neighborhood-
level variables. The Multilevel Project mainly used two neighborhood-level variables to
evaluate within-cluster homogeneity and across-cluster heterogeneity: linguistic isolation,
measured by the percentage of residents who speak English less than very well, and
homeownership, measured the by the percentage of housing units occupied by owners as
opposed to renters. Both variables were obtained at the level of census tract from the American

10
Census tract boundaries were based on Census 2010 and were downloaded from http://www.nhgis.org.
231

Community Survey 5-Year Estimate 2007-2011
11
. The decision of using these two variables was
based on a series of spatial analysis and principal component analysis over a wide range of
census data (e.g., percentage of Hispanic populations, foreign born populations, residential
stability, etc). Results of the principal component analysis indicate that linguistic isolation was
the strongest factor to differentiate the clusters on the social-demographic dimension, followed
by homeownership.
Procedure
Identify spatial distribution of participants
During the survey, participants were asked to provide their residential address
information specific to street level. The residential addresses were imported into an online
geographic information-processing platform
12
for batch geocoding and manual geocode
correction to ensure the desired quality of address information (Goldberg, 2013). Valid
addresses were then used to explore the spatial pattern of the survey respondents, supplemented
by the observation gained from a review of medical charts of patients from the LAC+USC
Women’s Clinic
13
. The geographic distribution of participants revealed three catchment areas
that are home to the majority of participants as of late May, 2013: East LA, Central LA, and
South LA.
Delineate initial clusters
A total of 22 neighborhood clusters (hereafter referred as “pre -identified clusters”) were
first drawn based on the valid residential addresses for 1035 participants recruited prior to June

11
Available at http://factfinder2.census.gov/
12
http://geoservices.tamu.edu/Services/Geocode/
13
The research team conducted a separate review of medical charts of 962 women who had Pap tests at the Los
Angeles County USC (LAC+USC) Women's Clinic in May, June and July of 2011. These data indicate that patients
at the LAC+USC Women’s Clinic reside in diverse locations throughout Los Angeles County. Furthermore, three
main geographic concentrations of patients were identified, including East LA, South LA, and Central LA.
232

2013. The research team also reviewed a number of sources in order to ensure that the clusters
were delineated in a way that approximated the historical and sociocultural identities of
established neighborhoods. These sources are: Healthy City (http://www.healthycity.org/ ),
Mapping LA (http://maps.latimes.com/neighborhoods), Metamorphosis Project
(http://www.metamorph.org/), land use maps (http://planning.lacounty.gov/luz), and natural
boundaries such as rivers, freeways and major streets. The text below describes how the
Metamorphosis Project, Mapping LA, and the natural boundaries were employed for cluster
development.
Metamorphosis Project. Headquartered at USC, the Metamorphosis Project is
dedicated to understanding “the transformation of urban community under the forces of
globalization, population diversity, and new communication technology through multi-level and
multi-method field analysis of communication flows that sustain and transform the social
textures of everyday place and cyberspace” (http://www.metamorph.org/). Since its inception,
the Metamorphosis Project has developed an archive of neighborhood boundaries based on not
only public data, but also residents’ perceptions as for what their neighborhoods or communities
are like.
Mapping L.A. Drawn and maintained by the Los Angeles Times (LA Times), the
Mapping LA project has become a growing resource about neighborhoods in Los Angeles
County, providing individual maps and information about demographics, crime, and education
for 272 neighborhoods. The building blocks of neighborhoods are census tracts obtained from
the U.S. Census Bureau, which are then supplemented with reader feedback. The neighborhood
boundaries visualized by Mapping LA guided the initial attempt of cluster development.
Moreover, Mapping LA was extensively used in combination with the research findings of the
233

Metamorphosis Project to guide the decision in refining the pre-identified clusters. For example,
both sources indicate that central LA is home to ethnically, culturally, and socioeconomically
distinctive neighborhoods. Such local variations in neighborhood characteristics would be
obscured if the level of analysis was at the level of LA County or a catchment area.
Natural boundaries. One of the major limitations of using census tracts to define a
neighborhood is that tracts do not always respect street patterns where, for example, tract
boundaries may cut through residential areas or do not line up with major streets. Thus,
adjustments were made during cluster delineation by excluding or including a particular census
tract so that the cluster boundaries maximally aligned with natural boundaries, such as rivers,
freeways, and major streets. The impact of land use was also taken consideration. That is,
census tracts were excluded if they were largely for business and industry use, or only scarce
residential census data were available for them.
Modifying Pre-Identified Clusters
Grouping Analysis
Using the Grouping Tool available for ArcGIS 10.2, a data-driven approach to regroup
the pre-identified clusters was adopted to acquire additional insights on the heterogeneity across
clusters. Given the number of groups to create, the Grouping Tool groups features (here, census
tracts where participants lived) in such a way that the features within each group are as similar to
one another, whereas the groups themselves are as different from one another on a number of
variables as possible. For the cluster development purpose, a stepwise analysis was performed
by adding at a time each of the following variables known to affect neighborhood belonging
234

(Kim & Ball-Rokeach, 2006a, 2006b): linguistic isolation, residential tenure
14
, and home
ownership.
Hot Spot analysis
Additionally, it was also of interest whether our pre-identified/original and regrouped
clusters (of census tracts) follow the clustering of linguistic isolation, residential tenure, and
homeownership. This question was explored by using the Hot Spot Analysis tool available at
ArcGIS 10.2. The tool calculates the Getis-Ord Gi* statistic (pronounced G-i-star) for each
feature (here, census tracts) in a spatial dataset. The analysis generates z-scores and p-values
that show where features with either high or low values cluster spatially. A statistically
significant hot spot is a feature that not only has a high value, but also is surrounded by other
features with high values. Results from the hotspot analysis allowed for refining clusters. For
example, there is a previously unassigned area that shares boundaries with, and is
demographically similar to Central Alameda as suggested in hotspot analysis. This area was
later incorporated into Central Alameda.
Finalizing Clusters
Based on the procedures delineated above, a total of 25 clusters were finalized as of
October 19, 2013. These clusters are visualized in Appendix C.

14
Measured as the tract-level percentage of households that moved before (Census 2010)
235

APPENDIX C: 25 NEIGHBORHOOD CLUSTERS DEFINED BY THE MULTILEVEL PROJECT

Figure C1. 25 neighborhood clusters defined by the Multilevel Project
236

APPENDIX D: SPATIAL DISTRIBUTION OF POPULATION CHARACTERISTICS OF THE STUDY AREA

Figure D1. Census-tract level percentage of linguistically isolated population in the study area
237

Figure D2. Census-tract level ethnic heterogeneity in the study area
238

Figure D3. Census-tract level percentage of Hispanic population in the study area
239

Figure D4. Census-tract level percentage of African American population in the study area
240

Figure D5. Census-tract level percentage of foreign-born population in the study area
241

Figure D6. Hotspot analysis of linguistic isolation in the study area
242

Figure D7. Hotspot analysis of ethnic heterogeneity in the study area
243

Figure D8. Hotspot analysis of percentage of Hispanic population in the study area
244

Figure D9. Hotspot analysis of African American population in the study area
245

Figure D10. Hotspot analysis of foreign-born population in the study area
246

APPENDIX E: COMMUNICATION RESOURCES IN THE STUDY AREA

Table E1. Neighborhood-Level Categories of Communication Resources for Neighborhood
Clusters in the Study Area

Neighborhood
Clusters
Categories
a
(N)
Church School Library Recreation
Organizations &
Social Services
1 Boyle Heights
31 14 3 20 15
2 Central Alameda
1 5 1 7 2
3 Crenshaw
46 11 4 23 40
4 East Adams
13 2 0 4 6
5 East Hollywood
18 4 2 5 30
6 East LA 37 16 3 11 49
7 El Monte
33 42 3 17 53
8 El Sereno
12 6 1 4 8
9 Florence
14 13 2 4 13
10 Highland Park
18 3 1 2 3
11 Historic South Central
23 5 0 2 24
12 Huntington Park
19 4 2 5 19
13 Koreatown
29 7 2 8 117
14 Lincoln Heights
10 4 1 9 12
15 Maywood
22 11 3 11 7
16 Montebello
15 10 1 6 8
17 Pasadena
5 10 1 7 20
18 Pico Union
11 1 0 7 9
19 Pomona
46 24 2 32 32
20 South Gate
35 13 1 5 5
21 Vermont Slauson
44 11 3 9 41
22 Vernon Main
4 6 1 4 21
23 Watts
64 19 2 13 52
24 Westlake
13 6 2 7 93
a
Categories refer to the types of services offered at each location of interest. A location can be
grouped into more than one category if multiple services are provided at the site. For illustration
purposes, locations considered in this table are for communication resources that strictly fall
within neighborhood clusters.

247

APPENDIX F: UNREPORTED STATISTICAL TESTS
Table F1. Hierarchical Linear Models of Descriptive Norms Regarding Never Having a Pap Test
for Noncompliant Participants: Individual-Level Covariates, Neighborhood Storytelling
Resources, and Neighborhood-Level Variables (Full Models)
Predictors
Model 1 Model 2 Model 3 Model 4
Coefficient (Standard Error)

Neighborhood Mean Descriptive Norms
41.108
(4.475)***
41.298
(4.457)***
41.043
(4.482)***
41.179
(4.483)***

Individual-Level Variables
Age

-.399
(.178)*
-.412
(.177)*
-.405
(.178)*
-.404
(.178)*
8th Grade and Above
(Dummy)
6.568
(3.642)
¶

5.854
(3.597)
¶

6.114
(3.599)
¶

5.974
(3.611)
¶

Health Care Coverage
(Dummy)
4.648
(2.751)
¶

4.617
(2.749)
¶

4.732
(2.752)
¶

4.767
(2.752)
¶

English Proficiency

.410
(2.154)
.742
(2.139)
.598
(2.141)
.623
(2.142)
Residential Tenure

-.071
(.155)
-.061
(.155)
-.069
(.155)
-.069
(.155)
First Generation Immigrants
(Dummy)
2.153
(4.237)
2.802
(4.199)
2.544
(4.204)
2.619
(4.205)
English Language Media

-1.873
(1.703)
-1.867
(1.700)
-1.793
(1.703)
-1.758
(1.704)
Interpersonal Discussion

1.383
(.478)**
1.351
(.478)**
1.372
(.478)**
1.369
(.478)**
Local/Ethnic Media

-1.693
(1.582)
-1.580
(1.579)
-1.630
(1.581)
-1.610
(1.580)
Community Org. 1.612
(1.734)
1.696
(1.733)
1.601
(1.736)
1.633
(1.733)

Neighborhood-Level Variables
Linguistic Isolation

.142
(.199)
-- -- --
Ethnic Heterogeneity

-- 12.090
(8.970)
-- --
Density of Communication Resources

-- -- -.194
(.450)
--
Density of Health Service Providers

-- -- -- -1.615
(2.812)
Level 2 variance 26.096 22.982 27.316 26.485
Level 1 variance 605.542 605.293 605.142 605.863
Chi-square 2.96* 2.08*
2.72* 3.04*
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001
248

Table F2. Hierarchical Linear Models of Descriptive Norms Regarding Never Having a Pap Test
for Noncompliant Participants: Individual-Level Covariates, ICSN, and Neighborhood-Level
Ethnic Heterogeneity (Full Model and Cross-Level Model)
Predictors
Model 1 Model 2
Coefficient (Standard Error)

Neighborhood Mean Descriptive Norms

42.381 (4.483)***

42.289 (4.480)***
Individual-Level Variables
Age

-.386 (.180)* -.385 (.179)*
8th Grade and Above (Dummy)

5.478 (3.633)

5.693 (3.627)

Health Care Coverage (Dummy)

4.733 (2.773)
¶
4.990 (2.772)
¶

English Proficiency

1.099 (2.134) .818 (2.138)
Residential Tenure

-.025 (.156) -.018 (.155)
First Generation Immigrants (Dummy)

1.639 (4.208)

1.482 (4.199)

English Language Media

-2.164 (1.690) -2.011 (1.689)
ICSN

.846 (.469)
¶
.759 (.472)
Neighborhood-Level Variables
Ethnic Heterogeneity

12.992 (9.065) 12.873 (9.163)
Cross-Level Interaction
Ethnic Heterogeneity ×ICSN

-3.002 (2.272)
Level 2 variance 23.571 25.161
Level 1 variance 617.790 613.895
Chi-square 2.13
¶
2.41
¶

¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

249

Table F3. Hierarchical Linear Models of Descriptive Norms Regarding Never Having a Pap Test
for Noncompliant Participants: Individual-Level Covariates, ICSN, and Neighborhood-Level
Density of Communication Resources (Full Model and Cross-Level Model)
Predictors
Model 1 Model 2
Coefficient (Standard Error)

Neighborhood Mean Descriptive Norms

42.139 (4.514)***

42.067 (4.512)***
Individual-Level Variables
Age

-.378 (.178)* -.372 (.180)*
8th Grade and Above (Dummy)

5.761 (3.636)

5.849 (3.641)

Health Care Coverage (Dummy)

4.860 (2.776)
¶
4.768 (2.784)
¶

English Proficiency

.933 (2.137) .981 (2.138)
Residential Tenure

-.032 (.156) -.034 (.156)
First Generation Immigrants (Dummy)

1.343 (4.214)

1.426 (4.216)

English Language Media

-2.076 (1.693) -2.126 (1.697)
ICSN

.830 (.469)
¶
.850 (.471)
¶

Neighborhood-Level Variables
Density of Communication Resources

-.178 (.458) -.184(.455)
Cross-Level Interaction
Density of Comm. Resources ×ICSN

.053 (.118)
Level 2 variance 28.116 27.186
Level 1 variance 618.254 618.439
Chi-square 3.03* 2.83*
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

250

Table F4. Hierarchical Linear Models of Descriptive Norms Regarding Never Having a Pap Test
for Noncompliant Participants: Individual-Level Covariates, ICSN, and Neighborhood-Level
Density of Health Service Providers (Full Model and Cross-Level Model)
Predictors
Model 1 Model 2
Coefficient (Standard Error)

Neighborhood Mean Descriptive Norms

42.266 (4.514)***

42.426 (4.498)***
Individual-Level Variables
Age

-.378 (.180)* -.349 (.181)*
8th Grade and Above (Dummy)

5.626 (3.648)

5.320 (3.651)

Health Care Coverage (Dummy)

4.896 (2.776)
¶
4.588 (2.787)

English Proficiency

.957 (2.137) 1.272 (2.149)
Residential Tenure

-.032 (.156) -.053 (.157)
First Generation Immigrants (Dummy)

1.422 (4.216)

1.559 (4.210)

English Language Media

-2.035 (1.695) -2.140 (1.696)
ICSN

.834 (.469)
¶
.804 (.469)
¶

Neighborhood-Level Variables
Density of Health Service Providers

-1.640 (2.857) -1.553 (2.811)
Cross-Level Interaction
Density of Health Services ×ICSN

.891 (.773)
Level 2 variance 28.841 26.324
Level 1 variance 617.536 616.726
Chi-square 3.33* 2.82*
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

251

Table F5. Hierarchical Linear Models of Media Recall Regarding Pap Tests for Noncompliant
Participants: Individual-Level Covariates, Neighborhood Storytelling Resources, and
Neighborhood-Level Variables (Full Models)
Predictors
Model 1 Model 2 Model 3 Model 4
Coefficient (Standard Error)
Neighborhood Mean Media Recall 1.632
(.161)***
1.632
(.161)***
1.625
(.160)***
1.634
(.159)***

Individual-Level Variables
Age

.010
(.006)
.010
(.006)
.010
(.006)
.010
(.006)
8th Grade and Above
(Dummy)
-.060
(.128)
-.060
(.127)
-.061
(.127)
-.079
(.126)
Health Care Coverage
(Dummy)
.192
(.096)

*
.192
(.096)

*
.193
(.096)

*
.197
(.096)

*
English Proficiency

-.082
(.075)
-.082
(.075)
-.081
(.075)
-.074
(.075)
Residential Tenure

-.005
(.005)
-.005
(.005)
-.005
(.005)
-.005
(.005)
First Generation Immigrants
(Dummy)
-.021
(.148)
-.021
(.147)
-.020
(.147)
-.007
(.147)
English Language Media

.021
(.060)
.021
(.060)
.023
(.060)
.027
(.060)
Interpersonal Discussion

.028
(.017)
¶

.028
(.017)
¶

.029
(.017)
¶

.029
(.017)
¶

Local/Ethnic Media

.147
(.055)**
.147
(.055)**
.146
(.055)**
.149
(.055)**
Community Org. .020
(.061)
.020
(.061)
.019
(.061)
.024
(.061)

Neighborhood-Level Variables
Linguistic Isolation

.000
(.008)
-- -- --
Ethnic Heterogeneity

-- -.019
(.366)
-- --
Density of Communication Resources

-- -- -.017
(.017)
--
Density of Health Service Providers

-- -- -- -.205
(.102)*
Level 2 variance .055 .055 .050 .040
Level 1 variance .763 .763 .763 .762
Chi-square 6.69** 6.58**
5.53** 4.12*
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

252

Table F6. Hierarchical Linear Models of Media Recall Regarding Pap Tests for Noncompliant
Participants: Individual-Level Covariates, ICSN, and Neighborhood-Level Linguistic Isolation
(Full Model and Cross-Level Model)
Predictors
Model 1 Model 2
Coefficient (Standard Error)
Neighborhood Mean Media Recall

1.613 (.160)*** 1.612 (.160)***
Individual-Level Variables
Age

.009 (.006) .009 (.006)
8th Grade and Above (Dummy)

-.057 (.128) - .062 (.128)
Health Care Coverage (Dummy)

.197 (.096)* .201 (.096)*
English Proficiency

-.101 (.075) -.103 (.075)
Residential Tenure

-.005 (.005) -.005 (.005)
First Generation Immigrants (Dummy)

.010 (.148) .013 (.148)
English Language Media

.038 (.059) .039 (.059)
ICSN

.047 (.016)** .046 (.016)**
Neighborhood-Level Variable
Linguistic Isolation

.001 (.008) .001 (.008)
Cross-Level Interaction
Linguistic Isolation ×ICSN

.001 (.002)
Level 2 variance .052 .051
Level 1 variance .770 .768
Chi-square 6.08** 6.02**
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

253

Table F7. Hierarchical Linear Models of Media Recall Regarding Pap Tests for Noncompliant
Participants: Individual-Level Covariates, ICSN, and Neighborhood-Level Ethnic Heterogeneity
(Full Model and Cross-Level Model)
Predictors
Model 1 Model 2
Coefficient (Standard Error)
Neighborhood Mean Media Recall

1.613 (.160)*** 1.613 (.161)***
Individual-Level Variables
Age

.009 (.006) .009 (.006)
8th Grade and Above (Dummy)

-.059 (.127) -.059 (.127)
Health Care Coverage (Dummy)

.197 (.096)* .197 (.096)*
English Proficiency

-.101 (.074) -.101 (.074)
Residential Tenure

-.005 (.005) -.005 (.005)
First Generation Immigrants (Dummy)

.011 (.147) .011 (.147)
English Language Media

.038 (.059) .038 (.059)
ICSN

.047 (.016)** .047 (.017)**
Neighborhood-Level Variable
Ethnic Heterogeneity

- .022 (.360) - .022 (.360)
Cross-Level Interaction
Ethnic Heterogeneity ×ICSN

- .002 (.079)
Level 2 variance .052 .052
Level 1 variance .770 .770
Chi-square 5.95** 5.95**
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

254

Table F8. Hierarchical Linear Models of Media Recall Regarding Pap Tests for Noncompliant
Participants: Individual-Level Covariates, ICSN, and Neighborhood-Level Density of
Communication Resources (Full Model and Cross-Level Model)
Predictors
Model 1 Model 2
Coefficient (Standard Error)
Neighborhood Mean Media Recall

1.606 (.160)*** 1.606 (.160)***
Individual-Level Variables
Age

.009 (.006) .009 (.006)
8th Grade and Above (Dummy)

-.060 (.127) -.060 (.127)
Health Care Coverage (Dummy)

.199 (.096)* .198 (.096)*
English Proficiency

-.100 (.074) -.099 (.074)
Residential Tenure

-.005 (.005) -.005 (.005)
First Generation Immigrants (Dummy)

.012 (.146) .013 (.147)
English Language Media

.040 (.059) .039 (.059)
ICSN

.047 (.016)** .047 (.016)**
Neighborhood-Level Variable
Density of Communication Resources

- .016 (.017) - .016 (.017)
Cross-Level Interaction
Density of Comm. Resources ×ICSN

.001 (.004)
Level 2 variance .047 .047
Level 1 variance .770 .770
Chi-square 5.05* 5.04**
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

255

Table F9. Hierarchical Linear Models of Media Recall Regarding Pap Tests for Noncompliant
Participants: Individual-Level Covariates, ICSN, and Neighborhood-Level Density of Health
Service Providers (Full Model and Cross-Level Model)
Predictors
Model 1 Model 2
Coefficient (Standard Error)
Neighborhood Mean Media Recall

1.615 (.158)*** 1.616 (.158)***
Individual-Level Variables
Age

.009 (.006) .010 (.006)
8th Grade and Above (Dummy)

-.079 (.127) -.086 (.127)
Health Care Coverage (Dummy)

.202 (.096)* .192 (.096)*
English Proficiency

-.093 (.074) -.082 (.074)
Residential Tenure

-.005 (.005) -.006 (.005)
First Generation Immigrants (Dummy)

.025 (.146) .032 (.146)
English Language Media

.044 (.059) .040 (.059)
ICSN

.048 (.016)** .048 (.016)**
Neighborhood-Level Variable
Density of Health Service Providers

- .201 (.101)* - .202 (.101)
Cross-Level Interaction
Density of Health Service ×ICSN

.032 (.027)
Level 2 variance .038 .038
Level 1 variance .769 .766
Chi-square 3.72* 3.84*
¶
p ≤ .10,*p ≤ .05, ** p ≤ .01, *** p ≤ .001

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Contagious: social norms about health in work group networks

PDF

A multilevel communication approach to understanding human trafficking prevention behaviors in Indonesia

PDF

Immigrant health in context: a communication ecology approach to understanding health behavior

PDF

Cervical cancer study in Malawi: a novel design for field investigation

PDF

The role of social control in promoting healthy eating behavior among Chinese immigrants: an ecological approach

Asset Metadata

Creator Zhao, Nan (author)

Core Title Understanding normative influence of neighborhoods: a multilevel approach to promoting Latinas’ cervical cancer prevention behaviors in urban ethnic communities

School Annenberg School for Communication

Degree Doctor of Philosophy

Degree Program Communication

Publication Date 09/14/2017

Defense Date 06/15/2015

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag cancer prevention,cervical cancer,communication infrastructural theory,descriptive norms,ethnicity,geographic information system,health communication,health disparities,hierarchical linear modeling,Hispanics,Latinos,neighborhoods,OAI-PMH Harvest,Pap test,perceived norms,public health,spatial analysis,structural equation modeling,urban health

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Murphy, Sheila T. (committee chair), Ball-Rokeach, Sandra J. (committee member), Chou, Chih Ping (committee member)

Creator Email nzhao@usc.edu,nznanzhao@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-182262

Unique identifier UC11274639

Identifier etd-ZhaoNan-3906.pdf (filename),usctheses-c40-182262 (legacy record id)

Legacy Identifier etd-ZhaoNan-3906.pdf

Dmrecord 182262

Document Type Dissertation

Format application/pdf (imt)

Rights Zhao, Nan

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA