Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Classically conditioned responses to food cues among obese and normal weight individuals: conditioning as an explanatory mechanism for excessive eating
(USC Thesis Other)
Classically conditioned responses to food cues among obese and normal weight individuals: conditioning as an explanatory mechanism for excessive eating
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
CLASSICALLY CONDITIONED RESPONSES TO FOOD CUES AMOUNG OBESE AND
NORMAL WEIGHT INDIVIDUALS: CONDITIONING AS AN EXPLANATORY
MECHANISM FOR EXCESSIVE EATING
BY
Kulwinder Singh, M.A.
______________________________________________________________________________
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PSYCHOLOGY)
May 2014
Copyright 2014 Kulwinder Singh
CLASSICAL CONDITIONING AND OBESITY 2
Acknowledgements
I am extremely grateful for the abundance of support I received during the five-year
journey that has culminated in this dissertation. First, I would like to thank my advisor and
mentor, Dr. Michael E. Dawson. I will never forget your enthusiasm for taking me on and
accepting me as your student during a time when I was in need. From when we first met during
first-year orientation, to our close collaborations on various projects and this dissertation, you
provided me with a model for developing a successful and worthwhile career. I greatly
appreciate the countless hours that you have spent with me to help mold me into the best student
and professional as possible. I would also like to thank Dr. Anne M. Schell for investing great
time and energy into my progression as a student and scholar. Your guidance in teaching me how
to develop strong experimental designs, utilize appropriate statistical tests, and produce written
work of a high quality was invaluable to my progression in graduate school.
I would also like the thank each of my dissertation committee members, including Drs.
Biing-Jiun Shen, Bob Knight, John Monterosso, and John Brekke for their participation and
ongoing support. Each of you has had a substantial influence on my progression as a graduate
student, and I thank you for your invaluable feedback and suggestions on my work. I’d like to
give extra appreciation to Dr. Biing-Jiun Shen. Without your belief that I would be a successful
graduate student, I would not have had any of the wonderful opportunities over the past five
years at USC. Dr. Larry Jamner, my undergraduate mentor, is also someone who I would like to
express my appreciation too. Among so many other potential undergraduate students, you gave
me what little time you had, cared about my progression as a student and scholar, and truly
believed that I would be able to achieve my professional and personal goals. I thank you for
being a major inspiration and role model in my life. I would also like to give special thanks to the
CLASSICAL CONDITIONING AND OBESITY 3
entire USC clinical psychology faculty, many of whom served as teachers, supervisors, and
colleagues. Over the past five years, many of you supported and guided me through times of
challenge as well as success and celebration.
I cherish the many friendships that blossomed during my time in graduate school. Drs.
Stacy Eisenberg and Uta Maeda were my lab-mates during my first two years at USC. From
when we met during graduate school interviews, and to this day, you both took me under your
wing and helped make my transition from college to graduate school as seamless as possible. I
greatly admire both of you and am excited to witness the great contributions that you have
already made and will continue to make in the field. Dr. Chris Courtney was my lab-mate in the
Dawson lab during my third and fourth year of graduate school. You played an instrumental role
in my transition into a new lab, an experience that can be stressful without the necessary support.
You showed patience and grace when providing me with support on any of the projects that I
was involved with, and for that I am truly appreciative. I would also like to acknowledge my
cohort-mates, Larissa Borofsky Del Piero, Illana Kellerman, John Keefe, and Caitlin Smith, for
providing me with the support that can only be given by life-long friends. Through the joy, grief,
anxiety, and basically whole spectrum of possible emotions, we survived and our bond has only
grown stronger. I want to personally thank each of you for giving me your trust and friendship,
and am eager and excited to see how we grow together in the future.
By directly witnessing how such excellent mentorship can influence one’s personal and
career trajectories, it has always been an important goal of mine to provide the highest possible
standard of mentorship to students who were once in my position. I’d like to thank each of the
undergraduate research assistants that have assisted me in my journey over these past five years.
Jessica Matlock, John Arrington, Kelsey Rupp, Danny Liu, Judy Chen, Sarah Sack, Emily
CLASSICAL CONDITIONING AND OBESITY 4
Kamen, Zara Butte, Brendan Rabideau, Andrew Payne, Sarah Barrett, Nethanya Cortez, Kathryn
DeYoung, and Jonathan Kashani, thank you for all of your hard work and dedication to the
projects that we collaborated on. I consider myself extremely fortunate to have worked with each
of you. Already, many of you have gone on to pursue careers in medicine, psychology, social
work, and entrepreneurship. It was a pleasure to have worked with you and I hope that I was able
to provide the same quality of mentorship that I have been so fortunate to receive.
From the beginning of my journey, a consistent source of strength and support for me has
been my family. At an early age, my parents faced extraordinary difficulties associated with
immigrating to a foreign country. Even though there were many challenges, you made the
sacrifices that were necessary to provide your children with the best educational opportunities
possible. It was only through your hard work, self-sacrifice, and love that I was able to make it to
this point. Thank you for always serving as a model of strength and courage for me. From when
we first met in college, no one has been closer to me than my beautiful wife, Aditi. There are so
many things that I can say to express my appreciation to you, but nothing would be quite enough.
From the day that you stepped into my life, I was transformed into the person that I always
wanted to be. Your unconditional love and support has been the foundation for my growth and
the fuel for my continued drive to achieve the goals I have set out for myself. Thank you for
giving me your love, strength, courage, and devotion. Thank you for allowing my successes to be
your successes, and yours mine. Thank you for always being by my side and cherishing the
person I am. We did this together, and it could not have happened without your love and support.
Finally, I thank God for all that I have been blessed with in my life.
CLASSICAL CONDITIONING AND OBESITY 5
Table of Contents
Acknowledgements 2-4
Table of Contents 5
Abstract 6
Introduction 7-37
Methods 38-50
Results 51-60
Discussion 61-75
References 76-89
Tables 90
Figures 91-101
Appendix 102-111
CLASSICAL CONDITIONING AND OBESITY 6
Abstract
As prevalence rates of obesity greatly increase in response to changing societal and
environmental landscapes, it has become clear that genetic influences cannot solely account for
recent global surges in weight. One possible mechanism that has been proposed to contribute to
the powerful effects of the environment on eating is associative learning. 121 obese, overweight,
and normal-weight students participated in a differential conditioning experiment using food
cues as conditioned stimuli (CSs) and a sip of chocolate milkshake combined with a picture of an
appetizing chocolate snack as unconditioned stimuli (UCS). Skin conductance and heart rate
interbeat-intervals were utilized as measures of autonomic conditioning, and self-report valence
changes and affective priming were used as measures of evaluative conditioning. It was
hypothesized that obese participants would show greater responses to the UCS and stronger
conditioning to food cues compared to normal-weight participants. It was also hypothesized that
autonomic conditioning would predict differential eating only for obese participants during a
bogus taste test task. Results demonstrated that participants in the obese group had larger
magnitude and more frequent responses to the UCS than did normal-weight participants. For
autonomic conditioning, only normal-weight participants who were sated showed autonomic
conditioning. For evaluative conditioning, obese and normal-weight participants’ demonstrated
conditioning based on the self-report valence ratings but not on the affective priming task.
Finally, autonomic conditioning predicted differential eating during the taste test only for obese
participants who were food deprived. This association was non-significant, albeit positive for
normal-weight participants who were food-deprived. These results suggest that processes of
Pavlovian classical conditioning may contribute to short-term increases in caloric consumption.
However, additional steps which are further elaborated upon, must be taken to determine
whether processes of classical conditioning may be a mechanism for weight gain and obesity.
CLASSICAL CONDITIONING AND OBESITY 7
Classically Conditioned Responses to Food Cues Among Obese and Normal-Weight
Individuals: Conditioning as an Explanatory Mechanism for Excessive Eating
Obesity rates in the United States have grown to alarming rates, as approximately 35.7%
of adults and 16.9% of children and adolescents in the general population are now obese (Ogden,
Carroll, Kit, & Flegal, 2012). The high prevalence of obesity in the general population has
produced a large societal burden. In the United States alone, the direct and indirect cost of
obesity is estimated at over 215 billion dollars per year (Hammond & Levine, 2010). In addition,
the increasing prevalence of chronic disease, including cardiovascular disease, type 2 diabetes,
stroke, and cancer, is closely tied to excessive weight gain. For example, approximately 90
percent of total type 2 diabetes cases are attributed to excess weight (Hossain, Kawar, & El
Nahas, 2007). Although recent evidence suggests a slowing or leveling off (Flegal, Carroll,
Ogden, & Curtin, 2010; Ogden, Carroll, Curtin, Lamb, & Flegal, 2010), current obesity
prevalence rates represent a twofold increase in the United States (Stein & Colditz, 2004) and a
threefold increase in developing countries (Hossain et al., 2007) since the 1980’s. By 2030,
obesity rates are expected to further increase, particularly in developing regions such as the
Middle East, South America, Pacific Islands, and Asia (WHO, https://apps.who.int/infobase).
Increasing obesity prevalence rates in developing countries have coincided with
“westernization,” a lifestyle that is typically characterized by increased sedentary behavior and
an abundance of highly caloric and palatable food available at little expense (Price, 2011). As
prevalence rates of overweight and obesity dramatically increase in response to changing societal
landscapes and urbanization, it has become clear that genetic factors cannot solely be held
accountable. Over the past 40 years, a great deal of effort has been expended to better understand
how environmental characteristics may influence eating. As a result, it has become more clear
CLASSICAL CONDITIONING AND OBESITY 8
that the environment has powerful influences in determining what, when, and how much we eat
on a daily basis (Herman & Polivy, 2008; Wansink, 2004).
Since environmental influences have been shown to have a profound impact on eating,
investigators have increasingly attempted to identify and understand the specific mechanisms
that may help explain this relationship. Associative learning, typically tested through classical
conditioning procedures, has been suggested to be a potential mechanism for excessive eating
(Birch, McPhee, Sullivan, & Johnson, 1989; Booth, 1977; Brunstrom, 2007; Gibson &
Brunstrom, 2007; Rozin & Zellner, 1985). Although associative learning cannot solely explain
the high prevalence of obesity in the United States and worldwide, it may help provide a clearer
understanding of the processes by which environmental influences can promote excessive eating.
Food Cues and Eating
Associative learning may potentially provide one explanatory pathway by which
environmental cues can influence eating. Prior to examining the validity of a learned model of
eating, it is prudent to first establish whether exposure to food and food-related cues can reliably
influence consummatory behavior. A food cue is any non-visceral stimulus that can signal that
food is available to be consumed (Schachter, 1971; Schachter & Gross, 1968). Food cues can be
of two types. Inherent food cues are directly related to the sensory characteristics of food or how
it is provided or presented (Wansink, 2004). Examples of inherent food cues that have been
shown to influence eating include aroma (Fedoroff, Polivy, & Herman, 2003), visibility (Painter,
Wansink, & Hieggelke, 2002), proximity (Wansink, Painter, & Lee, 2006), and portion size
(Levitsky, 2005). For example, Wansink et al. (2006) examined whether manipulations in
proximity and visibility influenced caloric consumption among 40 secretaries. The study was a 2
CLASSICAL CONDITIONING AND OBESITY 9
x 2 within-subjects design with repeated measures. Visibility (visible vs. hidden) and proximity
(proximate vs. distant) were the factors. Visibility was manipulated by placing chocolate in
bowls that were either clear or opaque. Proximity was manipulated by either placing chocolates
on the desks of participants so that they were immediately accessible or 2 m from the desks so
that they were less accessible. Data collection commenced for a period of four weeks during
which participants were tested within each condition while caloric consumption was measured.
Results revealed significant main effects for both proximity and visibility. Caloric consumption
significantly increased both when the chocolates were immediately accessible and clearly visible.
Furthermore, visibility and proximity significantly interacted such that individuals consumed
significantly more chocolates when they were visible and proximate compared to when less
proximate and less visible.
Arbitrary food cues refer to contextual stimuli that may be associated with eating but are
independent of any particular food. Examples of these cues include time on the clock (Schachter
& Rodin, 1974), the number of individuals present during a meal (de Castro, Brewer, Elmore, &
Orozco, 1990; de Castro & de Castro, 1989), the presence of family members (de Castro, 1994;
Laessle, Uhl, & Lindel, 2001), and arbitrary environmental stimuli such as lights and tones
(Birch et al., 1989). de Castro et al. (1990) tracked participant’s eating habits for seven days
using a daily-diary monitoring system. Results indicated that meal size and the number of people
present were positively correlated such that meal sizes increased concomitantly as the number of
individuals present during a meal increased. In a follow-up study, de Castro (1994) examined
meal patterns among 515 adult participants who agreed to track meal consumption, time of
occurrence, hunger, mood, and the presence of others using a daily-diary monitoring system.
Results demonstrated that self-reported consumption was significantly higher and length of time
CLASSICAL CONDITIONING AND OBESITY 10
eating was shorter when participants reported eating in the presence of family compared to when
with companions or alone. These studies provide evidence that social facilitation of a meal may
modify typical eating practices and influence individuals to increase caloric intake.
Food Cues and Obesity
Whether normal-weight, overweight, and obese individuals differ in their responsiveness
to food cues has been debated for decades, and remains a hotly-contested issue today. The
earliest work examining differential cue responding among obese and normal-weight individuals
was conducted by Stanley Schachter and colleagues (Schachter, 1971; Schachter & Gross, 1968;
Schachter & Rodin, 1974). Schachter’s externality theory of obesity suggests that obese
individuals are more susceptible to the influences of external cues related to eating and less
sensitive to internal cues and satiety signals than their normal-weight counterparts (Schachter,
1971). Thus, obese individuals are particularly likely to consume food when in the presence of
external cues.
A number of existing studies provide support for Schachter’s externality theory. For
example, Nisbett (1968) found that augmenting portion size increased caloric intake among the
obese, whereas normal-weight individuals ate the same amount irrespective of their portion size.
Ross (1974) found that manipulations of the visual and cognitive salience of food cues produced
greater consumption among obese participants in the high-salience condition than obese-low
salience and normal-weight participants in either condition. Social influences have also been
found to differentially impact eating among the obese. Relative to normal-weight children, obese
children eat more in the presence of important family members than when they are alone
(Laessle et al., 2001). In addition, overweight and obese participants regulate their eating by
CLASSICAL CONDITIONING AND OBESITY 11
consuming less food when in the presence of lean compared to overweight individuals, whereas
normal-weight participants eat similar amounts of food regardless of the weight status of their
company (de Luca & Spigelman, 1979; Salvy, Romero, Paluch, & Epstein, 2007). Other food
cues have been shown to influence eating, particularly for overweight and obese individuals. In a
sample of 42 school children aged 9-11 years, Halford, Gillespie, Brown, Pontin, and Dovey
(2004) found that overweight and obese children were more likely to recognize food-related
advertisements during television viewing than were lean children, whereas the groups did not
differ in their recognition of non-food related advertisements. Of greater importance, they also
found that the overall snack food intake of overweight and obese children was greater than that
of lean children when tested following advertisement viewing. These results were replicated
among a younger sample (5-7 years old) (Halford, Boyland, Hughes, Oliveira, & Dovey, 2007).
Although obese and overweight participants were more influenced by the food-related
advertisements than were normal-weight individuals, it should be noted that all three groups ate
significantly more after viewing the advertisements.
A number of other studies have also demonstrated that the eating behavior of both
normal-weight and overweight/obese individuals can be influenced by cue exposure. Schachter
and Gross (1968) discovered that manipulating a clock (an arbitrary cue) to inaccurately suggest
that dinner time was approaching increased caloric intake (number of crackers consumed) among
obese participants. Normal-weight participants were also affected by the time manipulation, but
instead of eating more as dinner time approached, they reduced consumption. As an explanation
of the results, the investigators stated that normal-weight participants often refused the crackers,
as they were concerned that increasing consumption during the experiment would spoil their
appetite for dinner. Although the normal-weight participants decreased consumption as a result
CLASSICAL CONDITIONING AND OBESITY 12
of the time manipulation, they nevertheless used the time to draw inferences about how much
they should have eaten.
Jansen et al. (2003) found a similar pattern of results. Following a preload, overweight
and normal-weight children were exposed to intense food smells (cue exposure condition) for 10
minutes. Results indicated that both groups were influenced by the exposure condition; following
the cue exposure condition, overweight children increased their caloric intake, whereas normal-
weight children decreased intake. To explain the decreased eating among normal-weight
participants, the investigators suggested that the initial preload may have primed overweight
children to eat more, whereas it may have produced sensory-specific satiety among the normal-
weight participants. The investigators also found that although both overweight and normal-
weight children experienced increased salivary flow in response to food cue exposure, the
difference between the groups was not significant. However, among the normal-weight children,
increased salivary flow did not predict caloric intake during a subsequent taste test, but the
association between salivation and caloric intake was significant in overweight children (r = .62),
demonstrating a potential link between physiological activity after food cue exposure and eating
behavior.
Two recent studies examined whether overweight and normal-weight individuals differed
across numerous outcomes when directly exposed to food cues. Both Tetley, Brunstrom, and
Griffiths (2009) and Ferriday and Brunstrom (2011) conducted cue exposure studies with the
sight and smell of pizza to examine whether overweight individuals were more likely to respond
to such cues than normal-weight individuals. Tetley et al. (2009) found that their participants on
the whole indicated more desire and cravings for pizza after exposure than before. Although
overweight and normal-weight participants did not differ in desire to eat and cravings for pizza
CLASSICAL CONDITIONING AND OBESITY 13
after exposure, the overweight participants showed greater increases in desired portion sizes than
normal-weight individuals after controlling for participants’ liking of pizza. It should be noted
that although the study sought to test differences between normal-weight and overweight
participants, there was a relatively low overall sample mean body-mass index (BMI) (M = 22.9,
SD = 2.6; a value in between 25 and 30 would be considered overweight and 30 or greater
obese), which may have influenced critical comparisons between the groups. Ferriday and
Brunstrom (2011) found increases in salivation, rated hunger, and consumed portion sizes after
exposure among all participants. Overweight participants had greater increases in desire to eat
and produced more saliva after cue exposure than did normal-weight participants. However, they
found no differences in consumed portion sizes or caloric consumption between the groups.
Although the previous studies provide varying degrees of support for the externality
hypothesis, other evidence is inconsistent with this hypothesis (Grinker, Hirsch, & Smith, 1972;
Price & Grinker, 1973; Rodin, Slochower, & Fleming, 1977). In a series of studies, Rodin et al.
(1977) tested responsiveness to food cues before and after weight loss among a predominantly
overweight sample of dieters. Results demonstrated only a weak (and non-significant)
relationship between degree of obesity and the degree to which participants responded to food
cues across a multitude of measures. In addition, weight loss did not influence food cue
responding. Although this study found no relationship between weight loss and externality, the
authors cite data (Milstein, 1980; Rodin & Slochower, 1976) supporting the notion that
responsiveness to food cues may be a characteristic that exists independently of obesity, and may
produce overweight and obesity in the appropriate (obesogenic) environment. In other words, it
is possible that some individuals possess a predisposition to eat excessively when they encounter
food cues, thereby placing them at a higher risk of weight gain. Although this predisposition can
CLASSICAL CONDITIONING AND OBESITY 14
exist across the spectrum of weight, these individuals nevertheless may be at a higher risk of
developing or maintaining obesity after prolonged and repeated cue exposure. This hypothesis
represents an intriguing possibility, given that many individuals do not drastically gain weight
despite continuous exposure to food cues that are pervasive in our modern food abundant
environments. All in all, although the literature is mixed, the majority of studies demonstrate that
the eating of normal-weight, overweight, and obese participants can be influenced by food cues,
with some studies suggesting that overweight and obese participants may be more likely to
increase consumption, whereas the eating behavior of normal-weight participants can be more
varied. In addition, studies also suggest that individuals who are responsive to food cues,
regardless of weight status, may be at risk for meal initiation, increased caloric consumption, and
perhaps even weight gain (Rodin & Slochower, 1976).
Attentional and Physiological Responses to Food Cues
Existing evidence suggests that obese individuals may show significant attentional biases
to food-related stimuli compared to normal-weight individuals. Castellanos et al. (2009) found
that obese participants demonstrated a significant gaze bias and longer gaze durations towards
pictures of food compared to normal-weight participants, suggesting an attentional bias for food
cues among the obese. Interestingly, group differences were discovered among the participants
who were sated, and not among food deprived individuals. Braet and Crombez (2003) also found
a significant attention bias to food cues among the obese. During a computerized version of the
Stroop task, obese children were markedly slower in naming the color of food words versus
control words compared to normal-weight children. In addition, the obese were not slower in
naming the color of negative-emotion words compared to normal-weight children, which may
reflect significant differences in attentional allocation specifically to food-related stimuli among
CLASSICAL CONDITIONING AND OBESITY 15
obese and normal-weight individuals. Altogether, these finding may be attributable to the greater
influence of external cues on the eating of the obese.
Food cue exposure has also been found to be associated with substantial autonomic and
endocrine activity across multiple physiological measures, including increased heart-rate (HR),
diastolic blood pressure (DBP), systolic blood pressure (SBP), skin conductance, and decreased
heart-rate variability (HRV) (Nederkoorn, Smulders, & Jansen, 2000; Vögele & Florin, 1997). In
addition, although evidence is limited in size, obese individuals demonstrate elevated autonomic
and endocrine activity when exposed to food cues compared to normal-weight individuals.
Epstein, Paluch, and Coleman (1996) found that obese participants displayed sustained
elevations in saliva after repeatedly tasting food compared to normal-weight participants. Vögele
and Florin (1997) demonstrated that binge eaters exhibited significantly higher blood-pressure
and skin conductance responses (SCRs) when exposed to food compared to non-binge eaters. In
this study, binge eaters had a significantly higher BMI compared to non-binge eaters. Although
the evidence is ambiguous, some evidence suggests that binge eating is far more common among
obese individuals, with prevalence rates more than double than those found among normal-
weight individuals (Smith, Marcus, Lewis, Fitzgibbon, & Schreiner, 1998). The authors also
found that HR during exposure to food cues predicted the amount of food eaten for all
participants, but especially for binge eaters, which suggests that autonomic responses as a result
of food cue exposure may influence subsequent caloric intake. Similar results have also been
found for obese and normal-weight participants. As previously stated, Jansen et al. (2003) found
that among normal-weight children, increased salivary flow after cue exposure did not predict
caloric intake during a subsequent taste test; however, the association between salivation and
caloric intake was highly significant in overweight children (r = .62, p < .01). Although only
CLASSICAL CONDITIONING AND OBESITY 16
correlations were tested, these studies suggest that autonomic and endocrine responses as a result
of food cue exposure may potentially serve as a mechanism for increased caloric consumption,
particularly among overweight or obese individuals.
Although normal-weight individuals experience increased autonomic and endocrine
activity during food cue exposure, existing evidence suggests that they do not increase caloric
consumption as a result. Stroebele and de Castro (2006) found that normal-weight participants
exhibited significantly higher mean HR when in the presence of food; however, HR was not
associated with increased caloric intake. Although the evidence is far from conclusive, the results
of these studies suggest that both normal-weight and obese individuals experience increased
physiological activity when in the presence of food cues; however, changes in physiological
indices may be predictive of subsequent caloric consumption only among the obese.
Learned Responses to Food Cues
Although the majority of existing literature demonstrates that food cues can have
substantial effects in determining what, when, and how much we eat, the processes that may
explain these effects remain poorly understood. Associative learning, typically tested using
classical conditioning procedures, has been suggested to help explain the relationship between
food cues and eating (Birch et al., 1989; Gibson & Brunstrom, 2007; Weingarten, 1983). In
addition, differences in behavioral, attentional, and physiological responses to food cues between
obese, overweight, and normal-weight individuals may reflect underlying group differences in
associative learning with food stimuli.
In a classical conditioning experiment, an external cue, which is intended to become a
conditioned stimulus (CS), is repeatedly paired with the presentation of food, the unconditioned
CLASSICAL CONDITIONING AND OBESITY 17
stimulus (UCS). After repeated pairings, exposure to the CS alone initiates appetitive behavior.
Differential conditioning, conducted with a within-subjects design, utilizes two different CSs
typically presented in a semi-random order during the associative phase of an experiment. One of
the two conditioned stimuli, the CS+, is consistently paired with the UCS; the second stimulus,
the CS−, can be paired with a low-intensity UCS (e.g., water) or presented in the absence of a
UCS. Participants’ conditioned responses (CRs) on CS+ trials are compared with their responses
on CS− trials, and conditioning is inferred on the basis of significant differential responding
(e.g., greater salivation) to CS+ than CS−. This design allows for increased confidence that
participants’ behavior change is a result of learning the specific CS-UCS contingency. Although
some studies have utilized between-subjects designs in which one group of participants receives
CS-UCS pairings and a control group receives unpaired presentations of the CS and the UCS, a
within-subjects design is the most optimal test for conditioning.
Conditioned Responses to Food Cues among Non-Human Species
Since the time of Pavlov (1927), evidence from animal studies has strongly suggested
that appetitive behavior in response to food cues can be classically conditioned to promote liking
and consumption, a finding which has been consistently replicated (Bitterman, Menzel, Fietz, &
Schäfer, 1983; Holland, 1980; Jenkins, Barrera, Ireland, & Woodside, 1978; Wasserman,
Franklin, & Hearst, 1974). In fact, studies have shown that even subtle differences in contextual
CSs can substantially influence eating in rats. Boggiano, Dorsey, Thomas, and Murdaugh (2009)
conducted a series of studies to examine the effects of subtle contextual CSs on conditioned
feeding. Rats were sated prior to presentation of the CSs. The authors discovered that after
consistently presenting appetizing, highly palatable food in the presence of a certain bedding
(CS+), rats increased chow consumption compared to when in the presence of another type of
CLASSICAL CONDITIONING AND OBESITY 18
bedding (CS−) that was never paired with food. The authors expressed surprise that a subtle
manipulation in a single environmental cue was enough to establish conditioning and
meaningfully influence caloric intake, a finding supported by existing literature demonstrating
the importance of contextual cues in appetitive conditioning processes (Petrovich, Ross,
Gallagher, & Holland, 2007). This finding is also noteworthy considering that the powerful
effects of contextual CSs on food consumption occurred above and beyond homeostatic-
motivated drives to eat, as rats were sated prior to food cue exposure.
In another series of studies with rats, Weingarten (1984a) demonstrated that a light and
buzzer (CS+) that had consistently been paired with food produced shorter latencies to engage in
a meal compared with a different light and buzzer (CS−) which had never been paired with food.
In a follow-up study, Weingarten (1984b) demonstrated that even after suppressing the effects of
the peripheral cholinergic system on the digestive system, and thus suppressing internally
motivated drives to eat, rats demonstrated conditioned feeding. Thus, discrimination learning to
appetitive stimuli was established in the absence of internal appetitive signals. Existing evidence
suggests that conditioned feeding among animals may instead be due to a specific component of
brain circuitry. Petrovich, Setlow, Holland, and Gallagher (2002) showed that after lesioning the
contralatral basolateral amygdalar-lateral hypothalamic (BLA-LHA) system, a critical
neuroanatomical feeding circuit particularly involved in meal initiation (Elmquist, Elias, &
Saper, 1999), food deprived rats still demonstrated Pavlovian discrimination learning with a tone
or noise as CS+ and delivery of food pellets as the UCS. However, when consumption tests
occurred one week after the initial learning period, the rats (who were now sated) failed to
demonstrate retention of conditioned feeding responses whereas control animals did show
significant retention. These results imply that separate brain circuits may be involved in
CLASSICAL CONDITIONING AND OBESITY 19
processes of acquisition and later retention of conditioned responses Altogether, the results from
animal studies are particularly impressive given the consistent finding that eating behavior can
come under stimulus control independent of internally motivated drives to eat.
Conditioned Responses to Food Cues among Humans
Conditioning Based on Measures of Consumption. The majority of studies using
differential conditioning designs have demonstrated appetitive conditioning with food among
humans, although the literature has not been entirely consistent. For example, a number of early
studies, primarily from the laboratory of Leanne Birch, demonstrated conditioned flavor
preferences among children (Birch, McPhee, Steinberg, & Sullivan, 1990; Johnson, McPhee, &
Birch, 1991; Kern, McPhee, Fisher, Johnson, & Birch, 1993). However, as noted by Birch et al.
(1990) other experiments conducted in their laboratory failed to demonstrate conditioning using
human participants. Brunstrom (2005) also noted an inconsistency in experimental studies of
appetitive conditioning. Although there are inconsistencies in the methodologies employed to
study differential appetitive conditioning, and the results gleaned from such experiments, a
sizeable literature exists that demonstrates differential conditioning with food among human
participants.
Two studies have demonstrated conditioning among human participants based on
measures of consumption. Coyle, Arnold, Goldberg-Arnold, Rubin, and Hall (2000) used
Pavlovian conditioning procedures to test whether infants would increase their intake of water
after an association was developed between an olfactory (CS) and their preferred formula (UCS).
Following the training phase, when the associated odor was later paired with water, infants
substantially increased water intake compared to pairing with a novel or no odor. Birch et al.
CLASSICAL CONDITIONING AND OBESITY 20
(1989) conducted a series of experiments to examine whether arbitrary CSs paired with food
during conditioning would promote meal initiation and caloric consumption during a subsequent
test phase. In the first study, the authors found that preschool children who consistently received
appetizing snacks (UCS) in the presence of a rotating red light and a song (CS+) increased
caloric consumption when later tested in the presence of the CS+ than when in the presence of a
distinctly different light and piece of music (CS−) that were never paired with food. Results
demonstrated increased caloric consumption only among children who were aware of the CS-
UCS contingency (n = 4). Children who were not aware of the CS-UCS contingency did not
show differential conditioning (n = 3). In addition, all children were sated prior to CS exposure,
indicating a differential effect of food cues on caloric consumption above and beyond
homeostatic-motivated drives to eat. However, the authors noted that conditioning was only
tested among a small sample size (n = 7). In the second study, the investigators employed similar
procedures to those of Study 1 except that they used a slightly larger sample size (n = 15) and
added latency to eat as a dependent measure. In addition, all participants correctly identified the
CS-UCS contingency. Results indicated that children ate more when exposed to the CS+
compared to the CS− (190 vs. 118 kcal), and initiated eating faster during CS+ than CS− trials
(approximately 90 s vs. 430 s). Thus, differential conditioning was established based on
measures of the latency of food intake initiation and the amount of intake as CRs. In addition, the
fact that conditioning was established using arbitrary neutral CSs (lights and music) is
impressive and is analogous to other seemingly arbitrary food cues (e.g., time on the clock) that
have been shown to influence eating.
The Birch et al. (1989) studies add support to a large body of evidence that indicates the
necessity of contingency awareness for the establishment of classical conditioning (Lovibond &
CLASSICAL CONDITIONING AND OBESITY 21
Shanks, 2002). In addition to classical conditioning with behavioral measures, contingency
awareness is also a requirement for “evaluative conditioning” (De Houwer, Thomas, & Baeyens,
2001). Thus, for studies testing the establishment of appetitive conditioning of food cues, it is
essential to differentiate between participants who are aware of the CS-UCS contingency from
those who are not.
Conditioning Based on Evaluative Valence Changes. Differential conditioning has
also been established based on measures of evaluative valence ratings, known as “evaluative
conditioning.” Evaluative conditioning is a type of Pavlovian classical conditioning that occurs
when the valence of a stimulus (the CS) is changed by pairing it with another valenced stimulus
(the UCS). In a typical evaluative conditioning paradigm, a CS is continuously paired with an
affectively positively or negatively valenced UCS. Evaluative conditioning occurs if the CS
adopts the affective valence of the UCS; that is, if a CS is rated by the participant more
positively after being paired with a positive UCS, or more negatively after being paired with a
negative UCS (Hofmann, De Houwer, Perugini, Baeyens, & Crombez, 2010; Levey & Martin,
1975; Martin & Levey, 1978). Evaluative conditioning allows for the examination of how likes
and dislikes (preferences) develop and change. Although some food preferences are genetically
determined (Poulton & Menzies, 2002), the majority are learned rather than innate (Rozin &
Millman, 1987). Evaluative conditioning is a likely way in which food preferences develop and
can be modified.
The learned valence in evaluative conditioning is commonly measured by asking
participants to verbally provide affective ratings of the CSs prior to conditioning. Subsequent to
conditioning, participants are again asked to provide affective ratings of the same stimuli.
Evaluative conditioning is evidenced if the reported valence of the CS significantly changes in
CLASSICAL CONDITIONING AND OBESITY 22
the direction of the UCS after continuous pairings with the negative or positive UCS. Although
self-report measures represent a simple method for testing evaluative responses, they are highly
susceptible to demand characteristics. For example, participants may rate the CS paired with a
positive UCS as more positive simply because they believe that is what the experimenter expects
or wants. In order to measure evaluative conditioning with more automatic measures that are less
susceptible to demand characteristics, recent studies have measured evaluative conditioning with
less obtrusive behavioral and psychophysiological measures.
A less obtrusive behavioral index of evaluative conditioning involves the measurement of
affective priming (Fazio, Sanbonmatsu, Powell, & Kardes, 1986). In affective priming,
participants view a sequence of stimulus pairs (e.g., words with positive or negative valence) in
which the congruence of the affective property of the first (prime) and the second (target) are
manipulated. Participants are instructed to identify the valence of the target as quickly as
possible and have been found to respond more quickly to the target when its valence is congruent
with the valence of the prime. In the evaluative conditioning paradigm, when the CS+ previously
paired with a positive UCS is used as a prime, motor responses to positively valenced targets are
faster than if the target is negatively or neutrally valenced, thus indicating that the CS+ acquired
positive valence (De Houwer, Hermans, & Eelen, 1998). The specific affective priming
procedure is described in further detail in the Methods section.
An interesting and relevant example of evaluative conditioning is conditioned taste
aversion (Garcia, Hankins, & Rusiniak, 1974; Rozin & Kalat, 1971). In order to demonstrate
conditioned taste aversion, the UCS is typically a food or other stimulus that produces nausea,
and the CS is an associated taste or flavor. After repeated pairings, or in some cases one pairing,
the affective valence of the previously neutral CS is rated significantly lower (Arwas, Rolnick, &
CLASSICAL CONDITIONING AND OBESITY 23
Lubow, 1989; Bernstein & Webster, 1980; Cannon, Best, Batson, & Feldman, 1983). Flavor-
flavor learning is another form of evaluative conditioning that employs food UCSs and is
particularly relevant for examining changes in food preferences. Flavor-flavor learning occurs if
a previously neutral flavor adopts the affective valence of another flavor with which it is
consistently paired (Capaldi, 1996). Evidence of flavor-flavor learning has been consistently
demonstrated in the literature. For example, in the original flavor-flavor learning paradigm,
Zellner, Rozin, Aron, and Kulish (1983) found that after pairing tea with sucrose, participants
rated the unsweetened tea as significantly more positive and desirable on the day of exposure and
one week later. Baeyens, Eelen, Van den Bergh, and Crombez (1990) later extended these results
and found evidence of evaluative learning for negative UCSs. Specifically, the investigators
found that a previously reported neutral flavor was rated as significantly more negative after
continuous pairings with polysorbate 20 compared to a neutral flavor that was not paired with the
negative UCS. Notably, the effects for the negative UCS were stronger than those for the
positive UCS. However, the experimenters suggested that the positive UCS was not as equally
liked among participants as the negative UCS was disliked, which likely influenced the degree of
conditioning and renders it difficult to compare evaluative responses across UCSs in the study.
In a sample of young children, Kern et al. (1993) tested whether participants would
acquire conditioned preferences for flavors paired with high-fat foods. Participants were
randomized to either a conditioning or control group. Each child was presented with samples of
five differently flavored yogurt drinks, and asked to rank their preferences for each flavor. The
flavors that fell in the middle of their rank-ordered preferences served as neutral CSs during the
conditioning trials. One neutral flavor, the CS+, was presented with a high-fat yogurt drink while
the other neutral CS− flavor was presented with the fat-free version. Results demonstrated that
CLASSICAL CONDITIONING AND OBESITY 24
for the conditioning group, children rated the CS+ more positive after conditioning than before.
There was no change in rating for the CS−. These effects were modulated by hunger such that
they were stronger among children who were hungry. Interestingly, during a two-month delayed
preference assessment, the children in the conditioning group continued to demonstrate greater
preferences for the CS+ compared to baseline preferences, whereas there still was no effect for
the CS−. These results suggest that, using self-reported valence ratings, flavor preferences can be
conditioned based on the postingestive consequences of high-fat intake, and may serve to
contribute to children’s preferences for and selection of high-fat foods.
Understanding how food preferences can change may allow for the development of
effective interventions to reduce excessive eating. For example, individuals’ liking of healthy
food can be increased by adding positive flavors (e.g., sugar) and gradually decreasing its
amount (Eertmans, Baeyens, & Van den Bergh, 2001). This procedure has been found to be
effective in increasing the liking of vegetables among college students (Capaldi, 1996). Two
recent studies have examined whether evaluative learning impacts food preferences and
behavior. Hollands, Prestwich, and Marteau (2011) randomly assigned participants to either an
experimental or control group. Participants in the experimental group were exposed to 100
conditioning trials in which pictures of appetizing snack food CSs were consistently paired with
aversive images of morbid obesity, whereas those in the control group were presented with a
blank screen. The primary outcome measures included implicit attitudes as measured by the
implicit association test (IAT) and whether participants opted to eat an appetizing snack or
healthy fruit after conditioning. Results indicated that participants in the experimental group
rated the appetizing snacks significantly more negative after conditioning and were also
significantly more likely to eat a healthy snack compared to a control group. Thus, changes in
CLASSICAL CONDITIONING AND OBESITY 25
food preferences as a result of evaluative conditioning increased the likelihood that participants
opted for healthy food over appetizing, high-fat food. Although the findings were promising,
consideration should be given to the possibility that participants’ food choice may be influenced
by demand characteristics. In addition, it would be important to test how long the effect of the
intervention persists. Nevertheless, this study suggests that the conditioned modification of
evaluative responses to food-related stimuli may represent an effective intervention for the
promotion of healthy eating.
Lebens et al. (2011) also tested whether evaluative conditioning modified food
preferences and behavior. 41 female participants in an experimental group were exposed to both
pictures of snack foods that were consistently paired with images of negatively-valenced female
bodily shapes and pictures of fruits paired with positively-valenced body shapes. For participants
in a control group, the CS-UCS pairings were random so that both the snack and fruit images
were followed by an equal number of positively- and negatively-valenced body shapes. In
addition, participants were given a virtual supermarket task in which they received an imaginary
budget of €15 (approximately 20.00 USD) and were asked to choose enough food and drink for
one day. Implicit attitudes about snack foods were measured using the IAT, while the calories
from the foods chosen were calculated and represented a dependent variable. Compared to the
control group, participants in the experimental group demonstrated greater negative associations
with snack foods. However, the total calories from the virtual foods indicated no differences
between the experimental and control groups. It is possible that the latter finding is due to the use
of a virtual environment to assess caloric consumption as opposed to testing participants’ actual
caloric consumption. Participants were also asked to choose enough food for only one day and
given a small budget. Given the restriction on the amount of food participants were instructed to
CLASSICAL CONDITIONING AND OBESITY 26
buy, it may not be surprising to find that there were no significant differences in calories of the
virtual foods chosen between the groups. Although the findings for caloric consumption are
mixed, all in all, these studies support to the notion that, through processes of evaluative
conditioning, our food attitudes and preferences (e.g., what we chose to eat) can be conditioned
and may significantly influence eating behavior.
Conditioned Responding to Food Cues among Obese Humans
The previous studies provide evidence that classically conditioned responses to food cues
influence eating and caloric intake. However, no studies have examined whether obese
individuals differ in conditioned responses to food cues compared to normal-weight individuals
based on measures of caloric consumption (Meyer (2012) compared conditioning between
normal-weight and overweight participants based on a proximal measure of salivation; see
Discussion section for further elaboration). As previously stated, obese individuals experience
substantial changes in behavioral, attentional, and physiological activity in response to food cues
compared to normal-weight individuals. In addition, the majority of evidence also indicates that
food cues more strongly influence meal initiation and overall caloric consumption among the
obese. The possibility exists that learned responses to food cues may meaningfully contribute to
an increased daily intake among the obese compared to normal-weight individuals.
In addition to studies comparing conditioned responding between obese and normal-
weight individuals based on measures of consumption, no studies have examined whether obese
individuals differ in conditioned evaluative responses to food cues compared to normal-weight
individuals. Understanding how food preferences develop, particularly among obese individuals,
may help explain maladaptive patterns of eating that ensue as a result of these learned
CLASSICAL CONDITIONING AND OBESITY 27
preferences. The development of food preferences for unhealthy, fatty foods may serve as a
particularly strong mechanism leading to excessive eating amongst obese individuals. Existing
evidence indicates that obese compared to normal-weight individuals develop strong preferences
for unhealthy foods, including meats, carbohydrate/fat sources (e.g., doughnuts, cookies, cake)
and sweet foods (Drewnowski, Kurth, Holden-Wiltse, & Saari, 1992). Rissanen et al. (2002)
also found that obese twins reported a significantly higher preference for fatty foods compared to
lean co-twins. Given that obese individuals demonstrate higher preferences for fatty foods, they
may subsequently be more likely to develop preferences for cues that are associated with
appetizing, albeit unhealthy food. If obese individuals show preferences for these cues, they may
also be more susceptible to initiate in a meal when in presence of these cues. If obese individuals
are more likely to rate neutral CSs as more likeable after being paired with a positive food-UCS,
modifying food preferences may represent an appealing intervention. Evaluative conditioning
would allow for a true test of this mechanism, particularly if truly neutral CSs are employed.
Hunger and Conditioning
Throughout history, learning has played an important role in determining the lifespan of
an organism. For our early ancestors, food preferences that developed as a result of learning were
essential in determining whether an organism was likely to survive. For example, humans and
animals learn quickly to reject foods that are bitter and astringent, as these characteristics are
usually signals that foods are poisonous and unsafe to eat. Conversely, humans and animals also
learn to ingest foods and liquids that are sweet-tasting. Sweetness is a sign of ripeness and high
sugar-content, and is a signal that food is likely safe to eat. Learned preferences to consume food
high in sugar content and avoid foods with bitter, astringent qualities likely protected our
ancestors from sickness and early death. In addition to guiding food preferences, learning has
CLASSICAL CONDITIONING AND OBESITY 28
also been instrumental in determining how much food was consumed. For much of human
history, humans have experienced extended periods of food deprivation. Learning to consume
foods high in fat and caloric content in excess of satiety whenever possible was an adaptive
behavior, as these opportunities were few and far between. Today, the need to eat in excess of
satiety has diminished, as current advancements in agricultural efficiency and manufacturing
have made high-energy, highly-palatable food available to the masses at little cost. Learned
processes that motivated the consumption of high-energy food in excess of satiety, which was
once an evolutionarily adaptive behavior that directly increased survivability, is now a potential
risk factor for overeating and the establishment of negative eating patterns in today’s food
abundant societies.
The previous studies provide evidence that both obese and normal-weight participants
show significant behavioral, attentional, and physiological responding when exposed to food
cues. Existing evidence also suggests that hunger may be an important factor that influences food
cue responding. Evidence from neuroimaging studies suggest that neurological responses to
pictures of food are substantially modulated by hunger. LaBar et al. (2001) found that during
exposure to food-related visual stimuli, food deprived participants demonstrated significantly
greater activation in the amygdala, parahippocampal gyrus, and anterior fusiform gyrus
compared to sated participants. Interestingly, deprivation influenced activation in areas of the
brain that are believed to mediate classical conditioning (e.g., amygdala and parahippocampal
gyrus). Physiological responses to food cues are also influenced by hunger. Drobes et al. (2001)
found that participants who were food deprived for 24 hours had significantly higher SCRs to
pictures of food compared with individuals who were not food deprived. In addition, food cues
CLASSICAL CONDITIONING AND OBESITY 29
elicited significant HR deceleration among participants who were food deprived for 24 hours, but
not amongst those who were sated.
Evidence also suggests that hunger may differentially influence food cue responding
among obese and normal-weight individuals, and may be especially influential for normal-
weight individuals (Nijs, Muris, Euser, & Franken, 2010; Schachter & Gross, 1968). A number
of studies have demonstrated that normal-weight individuals who are hungry show increased
attentional allocation to food-related stimuli compared to normal-weight, sated individuals
(Placanica, Faunce, & Soames Job, 2002; Stockburger, Schmälzle, Flaisch, Bublatzky, &
Schupp, 2009). In addition, hunger may also differentially influence food cue responding
between obese and normal-weight individuals. Nijs et al. (2010) found that both normal-weight
and obese participants demonstrated significant increases in attentional allocation (as measured
by an increased amplitude of the P300 event-related potential) to food-related stimuli compared
to neutral stimuli. However, normal-weight participants only showed an attentional bias to food
cues if they were hungry, whereas obese participants had an attentional bias regardless of hunger
state. Other studies also provide evidence that hunger differentially influences food cue
responding among obese and normal-weight individuals. Castellanos et al. (2009) demonstrated
that among sated participants, the obese continued to show increased gaze duration to food
images compared to neutral images, whereas normal-weight participants did not continue to
show a gaze bias to food images. In addition, among hungry participants, a group difference was
not observed, as obese and normal-weight individuals both demonstrated increased gaze duration
to food-related images compared to neutral images. These studies provide evidence that hunger
may significantly influence food cue responding among obese and normal-weight individuals, as
both groups demonstrate attentional biases to food cues when hungry; however, obese
CLASSICAL CONDITIONING AND OBESITY 30
individuals continue to show these biases after reaching satiation, whereas normal-weight
individuals do not.
In addition to differences in attentional allocation, evidence suggests that hunger may
also influence caloric consumption in response to food cue exposure. Normal-weight individuals
significantly increase caloric consumption when exposed to food cues if they are hungry than
when sated. Kauffman, Herman, and Polivy (1995) found that normal-weight participants who
were food deprived showed significant increases in caloric consumption when given ad libitum
access to food compared to sated participants. However, evidence also suggests that normal-
weight participants significantly reduce consumption when sated, a finding that generally does
not hold true for the obese. Although this notion is still highly debated, it may reflect the
increased responsiveness to internal cues among normal-weight individuals (Goldman, Jaffa, &
Schachter, 1968) and decreased sensitivity to internal cues among obese individuals (Schachter,
1968). As previously demonstrated, the eating behavior of the obese may be guided more by
external, environmental cues to eat (Schachter, Goldman, & Gordon, 1968). The possibility
exists that obese individuals who are less influenced by internal cues of satiation, may be more
likely to initiate in a meal and continue eating regardless of hunger state compared to normal-
weight individuals. This may be true particularly when the obese are in an environment saturated
with external food cues. During satiation, since internal cues are absent, normal-weight
individuals will likely reduce consumption, as they may be more sensitive to internal cues. Since
obese individuals respond to food cues even after reaching satiation, they will be more likely to
maintain caloric consumption upon reaching satiety.
Although evidence clearly indicates that hunger differentially influences food cue
responding among obese and normal-weight individuals, studies have not examined the role of
CLASSICAL CONDITIONING AND OBESITY 31
hunger in conditioned responses between obese and normal-weight individuals. It is plausible
that conditioned responses to food cues may be less influenced by hunger in obese compared to
normal-weight individuals. Normal-weight participants may show stronger conditioning when
they are hungry than when sated. In the obese, hunger may have less of an influence on
conditioned responses to food cues. Although studies have not compared differences in
conditioning across satiation state, it is plausible that hunger may be an important variable that
differentially influences conditioning among obese and normal-weight individuals.
Clinical Implications
Providing evidence that obese and normal-weight individuals differ in learned responses
to food cues, and that individual-differences in learning predicts differential eating would have
significant clinical implications for the treatment of obesity. If these results are found, learning
theory implies that repeated exposure to food cues in the absence of consumption would reduce
or extinguish the conditioned response (e.g., salivation, other autonomic responding). This type
of “food cue exposure” treatment can be considered a cognitive-behavioral intervention for
reducing conditioned responding through repeated exposure to cues in the absence of pleasurable
effects and appetitive behavior. This treatment may have substantial implications for intervention
efforts aiming to reduce unhealthy eating. If food cue exposure treatments do in fact produce
reliable reductions in eating, they would represent a useful treatment component in a weight
reduction program.
Although there is a paucity of evidence, a recent study evaluated the effects of food cue
exposure on a number of important outcomes among an overweight but otherwise healthy
sample. Boutelle et al. (2011) recruited children to participate in an 8 week food exposure
CLASSICAL CONDITIONING AND OBESITY 32
treatment (7 food exposure sessions). Children were asked to bring highly-craved foods to the
laboratory at each visit. During exposures, children provided ratings of their cravings while
looking at the food, holding the food, smelling the food, and after taking two bites of the food
(but not swallowing). Children were asked to give ratings of their cravings at 30 second intervals
for 15 minutes. After cravings were reported to be at a 2 or lower (on a five-point scale),
exposures were discontinued. In addition, participants were encouraged to use cognitive coping
strategies that were previously learned in past treatment sessions to “ride the craving wave.” The
treatment produced significant reductions in eating in the absence of hunger, subjective bulimic
episodes (feelings of loss of control without objectively large food consumption), and the
number of overeating episodes immediately after and six months post-treatment. All outcomes
remained significant 12 months post-treatment except for eating in the absence of hunger.
However, the treatment did not produce significant reductions in caloric intake or BMI at any
point during post-treatment follow-up.
Although the treatment produced several positive outcomes, a number of procedural
limitations may account for the non-significant findings for caloric intake and BMI. First, the
treatment offered only seven exposure sessions, which may not have been enough to
substantially reduce or extinguish conditioned responding to food, particularly if conditioning
with food UCSs is biologically prepared. Given that biologically prepared learning is particularly
resistant to extinction, a more effective intervention procedure may involve counterconditioning
with negative UCSs (e.g., aversive taste) as opposed to simple non-reinforcement of the
conditioned response. Second, treatment was only conducted within a laboratory setting instead
of the participant’s daily environment. It is possible that treatment gains made in the laboratory
may not have generalized to a participant’s home environment in which external cues are both
CLASSICAL CONDITIONING AND OBESITY 33
unique and plentiful. Third, participants were only exposed to the sensory qualities of the food
(e.g., proximity, smell, taste) without being exposed to subtle arbitrary food cues that may have
predicted the presentation or consumption of food. Providing evidence that subtle environmental
cues associated with food influence meal initiation and caloric consumption would encourage
clinicians to conduct an individualized, in-depth examination to identify these cues and
incorporate them into cue exposure treatment in the patient’s normal environment in addition to
the clinical setting. For example, during an individualized interview, the patient may reveal that
he/she typically consumes a favorite desert (e.g., ice cream) while watching TV each night after
dinner. During cue exposure, the clinician may ask the patient to retrieve the ice cream container
at the time in which the patient typically consumes the snack. The clinician may also ask the
patient to use the same dishware and sit in the same location while watching TV. The patient
would then be asked to open the container of ice cream, smell and taste the ice cream without
taking a bite. Consistent with a counter-conditioning model, an aversive taste could be mixed
with the ice cream which would then be consumed by the patient.
It should also be noted that Boutelle et al. (2011) did not include a group of overweight
and obese children who did not receive the cue exposure treatment. It is possible that the finding
of non-significant changes in BMI at follow-up may have been due to the success of the
treatment in preventing further increases in weight. However, because there were no obese
participants who did not receive the treatment, this hypothesis could not be tested. Future studies
examining the effects of cue exposure should aim to institute a prolonged cue exposure treatment
across many sessions, including sessions within participants’ home and potentially even school
environments. In addition, a control group of obese participants who do not receive the treatment
CLASSICAL CONDITIONING AND OBESITY 34
should be considered in order to test whether a cue exposure treatment may be successful in
preventing weight gain.
If obese and normal-weight individuals differ in evaluative responses to food cues, a
number of intervention strategies could be considered. For example, individuals’ liking for
healthy food can be increased by adding positive flavors (e.g., sugar) and gradually decreasing
its amount (Eertmans et al., 2001). This procedure has been found to be effective in increasing
the liking of vegetables among college students (Capaldi, 1996). Among obese individuals, food
preferences for positively-valenced healthy foods can be decreased by adding disliked flavors
(e.g., Polysorbate-20) to foods. These procedures would reflect the basic principles of evaluative
conditioning, and may represent interventions that produce successful reductions in unhealthy
eating and weight gain.
Aims and Hypotheses
Figure 1 provides an illustration of the hypothesized results for specific aims 2 and 3.
Specific Aim 1: To examine UCRs in normal-weight and obese individuals.
Hypothesis 1a: Obese participants will demonstrate a significantly higher frequency of UCRs
and show greater average UCR magnitudes than normal-weight participants.
Hypothesis 1b: Participants in the food deprived group will demonstrate a significantly higher
frequency of UCRs and show greater average UCR magnitudes than normal-weight participants.
Hypothesis 1c: The magnitude of skin conductance responding to the UCR will predict
differential and total calories consumed during the taste test among obese participants, but not
for normal-weight participants. These effects will be moderated by satiation state such that skin
CLASSICAL CONDITIONING AND OBESITY 35
conductance responding to the UCS will be a particularly strong predictor of differential and
total calories consumed among obese participants in the food deprived group.
Specific Aim 2: To examine classically conditioned responses to food cues in normal-weight and
obese individuals.
Hypothesis 2a: Obese and normal-weight participants will show differential autonomic
conditioning, as evidenced by increased skin conductance and HR deceleration to a neutral sound
consistently paired with food (CS+) compared to a sound never paired with food (CS-).
Hypothesis 2b: Obese and normal-weight participants will also demonstrate evaluative
conditioning as evidenced by greater changes in self-reported CS positive valence for CS+
compared to CS- and faster response latencies for congruent vs. incongruent trials (see affective
priming procedure in Methods section for further explanation).
Hypothesis 3a: Food deprived participants will demonstrate significantly greater differential
autonomic responding to CS+ than CS- compared to sated participants.
Hypothesis 3b: Food deprived participants will demonstrate greater changes in self-reported CS
positive valence for CS+ compared to CS- following conditioning and demonstrate faster
response latencies for congruent vs. incongruent trials than sated participants.
Hypotheses 4a: Obese participants in the sated group will demonstrate significantly greater
differential autonomic responding to CS+ than CS- compared to normal-weight participants in
the sated group.
Hypothesis 4b: Obese participants in the sated group will also demonstrate greater changes in
self-reported CS positive valence for CS+ compared to CS- following conditioning and
CLASSICAL CONDITIONING AND OBESITY 36
demonstrate faster response latencies for congruent vs. incongruent trials than normal-weight
participants in the sated group.
Hypothesis 5a: Obese participants in the food deprived group will not show greater autonomic
responding to CS+ versus CS- compared to normal-weight participants in the food deprived
group.
Hypothesis 5b: Obese participants in the food deprived group will not demonstrate greater
changes in self-reported CS positive valence to CS+ than CS- compared to normal-weight
participants in the food deprived group. In addition, obese participants in the food deprived
group will not demonstrate faster response latencies for congruent vs. incongruent trials
compared to normal-weight participants in the food deprived group.
Specific Aim 3: To examine whether conditioned responses to food cues influence caloric
consumption.
Hypothesis 6: Obese participants in the sated group will consume significantly more calories
compared to normal-weight participants in the sated group during CS+ compared to CS-
presentation.
Hypothesis 7: Obese and normal-weight participants in the food deprived groups will not differ
in caloric consumption during CS+ compared to CS- presentation.
Hypothesis 8: Among obese participants, autonomic conditioning will significantly predict
differential eating during the taste test task. This effect will be particularly pronounced among
obese participants in the food deprived group. Among normal-weight participants, autonomic
CLASSICAL CONDITIONING AND OBESITY 37
conditioning will not significantly predict differential eating during the taste test task, regardless
of satiation group assignment.
CLASSICAL CONDITIONING AND OBESITY 38
Methods
Participants
121 undergraduate students participated in the study. Participants were recruited through
on-campus solicitation and the psychology student subject pool. Participants were 26 male
(21.5%) and 95 female (78.5%) students between the ages of 18 and 31 (M = 20.12, SD = 1.84)
who met the study inclusion criteria. These criteria included: chocolate likers (>7 on a scale of 1-
10), no eating restrictions, no food allergies, and native English speakers. Inclusion criteria were
chosen based on their potential influences on the conditioning or taste test task procedures. A
number of additional inclusionary criteria were included based on their potential influences on
the autonomic nervous system (Mills & Dimsdale, 1991; Straneva, Hinderliter, Wells, Lenahan,
& Girdler, 2000). These criteria included: no past or current history of chronic illness, not
currently taking prescription medication known to influence the cardiovascular system, and no
smoking within the past year. Eligible participants were instructed to abstain from alcohol and
strenuous physical exercise 12 hours before their study appointment, as these variables may also
influence autonomic activity.
Stimuli and Measures
Food Cue Stimuli. Two 57 dB arbitrary sounds (600 Hz, white noise) were used as CSs.
The UCS was 3.5 cubic centimeters (cc) of chocolate milkshake delivered through plastic tubing
(PETCO, San Diego, CA) into the participant’s mouth combined with simultaneous presentation
of an appetizing chocolate snack picture. Prior to delivery, the milkshake was stored in a 60cc
hypodermic syringe (Terumo Medical Corporation, Tokyo, Japan), and automatically delivered
using an electrical fluid pump (New Era Pump Systems, Inc, NY). During conditioning, CSs
CLASSICAL CONDITIONING AND OBESITY 39
were presented binaurally using over-the-head style headphones eight seconds prior to the
administration of a four-second UCS for CS+ trials or in absence of a UCS for CS- trials. The
CSs and visual UCS were presented using PowerPoint software (Microsoft, WA). Participants
were presented with 10 trials of each CS (10 CS+ trials, 10 CS- trials; 20 total trials).
Psychophysiological Measures. In order to measure conditioned responses to food
stimuli, participants’ HR and SCRs were measured. Contact Precision Instruments equipment
with a 24-bit digital amplifier was used to record SCR and HR at a sampling rate of 1000 Hz.
Data were collected with 8 mm reusable silvers-silver chloride cup electrodes filled with a 0.05
NaCl paste. For skin conductance, electrodes were attached to the volar surface of the distal
phalanges of the first and second fingers on the non-dominant hand (Dawson, Schell, & Filion,
2007). Heart rate was recorded with use of a Lead II electrode placement (Cacioppo, Tassinary,
& Berntson, 2007). Although HR and skin conductance have primarily been utilized in fear
conditioning, evidence also suggests that these measures can also be indicators of autonomic
conditioning in an appetitive conditioning paradigm. Specifically, when highly appetitive stimuli
are employed, cardiac deceleration and increased skin conductance to the CS+ compared to CS-
are likely to result (Stamps & Porges, 1975).
Evaluative Conditioning
Evaluative conditioning was tested with two measures. First, changes in self-reported CS
valences were measured. Participants were asked to rate CS valence on a scale of -100 to 100
before and after conditioning. Evaluative conditioning was evidenced if participants
demonstrated greater changes in self-reported positive valence for CS+ than CS-. Although self-
report valence assessments are frequently employed in studies of evaluative conditioning, one
CLASSICAL CONDITIONING AND OBESITY 40
potential limitation of this methodology is that it is vulnerable to the effects of demand
characteristics (Dawson, Rissling, Schell, & Wilcox, 2007). Specifically, participants may be
more likely to rate CSs paired with negative UCSs as negative and CSs paired with positive
UCSs as positive because they believe that is what the experimenter wants. To protect against
these effects, the affective priming paradigm (Fazio et al., 1986) can be utilized as an alternative
or complementary methodology. In affective priming, a series of either positive or negative
target stimuli are presented. Participants are instructed to evaluative each target stimuli as
“positive” or “negative” as quickly as possible. Each target stimulus is preceded by a prime
stimulus that is positive, negative, or neutral. Although participants are instructed to ignore the
prime stimulus, evidence indicates that the time to evaluative the target stimulus is dependent on
the valence of the primes. If targets are preceded by a congruently valenced prime (positive
target preceded by positive prime), the response latencies are significantly shorter than on
incongruently valenced primes (positive target preceded by negative prime). Thus, the affective
priming procedure represents a simple nonverbal method for examining changes in CS valence
independent of the effects of demand characteristics.
The affective priming procedure consisted of six practice trials followed by 40 test trials.
Each trial was preceded with a symbol that oriented participants attention to the center of the
screen and provided warning that the prime was about to occur. The primes were presented for
200 ms and were either the CS+ sound or CS- sound. Following the auditory prime stimulus, the
target stimulus was presented 100 ms after the offset of the prime. Thus, the stimulus onset
asynchrony (SOA) was 300 ms, an optimal SOA duration to detect affective priming effects
(Hermans, Spruyt, & Eelen, 2003). The target stayed on the screen for 2000 ms or until the
participant gave a response. To indicate positive ratings of words, participants pressed the letter
CLASSICAL CONDITIONING AND OBESITY 41
“Q” on the keyboard, and to indicate negative ratings, participants pressed “P”. All response
latencies shorter than 200 ms or longer than 1500 ms were excluded to reduce the influence of
outlier responses (Hermans, Vansteenwegen, Crombez, Baeyens, & Eelen, 2002). The intertrial
interval was 2000 ms. Target stimuli were positively valenced adjectives (e.g., handsome,
hopeful, admired) and negatively valenced adjectives (e.g., rude, toxic, violent). Mean reaction
times in ms were calculated based on the combination of prime and target valence, resulting in
four variables: congruent (CS+ with positive target), incongruent (CS+ with negative target),
neutral/positive (CS- with positive target), and neutral/negative (CS- with negative target).
The 40 total affective priming trials consisted of 1) 10 congruent trials (CS+ with positive
target), 2) 10 incongruent trials (CS+ with negative target), 3) 10 neutral trials with positive
targets (CS- with positive target), and 4) 10 neutral trials with negative targets (CS- with
negative target). Each positively and negatively valenced adjective was selected from the
Affective Norms for English Words (Bradley & Lang, 1999), a list of approximately 600 words
normed on dimensions of arousal, valence, dominance, and frequency in a large undergraduate
sample. The trial sequence was created using a random number generator. In addition, trials were
randomized by groups of four using a block randomization method. Evaluative conditioning was
evidenced if reaction times to congruent trials were significantly faster than reaction times to
incongruent or neutral trials.
Bogus Taste Test
A bogus taste test task was employed to measure conditioning based on measurements of
caloric consumption. The taste test was previously developed by Schachter et al. (1968).
Schachter et al. (1968) designed this procedure to measure differences in caloric consumption
CLASSICAL CONDITIONING AND OBESITY 42
after prolonged exposure to food cues. Participants engaging in the taste test task were instructed
to rate different dimensions of crackers (e.g., salty, cheesy, garlicky, etc…) listed on a sheet.
Participants were then instructed to consume as many crackers as they preferred in order to
obtain the most accurate assessment of the food characteristics as possible. Participants were
given two five-minute segments for the taste test. During the taste test, participants’ believed that
they were left alone to complete the task. However, unbeknownst to them, the experimenters
were observing and counting the number of crackers consumed through a one-way mirror. This
methodology was created to compare caloric consumption between obese and normal-weight
participants, but also circumvent the possibility that the eating habits of some participants,
particularly the obese, may be driven by the presence of the experimenter. In the current study,
the procedures of Schachter et al. (1968) were modified to measure differences in caloric
consumption between obese and normal-weight participants after the conditioning procedure.
For the bogus taste-test task in the current study, participants were told that the purpose
of the task was to test their ability to multitask with foods and sounds. Equally-sized snacks were
presented to reduce the impact of portion size on participants’ eating. Participants were first
asked to rate the snacks on a number of food qualities (e.g., crunchy, smooth, sweet) on a 1-10
scale. While tasting the foods and making their ratings, participants were also asked to attend to
sounds that were presented and make assessments of whether the pitches of the sounds changed
during the test. The two sounds that were presented were either the CS+ or CS- sound, and were
presented one after the other in a counterbalanced order. The experimenters informed
participants that the pitch could have changed frequently throughout the procedure, or may not
have changed at all, and that this would be due to the randomization process of the sound files.
However, unbeknownst to the participant, the pitches of the sounds did not change at any point
CLASSICAL CONDITIONING AND OBESITY 43
during the taste-test. The experimenters encouraged participants to consume as many snacks as
they preferred in order to obtain the most accurate ratings. All participants indicated post-
experimentally that they were not aware of the true purpose of the taste test task.
Procedure
Pre-Experimental Phase. Prior to experimentation, study personnel contacted
participants to confirm that they met inclusion criteria for study participation. After confirming
eligibility, participants were randomized into either the sated or food deprived groups.
Participants in the sated group were instructed to eat a full meal the morning of their
appointment, while participants in the food deprived group were instructed to abstain from eating
after 7 PM (or 8 PM for participants in the 10:30 AM timeslot) the night prior to their
appointment. In order to encourage honest reporting of consumption and increase the likelihood
that participants followed the instructions provided to them, all participants were told that the
experimenters possessed equipment to assess for the previous day’s caloric consumption.
Following group assignment, participants were randomized across a number of variables.
First, participants were randomized to one of two trial sequences. The trial sequences, which
determined the order of stimulus presentation during the conditioning task, were created using a
random number generator. A block randomization procedure was employed to create the first
trial sequence. For the second sequence, the order of stimulus presentation from the first trial
sequence was reversed. Two separate trial sequences were created to control for the possible
effects of trial order on conditioning. As a further control for trial sequence effects, the trial
sequences were created with the restraint that the ratio of CS+ trials following CS+ trials was
approximately equal to the ratio of CS+ trials following CS- trials, and vice versa for CS- (Singh,
CLASSICAL CONDITIONING AND OBESITY 44
Dawson, Schell, Courtney, & Payne, 2013). Second, the two sounds (600 Hz vs. white noise)
that served as CS+ were randomized for each participant. Finally, participants were also
randomized with respect to the order of CS presentation during the bogus-taste test.
Experimental Phase. The study was conducted between the hours of 9:00 A.M. and
12:00 P.M in a well-lit, sound-proofed room (subject room). The experiment was conducted
within this timeframe as a means to equalize as best as possible the time since last food
consumption for each participant in their respective satiation groups. Experimenters were
stationed in an adjacent room (experimenter room) to monitor psychophysiological recording
equipment and oversee stimulus presentation. When participants arrived to the laboratory, they
were led to the subject room and seated in front of a computer monitor that later presented the
stimuli for the conditioning experiment. Informed consent was obtained and questionnaires were
given to assess for 24-hour caloric intake, feelings of hunger, and demographic information.
Participants were also asked to pick six favored snacks out of 13 possibilities, and rank them on a
scale of one to six. The 13 snacks were chosen on the basis of 1) chocolate being a primary
ingredient, and 2) the energy density of the snacks (cal/grams) being approximately equal (no
difference > 1.0 cal/grams). The latter requirement was necessary as it allowed for a control of
the possible influences of participants reaching satiation more quickly as a result of highly
energy-dense foods. Without this critical control, external characteristics of the food may have
been more likely to influence caloric intake. In addition, equal amounts of each snack (~ 100 g)
were presented during the taste test to control for the possible effects of portion size on eating
(Levitsky, 2005). Participants also indicated whether any of the snacks were especially more
favored than the rest. Snacks that were endorsed as highly favored were eliminated from the
choices, and participants were asked to pick another snack. This important procedure controlled
CLASSICAL CONDITIONING AND OBESITY 45
for the possibility that participants eating was driven by food preferences as opposed to a
differential CS effect.
After participants completed the brief questionnaires, they provided pre-conditioning
affective ratings for the CSs which were presented auditorily through the headphones.
Participants were then instrumented with five electrodes for HR and SCR measurement.
Following instrumentation, participants were asked to rest during a three-minute baseline period
while experimenters prepared the milk shake. Primary ingredients for the milk shake included: 1
standard scoop of vanilla ice cream, one ounce of Hershey’s chocolate syrup, and two ounces of
whole milk. Following the baseline period, the experimenters provided oral instructions for the
conditioning portion of the experiment (instructions can be viewed in the Appendix).
For the autonomic conditioning procedure, participants were instructed to try to identify
the relationship between the sounds, images, and milkshake. To reduce the occurrence of
excessive signal noise in the autonomic readings, participants were also instructed to hold the
tubing in their mouth using their hands at all times while staying as still as possible. While
participants engaged in the conditioning portion of the experiment, experimenters pre-measured
and recorded the weight of the six snacks previously endorsed by the participant. After preparing
the food, each snack was set aside for later use during the bogus taste test task. Upon completion
of the conditioning procedure, participants were immediately given a post-experimental
questionnaire to assess their awareness of the CS-UCS contingency.
Following the autonomic conditioning phase, participants were again presented with each
CS and provided self-report affective ratings for both sounds. After making their ratings,
participants were led into the experimenter room and seated in front of a computer monitor. They
CLASSICAL CONDITIONING AND OBESITY 46
were then read the instructions for the affective priming procedure (see Appendix). After the
instructions were presented, participants were left alone to complete the task.
After completing the affective priming task, participants were led back into the subject
room and asked to wait. At this point, the experimenter placed three of the six pre-measured
snacks in the area where the affective priming procedure was conducted. Participants were then
brought back into the experimenter room and provided with the instructions for the taste-test task
(see Appendix). If participants did not indicate that they had questions, the initial five-minute
portion of the taste-test task was initiated. Upon conclusion of the five minute segment,
participants were again led into the subject room and asked to wait while the experimenter
prepared the rest of the snacks for the second five-minute portion of the taste test task.
Participants were then led back into the experimenter room and completed the remaining task.
Post-Experimental Phase. Following the experimental procedure, the experimenter
assessed participants’ awareness of the study procedure and purpose. The experimenter also
measured participants’ weight, height, body-mass index (BMI) waist circumference (WC), and
hip circumference (HC). Finally, participants were asked to complete brief self-report
assessments of their eating habits and thanked for their participation.
Data Scoring and Analysis
The study employed a 2 x 2 x 2 mixed factorial design consisting of four groups. The two
between-subjects factors were Satiation Group (food deprived vs. sated) and Weight Group
(obese vs. normal-weight). The within-subjects factor was the CS Condition (CS+ consistently
followed by chocolate/picture UCS vs. CS- never followed by chocolate/picture UCS) for
CLASSICAL CONDITIONING AND OBESITY 47
autonomic conditioning and evaluative conditioning, and was Trial Type (congruent vs.
incongruent) for evaluative conditioning based on affective priming.
Data Scoring and Reduction
To examine UCRs, the skin conductance magnitude and frequency of responses to the
milkshake UCS were analyzed. IBIs to the UCS were not recorded, and thus, not reported in the
current study. Two primary autonomic measures were used for conditioning. For skin
conductance, the largest increase in skin conductance magnitude that began between 1 and 8
seconds following CS onset was measured. This method of skin conductance scoring, has been
demonstrated to be an optimal method of skin conductance scoring compared to the more
traditional convention of separating the skin conductance response into early and late interval
responses during the CS (Pineles, Orr, & Orr, 2009).
Analysis of heart rate is more complex than skin conductance, as heart rate can accelerate
due to defensive responses or decelerate due to orienting (Graham & Clifton, 1966; Hugdahl,
1995). For HR, electrocardiograms were screened and heart beats were identified (R-waves). The
inter-beat-interval (IBI) was scored as the temporal difference between successive R-waves in
the ECG signal. IBIs were used as the dependent variable instead of heart rate because of a
lowered susceptibility to artifact due to differences in baseline values (Stern, Ray, & Quigley,
2001). A window of 2 seconds prestimulus onset to 8 seconds beginning at stimulus onset was
scored. For the poststimulus value, an average IBI value across 5-8 seconds was calculated for
each trial. This 5-8 second period represents the period of time with the largest deceleration in an
8 second CS-UCS interval, and is also the most sensitive indicator of conditioning (Graham,
1979; Hodes, Cook, & Lang, 2007). A difference score between the 2 second mean pre-stimulus
CLASSICAL CONDITIONING AND OBESITY 48
IBI for each trial and the mean poststimulus IBI between 5-8 seconds was computed for each
trial. Differential conditioning was indexed by longer IBIs during the CS+ trials than during CS-
trials.
As previously stated, for affective priming, mean reaction times in ms were calculated for
congruent, incongruent, and neutral trials.
Analyses for Specific Aim 1. 2 (obese vs. normal-weight) x 2 (food deprived vs. sated)
ANOVAs were utilized to examine whether groups differed in their self-reported hunger and
hedonic ratings of the UCS (e.g., milkshake). Significant effects were followed up with t-tests.
To compare UCRs between the groups, two 2 (obese vs. normal-weight) x 2 (food deprived vs.
sated) ANOVAs were conducted separately for the UCR frequency and UCR magnitude
(hypothesis 1a and 1b). ANOVAs were followed up by t-tests to examine the nature of the
effects. Finally, to test whether autonomic responding to the UCS was a predictor of differential
and total caloric intake during the taste test, separate linear multiple regression analyses were
conducted for each group (hypothesis 1c). The predictor variable was the magnitude of the skin
conductance response to the UCS, and the dependent variable was either differential
consumption or total caloric consumption during the taste test.
Analyses for Specific Aim 2. Separate analyses were conducted for the hypotheses
related to conditioning (specific aim 2). In addition, analyses were also conducted separately for
autonomic and evaluative dependent variables. Similar mixed-design ANOVAs and t-tests were
utilized for all dependent variables. For autonomic conditioning and evaluative conditioning
based on the self-report valence ratings of the CSs, 2 (CS+ vs CS-) x 2 (obese vs. normal-weight)
x 2 (food deprived vs. sated) mixed ANOVAs were conducted to examine each hypothesis. For
CLASSICAL CONDITIONING AND OBESITY 49
evaluative conditioning based on affective priming, 2 (congruent vs. incongruent) x 2 (obese vs.
normal-weight) x 2 (food deprived vs. sated) ANOVAs were conducted. Paired-samples t-tests
were conducted separately within each group to examine whether obese and normal-weight
participants demonstrated greater autonomic responding to CS+ than CS-, greater changes in
self-reported valence for CS+ than CS-, and faster response latencies to congruent vs.
incongruent trials (hypotheses 2a and 2b). Independent-samples t-tests were also conducted to
examine whether food deprived participants demonstrated greater differential autonomic
responding to CS+ than CS-, greater changes in self-reported positive valence for CS+ than CS-,
and faster response latencies to congruent vs. incongruent trials compared to sated participants
(hypothesis 3a and 3b). The CS x Satiation State x Weight interaction from the ANOVA and
follow-up t-tests also examined whether obese participants in the sated group demonstrated
greater differential autonomic responding to CS+ than CS-, greater changes in self-reported
positive valence for CS+ than CS-, and faster response latencies to congruent vs. incongruent
trials compared to normal-weight participants in the sated group (hypotheses 4a and 4b). This
difference will not be observed in the deprived normal-weight and obese groups (hypothesis 5a
and 5b).
Analyses for Specific Aim 3. Analyses for the bogus taste test were similar to the
previously stated analysis plan for conditioning. The dependent variable was the difference in
calories consumed after conditioning between the CS+ and CS-. A 2 (CS+ vs. CS-) x 2 (obese
vs. normal-weight) x 2 (food deprived vs. sated) ANOVA was conducted to examine specific
hypotheses. Paired-samples t-tests were conducted separately for each group to examine whether
individual groups demonstrated greater caloric consumption to the CS+ than CS-. Finally,
CLASSICAL CONDITIONING AND OBESITY 50
significant ANOVA effects were followed up with independent-samples t-tests to examine the
nature of significant effects (hypotheses 6 and 7).
To examine whether autonomic conditioning significantly predicted differential eating
during the taste test task, two multiple linear regression analyses were conducted separately
within each group (hypothesis 8). The outcome variable was the difference in calories consumed
during the taste test between CS+ and CS-, while the predictors were the difference in skin
conductance and IBIs between the CS+ and CS-.
Power Analysis. Power analysis was conducted using G power software (Buchner et al.,
2009) to determine the necessary sample size to detect main effects and interaction effects. The
power analysis specified a medium effect size (f = .25), an alpha level of .05, and power of .80.
The results of the power analysis revealed a target sample size of 128 participants to detect the
effects for between-subjects group comparisons. Power analysis was also conducted to determine
the necessary sample size to detect significant within-groups effects. Such analyses were
required to test for conditioning within each group. The power analysis specified a medium
effect size (.5), an alpha level of .05, and a power of .80. The results of the power analysis
revealed a target sample size of 10 participants within each group to detect the effects of within-
subjects comparisons.
CLASSICAL CONDITIONING AND OBESITY 51
Results
Participant Characteristics
A total of 121 undergraduate students participated in the study. Sample characteristics are
presented in Table 1. The sample was ethnically diverse, with 41 (33.9%) Caucasian, 41 (33.9%)
Asian American, 19 (15.7%) Hispanic/Latino, 6 (5%) African American, and 13 (10.7%)
reporting as “other.” On the basis of BMI, 60 (49.6%) of the sample was normal-weight (BMI ≤
25), 14 (11.6%) overweight (BMI ≥ 25 and < 30), and 47 (38.8%) obese (BMI ≥ 30). For the
total sample, the mean BMI was 28.7. Mean BMI for normal-weight participants was 21.61,
27.26 for overweight participants, and 38.9 for obese participants. 58 (47.9%) participants were
randomly assigned to the food deprived group, while 63 (52.1%) participants were assigned to
the sated group. Out of the 60 participants who were normal-weight, 31 (51.7%) were in the food
deprived group, and 29 (48.3%) were in the sated group. For the 14 participants who were
overweight, 5 (35.7%) were in the food deprived group, and 9 (64.3%) were in the sated group.
Out of the 47 participants in the obese group, 22 (46.8%) were in the food deprived group, and
25 (53.2%) were in the sated group. Originally, overweight participants were to be grouped with
normal-weight participants in primary analysis. However, the decision was made to separate the
groups and exclude overweight participants from analysis to maximize the effect of the
difference in BMI between the normal-weight and obese participants. This decision also
accounted for the unequal satiation group assignment among overweight participants.
For the satiation groups (food deprived vs. sated), there was a significant difference in the
amount of time since previous food consumption (t = -25.28, p < .001), as participants in the
food deprived group reported an average of 14.08 hours since last consumption, and participants
CLASSICAL CONDITIONING AND OBESITY 52
in the sated group reported an average time of 1.23 hours since last consumption. There was also
a significant difference in self-reported hunger between the groups (t = -12.41, p < .001),
indicating that participants in the food deprived group reported significantly greater hunger than
participants in the sated group.
Out of the total 121 participants, 106 (87.6%) participants correctly identified the CS-
UCS contingency, and thus were defined as having contingency awareness. 15 (12.4%)
participants did not correctly identify the CS-UCS contingency, and thus were defined as not
having contingency awareness. Only participants who correctly identified the CS-UCS
contingency are included in the reported analyses.
UCR Responding
Sample Distribution of UCRs. The number of UCRs from the total 10 CS+ trials and
the number of UCRs on the first 5 CS+ trials are displayed in Figure 2 separately for the total
sample and the sample restricted to only aware participants. The number of UCRs on the first 5
CS+ trials was separately examined due to the possibility that participants habituated to the
UCSs before the end of the conditioning portion of the experiment. For the total number of
UCRs on the 10 CS+ trials for the total sample, 16 (13.2%) participants had no UCRs, 55
(45.4%) participants had between 1 and 3 UCRs, and 50 (41.3%) participants had 4 or more
UCRs. For the number of UCRs on the first 5 CS+ trials, 45 (37.2%) participants had 0 or 1
UCRs, 26 (21.5%) participants had 2 UCRs, and 50 (41.3%) participants had 3 or more UCRs.
Participants were classified as UCR responders if they showed 2 or more UCRs on the first 5
trials. For the results reporting conditioning, only UCR responders were included in analyses.
Thus after excluding participants who were unaware of the CS-UCS contingency and were not
CLASSICAL CONDITIONING AND OBESITY 53
UCR responders, there was a total sample of 69 (57%) participants. Sample characteristics for
the final sample can be viewed in Table 1.
Group Comparisons in UCRs. To compare skin conductance UCR responding across
the groups, a series of 2 (obese vs. normal-weight) x 2 (food deprived vs. sated) ANOVAs were
conducted for each UCR measure, as can be seen in Figure 3. The two UCR measures included
the skin conductance magnitude and number of responses. For the UCR magnitude measure, the
ANOVA revealed a significant Weight Group effect (F(1, 103) = 5.59, p = .002) but no
significant Satiation Group or Weight Group x Satiation Group effect. The results for the UCR
frequency measure were similar to those of the UCR magnitude measure. For the UCR
frequency measure, the ANOVA revealed a significant Weight Group effect (F(1, 103) = 7.75, p
= .006), but no significant Satiation Group or Weight Group x Satiation Group effect. The
significant Weight Group main effects for the UCR magnitude and frequency measures reflect
the fact that obese participants demonstrated significantly greater mean skin conductance
magnitude responses and showed a significantly greater frequency of responses to the UCS than
did normal-weight participants. In addition, the percentage of participants who responded to the
UCS was significantly greater for obese than normal-weight participants (χ
2
= 4.31, p = .045).
Thus, hypotheses 1a was supported. Interestingly, although obese participants had larger average
and more frequent responses to the UCS than normal-weight participants, they rated the
palatability of the UCS (e.g., milkshake) significantly lower than normal-weight participants
(t(105) = -2.09, p = .039).
When the amount eaten during the experiment was compared between the groups, there
was no difference in the amount of food eaten between obese and normal-weight participants.
Interestingly, obese participants reported lower levels of hunger, albeit the effect was marginal
CLASSICAL CONDITIONING AND OBESITY 54
(t(105) = 1.76, p = .08). Even though obese participants consumed a similar amount of calories
during the experiment as normal-weight participants, they reported lower levels of hunger and
rated the appetizing milkshake UCS as being less palatable.
For comparisons in UCR responding between the satiation groups, there was no
significant difference in skin conductance responding across either measure between the two
groups. Therefore, hypothesis 1b was not supported. In addition, the rating for milkshake
palatability did not significantly differ across participants who were either food deprived or
sated.
To examine whether skin conductance responding to the UCS was a predictor of
differential or total caloric consumption during the taste test, separate multiple linear regressions
were conducted for each group. Results demonstrated that, for all participants, responding to the
UCS did not predict differential or total caloric intake. However, for obese participants as a
whole, the skin conductance response to the UCS significantly predicted total caloric intake
during the taste test (b = 26.59, β = .35, t(30) = 2.02, p = .05) but not differential eating. There
were no effects for obese participants when separated into either the sated or food deprived
groups. In addition, no significant effects were observed for normal-weight participants as a
whole or when divided up into either group. Thus, hypothesis 1c was partially supported.
Differential Conditioning
Autonomic Conditioning. For both autonomic measures, a 2 (CS+ vs. CS-) x 2 (obese
vs. normal-weight) x 2 (food deprived vs. sated) ANOVA was conducted. As seen in Figure 4,
results demonstrated no significant CS, CS x Weight Group, CS x Satiation Group, and CS x
Weight Group x Satiation Group effects for either measure (all p’s > .05). ANOVAs were then
CLASSICAL CONDITIONING AND OBESITY 55
followed up with paired-samples t-tests to examine differential conditioning separately for each
group. Results demonstrated no evidence of differential autonomic conditioning among obese
participants as a whole and obese participants in either the food deprived or sated groups for
either autonomic measure. It should be noted, though, that obese participants in the sated group
showed larger IBIs to CS+ than CS-, albeit the effect was marginal (t(18) = 1.78, p = .093).
Among normal-weight participants, there was significant differential conditioning as measured
by skin conductance (t(30) = 2.238, p = .033). Normal-weight participants in the sated group
demonstrated significant differential conditioning using the skin conductance measure (t(15) =
2.42, p = .029); however, normal-weight participants in the food deprived group did not show a
significant effect across any autonomic measure. Thus, hypothesis 2a was partially supported, as
normal-weight participants demonstrated differential autonomic conditioning, although obese
participants failed to demonstrate the effect.
Paired-samples t-tests were also conducted separately for participants in the food
deprived and sated groups to examine whether participants responded more to CS+ than CS- and
therefore demonstrated autonomic conditioning. Results demonstrated that participants in the
food deprived group did not show significantly greater responding to CS+ than CS- across either
autonomic measure (all p’s > .05). These results were also observed for participants who were in
the sated group. Because the ANOVA demonstrated no significant CS x Satiation Group effect,
there was no significant difference between the groups in conditioning across any autonomic
measure. Therefore, hypothesis 3a was not supported. In addition, because the ANOVA also
demonstrated no significant CS x Weight Group x Satiation Group interaction effect across
either autonomic measure, there was no significant difference between the groups regardless of
CLASSICAL CONDITIONING AND OBESITY 56
their satiation state on conditioning. Thus, hypothesis 4a was not supported, while hypothesis 5a
was.
Evaluative Conditioning. A 2 (CS+ vs. CS-) x 2 (obese vs. normal-weight) x 2 (food
deprived vs. sated) ANOVA was conducted to examine whether groups differed in their self-
report evaluations of the CS valences. As can be seen in Figure 5, the ANOVA demonstrated a
significant CS main effect (F(1, 58) = 24.22, p < .001), but no significant, CS x Weight Group,
CS x Satiation Group, or CS x Weight Group x Satiation Group effects (all p’s > .05). However,
it should be noted that the CS x Weight Group effect was marginally significant (F(1, 58) = 3.81,
p = .056). A series of t-tests were conducted separately for each group to examine the occurrence
of evaluative conditioning. The obese group as a whole rated the CS+ significantly higher after
conditioning than before, an effect not seen for the CS- (t(30) = 3.33, p = .002). When the obese
group was divided into those who were sated and those who were food deprived, the change in
evaluation for the CS+ was seen for the sated group (t(18) = 2.89, p = .01), but not for the
deprived group. The normal-weight group as a whole rated the CS+ significantly higher after
conditioning than before, an effect not seen for the CS- (t(30) = 3.98, p < .001). When the
normal-weight group was divided into the sated and food deprived subgroups, the effect was
seen among both the food deprived group (t(14) = 2.94, p = .011) and sated group (t(15) = 2.94,
p = .01).
For affective priming, one participant was excluded from analyses due to an error in the
programming during data collection. Thus, there was a total sample of 68 participants for this
analysis. For the primary analyses, a 2 (congruent vs. incongruent) x 2 (obese vs. normal-weight)
x 2 (food deprived vs. sated) ANOVA was conducted to examine whether groups differed in
their responses to congruent and incongruent trials for affective priming. As seen in Figure 6,
CLASSICAL CONDITIONING AND OBESITY 57
results from the ANOVA demonstrated no significant Trial Type, Trial Type x Weight Group,
Trial Type x Satiation Group, or Trial Type x Weight Group x Satiation Group effects (all p’s >
.05). Paired-samples t-tests were also conducted to examine the presence of evaluative
conditioning separately for each group. Across all groups, results indicated that there was no
significant difference in the mean response times between congruent and incongruent trials (all
p’s > .05). Although both obese and normal-weight participants did show greater changes in self-
reported CS positive valence for CS+ compared to CS-, there was no significant difference in
mean response latencies to congruent vs. incongruent trials within each group. Therefore, based
on the results from affective priming, hypothesis 2b was not supported. However, based on the
self-reported positive valence changes, hypothesis 2b was supported. In addition, because there
was not a significant Trial Type x Satiation Group effect, there was no significant difference
between the satiation groups in their mean response times to congruent vs. incongruent trials.
Thus, hypothesis 3b was not supported. In addition, because the ANOVA also demonstrated no
significant Trial Type x Weight Group x Satiation Group interaction effect across either
autonomic measure, there was no significant difference between the weight groups regardless of
their satiation state on their response times to affective priming trials. Therefore, hypothesis 4b
was not supported, while hypothesis 5b was.
Behavioral Conditioning. Whether participants’ differed in conditioning based on
differential caloric consumption during the taste test was also tested. A 2 (CS+ vs. CS-) x 2
(obese vs. normal-weight) x 2 (food deprived vs. sated) ANOVA demonstrated no significant
CS, CS x Weight Group, CS x Satiation Group, and CS x Weight Group x Satiation Group
effects (all p’s > .05), as can be seen on Figure 7. Separate t-tests for each group also indicated
no significant conditioning among any of the groups (all p’s > .05). Because the CS x Weight
CLASSICAL CONDITIONING AND OBESITY 58
Group x Satiation Group interaction effect was not significant, it must be concluded that obese
participants in the sated group did not differ in their calories consumed during CS+ than CS-
compared to normal-weight participants in the sated group. Thus, hypothesis 6 was not
supported. In addition, obese participants in the food deprived group also did not differ in
calories consumed during CS+ than CS- compared to normal-weight participants in the food
deprived group. As a result, hypothesis 7 was supported.
Autonomic Conditioning Predicting Differential Eating
Although groups on the whole did not demonstrate significant differential conditioning,
there were participants in all groups who did in fact show greater autonomic responding to CS+
than CS-. It was predicted that among obese participants, greater autonomic conditioning would
significantly predict differential eating during the taste test. It was also predicted that this effect
would not be significant among normal-weight participants. A series of multiple linear
regression analyses were conducted in each group to determine whether autonomic conditioning
predicted differential caloric consumption during the taste-test task. Separate analyses were
conducted for both autonomic measures. Results indicated that for the entire sample of
contingency-aware participants who demonstrated UCRs (n = 69), autonomic conditioning based
on skin conductance predicted differential eating on the taste test (b = 10.578, β = .247, t(68) =
2.09, p = .041). When analyses were conducted separately for participants in the obese group,
autonomic conditioning based on either measure did not significantly predict differential eating.
However, as seen in Figure 8, autonomic conditioning based on skin conductance significantly
predicted differential eating (b = 16.31, β = .60, t(11) = 2.35, p = .041) among obese participants
in the food deprived group. As can be seen in Figure 9, there was no significant prediction
among obese participants in the sated group for either autonomic measure (p’s > .05).
CLASSICAL CONDITIONING AND OBESITY 59
Among normal-weight participants as a whole, autonomic conditioning did not predict
differential eating with either measure. As can be seen in Figures 10 and 11, autonomic
conditioning did not significantly predict differential eating among food deprived or sated
participants in the normal-weight group. However, it should be noted that for normal-weight
participants in the food deprived group, the direction of the association was positive. Therefore,
the correlation coefficients representing the association between autonomic conditioning and
differential eating on the taste test were calculated and compared for obese participants in the
food deprived group and normal-weight participants in the food deprived group. This
comparison was made to test whether the strength of the association was significantly different
for obese compared to normal-weight participants in the food deprived group (Cohen & Cohen,
1983). Results indicated that there was no significant difference in the association between
autonomic conditioning and differential eating on the taste test for obese and normal-weight
participants in the food deprived group (p = .54). Although autonomic conditioning among obese
participants in the food deprived group significantly predicted differential eating during the taste
test task, but not for normal-weight participants, the strength of the association was not
significantly different between the groups. Thus, hypothesis 8 was partially supported.
Further analyses were also conducted to examine whether participants who showed
greater average responding to CS+ than CS- (“conditioners”) differed in their differential and
total eating compared to participants who showed greater average responding to CS- than CS+
(“non-conditioners”). The autonomic measure used to differentiate participants’ responses to
CS+ and CS- was skin conductance. Results demonstrated that conditioners showed greater total
eating compared to non-conditioners (M = 29.13, SD = 13.53 for conditioners; M = 25.88, SD =
13.88 for non-conditioners), however the difference was not significant (t(43) = .795, p = .430).
CLASSICAL CONDITIONING AND OBESITY 60
For differential eating, there was also no significant difference between the groups (M = -.75, SD
= 9.36 for conditioners; M = -2.04, SD = 4.75 for non-conditioners), although the difference was
again in the expected direction (t(43) = .585, p = .562).
CLASSICAL CONDITIONING AND OBESITY 61
Discussion
In this study, comparisons between food deprived and sated obese and normal-weight
participants on conditioned responding to food cue stimuli were conducted. The principle
findings of the current study are: 1) Participants in the obese group had larger average magnitude
and more frequent responses to the UCS than did normal-weight participants. This effect was not
moderated by satiation group status. Autonomic responding to the UCS predicted increased total
caloric consumption during the taste test only for obese participants in the food deprived group;
2) Based on autonomic conditioning, participants in the obese group who were either food
deprived or sated did not demonstrate significant differential conditioning. Normal-weight
participants as a whole and those in the sated group demonstrated significant differential
autonomic conditioning; 3) Participants in the obese and normal-weight groups did not differ in
their self-report evaluations of the CS valences. This effect was not moderated by satiation group
status. All participants except for those in the obese food deprived group demonstrated
evaluative conditioning; 4) Based on affective priming, however, where no groups differed in
their mean response times to congruent vs. incongruent trials, evaluative conditioning was not
supported; and 5) autonomic conditioning predicted differential eating only for obese
participants in the food deprived group; however, the association was also positive for the
normal-weight group (though non-significant). The strength of the association did not
significantly differ for obese and normal-weight participants in the food deprived group. In the
following section, the interpretations of the results for each of the principle findings are
discussed in greater detail.
CLASSICAL CONDITIONING AND OBESITY 62
Group Differences in UCRs
In the current study, obese participants showed larger magnitude and more frequent skin
conductance responses to the UCS than did normal-weight participants. Autonomic responding
to the UCS predicted the total amount of calories consumed during the taste test only for obese
participants who were food deprived. Although the milkshake UCS may not be classified as only
a food cue (as opposed to just food), this finding lends support to the notion that obese
individuals are more likely to be influenced by food and food-related cues compared to normal-
weight individuals (de Luca & Spigelman, 1979; Halford et al., 2007; Halford et al., 2004;
Jansen et al., 2003; Nisbett, 1968; Salvy et al., 2007; Schachter & Gross, 1968). Although studies
have tested differences in food cue responding between obese and normal-weight individuals
across behavioral measures, there have been very few studies that have compared autonomic
responses during food cue exposure between the groups. Consistent with the general findings
from the current study, Epstein et al. (1996) found that obese participants sustained elevations in
saliva after repeatedly tasting food compared to normal-weight participants. However, Esteves,
Arriaga, Carneiro, and Flykt (2010) found that their participants as a whole did not show
elevated skin conductance responses when viewing pictures of food than when viewing pictures
of other emotionally valent stimuli, although they did not compare responses of different weight
groups.
Another key difference between the two studies is that Esteves et al. (2010) measured
participants’ autonomic responses while they were viewing pictures of foods, whereas Epstein et
al. (1996) provided participants with visual, gustatory, and olfactory inputs of actual food during
exposure. Jansen et al. (2003) found that both overweight and normal-weight children showed
increased salivation when exposed to intense food smells. Taken together, these studies suggest
CLASSICAL CONDITIONING AND OBESITY 63
that individuals’ autonomic responses as a whole may be particularly activated by the “inherent”
qualities of food (e.g., aroma, visibility, proximity), as opposed to the “arbitrary” characteristics
(e.g., time on the clock, individuals present during a meal). The results from the current study
also suggest that obese individuals may be particularly activated when exposed to the inherent
qualities of food compared to normal-weight individuals. Importantly, previously-mentioned
studies have also demonstrated that obese individuals are more likely to be influenced to eat by
the arbitrary qualities of food than are normal-weight individuals.
Although Epstein et al. (1996) demonstrated differential autonomic activity during food
exposure between obese and normal-weight individuals, the primary measure used was
salivation. The current study is the only one to have demonstrated that obese participants show
more frequent and elevated skin conductance responses than normal-weight participants, and
thus, also directly implicates influences related to the sympathetic nervous system in responding
to food cues. A number of potential explanations exist that may help explain why groups differed
in their skin conductance responses to the food UCS. First, skin conductance has a long history
of use in psychophysiological research as an indicator of arousal and salience (Fowles, 1986).
Specifically, a highly aroused state, whether induced by positive or negative stimuli, is
associated with increases in skin conductance responding. The greater skin conductance
responding to the appetitive UCS among obese participants may be a result of the UCS
presenting as a more arousing or salient stimulus for obese compared to normal-weight
participants. Another potential explanation may involve the hormone insulin, which is released
into the bloodstream during food consumption to absorb excess glucose. Without insulin, glucose
levels would build up in the bloodstream to levels that are toxic to the human body. In addition
to absorbing glucose, insulin also has sympatho-activating effects on the central nervous system.
CLASSICAL CONDITIONING AND OBESITY 64
Specifically, higher circulating insulin in associated with an elevated sympathetic nervous
system (SNS) response (Rowe et al., 1981). Existing evidence suggests that insulin release
during food consumption is higher among obese and overweight individuals (Troisi et al., 1991).
Thus, the greater skin conductance response to the milkshake UCS may reflect a greater release
of insulin into the bloodstream, and thus, a higher SNS response during consumption among
obese participants.
An additional finding gleaned from the UCR comparison data was that while obese
participants demonstrated greater responding to the UCS, they also reported that the milkshake
UCS was significantly less appetizing compared to the reports of normal-weight participants.
These seemingly contradictory findings may have two possible explanations. First, all
participants had knowledge that the study in some way involved food (although not that it was a
study comparing differences in food cue conditioning between obese and normal-weight
individuals). Obese participants may have been more sensitive to the aspects of the study that
directly involved food, and as a result, may have intentionally underreported the palatability of
the milkshake UCS. A second potential explanation of the results is that appraisals of the UCS in
processes of appetitive conditioning may not be as closely linked to autonomic responding to the
UCS as has been demonstrated in fear conditioning. Hermann, Ziegler, Birbaumer, and Flor
(2002) found that participants showed greater electromyographic responses to a UCS that they
rated as being more aversive (e.g., noxious odor) compared to a less aversive UCS (e.g., odorless
air puff). These findings are in contrast to the results of the current study, where participants who
on averaged rated the pleasantness of the UCS higher demonstrated lower autonomic responding
to the UCS, whereas participants who on average rated the pleasantness of the UCS lower
CLASSICAL CONDITIONING AND OBESITY 65
demonstrated higher autonomic responses to the UCS. These separate findings may reflect the
distinct differences between processes involved in aversive and appetitive conditioning.
Implications. The fact that obese participants demonstrated greater skin conductance
responding to the UCS, and that skin conductance responding to the UCS was a strong predictor
of total caloric intake is a key finding and is an important contribution to the existing literature.
This finding provides a framework for future studies to examine why autonomic activity
specifically linked to the SNS may be associated with an increased risk of eating only for
individuals who are obese. Such a study may implicate the SNS as a major contributor to weight
gain and obesity.
Autonomic and Evaluative Conditioning
In the current study, it was hypothesized that all individuals, regardless of their weight or
satiation status, would demonstrate significant differential conditioning to the food cue stimulus.
In addition, it was also hypothesized that obese individuals would show particularly greater
conditioning across behavioral, autonomic, and evaluative measures. The results demonstrated
that no groups showed differential conditioning based on the behavioral measure (the post-
conditioning bogus taste test task where calories consumption was measured). When autonomic
conditioning was tested separately in each group, normal-weight participants as a whole and in
the sated group showed significant differential conditioning. In addition, participants in all
groups (except obese food deprived) showed evaluative conditioned based on the self-report
measures, but no group demonstrated conditioning based on affective priming.
Only one study has been conducted comparing autonomic responding to food cues
between weight groups. Meyer (2012) recently conducted a study examining whether overweight
CLASSICAL CONDITIONING AND OBESITY 66
participants (n = 25) differed in the acquisition and extinction of conditioned salivary responses
compared to lean participants (n = 20). For the conditioning procedure, one visual cue (CS+) was
paired with an appetizing chocolate milkshake UCS, while another visual cue (CS-) was paired
with tasteless water. The frequency of swallows was utilized as the dependent variable. Results
demonstrated that overweight participants demonstrated a significantly greater frequency of
swallows to the cue associated with the milkshake compared with the cue associated with the
tasteless water. Normal-weight participants did not show differential conditioning. The results
for the extinction tests were in the expected direction, although not significant.
The Meyer (2012) study was procedurally similar to the current study yet gleaned both
similar and contrasting results. For example, both studies found significant differential
conditioning among overweight participants (although the effects were marginally significant in
the current study, p = .093). On the other hand, Meyer (2012) found no evidence of differential
conditioning among normal-weight participants, whereas the current study did. Both studies
shared a number of procedural similarities. For example, a chocolate milkshake UCS was
utilized in a differential conditioning procedure. In addition, differential conditioning between
weight groups was compared. Although the studies were similar in procedures employed, there
were also key differences. For example, the CSs utilized were of a dissimilar nature, with the
Meyer (2012) study using visual CSs, and the current study utilizing auditory CSs. In addition,
Meyer (2012) did not test the acquisition of conditioned responding among participants who
were considered obese, whereas the current study did. The Meyer (2012) study also only utilized
one autonomic measure of conditioning, whereas the current study tested conditioning across
multiple autonomic, behavioral, and evaluative measures. Finally, Meyer (2012) only tested the
acquisition of differential autonomic conditioning, and not whether autonomic conditioning was
CLASSICAL CONDITIONING AND OBESITY 67
predictive of caloric consumption. All in all, the Meyer (2012) study and the current study are
the only experimental investigations comparing differences in differential conditioning among
participants differing in their weight. The contrasting findings between the studies highlight the
critical need for additional studies examining processes of appetitive Pavlovian conditioning, and
how such processes may influence eating behavior.
In the current study, the finding that normal-weight participants showed significant
differential conditioning whereas obese participants did not was surprising and unexpected. A
potential explanation of the results may involve the dependent variables employed in the current
study to test autonomic conditioning. In the Meyer (2012) study, the investigators utilized a
direct measure related to digestion and gastric activity (e.g., the number of saliva swallows
during CS exposure). However, the current study did not include any measures of these
physiological systems as indicators of conditioned responses. For example, the
electrogastrogram, which records the electrical signals reflecting peristalsis movements of the
stomach musculature, can be employed as a measure of differential conditioning. An interesting
study design may involve the incorporation of gastrointestinal and central nervous system
measures when comparing individuals’ conditioned responses.
The results from evaluative conditioning also proved mixed. Based on the self-report
valence change measures, all groups (except for obese participants in the food deprived group)
demonstrated evaluative conditioning. However, based on the affective priming measure, which
is generally considered an experimental methodology that leads to a stronger test of evaluative
conditioning, there was no evidence of conditioning within any group. These results can be
interpreted in a number of ways. First, participants may have shown a significant evaluative
conditioning effect based on the self-report valence measures primarily due to demand
CLASSICAL CONDITIONING AND OBESITY 68
characteristics. A pre- and post-evaluation of the CSs may have created awareness among
participants as to the nature of the experimental task, and thus may have influenced participants’
evaluative ratings. In addition, the use of auditory prime stimuli during affective priming may
also help explain the results. The auditory tones, which were presented for 200 ms during
affective priming, may have differed in their perceived sound compared to when they were
presented continuously for eight seconds over 20 trials during conditioning. Auditory differences
between the sounds employed during affective priming and conditioning may have reduced any
potential conditioning effect.
There were also non-significant differences found in each group for the differential eating
(taste test) task. In fact, aside from normal-weight participants in the food deprived group, all
participants showed greater eating during CS- than CS+ (although the difference between the
amount eaten during the two conditions was non-significant). This finding may be explained by
the fact that the test of differential eating only occurred over one session and was relatively short
in duration. Birch et al. (1989) found differential eating effects in a conditioning paradigm that
occurred over a timespan of five weeks. Participants in their sample participated in a series of ten
pairs of conditioning trials, one trial per day, two pairs of trials per week. Kern et al. (1993)
conducted their conditioning tests for 2 days each week over a period of 6 weeks. Coyle et al.
(2000) had a three-day acquisition phase, followed by a test day. There was also a retention test 6
days after the test day. In the current study, participants were only given two five-minute periods
during which they heard both the CS+ and CS- tones presented continuously. Similar to the
studies above, it would have been interesting to examine whether the effects of conditioning may
have persisted across a longer duration of time from the period of acquisition. Given that
conditioning was only tested over one session for a total period of 10 minutes, it is possible that
CLASSICAL CONDITIONING AND OBESITY 69
the total duration for the test of conditioning may not have been sufficient to detect a significant
differential conditioning effect using a measure based on consumption.
Implications. The results from the current study imply that obese and normal-weight
individuals may not differ in their conditioned responding to food cues. A comprehensive set of
autonomic, evaluative, and behavioral measures were used to test conditioning. Based on this
finding alone, one might conclude that conditioning may not be a mechanism of excessive
eating, given that the groups did not differ in their conditioned responding. However, a recent
study utilizing gastrointestinal activity as an indicator of conditioning and found that overweight
participants did in fact show greater conditioning than normal-weight participants (Meyer, 2012).
Thus, future studies are recommended to employ a multi-measures approach when comparing
conditioning between normal-weight, overweight, and obese individuals, and to directly measure
gastrointestinal activity as an indicator of conditioning. In addition, future studies should
examine whether autonomic conditioning may predict differential eating. This was the first
conditioning study to include such a procedure.
Autonomic Conditioning as a Predictor of Differential Eating
Although only one group showed differential autonomic conditioning, and the results for
evaluative conditioning were mixed, perhaps one of the more interesting findings gleaned from
the current study was that autonomic conditioning was a strong predictor of differential eating,
particularly among obese participants who were food deprived. This effect was non-significant
for normal-weight participants in either group, although the direction of the association was
positive and did not differ significantly from obese participants. Although Jansen et al. (2003)
and Vögele and Florin (1997) didn’t conduct conditioning experiments, both studies found that
CLASSICAL CONDITIONING AND OBESITY 70
autonomic activity during food cue exposure predicted later caloric consumption. Jansen et al.
(2003) demonstrated that there was only an effect among overweight children, and no such
linkage for normal-weight children. The results from the current study further extend the findings
of Jansen et al. (2003) and Vögele and Florin (1997) by showing similar effects in processes of
classical conditioning. In addition, the effects were demonstrated using skin conductance,
whereas the results from the prior studies were based on saliva and HR as autonomic measures.
A particularly interesting set of findings is that, among obese participants, autonomic
responding to the UCS predicted later total caloric consumption for the group as a whole
(hypothesis 1c), and autonomic conditioning predicted differential eating during the taste test
when food deprived (hypothesis 8). There were no such effects for normal-weight participants.
These results suggest that obese individuals may be at a particularly increased risk of eating
through two distinct, but related pathways: 1) mere exposure to food stimuli and 2) through
processes of associative learning. These two factors may interact and produce a compounded
profile of risk for excessively eating among obese individuals, particularly so when food
deprived. An additional possibility is that autonomic conditioning may predict differential caloric
consumption among obese individuals because they demonstrate a particularly larger autonomic
response to the UCS. However, this possibility is less likely give that the magnitude of skin
conductance responding to the UCS was not significantly associated with the strength of
conditioning (although the relationship was positive).
The fact that autonomic conditioning predicted differential eating in groups that did not
show differential conditioning overall highlights the important possibility that within each group,
whether normal-weight or obese, there were participants who showed larger conditioned
responses (strong conditioners) compared to others who showed a smaller difference in
CLASSICAL CONDITIONING AND OBESITY 71
responses to CS+ and CS- (weak conditioners). Thus, it is likely that the degree of
conditionability to food cues does not substantially differ between individuals along the weight
spectrum; however, as previously mentioned, conditionability may exist as a type of trait
characteristic that predisposes individuals to gain weight. An individual possessing a potential
“conditionability” trait may be at especially greater risk of weight gain in an environment
saturated with food cues (Milstein, 1980; Rodin & Slochower, 1976; Rodin et al., 1977). The
finding that greater autonomic conditioning was a strong predictor of differential eating lends
some support to this hypothesis. In addition, the fact that “conditioners” consumed a higher
average number of calories during the experiment than “non-conditioners,” albeit the difference
was not significant, also provides additional support. Although the results of the current study
provide some support for a “conditionability trait hypothesis”, the optimal test of conditioning
would include measuring participants’ (preferably young children) conditioned responses to food
cues, distinguishing between high, low, and non-conditioners, and prospectively following the
groups to compare differences in the amount of weight gained. If high conditioners demonstrated
a greater degree of weight gain compared to low- or non-conditioners, a conditionability trait
hypothesis would receive stronger support.
Within a potential conditionability trait hypothesis framework, individuals who possess
such a trait would conceivably be more likely to gain weight over time, and thus be more likely
to be overweight and obese in early and middle adulthood (assuming a temporally stable trait).
Given that autonomic conditioning was the strongest predictor of differential eating among obese
participants in the food deprived group, it is possible that these participants did in fact possess
such a trait to a greater extent than normal-weight participants. Alternatively, obese and normal-
weight participants may have possessed such a trait to the same degree, but may have predicted
CLASSICAL CONDITIONING AND OBESITY 72
differential eating only for the obese individuals (normal-weight participants may have been able
to better control their impulses to eat). In addition, since these effects were only observed for
participants who were food deprived, it is also possible that a “conditionability” trait may be a
particularly strong risk factor for eating while an organism is in a state of food deprivation.
However, such a conclusion cannot be fully determined by the results of the current study, as
should consider such an experimental design to fully determine whether conditionability may
serve as a mechanism for excessive unhealthy eating, and potentially weight gain.
Implications. The fact that autonomic conditioning significantly predicted differential
eating during the taste test for obese participants in the food deprived group, but not for normal-
weight participants, supports the notion that conditioning may be a mechanism for excessive
eating. However, as previously mentioned, this result must be interpreted with caution given that
the relationship between conditioning and eating was positive for normal-weight participants and
the strength of the association (r
2
) did not significantly differ compared to obese participants.
Nevertheless, these results set an important foundation for future studies to examine whether
individuals who show higher conditioned responses prospectively gain weight at a greater rate
than individuals who do not condition. In addition, because autonomic conditioning was
predictive of differential eating especially for obese individuals who were food deprived, the
results of the current study would lend some support to the use of food cue exposure intervention
strategies.
Limitations and Strengths
The current study had a number of limitations that should be considered when
interpreting the results. First, as previously mentioned, there were no measures of conditioning
CLASSICAL CONDITIONING AND OBESITY 73
based on gastrointestinal activity. Gasotrointestinal activity may have proved a more sensitive
measure of conditioning to food cues rather than measures of the sympathetic nervous system.
Second, the conditioning and test phases were conducted in only one session over 20 minutes,
which may have prevented the establishment of a conditioned response or have been too short in
duration to detect a significant conditioning effect based on a measure of differential
consumption. As previously noted, studies with longer acquisition and testing periods have found
successful differential conditioning of food cues. Finally, weight changes were not measured
prospectively. This is an important limitation, as an optimal test of conditioning as a mechanism
of excessive eating and weight gain would be to compare weight changes among participants
classified as “high conditioners” vs. “low conditioners” and “non-conditioners”.
Despite the limitations of the current investigation, this study has a number of strengths
and makes important contributions to the existing studies of conditioning to food cues. First,
participants’ conditioned responding to the food cues was tested in a large sample and across a
wide range of BMIs, with over half the total sample of participants meeting criteria for
overweight or obesity. The spectrum of BMIs in the obese group was also well-distributed. The
current study also highlights the critical importance of manipulating participants’ satiation state.
Conditioned responses between participants both food deprived and sated were compared, and
results frequently differed for participants in the two satiation groups. This was the first
conditioning study to actively manipulate satiation state as opposed to either satiating or
depriving all participants in the sample.
The current study utilized the most comprehensive assortment of autonomic, evaluative,
and behavioral indicators of conditioning. In addition, multiple measures were used for each
category of conditioning, as there were two measures for autonomic conditioning (e.g. heart rate,
CLASSICAL CONDITIONING AND OBESITY 74
skin conductance) and two measures for evaluative conditioning (self-report valence changes,
affective priming). The study also utilized a strong experimental design. For example, essential
variables were counterbalanced (e.g., trial order, CS presentation, snack presentation),
participants were selected based on strict inclusionary criteria (e.g., free of characteristics that
would impact autonomic responding, chocolate-likers to maximize response to UCS), and
external influences affecting conditioning were anticipated and minimized (e.g., reduced the
impact of food preferences and differences in caloric density on conditioning, presented equally-
sized portions of each snack during the differential eating task). Thus, increased confidence can
be given to the fact that conditioning was tested across a representative sample using a strong
experimental design.
Conclusions
The findings of the current study suggest a number of important implications regarding
associative learning as a mechanism of excessive eating and also point to a number of important
future research directions. First, autonomic conditioning in the obese food deprived group
significantly predicted differential eating during the taste test. Second, obese participants
demonstrated larger magnitude and more frequent skin conductance responses to the UCS, and
skin conductance responding to the food UCS predicted the total amount of calories consumed
during the experiment. The results of the current study raise the possibility that obese individuals
may be at an elevated risk of overeating through the compounded effects of classical
conditioning and through mere exposure to food and its related stimuli. Such effects were only
found with the skin conductance autonomic measure, and thus, implicates the SNS an important
influence on eating behavior. Third, “conditioners” demonstrated greater total eating during the
experiment that “non-conditioners” (although the difference was non-significant), raising the
CLASSICAL CONDITIONING AND OBESITY 75
possibility of a “conditionability trait hypothesis.” It is recommended that future studies examine
whether “high conditioners” show greater prospective total and differential caloric consumption
over time compared to “low conditioners” or “non-conditioners.” Fourth, the effects of
conditioning were generally stronger for participants who were food deprived than those who
were sated. Thus, the findings highlight the necessity of manipulating satiation state when
comparing group differences in conditioning. Fifth, conditioned responding to food cues was
found to differ among individuals across the weight spectrum. Future comparisons of
conditioning to food cues should include a representative sample of individuals across the weight
spectrum of normal-weight, overweight, and obese. Sixth, the results also differed based on the
indicator of conditioning used, with autonomic conditioning and self-report valence changes
being the most sensitive indicators of conditioning. Thus, future studies testing conditioning to
food cues should utilize a comprehensive set of measures, including autonomic, evaluative, and
behavioral indicators of conditioning. For autonomic conditioning, measures of the
gastrointestinal system as well as the central nervous system should be considered. Finally, the
results from the current study provide limited support for the use of food cue exposure among
obese individuals. However, additional research must be conducted to examine whether
associative learning is truly a reliable predictor for long-term changes in weight.
CLASSICAL CONDITIONING AND OBESITY 76
References
Arwas, S., Rolnick, A., & Lubow, R. E. (1989). Conditioned taste aversion in humans using
motion-induced sickness as the US. Behaviour Research and Therapy, 27(3), 295-301.
Baeyens, F., Eelen, P., Van den Bergh, O., & Crombez, G. (1990). Flavor-flavor and color-flavor
conditioning in humans. Learning and Motivation, 21(4), 434-455.
Bernstein, I. L., & Webster, M. M. (1980). Learned taste aversions in humans. Physiology &
Behavior, 25(3), 363-366.
Birch, L. L., McPhee, L., Steinberg, L., & Sullivan, S. (1990). Conditioned flavor preferences in
young children. Physiology & Behavior, 47(3), 501-505.
Birch, L. L., McPhee, L., Sullivan, S., & Johnson, S. (1989). Conditioned meal initiation in
young children. Appetite, 13(2), 105-113.
Bitterman, M., Menzel, R., Fietz, A., & Schäfer, S. (1983). Classical conditioning of proboscis
extension in honeybees (Apis mellifera). Journal of Comparative Psychology, 97(2), 107-
119.
Boggiano, M. M., Dorsey, J. R., Thomas, J. M., & Murdaugh, D. L. (2009). The Pavlovian
power of palatable food: Lessons for weight-loss adherence from a new rodent model of
cue-induced overeating. International Journal of Obesity, 33(6), 693-701.
Booth, D. A. (1977). Satiety and appetite are conditioned reactions. Psychosomatic Medicine,
39(2), 76-81.
Boutelle, K. N., Zucker, N. L., Peterson, C. B., Rydell, S. A., Cafri, G., & Harnack, L. (2011).
Two novel treatments to reduce overeating in overweight children: A randomized
controlled trial. Journal of Consulting and Clinical Psychology, 79(6), 759-771.
CLASSICAL CONDITIONING AND OBESITY 77
Bradley, M. M., & Lang, P. J. (1999). Affective norms for English words (ANEW): Instruction
manual and affective ratings: Technical Report C-1, The Center for Research in
Psychophysiology, University of Florida.
Braet, C., & Crombez, G. (2003). Cognitive interference due to food cues in childhood obesity.
Journal of Clinical Child and Adolescent Psychology, 32(1), 32-39.
Brunstrom, J. M. (2005). Dietary learning in humans: directions for future research. Physiology
& Behavior, 85(1), 57-65.
Brunstrom, J. M. (2007). Associative learning and the control of human dietary behavior.
Appetite, 49(1), 268-271.
Cacioppo, J. T., Tassinary, L. G., & Berntson, G. G. (2007). Handbook of Psychophysiology (3rd
ed.). New York: NY: Cambridge University Press.
Cannon, D. S., Best, M. R., Batson, J. D., & Feldman, M. (1983). Taste familiarity and
apomorphine-induced taste aversions in humans. Behaviour Research and Therapy,
21(6), 669-673.
Capaldi, E. D. (1996). Conditioned food preferences. In E. D. Capaldi (Ed.), Why We Eat What
We Eat (pp. 53-80). Washington, DC: American Psychological Association (APA).
Castellanos, E. H., Charboneau, E., Dietrich, M. S., Park, S., Bradley, B. P., Mogg, K., &
Cowan, R. L. (2009). Obese adults have visual attention bias for food cue images:
Evidence for altered reward system function. International Journal of Obesity, 33(9),
1063-1073.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the
behavioral sciences. Hillsdale, NJ: Erlbaum.
CLASSICAL CONDITIONING AND OBESITY 78
Coyle, S., Arnold, H. M., Goldberg-Arnold, J. S., Rubin, D. C., & Hall, W. (2000). Olfactory
conditioning facilitates diet transition in human infants. Developmental Psychobiology,
37(3), 144-152.
Dawson, M. E., Rissling, A. J., Schell, A. M., & Wilcox, R. (2007). Under what conditions can
human affective conditioning occur without contingency awareness? Test of the
evaluative conditioning paradigm. Emotion, 7(4), 755-766.
Dawson, M. E., Schell, A. M., & Filion, D. L. (2007). The Electrodermal System. In J. T.
Cacioppo, L. G. Tassinary & G. G. Bernston (Eds.), Handbook of Psychophysiology (3rd
ed., pp. 159-181). New York, NY: Cambridge University Press.
de Castro, J. M. (1994). Family and friends produce greater social facilitation of food intake than
other companions. Physiology & Behavior, 56(3), 445-455.
de Castro, J. M., Brewer, E. M., Elmore, D. K., & Orozco, S. (1990). Social facilitation of the
spontaneous meal size of humans occurs regardless of time, place, alcohol or snacks.
Appetite, 15(2), 89-101.
de Castro, J. M., & de Castro, E. S. (1989). Spontaneous meal patterns of humans: Influence of
the presence of other people. The American Journal of Clinical Nutrition, 50(2), 237-247.
De Houwer, J., Hermans, D., & Eelen, P. (1998). Affective and identity priming with
episodically associated stimuli. Cognition & Emotion, 12(2), 145-169.
De Houwer, J., Thomas, S., & Baeyens, F. (2001). Association learning of likes and dislikes: A
review of 25 years of research on human evaluative conditioning. Psychological Bulletin,
127(6), 853-869.
CLASSICAL CONDITIONING AND OBESITY 79
de Luca, R. V., & Spigelman, M. N. (1979). Effects of models on food intake of obese and non-
obese female college students. Canadian Journal of Behavioural Science/Revue
Canadienne Des Sciences du Comportement, 11(2), 124-129.
Drewnowski, A., Kurth, C., Holden-Wiltse, J., & Saari, J. (1992). Food preferences in human
obesity: Carbohydrates versus fats. Appetite, 18(3), 207-221.
Drobes, D. J., Miller, E. J., Hillman, C. H., Bradley, M. M., Cuthbert, B. N., & Lang, P. J.
(2001). Food deprivation and emotional reactions to food cues: Implications for eating
disorders. Biological Psychology, 57(1), 153-177.
Eertmans, A., Baeyens, F., & Van den Bergh, O. (2001). Food likes and their relative importance
in human eating behavior: Review and preliminary suggestions for health promotion.
Health Education Research, 16(4), 443-456.
Elmquist, J. K., Elias, C. F., & Saper, C. B. (1999). Hypothalamic control of body weight.
Neuron, 22, 221-232.
Epstein, L. H., Paluch, R., & Coleman, K. J. (1996). Differences in salivation to repeated food
cues in obese and nonobese women. Psychosomatic Medicine, 58(2), 160-164.
Esteves, F., Arriaga, P., Carneiro, P., & Flykt, A. (2010). Emotional responses (verbal and
psychophysiological) to pictures of food stimuli. Psicologia, 24(2), 89-111.
Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986). On the automatic
activation of attitudes. Journal of Personality and Social Psychology, 50(2), 229-238.
Fedoroff, I., Polivy, J., & Herman, C. (2003). The specificity of restrained versus unrestrained
eaters' responses to food cues: General desire to eat, or craving for the cued food?
Appetite, 41(1), 7-13.
CLASSICAL CONDITIONING AND OBESITY 80
Ferriday, D., & Brunstrom, J. M. (2011). ‘I just can’t help myself’: Effects of food-cue exposure
in overweight and lean individuals. International Journal of Obesity, 35(1), 142-149.
Flegal, K. M., Carroll, M. D., Ogden, C. L., & Curtin, L. R. (2010). Prevalence and trends in
obesity among US adults, 1999-2008. JAMA: The Journal of the American Medical
Association, 303(3), 235-241.
Fowles, D. C. (1986). The eccrine system and electrodermal activity. Psychophysiology:
Systems, Processes, and Applications, 1, 51-96.
Garcia, J., Hankins, W. G., & Rusiniak, K. W. (1974). Behavioral regulation of the milieu
interne in man and rat. Science, 185(4154), 824-831.
Gibson, E. L., & Brunstrom, J. M. (2007). Learned influences on appetite, food choice, and
intake: Evidence in human beings. In T. Kirkham & S. J. Cooper (Eds.), Appetite and
Body Weight: Integrative Systems and the Development of Anti-Obesity Drugs (pp. 271-
300). London: UK: Elsevier Ltd.
Goldman, R., Jaffa, M., & Schachter, S. (1968). Yom Kippur, Air France, dormitory food, and
the eating behavior of obese and normal persons. Journal of Personality and Social
Psychology, 10(2), 117-123.
Graham, F. K. (1979). Distinguishing among orienting, defense, and startle reflexes. In H. D.
Kimmel, van Olst, E.H., Orlebeke, J.F. (Ed.), The Orienting Reflex in Humans (pp. 137-
168). Hillsdale, NJ: Lawrence Erlbaum and Associates.
Graham, F. K., & Clifton, R. K. (1966). Heart-rate change as a component of the orienting
response. Psychological Bulletin, 65(5), 305.
CLASSICAL CONDITIONING AND OBESITY 81
Grinker, J., Hirsch, J., & Smith, D. V. (1972). Taste sensitivity and susceptibility to external
influence in obese and normal weight subjects. Journal of Personality and Social
Psychology, 22(3), 320-325.
Halford, J. C., Boyland, E. J., Hughes, G., Oliveira, L. P., & Dovey, T. M. (2007). Beyond-brand
effect of television (TV) food advertisements/commercials on caloric intake and food
choice of 5–7-year-old children. Appetite, 49(1), 263-267.
Halford, J. C., Gillespie, J., Brown, V., Pontin, E. E., & Dovey, T. M. (2004). Effect of television
advertisements for foods on food consumption in children. Appetite, 42(2), 221-225.
Hammond, R. A., & Levine, R. (2010). The economic impact of obesity in the United States.
Diabetes, Metabolic Syndrome and Obesity, 3, 285-295.
Herman, C. P., & Polivy, J. (2008). External cues in the control of food intake in humans: The
sensory-normative distinction. Physiology & Behavior, 94(5), 722-728.
Hermann, C., Ziegler, S., Birbaumer, N., & Flor, H. (2002). Psychophysiological and subjective
indicators of aversive Pavlovian conditioning in generalized social phobia. Biological
Psychiatry, 52(4), 328-337.
Hermans, D., Spruyt, A., & Eelen, P. (2003). Automatic affective priming of recently acquired
stimulus valence: Priming at SOA 300 but not at SOA 1000. Cognition & Emotion,
17(1), 83-99.
Hermans, D., Vansteenwegen, D., Crombez, G., Baeyens, F., & Eelen, P. (2002). Expectancy-
learning and evaluative learning in human classical conditioning: Affective priming as an
indirect and unobtrusive measure of conditioned stimulus valence. Behaviour Research
and Therapy, 40(3), 217-234.
CLASSICAL CONDITIONING AND OBESITY 82
Hodes, R. L., Cook, E. W., & Lang, P. J. (2007). Individual differences in autonomic response:
Conditioned association or conditioned fear? Psychophysiology, 22(5), 545-560.
Hofmann, W., De Houwer, J., Perugini, M., Baeyens, F., & Crombez, G. (2010). Evaluative
conditioning in humans: A meta-analysis. Psychological Bulletin, 136(3), 390-421.
Holland, P. C. (1980). Influence of visual conditioned stimulus characteristics on the form of
Pavlovian appetitive conditioned responding in rats. Journal of Experimental
Psychology: Animal Behavior Processes, 6(1), 81-97.
Hollands, G. J., Prestwich, A., & Marteau, T. M. (2011). Using aversive images to enhance
healthy food choices and implicit attitudes: An experimental test of evaluative
conditioning. Health Psychology, 30(2), 195-203.
Hossain, P., Kawar, B., & El Nahas, M. (2007). Obesity and diabetes in the developing world--a
growing challenge. The New England Journal of Medicine, 356(3), 213-215.
Hugdahl, K. (1995). Psychophysiology: The Mind-Body Perspective. Cambridge, MA: Harvard
University Press.
Jansen, A., Theunissen, N., Slechten, K., Nederkoorn, C., Boon, B., Mulkens, S., & Roefs, A.
(2003). Overweight children overeat after exposure to food cues. Eating Behaviors, 4(2),
197-209.
Jenkins, H. M., Barrera, F., Ireland, C., & Woodside, B. (1978). Signal-centered action patterns
of dogs in appetitive classical conditioning. Learning and Motivation, 9(3), 272-296.
Johnson, S. L., McPhee, L., & Birch, L. L. (1991). Conditioned preferences: Young children
prefer flavors associated with high dietary fat. Physiology & Behavior, 50(6), 1245-1251.
Kauffman, N. A., Herman, C. P., & Polivy, J. (1995). Hunger-induced finickiness in humans.
Appetite, 24(3), 203-218.
CLASSICAL CONDITIONING AND OBESITY 83
Kern, D. L., McPhee, L., Fisher, J., Johnson, S., & Birch, L. L. (1993). The postingestive
consequences of fat condition preferences for flavors associated with high dietary fat.
Physiology & Behavior, 54(1), 71-76.
LaBar, K. S., Gitelman, D. R., Parrish, T. B., Kim, Y. H., Nobre, A. C., & Mesulam, M. (2001).
Hunger selectively modulates corticolimbic activation to food stimuli in humans.
Behavioral Neuroscience, 115(2), 493-500.
Laessle, R. G., Uhl, H., & Lindel, B. (2001). Parental influences on eating behavior in obese and
nonobese preadolescents. International Journal of Eating Disorders, 30(4), 447-453.
Lebens, H., Roefs, A., Martijn, C., Houben, K., Nederkoorn, C., & Jansen, A. (2011). Making
implicit measures of associations with snack foods more negative through evaluative
conditioning. Eating Behaviors, 12(4), 249-253.
Levey, A. B., & Martin, I. (1975). Classical conditioning of human ‘evaluative’ responses.
Behaviour Research and Therapy, 13(4), 221-226.
Levitsky, D. A. (2005). The non-regulation of food intake in humans: Hope for reversing the
epidemic of obesity. Physiology & Behavior, 86(5), 623-632.
Lovibond, P. F., & Shanks, D. R. (2002). The role of awareness in Pavlovian conditioning:
Empirical evidence and theoretical implications. Journal of Experimental Psychology:
Animal Behavior Processes, 28(1), 3-26.
Martin, I., & Levey, A. B. (1978). Evaluative conditioning. Advances in Behaviour Research and
Therapy, 1(2), 57-101.
Meyer, M. (2012). Pavlovian conditioning of food cues: Do acquisition and extinction of learned
responses differ between overweight and lean individuals? Alliant International
University, San Diego, CA.
CLASSICAL CONDITIONING AND OBESITY 84
Mills, P. J., & Dimsdale, J. E. (1991). Cardiovascular reactivity to psychosocial stressors. A
review of the effects of beta-blockade. Psychosomatics, 32(2), 209-220.
Milstein, R. (1980). Responsiveness in newborn infants of overweight and normal weight
parents. Appetite, 1(1), 65-74.
Nederkoorn, C., Smulders, F. T. Y., & Jansen, A. (2000). Cephalic phase responses, craving and
food intake in normal subjects. Appetite, 35(1), 45-55.
Nijs, I. M. T., Muris, P., Euser, A. S., & Franken, I. H. A. (2010). Differences in attention to
food and food intake between overweight/obese and normal-weight females under
conditions of hunger and satiety. Appetite, 54(2), 243-254.
Nisbett, R. E. (1968). Determinants of food intake in obesity. Science, 159(3820), 1254-1255.
Ogden, C. L., Carroll, M. D., Curtin, L. R., Lamb, M. M., & Flegal, K. M. (2010). Prevalence of
high body mass index in US children and adolescents, 2007-2008. JAMA: The Journal of
the American Medical Association, 303(3), 242-249.
Ogden, C. L., Carroll, M. D., Kit, B. K., & Flegal, K. M. (2012). Prevalence of obesity in the
United States, 2009-2010: US Department of Health and Human Services, Centers for
Disease Control and Prevention, National Center for Health Statistics.
Painter, J. E., Wansink, B., & Hieggelke, J. B. (2002). How visibility and convenience influence
candy consumption. Appetite, 38(3), 237-238.
Pavlov, I. P. (1927). Conditioned reflexes: An investigation of the physiological activity of the
cerebral cortex. London: UK: Oxford University Press.
Petrovich, G. D., Ross, C. A., Gallagher, M., & Holland, P. C. (2007). Learned contextual cue
potentiates eating in rats. Physiology & Behavior, 90(2), 362-367.
CLASSICAL CONDITIONING AND OBESITY 85
Petrovich, G. D., Setlow, B., Holland, P. C., & Gallagher, M. (2002). Amygdalo-hypothalamic
circuit allows learned cues to override satiety and promote eating. The Journal of
Neuroscience, 22(19), 8748-8753.
Pineles, S. L., Orr, M. R., & Orr, S. P. (2009). An alternative scoring method for skin
conductance responding in a differential fear conditioning paradigm with a long-duration
conditioned stimulus. Psychophysiology, 46(5), 984-995.
Placanica, J. L., Faunce, G. J., & Soames Job, R. (2002). The effect of fasting on attentional
biases for food and body shape/weight words in high and low eating disorder inventory
scorers. International Journal of Eating Disorders, 32(1), 79-90.
Poulton, R., & Menzies, R. G. (2002). Fears born and bred: Toward a more inclusive theory of
fear acquisition. Behaviour Research and Therapy, 40(2), 197-208.
Price, J. M., & Grinker, J. (1973). Effects of degree of obesity, food deprivation, and palatability
on eating behavior of humans. Journal of Comparative and Physiological Psychology,
85(2), 265-271.
Price, R. A. (2011). Genes and Human Obesity. In R. Ahima (Ed.), Metabolic Basis of Obesity
(pp. 127-138). New York, NY: Springer.
Rissanen, A., Hakala, P., Lissner, L., Mattlar, C. E., Koskenvuo, M., & Ronnemaa, T. (2002).
Acquired preference especially for dietary fat and obesity: A study of weight-discordant
monozygotic twin pairs. International Journal of Obesity, 26(7), 973-977.
Rodin, J., & Slochower, J. (1976). Externality in the nonobese: Effects of environmental
responsiveness on weight. Journal of Personality and Social Psychology, 33(3), 338-344.
CLASSICAL CONDITIONING AND OBESITY 86
Rodin, J., Slochower, J., & Fleming, B. (1977). Effects of degree of obesity, age of onset, and
weight loss on responsiveness to sensory and external stimuli. Journal of Comparative
and Physiological Psychology, 91(3), 586-597.
Ross, L. (1974). Effects of manipulating salience of food upon consumption by obese and
normal eaters. In S. Schacter, Rodin, J. (Ed.), Obese Humans and Rats (pp. 43-52).
Potomac, MD: Lawrence Earlbaum Associates.
Rowe, J. W., Young, J. B., Minaker, K. L., Stevens, A. L., Pallotta, J., & Landsberg, L. (1981).
Effect of insulin and glucose infusions on sympathetic nervous system activity in normal
man. Diabetes, 30(3), 219-225.
Rozin, P., & Kalat, J. W. (1971). Specific hungers and poison avoidance as adaptive
specializations of learning. Psychological Review, 78(6), 459-486.
Rozin, P., & Millman, L. (1987). Family environment, not heredity, accounts for family
resemblances in food preferences and attitudes: A twin study. Appetite, 8(2), 125-134.
Rozin, P., & Zellner, D. (1985). The role of Pavlovian conditioning in the acquisition of food
likes and dislikes. Annals of the New York Academy of Sciences, 443(1), 189-202.
Salvy, S. J., Romero, N., Paluch, R., & Epstein, L. H. (2007). Peer influence on pre-adolescent
girls’ snack intake: Effects of weight status. Appetite, 49(1), 177-182.
Schachter, S. (1968). Obesity and eating. Science, 161(3843), 751-756.
Schachter, S. (1971). Some extraordinary facts about obese humans and rats. American
Psychologist, 26(2), 129-144.
Schachter, S., Goldman, R., & Gordon, A. (1968). Effects of fear, food deprivation, and obesity
on eating. Journal of Personality and Social Psychology, 10(2), 91-97.
CLASSICAL CONDITIONING AND OBESITY 87
Schachter, S., & Gross, L. P. (1968). Manipulated time and eating behavior. Journal of
Personality and Social Psychology, 10(2), 98-106.
Schachter, S., & Rodin, J. (1974). Obese humans and rats. Washington, DC: Erlbaum Halstead.
Singh, K., Dawson, M. E., Schell, A. M., Courtney, C. G., & Payne, A. F. H. (2013). Can human
autonomic classical conditioning occur without contingency awareness? The critical
importance of the trial sequence. Biological Psychology, 93, 197-205.
Smith, D. E., Marcus, M. D., Lewis, C. E., Fitzgibbon, M., & Schreiner, P. (1998). Prevalence of
binge eating disorder, obesity, and depression in a biracial cohort of young adults. Annals
of Behavioral Medicine, 20(3), 227-232.
Stamps, L. E., & Porges, S. W. (1975). Heart rate conditioning in newborn infants: Relationships
among conditionability, heart rate variability, and sex. Developmental Psychology, 11(4),
424-431.
Stein, C. J., & Colditz, G. A. (2004). The epidemic of obesity. Journal of Clinical Endocrinology
& Metabolism, 89(6), 2522-2525.
Stern, R. M., Ray, W. J., & Quigley, K. S. (2001). Psychophysiological recording: Oxford
University Press.
Stockburger, J., Schmälzle, R., Flaisch, T., Bublatzky, F., & Schupp, H. T. (2009). The impact of
hunger on food cue processing: An event-related brain potential study. NeuroImage,
47(4), 1819-1829.
Straneva, P., Hinderliter, A., Wells, E., Lenahan, H., & Girdler, S. (2000). Smoking, oral
contraceptives, and cardiovascular reactivity to stress. Obstetrics & Gynecology, 95(1),
78-83.
CLASSICAL CONDITIONING AND OBESITY 88
Stroebele, N., & de Castro, J. M. (2006). Influence of physiological and subjective arousal on
food intake in humans. Nutrition, 22(10), 996-1004.
Tetley, A., Brunstrom, J., & Griffiths, P. (2009). Individual differences in food-cue reactivity.
The role of BMI and everyday portion-size selections. Appetite, 52(3), 614-620.
Troisi, R. J., Weiss, S. T., Parker, D. R., Sparrow, D., Young, J. B., & Landsberg, L. (1991).
Relation of obesity and diet to sympathetic nervous system activity. Hypertension, 17(5),
669-677.
Vögele, C., & Florin, I. (1997). Psychophysiological responses to food exposure: An
experimental study in binge eaters. International Journal of Eating Disorders, 21(2),
147-157.
Wansink, B. (2004). Environmetal factors that increase the food intake and consumption volume
of unknowning consumers. Annual Review of Nutrition, 24, 455-479.
Wansink, B., Painter, J. E., & Lee, Y. K. (2006). The office candy dish: Proximity's influence on
estimated and actual consumption. International Journal of Obesity, 30(5), 871-875.
Wasserman, E. A., Franklin, S. R., & Hearst, E. (1974). Pavlovian appetitive contingencies and
approach versus withdrawal to conditioned stimuli in pigeons. Journal of Comparative
and Physiological Psychology, 86(4), 616-627.
Weingarten, H. P. (1983). Conditioned cues elicit feeding in sated rats: A role for learning in
meal initiation. Science, 220(4595), 431-433.
Weingarten, H. P. (1984a). Meal initiation controlled by learned cues: Basic behavioral
properties. Appetite, 5(2), 147-158.
Weingarten, H. P. (1984b). Meal initiation controlled by learned cues: Effects of peripheral
cholinergic blockade and cholecystokinin. Physiology & Behavior, 32(3), 403-408.
CLASSICAL CONDITIONING AND OBESITY 89
Zellner, D. A., Rozin, P., Aron, M., & Kulish, C. (1983). Conditioned enhancement of human's
liking for flavor by pairing with sweetness. Learning and Motivation, 14(3), 338-350.
CLASSICAL CONDITIONING AND OBESITY 90
Table 1
Sample Characteristics
Total Sample (N = 121) Final Sample (N = 69)
Demographics N (%) N (%)
Gender --------------- ---------------
Female 95 (78.5%) 52 (75.4%)
Male 26 (21.5%) 17 (24.6%)
Age (yrs.) 20.12 ± 1.84 20.14 ± 2.04
Ethnicity --------------- ---------------
Caucasian 41 (33.9%) 28 (40.6%)
Asian American 41 (33.9%) 19 (27.5%)
Hispanic/Latino 19 (15.7%) 14 (20.3)
African American 6 (5%) 1 (1.4%)
Other
Satiation Group
Food deprived
Sated
Weight Group
Normal-weight
Food deprived
Sated
Overweight
Food deprived
Sated
Obese
Food deprived
Sated
13 (10.7%)
N (%)
58 (47.9%)
63 (52.1%)
N (%)
60 (49.6%)
31 (51.7%)
29 (48.3)
14 (11.6%)
5 (35.7%)
9 (64.3%)
47 (38.8%)
22 (46.8%)
25 (53.2%)
6 (8.7%)
---------------
30 (43.5%)
39 (56.5)
---------------
31 (44.9%)
15 (48.4%)
16 (51.6%)
7 (10.1%)
3 (42.9%)
4 (57.1%)
31 (44.9%)
12 (38.7%)
19 (61.3%)
Contingency Awareness N (%) ---------------
Contingency Aware 106 (87.6%) 69 (100%)
Contingency Unaware 15 (12.4%) 0
CLASSICAL CONDITIONING AND OBESITY 91
Figure 1. Hypothesized Results for Specific Aims 2 and 3.
Obese Normal-weight
Autonomic,
Evaluative, and
Behavioral
Conditioning
Hypothesized Results
Food Deprived
Sated
CLASSICAL CONDITIONING AND OBESITY 92
Figure 2. UCRs for 10 Total CS+ Trials and First 5 CS+ Trials for Total Sample and Aware
Sample.
CLASSICAL CONDITIONING AND OBESITY 93
Figure 3. Average UCR Magnitudes and Frequencies for Each Group.
CLASSICAL CONDITIONING AND OBESITY 94
Figure 4. Mean SCR and IBI responses to CS+ and CS- for Each Group.
Note. The values on the vertical axis are reversed for display purposes, so a deceleration appears as a downward
deflection (positive values indicate deceleration).
Mean HR (5-8 s) – Poststimulus – Prestimulus
IBI Change Scores
CLASSICAL CONDITIONING AND OBESITY 95
Figure 5. Self-report Valence Changes from Postconditioning – Preconditioning for CS+ and
CS-.
Note. Values above the horizontal axis indicate positive valence changes for the CSs after conditioning than before,
whereas values below the horizontal axis indicate negative valence changes for the CSs after conditioning than
before.
CLASSICAL CONDITIONING AND OBESITY 96
Figure 6. Mean Response Times to Congruent and Incongruent Trials for Each Group.
Note. Higher values on the vertical axis indicate longer response times. Larger values for incongruent trials than
congruent trials would be consistent with an evaluative conditioning effect.
CLASSICAL CONDITIONING AND OBESITY 97
Figure 7. Mean Calories Consumed for CS+ and CS- for Each Group.
CLASSICAL CONDITIONING AND OBESITY 98
Figure 8. Autonomic Conditioning Predicting Differential Eating for Obese/Food Deprived.
CLASSICAL CONDITIONING AND OBESITY 99
Figure 9. Autonomic Conditioning Predicting Differential Eating for Obese/Sated.
CLASSICAL CONDITIONING AND OBESITY 100
Figure 10. Autonomic Conditioning Predicting Differential Eating for Normal-weight/Food
Deprived.
CLASSICAL CONDITIONING AND OBESITY 101
Figure 11. Autonomic Conditioning Predicting Differential Eating for Normal-weight/Sated.
CLASSICAL CONDITIONING AND OBESITY 102
Appendix
Figure 12. Experimenter Instructions.
Conditioning Instructions
Now the first part of the experiment begins. During this phase, you will hear the two sounds that
you previously heard while you were making your ratings. Just to remind you, the two sounds
were a continuous low pitch sound and a fuzzy white noise that sounds like static. You will also
get a combination of a chocolate snack picture on the screen in front of you combined with a
tasty chocolate milkshake flavor delivered through the tubing. There is a relationship between
the kinds of sounds that you hear and whether or not you get the picture and milkshake
combination. Your job is to figure out what the relationship is. It is important to know that for
most people, if they pay attention very closely they can figure out the relationship. However, you
will most likely only be able to figure it out if you pay very close attention.
Affective Priming Instructions
Now this is the computerized task I was telling you about earlier. For this task, you will hear the
sounds (the continuous sound and the white noise static sound) that you previously heard in the
other room through these headphones. When you begin, you will first have the opportunity to
read the instructions. Please read these instructions very carefully as it can be confusing. Next,
six practice trials will be presented to make sure that you know what you are doing. Then, you
will see the instructions presented again. Finally, the test phase will begin. If at any point you
don’t know what you are doing, please call out for me and I can come help.
Taste-Test Task Instructions
For this task, we are testing your ability to multi-task with sounds and food. We will ask that you
pay attention to two things during this task. First, you will hear the sounds that you just heard in
the other room. We ask that you make a tick mark whenever you hear changes in the pitch of the
sound. It is important to know that for some participants, the pitch may change frequently and for
others the pitch may not change at all. This is determined by a randomized process in which the
computer selects a random file to play for you. At the same time, we want you to taste each of
three foods that you chose previously and rate each snack on a 1-10 scale on the list of food
characteristics below. What’s most important to us is the accuracy of the ratings that you are
making. Since we want you to make the most accurate ratings possible, feel free to eat as much
as you want in order to obtain the most accurate ratings. Finally, if you have time you can also
indicate which snacks are most characteristic of each of the food characteristics below.
CLASSICAL CONDITIONING AND OBESITY 103
Figure 13. Telephone Screening Questionnaire.
Participant ID Number: ___________
1. If you are currently taking any medications, please provide the names of each below.
______________________________________________________________________________
______________________________________________________________________________
______________________________________________________________________________
2. Are you currently dieting?
____ Yes ____ No
3. Have you ever suffered, or are currently suffering from any chronic conditions? (e.g.,
heart failure, atherosclerosis, diabetes, cancer, chronic pain, endocrine disorders, respiratory
disorders)
______________________________________________________________________________
______________________________________________________________________________
______________________________________________________________________________
4. Do you currently smoke cigarettes or use any other products that contain nicotine (e.g.,
cigars, chewing tobacco, nicotine patches)?
____ Yes ____ No
5. On a scale of 1-10, how much do you like chocolate?
____________
6. Approximately when do you usually eat breakfast?
_____________
7. Are you allergic to dairy, gluten, or nuts?
____ Yes ____ No
8. Are you 18 or over?
____ Yes ____ No
9. Are you a native English speaker?
____ Yes ____ No
CLASSICAL CONDITIONING AND OBESITY 104
Figure 14. Demographic and Health History Form.
PARTICIPANT #: ________________
GROUP: ___________________
Demographic and Health History Form
Please answer the following questions to the best of your ability.
Demographics
Male ________ Female ________
Date of Birth _________________
Age ______________
Right Handed ________ Left Handed ________
Ethnicity: _________________
Food Consumption
Have you consumed any food or drink in the past 24 hours?
_____ Yes No ______
If yes, please describe:
______________________________________________________________________________
______________________________________________________________________________
______________________________________________________________________________
Approximately how many hours has it been since you last ate a meal?
___________ hrs
CLASSICAL CONDITIONING AND OBESITY 105
On a scale from 1-10, how hungry do you currently feel?
1 2 3 4 5 6 7 8 9 10
Very full Satisfied Extremely Hungry
On a scale from 1-10, how much do you like chocolate?
1 2 3 4 5 6 7 8 9 10
Not at all It’s ok I love it!
Alcohol Consumption
Have you consumed any alcohol in the last 12 hours? ______________
Medication
Are you currently taking any medications? ________
Please specify ______________ Last taken _______________ Dosage ______________
Sleep
How many hours of sleep did you get last night? _______________
Caffeine
Have you had any caffeine in the last 12 hours? Yes ________ No ________
If yes, please specify (last taken, amount): _____________________________________
Hearing
Do you have any problems with your hearing? Yes ________ No ________
CLASSICAL CONDITIONING AND OBESITY 106
Figure 15. Favorite Foods Form.
Please circle 6 of your favorite foods from the list below, and rank which are your favorite (1 =
most favorite, 6 = least favorite).
Foods Rank
3 Musketeers
Brownies
Chocolate Muffin
Crunch
Famous Amos
Hershey Chocolate
M&Ms
Milky Way
Oreos
Milk-Chocolate Covered Pretzels
Teddy Grahams
Chocolate-covered Raisins
Do you like any of foods above much more than the rest? If so, please indicate:
_______________________________________________________________________
_______________________________________________________________________
_______________________________________________________________________
_______________________________________________________________________
CLASSICAL CONDITIONING AND OBESITY 107
Figure 16. Preconditioning and Postconditioning Self-Report Valence Rating Form.
CLASSICAL CONDITIONING AND OBESITY 108
Figure 17. Food Weight Form.
M&Ms (phase___, CS___): Chocolate Muffin (phase___, CS___):
Pre taste-test grams: __________________ Pre taste-test grams: __________________
Post taste-test grams: _________________ Post taste-test grams: _________________
Change: ______________________ Change: ______________________
Oreos (phase___, CS___): Hershey Chocolate (phase___, CS___):
Pre taste-test grams: __________________ Pre taste-test grams: __________________
Post taste-test grams: _________________ Post taste-test grams: _________________
Change: ______________________ Change: ______________________
Crunch (phase___, CS___): Chocolate Covered Pretzels (phase___, CS___):
Pre taste-test grams: __________________ Pre taste-test grams: __________________
Post taste-test grams: _________________ Post taste-test grams: _________________
Change: ______________________ Change: ______________________
Milky Way (phase___, CS___): Chocolate Covered Raisins (phase___, CS___):
Pre taste-test grams: __________________ Pre taste-test grams: __________________
Post taste-test grams: _________________ Post taste-test grams: _________________
Change: ______________________ Change: ______________________
3 Musketeers (phase___, CS___): Brownies (phase___, CS___):
Pre taste-test grams: __________________ Pre taste-test grams: __________________
Post taste-test grams: _________________ Post taste-test grams: _________________
Change: ______________________ Change: ______________________
Famous Amos (phase___, CS___): Teddy Grahams (phase___, CS___):
Pre taste-test grams: __________________ Pre taste-test grams: __________________
Post taste-test grams: _________________ Post taste-test grams: _________________
Change: ______________________ Change: ______________________
Total Calories Consumed: CS+ ______g
Total Calories Consumed: CS-_______g
CLASSICAL CONDITIONING AND OBESITY 109
Figure 18. Post-Experimental Questionnaire (PEQ).
Post-Experimental Questionnaire
1. The milkshake/picture combination was predicted by:
a) The fuzzy white-noise that sounded like static
b) The continuous low-pitch tone
c) It was not systematic
d) I could not tell
2. How certain are you?
a) Completely certain
b) Fairly certain
c) Fairly uncertain
e) Completely uncertain
3. Please rate how you felt about the chocolate milkshake on a scale of 0 to 6:
0 1 2 3 4 5 6
Very bad Neutral (its ok) Very good
CLASSICAL CONDITIONING AND OBESITY 110
Figure 19. Taste-Test Instruction Form.
Multitasking Test
For this task, we are testing your ability to multi-task with sounds and food. We will ask that you
pay attention to two things during this task. First, you will hear the sounds that you just heard in
the other room. We ask that you make a tick mark whenever you hear changes in the pitch of the
sound. It is important to know that for some participants, the pitch may change frequently and for
others the pitch may not change at all. This is determined by a randomized process in which the
computer selects a random file to play for you. At the same time, we want you to taste each of
three foods that you chose previously and rate each snack on a 1-10 scale on the list of food
characteristics below. What’s most important to us is the accuracy of the ratings that you are
making. Since we want you to make the most accurate ratings possible, feel free to eat as much
as you want in order to obtain the most accurate ratings. Finally, if you have time you can also
indicate which snacks are most characteristic of each of the food characteristics below.
TASK 1:
Tick mark:______________________________________________________
Total:_____________
Task 2:
Please rate each snack on a 1-10 scale for each of the following dimensions:
Crunchy Sweet Smooth Soft Hard Creamy Dry Filling
Snack 1:
( )
Snack 2:
( )
Snack 3:
( )
TASK 2:
Which snack is the most:
Crunchy: ______________________ Hard: _____________________________
Sweet: ________________________ Creamy: ___________________________
Smooth: _______________________ Dry: ______________________________
Soft: __________________________ Filling: ____________________________
CLASSICAL CONDITIONING AND OBESITY 111
Figure 20. Physical Characteristics Form.
Physical Characteristics
Height: _______in
Weight: _______lbs
Waist Circumference: _______cm
Waist Circumference (x2): _______cm
Waist Circumference (avg): _______cm
Hip Circumference: _______cm
Hip Circumference (x2): _______cm
Hip Circumference (avg): _______cm
Waist-to-Hip Ratio: _______
BMI: _______
NOTES ABOUT PROCEDURE (e.g., artifacts, on trials, deviations from procedure, general
observations):
______________________________________________________________________________
______________________________________________________________________________
______________________________________________________________________________
______________________________________________________________________________
______________________________________________________________________________
______________________________________________________________________________
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
The acute impact of glucose and sucralose on food decisions and brain responses to visual food cues
PDF
Food choice dimensions and the relationship with BMI
PDF
Self-reported and physiological responses to hate speech and criticisms of systemic social inequality: an investigation of response patterns and their mediation…
Asset Metadata
Creator
Singh, Kulwinder
(author)
Core Title
Classically conditioned responses to food cues among obese and normal weight individuals: conditioning as an explanatory mechanism for excessive eating
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Psychology
Publication Date
07/10/2014
Defense Date
05/13/2014
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
appetitive condtioning,classical conditioning,differential autonomic conditioning,food cues,OAI-PMH Harvest,obesity
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Dawson, Michael Edward (
committee chair
), Bob Knight (
committee member
), Brekke, John S. (
committee member
), Monterosso, John R. (
committee member
), Schell, Anne M. (
committee member
), Shen, Biing-Jiun (
committee member
)
Creator Email
kulwinds@usc.edu,kulwindss@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-434361
Unique identifier
UC11287763
Identifier
etd-SinghKulwi-2626.pdf (filename),usctheses-c3-434361 (legacy record id)
Legacy Identifier
etd-SinghKulwi-2626.pdf
Dmrecord
434361
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Singh, Kulwinder
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
Tags
appetitive condtioning
classical conditioning
differential autonomic conditioning
food cues
obesity