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
/
When and how infants discriminate between declaratives and interrogatives
(USC Thesis Other)
When and how infants discriminate between declaratives and interrogatives
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
WHEN AND HOW INFANTS DISCRIMINATE BETWEEN DECLARATIVES AND
INTERROGATIVES
by
Susan Geffen
________________________________________________________________________
A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
[PSYCHOLOGY]
August 2014
Copyright 2014 Susan Geffen
ii
Table of Contents
List of Tables ....................................................................................................................................... iv
List of Figures ...................................................................................................................................... vi
Abstract .............................................................................................................................................. viii
Acknowledgments................................................................................................................................. x
Introduction ......................................................................................................................................... 12
Experiment 1: Infants’ Ability to Distinguish between Declaratives and Interrogatives ................... 22
Methods ..................................................................................................................................... 22
Results ........................................................................................................................................ 27
Discussion ................................................................................................................................... 40
Experiment 2: Corpus Analysis of Prosodic Characteristics of Infant-Directed Speech .................... 44
Methods ..................................................................................................................................... 44
Results ........................................................................................................................................ 50
Discussion ................................................................................................................................... 64
Experiment 3: Adults Can Distinguish between Resynthesized Declaratives and Interrogatives ...... 67
Methods ..................................................................................................................................... 69
Results ........................................................................................................................................ 70
Discussion ................................................................................................................................... 73
Experiment 4: Infants’ Ability to Distinguish Resynthesized Declaratives and Interrogatives ......... 77
Methods ..................................................................................................................................... 78
Results ........................................................................................................................................ 79
Discussion ................................................................................................................................... 82
iii
Experiment 5: Infants’ Ability to Distinguish Declaratives and Polar Interrogatives Based on Word
Order Information Alone .............................................................................................................. 85
Experiment 5a ..................................................................................................................................... 86
Methods ..................................................................................................................................... 86
Results ........................................................................................................................................ 92
Discussion ................................................................................................................................... 94
Experiment 5b ..................................................................................................................................... 99
Methods ................................................................................................................................... 100
Results ...................................................................................................................................... 101
Discussion ................................................................................................................................. 105
General Discussion ........................................................................................................................... 107
References ......................................................................................................................................... 119
iv
List of Tables
Table 1. Familiarization and Test Sentences for Both Groups .......................................................... 24
Table 2. Beta Coefficient and Standard Error by Factor ................................................................... 30
Table 3. Beta Coefficient and Standard Error by Factor ................................................................... 34
Table 4. Beta Coefficient and Standard Error by Factor ................................................................... 37
Table 5. Frequency of Utterance Types .............................................................................................. 46
Table 6. Mean and Standard Deviation by Position in Utterance ...................................................... 55
Table 7. Mean Difference Scores (Standard Error) and Pitch Excursion (Standard Error) ............. 56
Table 8. Mean and Standard Deviation by Position in Utterance ...................................................... 58
Table 9. Mean Difference Scores (Standard Error) and Pitch Excursion (Standard Error) ............. 59
Table 10. Loadings of Individual Factors for Principal Components Analysis. ................................ 63
Table 11. Odds Ratio (OR) and 95% Confidence Intervals (CI) for Sentence Type Identification
Using Logistic Regression across Models .................................................................................... 73
Table 12. Chi Square, Degrees of Freedom and p-Values. ................................................................ 73
Table 13. Beta Coefficient and Standard Error by Factor .................................................................. 80
Table 14. Duration Difference Scores Averaged Over Sentence Type and Familiarization Type,
Calculated at the Word Level and Syllable Level. Difference Scores Were Computed by
Subtracting the Duration of the Second Position from the Duration of the First Position .......... 90
Table 15. T Statistics and Associated p Values for Comparisons of Duration Difference Scores
between Familiarization and Test Sentence. Comparisons Are Reported for Analyses at the Word
and Syllable Level ......................................................................................................................... 91
Table 16. Familiarization and Test Sentences for Both Groups ....................................................... 101
v
Table 17. Beta Coefficient and Standard Error by Factor ............................................................... 104
vi
List of Figures
Figure 1. Mean looking time duration for novel and familiar trials as a function of familiarization
group for 7-month-olds in Experiment 1. Error bars represent +1 standard error of the mean. .. 31
Figure 2. Mean difference score (Interrogative-Declarative Trial) regardless of familiarization type
for 7-month-olds in Experiment 1. Positive score indicates an interrogative preference, negative
score indicates a declarative preference. Error bars represent +1 standard error of the mean. .... 31
Figure 3. Mean looking time duration for novel and familiar trials as a function of familiarization
group for 9-month-olds in Experiment 1. Error bars represent +1 standard error of the mean. .. 34
Figure 4. Mean difference score (Interrogative-Declarative Trial) regardless of familiarization type
for 9-month-olds in Experiment 1. Positive score indicates an interrogative preference, negative
score indicates a declarative preference. Error bars represent +1 standard error of the mean. ... 35
Figure 5. Mean looking time duration for novel and familiar trials as a function of familiarization
group for combined age groups in Experiment 1. Error bars represent +1 standard error of the
mean. ............................................................................................................................................. 37
Figure 6. Mean difference score (Interrogative-Declarative Trial) regardless of familiarization type
for 7-month-olds in Experiment 1. Positive score indicates an interrogative preference, negative
score indicates a declarative preference. Error bars represent +1 standard error of the mean. .... 38
Figure 7. Scree plot of the principal components of Sentence-Type categorization. Shows the
variance of the first 10 components and supports a four component analysis.............................. 62
Figure 8. Mean looking time duration for novel and familiar trials as a function of familiarization
group for 7-month-olds in Experiment 4. Error bars represent +1 standard error of the mean. .. 81
vii
Figure 9. Mean difference score (Interrogative-Declarative Trial) regardless of familiarization type
for 7-month-olds in Experiment 4. Positive score indicates an interrogative preference, negative
score indicates a declarative preference. Error bars represent +1 standard error of the mean. ... 82
Figure 10. Boxplot depicting difference scores: Median listening times to novel stimuli subtracting
out median listening times to familiar stimuli calculated by subject, including both
familiarization groups. .................................................................................................................. 93
Figure 11. Mean looking time duration for novel and familiar trials as a function of familiarization
group for 7-month-olds in Experiment 5b. Error bars represent +1 standard error of the mean.
..................................................................................................................................................... 104
Figure 12. Mean difference score (Interrogative-Declarative Trial) regardless of familiarization type
for 7-month-olds in Experiment 5b. Positive score indicates an interrogative preference, negative
score indicates a declarative preference. Error bars represent +1 standard error of the mean. .. 105
viii
Abstract
To determine when and how infants begin to distinguish between declaratives and
interrogatives, five experiments were carried out with typically developing American English-
learning 7- to 12-month-olds. In Experiment 1, although there were no significant results for
discrimination for either age group, there was a trend towards discrimination between
declaratives and polar interrogatives. Additionally, the combined age group demonstrated a
significant interrogative preference, suggesting infants can distinguish between sentence types.
Experiments 2-5 evaluated the contribution of prosodic and word-order information in infants’
ability to distinguish between sentence types. Given infants’ early sensitivity to prosodic
information, I hypothesized that infants could use prosody to make initial sentence-type
distinctions. Accordingly, Experiment 2 analyzed prosodic measures across penultimate and
final syllables in infant-directed wh-questions, polar interrogatives, and declaratives. The results
showed that declaratives and polar interrogatives differed on several dimensions, but declaratives
and wh-questions did not, similar to adult-directed speech. Thus, while prosody is not likely to
aid discrimination of declaratives from wh-questions, infant-directed speech provides a learner
with sufficient prosodic information to distinguish declaratives and polar interrogatives.
Experiment 3 tested adults’ ability to make sentence-type distinctions in the absence of
lexical information, before testing infants. The results showed that adults were able to
distinguish between declaratives and polar interrogatives when only prosody was available and
suggests that the limited information provided in the resynthesized stimuli is sufficient for
distinguishing between sentence types. This suggests infants should be able to do the same,
although age of initial distinction remains to be determined.
ix
Experiment 4 evaluated 7-month-olds’ ability to make sentence-type distinctions in the
absence of lexical information. The results showed that despite the availability of prosodic
information, infants did not use prosody alone to distinguish between sentence types. This
suggests that infants may require other types of information to distinguish between sentence
types. Given the syntactic differences between declaratives and polar interrogatives (e.g. AUX-
inversion), word order is a plausible alternative.
Experiment 5 assessed word-order cues by presenting 7- and 12-month-olds with
sentences that had flattened intonation and different word order. Experiment 5a tested 12-
month-olds’ ability to distinguish between sentence types when only word-order information was
available. The results showed that 12-month-olds discriminated sentence types from word order
alone. These results suggest that word order may be sufficient for distinguishing between
declaratives and polar interrogatives.
Experiment 5b tested 7-month-olds’ ability to distinguish between sentence types based
on word order information alone. Infants did not show a reliable looking time difference,
suggesting that they did not distinguish between declaratives and polar interrogatives based on
word order alone. This dissertation provides evidence of infants’ ability to distinguish between
declaratives and polar interrogatives, manifesting as an interrogative preference. Seven-month-
olds did not distinguish between sentence types based on prosodic or word order information
alone, but 12-month-olds did make distinctions based on word order alone. One limitation was
the exclusion of wh- questions from the infant experiments. Future studies should evaluate the
generalizability of these findings to wh- questions, as well as specific distributional features
infants are attending to.
x
Acknowledgments
This dissertation would not been possible without the help of numerous people.
First, I would like to thank my adviser, Toby Mintz, who has provided endless feedback to create
a much more polished final product than I could have ever managed on my own. I would like to
thank the members of my qualifying and dissertation committees for providing guidance and
insight, as well as asking the challenging questions that helped make this dissertation possible:
JoAnn Farver, for helping me work on my writing style, to convey my work in a clear and
concise package; Justin Wood, who constantly encouraged me to look at the bigger picture; and
Megha Sundara and Khalil Iskarous for helping me to better understand the linguistic
implications and applications of my work.
I would also like to thank several other faculty members who have been instrumental in
my graduate school career: Frank Manis, for being a wonderful teaching role model and mentor,
helping to shape me into a more well-rounded researcher, instructor and member of the
psychology community; and Elsi Kaiser, whose classes and lab meetings helped me to better
understand the intersection between psychology and linguistics.
I would like to thank my lab mates, Vivian Jia Li and Felix Hao Wang, for listening to
multiple variations on the same topic, and continuing to ask new questions. I would also like to
thank all of the RA’s I’ve worked with for all of their invaluable help in recording stimuli,
running experiments, analyzing data, reading drafts, etc.: Conor Frye, Mor Ghermezi, Heidi
Mettler, Amanda Lutz, Caroline Bondurant, Josh Greenberger, Julia Marcheque, Stephanie
Seeman, Megan Goldring, and Ryan Emhoff.
I would like to thank Iris Chuoying Ouyang for helping me use PSOLA scripts to modify
my stimuli, saving me from having to manually modify numerous sentences. I would like to
xi
thank David Li for introducing me to ProsodyPro, which helped me to more efficiently collect
the acoustic data I needed. I would also like to thank Hao Wang for helping me figure out how
to navigate the CHILDES database more proficiently.
I’m indebted to Kevin Petway and Nick Jackson for their invaluable help in running the
numerous analyses necessary to make this dissertation reality. Thank you for teaching me new
statistical methods in ways that actually made sense, and helping me to at least partly enjoy the
process.
I would also like to acknowledge the University of Southern California for providing a
Morkovin fellowship, a Dissertation Completion Fellowship, and several teaching assistantships.
I also appreciate the constant support of the psychology department staff, including Sandy
Medearis, Irene Takaragawa and Twyla Ponton. I am especially grateful to Maria Nehlsen for
always having a smile and a hug ready when I needed it the most. Also, I would like to
acknowledge the National Science Foundation who, via a Doctoral Dissertation Research
Improvement Grant, partially funded my dissertation.
Finally, this dissertation would not have been possible without the support of my family
and friends. I am indebted to Andrea McColl for reading numerous drafts and providing
rewording suggestions and necessary distractions. I’m not sure I would’ve made it through this
process without it. Also many thanks to Mary Washburn for making numerous cups of tea,
providing a sounding board for fitting all the puzzle pieces together, and reminding me that I do
actually know what I’m doing.
12
Introduction
Distinguishing between declaratives and interrogatives is an important ability for
language learners. In English and many other languages, interrogatives generally involve
syntactic structures and word order patterns that are different from declaratives. Interrogatives
are often characterized by syntactic properties, such as the use of wh- words (e.g., who, what)
and auxiliary inversion (e.g., Will Becky be home for supper?) or the appearance of do — called
do-support — if no auxiliary is present (Kate likes chocolate becomes Does Kate like
chocolate?). Distinguishing declaratives and interrogatives is thus critical for learning the
grammatical properties of the two sentence types. More broadly, the ability of learners to
distinguish utterance types is required and assumed by many acquisition theories in which
grammatical analyses are first carried out on canonical structures, such as simple transitive,
declarative sentences (Pinker, 1984). Since interrogatives and declaratives each account for a
substantial portion of infant- and child-directed speech (44% and 30% respectively; Newport,
1977), learners’ ability to differentiate utterance types in their input is critical for correctly
categorizing the majority of utterances. Yet, although many language acquisition theories assume
that learners can differentiate these sentence types in the early stages of syntax acquisition, we
know little about how and when this ability develops. Therefore, in the current project, a series
of five studies was conducted with 7- to 12-month-olds to determine when infants begin to
distinguish between declaratives and polar interrogatives, and how prosody and word
order/distributional information may help infants to make these sentence-type distinctions.
Previous research on sentence-type discrimination
Soderstrom, Ko, and Nevzorova (2010) examined when listeners can differentiate
between sentence types. In a series of studies, they assessed the ability of English-learning
13
infants aged 4.5 and 24 months to distinguish the pitch contours that characterize declaratives
and several types of interrogatives. Using a habituation paradigm, one experiment contrasted
declarative statements (e.g., The puppy is cute) and declarative questions (e.g., The puppy is
cute?), which primarily differed in the final pitch contour: declaratives end with final falling or
flat intonation while declarative questions end with a final rise. The results showed that infants
demonstrated a preference for interrogatives when evaluated on the full speech signal, regardless
of habituation sentence type (declarative statement or declarative question), and a preference for
the familiar sentence type when tested on low-pass filtered versions of the same sentences in a
second experiment.
The overall interrogative preference is difficult to interpret since prosody may not be
responsible for the result. For example, infants may pay more attention to the interrogative
stimuli because they do not match the typical correspondence between interrogative word order
and a final rise. Additionally, infants’ ability to distinguish between low-pass filtered versions of
the stimuli may have been due to the fact that prosody was the only cue provided. It is important
to establish whether infants are using prosody as opposed to word order to discriminate between
interrogatives and declaratives because, if they are using prosody they may be able to use non-
word order cues (e.g. lexical selection, word frequency) to discriminate between other sentence
types as well. This raises further questions regarding the sensitivity of this prosodic awareness
and its generalizability to other interrogative types.
A third experiment in Soderstrom et al. (2010) tested infants’ ability to distinguish
between low-pass filtered versions of declaratives and wh- questions. Infants did not show
significant differences in looking time between familiar and novel trials, although there was a
trend towards a novelty preference in the interrogative habituation group.
14
Collectively, the results of Soderstrom et al.’s (2010) studies demonstrate that infants can
distinguish between declarative statements and declarative questions, which share the same word
order and differ only in their prosodic contours. This finding suggests that prosody can be a
useful cue for sentence-type discrimination when it is the only cue available.
Prosody is comprised of the suprasegmental features of speech, namely fundamental
frequency (perceived as pitch), duration, and intensity (perceived as loudness). These three
aspects of prosody allow speakers to parse fluent speech and learn the structure of rhythmic and
intonational cues relevant to their native language (Johnson & Jusczyk, 2001; Thiessen &
Saffran, 2003). There is a wealth of evidence demonstrating infants’ sensitivity to prosodic
information from an early age. Infants can use prosody to distinguish their native language from
other foreign languages and make distinctions between rhythmic classes of languages
1
(Mehler,
Jusczyk, Lambertz, Halsted, Bertoncini, & Amiel-Tison, 1988; Moon, Cooper, & Fifer, 1993;
Nazzi, Bertoncini, & Mehler, 1998). Prosodic cues such as rhythm and stress patterns may help
infants initially attend to the patterns native to their language and segment words from fluent
speech. For example, word-initial stress often signals word boundaries in English (Jusczyk,
Cutler & Redanz, 1993). In addition, infants can use their sensitivity to prosodic cues to identify
syntactic unit boundaries including clauses (e.g. Hirsh-Pasek, Kemler Nelson, Jusczyk, Cassidy,
Druss, & Kennedy, 1987; Kemler Nelson, Hirsh-Pasek, Jusczyk and Cassidy, 1989) and phrasal
boundaries (Gerken, Jusczyk, & Mandel, 1994; Jusczyk, Hirsh-Pasek, Kemler Nelson, Kennedy,
1
Languages can generally be divided into three rhythmic classes based on how the language rhythmically divides
time: stress-timed (e.g. English), syllable-timed (e.g. French) and mora-timed languages (e.g. Japanese)
(Abercrombie, 1967; Pike, 1945). In syllable-timed languages, like French, the duration between each syllable is
equal regardless of stress (e.g. L’ecole est fermée le dimanche) whereas in English, the duration between stressed
syllables is equal regardless of syllables (e.g. School is closed on Sundays.). For languages like Japanese, rhythmic
timing is based on the mora, which is a rhythmic unit equal to the duration of a short vowel (e.g./kookoo/, which has
two long vowels, has as many morae as /tokonoma/; Warner & Arai, 2001).
15
Woodward & Piwoz, 1992). Based on these studies, it is reasonable to expect that infants begin
distinguishing between sentence types using some form of prosodic cues.
A further reason to believe infants use prosody to distinguish between sentence types is
that adults and children do. Perceptual studies with English-speaking adults have demonstrated
that listeners rely on final pitch contours as an important perceptual cue for sentence-type
discrimination (American English- Săfárŏvá & Swerts, 2004; Australian English- Wales &
Taylor, 1987). Săfárŏvá and Swerts (2004) found that American English speakers were more
likely to identify utterances as declarative questions if the utterances ended with a final rise.
Production studies with children have demonstrated a shift in the types of prosodic cues they use
across development. For example, Patel and Grigos (2006) found that 11-year-olds primarily
used fundamental frequency to signal the difference between sentence types, 7-year-olds
contrasted declaratives and interrogatives using a combination of fundamental frequency,
duration and intensity, and 4-year-olds signaled interrogatives with increased syllable duration.
This demonstrates that older children use multiple types of prosodic cues to distinguish sentence
types. Based on these findings, it is possible that infants may rely on several prosodic cues for
perceptually distinguishing between declaratives and interrogatives.
However, prosody is naturally learned concurrently with lexical information, so it is
equally important to evaluate the role of word order/distributional information. So far, no study
has evaluated the role of word order information alone in sentence-type discrimination.
Soderstrom et al. (2010) utilized declarative questions to control for word order effects in their
first two studies, and used low-pass filtered wh- questions in their third experiment, effectively
eliminating word order information. Word order differences between declaratives and
16
interrogatives can only be controlled with the declarative question contrast, so the relative
influence of each type of cue needs to be evaluated.
At first glance it might seem that differences in word order patterns could be useful to
learners in distinguishing declaratives and interrogatives. Interrogatives often begin with
auxiliaries or wh- words, and do not match the canonical English word order (subject-verb-
object), both potentially useful positional cues. However, mechanisms that rely on word order
cues for forming initial utterance categories could lead to overly fine-grained and grammatically
irrelevant distinctions. For example, while an utterance-initial wh- word could be used to
distinguish interrogatives from declaratives, generally grouping utterances based on initial word
types might create distinctions between utterances that start with determiners vs. pronouns
despite their shared underlying structure (e.g. The boy saw the girl vs. You saw the girl). In
addition, different interrogative types have different characteristic word order patterns. Whereas
wh-questions begin with wh- words, polar interrogatives begin with an auxiliary verb: Can you
hand me that cup? Moreover, this type of distributional information presupposes having lexical
categories, such as “wh-word,” and “auxiliary,” that are presumably not part of young infants’
representations of utterances. Thus, it is doubtful that these distributional patterns alone would
be sufficient as an initial source of sentence-type information.
Previous research on age of initial discrimination
As well as addressing the types of cues infants could be using to distinguish between
sentence types, prior studies have evaluated when infants begin to make this distinction.
Soderstrom et al. (2010) evaluated infants between 4.5 and 24 months as a single group, making
it difficult to determine when the ability to distinguish between sentence types first develops.
Geffen (2010) used a modified Visual Habituation paradigm (Houston, Horn, Qi, Ting, & Gao,
17
2007) to test 7- and 9-month-olds’ ability to distinguish between declaratives and interrogatives.
While 7-month-olds demonstrated a strong preference for the sentence type they were
familiarized to, 9-month-olds showed a trend towards a preference for the non-familiarization
sentence type. This suggests that by seven months, infants can distinguish between declaratives
and polar interrogatives.
A follow-up study (Geffen & Mintz, 2011) tested 7-month-olds’ ability to distinguish
between declaratives and polar interrogatives when presented with the full speech signal using
the Headturn Preference Procedure (HPP; Kemler Nelson, Jusczyk, Mandel, Myers, Turk &
Gerken, 1995). Infants listened significantly longer to familiarized sentence types, suggesting
that by seven months, infants can distinguish between declaratives and polar interrogatives.
However, the familiarization and test stimuli were confounded for both declarative and
interrogative stimuli. There was greater utterance-initial lexical overlap between familiarization
and test items of the same sentence type compared to cross type, and this overlap was
confounded with familiarity. The majority of interrogative stimuli began with either can or did,
while four of the declarative familiarization items and one of the test items began with we. This
could mean that infants may not have discriminated between sentence types, but between
sentences that began with can or did versus we. This was less of an issue for Geffen’s (2010)
work because although four out of nine declaratives began with the, there was greater initial-
word variability in the interrogative stimuli. However, there was another potential confound in
that the interrogative stimuli in Geffen (2010) consisted of a mix of interrogative types (polar
interrogatives, wh- questions, and a declarative question). Thus, the current experiments aim to
confirm the findings that 7-month-olds can distinguish between declaratives and polar
18
interrogatives and further investigate the source of information infants attend to in discriminating
interrogatives from declaratives using a more controlled set of stimuli.
It is reasonable to hypothesize that infants as young as seven months can discriminate
between sentence types for several reasons. First, there is evidence that 7-month-olds are able to
segment words from, and detect words in, continuous speech using statistical, prosodic, and
phonological cues (Curtin, Mintz & Christiansen, 2005; Johnson & Jusczyk, 2001; Jusczyk &
Aslin, 1995; Thiessen & Saffran, 2003, 2007). This, then, is an age at which infants could start
noticing and analyzing word sequences in utterances. In addition, cross-linguistic evidence
suggests that infants are sensitive to prosodic properties that distinguish between sentence types
around seven months. Best, Levitt and McRoberts (1991) found that 6- to 8-month-olds
discriminated the prosodic contrast between exclamations and wh- questions in both English and
Spanish. Frota, Butler and Vigario (2014) tested infants learning European Portuguese, a
language which distinguishes between sentence types using only prosodic information, and found
that infants as young as five to six months can distinguish between declaratives and polar
interrogatives. Thus, infants may be able to use prosody at seven months to make initial
sentence-type distinctions (e.g. between declaratives and polar interrogatives), allowing them to
learn the correlated word order/distributional properties of each sentence type. Combined with
previous research (Geffen, 2010; Geffen & Mintz, 2011), seven months seems to be a logical age
to start examining the sentence-type discrimination ability.
Current Research
This dissertation builds on prior work (Geffen, 2010; Geffen & Mintz, 2011) that
provided preliminary evidence for 7-month-olds’ ability to distinguish declaratives from
interrogatives and broadly addresses two questions: 1) When do infants begin to distinguish
19
between declaratives and (polar) interrogatives and 2) What types or combinations of cues do
they use to make these distinctions? A series of five experiments tested two hypotheses: 1) that
infants initially differentiate declaratives and (polar) interrogatives using prosodic cues, allowing
them to then learn the correlated word order/distributional patterns that also distinguish these
sentence types and 2) that older infants can distinguish sentences based on lexical information
alone. Using the Head Turn Preference Procedure (HPP), infants were familiarized to either sets
of declaratives or sets of polar interrogatives, then tested on their ability to discriminate the
familiarized type from the non-familiarized type of sentence.
Addressing the first question, Experiment 1 replicated and extended Geffen and Mintz’s
(2011) findings, examining 7- and 9-month-olds to determine when infants first demonstrate
discrimination between declaratives and polar interrogatives. I hypothesized that Experiment 1
would replicate Geffen and Mintz’s (2011) findings with 7-month-olds and extend the work to 9-
month-olds. Geffen (2010) found a suggestion that 9-month-olds can distinguish between
declaratives and polar interrogatives. Given those findings, I hypothesized that 9-month-olds
could distinguish between declaratives and polar interrogatives. Therefore, Experiment 1
attempts to narrow the age range evaluated in Soderstrom et al. (2010) and addresses the stimuli
confound in Geffen and Mintz (2011).
Addressing the second question, Experiments 2-5 evaluated the types of information
infants might be using to make sentence-type distinctions and the age at which they begin using
those cues. Given infants’ early sensitivity to prosodic information (Mehler et al., 1988; Moon et
al., 1993; Nazzi et al., 1998), it was reasonable to hypothesize that infants may use prosody to
make initial sentence-type distinctions. While prosody distinguishes declaratives from polar
interrogatives but not wh-questions in adult-directed speech (Hedberg, Sosa & Fadden, 2004), it
20
remains unclear whether infant-directed speech shows the same patterns. Experiment 2
evaluated several prosodic dimensions that could distinguish between sentence types. I
hypothesized that infant-directed speech would show the same patterns as adult-directed speech
for declaratives, polar interrogatives and wh- questions.
Based on the results of Experiment 2, it was important to evaluate whether infants could
use prosodic information alone to make sentence-type distinctions. Stimuli were resynthesized,
replacing segmental information from the original sentences with a vowel-like humming sound.
Resynthesis provides an alternative to low-pass filtering, although one which has been primarily
tested on adults. A pilot study (Experiment 3) investigated whether adults could discriminate
resynthesized sentence types based on prosodic cues alone, before testing infants on the same
stimuli. Based on the results of Experiment 3, Experiment 4 evaluated infants’ ability to
distinguish between the resynthesized sentence types. Given infants’ early prosodic sensitivity, I
hypothesized that 7-month-olds may be capable of discriminating sentence types based on
prosodic information alone.
However, there is another potential source of sentence type information in word order
patterns. As previously discussed, many interrogatives have distinct word order properties such
as auxiliary inversion and wh-fronting. Although 2-month-olds’ demonstrate sensitivity to
simple word order sequences when presented with sentential prosody (Mandel, Kemler-Nelson
& Jusczyk, 1996), there is little additional evidence that infants attend to sequential word order
information prior to a year. By 12 months, infants use lexical distributional patterns to
categorize novel words (Mintz, 2006) as well as to distinguish novel grammatical and
ungrammatical utterances (Gómez & Gerken, 1999). Therefore, 12-month-olds may be able to
categorize utterances based on lexical distributional information. Experiment 5a evaluated 12-
21
month-old’ ability to use word order alone to distinguish declaratives and interrogatives. I
hypothesized that 12-month-olds would be able to distinguish between sentence types using
word order alone. In addition, given the stimuli confound in Geffen and Mintz (2011), it was
important to determine whether infants were relying at all on word order information to
distinguish between sentence types. Experiment 5b evaluated 7-month-olds’ ability to use word
order information alone. Based on Geffen and Mintz’s (2011) results, I hypothesized that 7-
month-olds might be able distinguish between sentence types using word order alone, although it
is unlikely.
22
Experiment 1: Infants’ Ability to Distinguish between Declaratives and Interrogatives
This study provides support that infants can distinguish between declaratives and polar
interrogatives. Geffen and Mintz (2011) provided preliminary evidence that 7-month-olds can
make sentence-type distinctions. However, there was a lexical confound between familiarization
and test items. This study replicated Geffen and Mintz (2011) while addressing the confound. In
addition, Geffen (2010) found that 7- and 9-months showed opposite patterns of preference.
Therefore, Experiment 1 evaluated whether 7- and 9-month-olds could distinguish between
declaratives and polar interrogatives, and whether there were age-related differences in patterns
of preference.
Based on prior work (Geffen, 2010; Geffen & Mintz, 2011), I hypothesized that both age
groups would be able to distinguish between the sentence types.
Methods
Subjects:
Fifty-seven typically developing English-learning infants were recruited. Data from two
subjects were excluded from the final analysis due to less than 50% daily English input (1) and
multiple interruptions during test phase
2
(1). From the remaining subjects, three more were
eliminated due to failure to meet basic inclusion criteria: four or more trials with listening times
1) shorter than 2000 ms, 2) greater than (1.5*InterQuartileRange + 3rdQuartile) or 3) less than
(1stQuartile - 1.5*InterQuartileRange) (three 9-month-olds). Data from 52 subjects were
analyzed [twenty-four 7-month-olds (M = 7.05 months, range: 6.63 – 7.36 months, 9 female) and
2
Had to pause the study several times for feeding breaks.
23
twenty-eight 9-month-olds (M = 8.79 months, range: 8.5 – 9.5 months, 10 female)]. Twenty-six
infants (twelve 7-month-olds, fourteen 9-month-olds) were randomly assigned to the declarative
familiarization group and 26 infants (twelve 7-month-olds, fourteen 9-month-olds) to the
interrogative familiarization group (see below). Subjects were recruited from county birth
records, contacted first by letter, and then by phone or e-mail to schedule an appointment.
Stimuli and Design:
Stimuli consisted of 10 declaratives and 10 interrogatives. Eight sentences of each type
served as familiarization sentences and the remaining two sentences of each type served as test
sentences. A given subject only heard one type of familiarization sentence, but all subjects heard
the same test sentences. Sentences varied in terms of individual initial words. Specifically, no
more than two of the familiarization items shared the same initial word, and the initial words of
test items were different from every initial word of familiarization items. Materials were also
controlled at the level of initial sounds. For example, one test exemplar each of declaratives and
interrogatives began with a [k], as did one and only one sentence in each familiarization set. The
other test exemplars began with a segment that did not occur as an initial segment in any
familiarization sentence. Thus, the utterance-initial properties of test items were equal in
comparison to both sets of familiarization utterances, ensuring that any effects of sentence-type
familiarity could not be attributed to surface-level similarities between specific test items and
familiarization items. In addition, the number of obstruents was approximately equal for
declaratives and interrogatives in both familiarization and test items (this will be important for
Experiments 3 and 4). Finally, stimuli were matched on number of syllables and approximate
duration. See Table 1 for a full list of stimuli.
24
Stimuli were recorded by a female native English speaker in an infant-directed register,
which exaggerates prosodic dimensions and is intrinsically appealing to infants (Fernald, 1984).
All recordings were made in a sound-attenuating booth using an EDIROL UA-25 USB Digital
Audio / MIDI Recording interface. Sentences were digitized at a sampling rate of 44,100 Hz. As
in previous studies (e.g. Geffen & Mintz, 2011), there were average pitch peak differences
between sentence types on the final syllable.
Table 1.
Familiarization and Test Sentences for Both Groups.
Declarative Familiarization Interrogative Familiarization
Familiarization
Chips should be eaten with salsa.
Tommy hurt his knee yesterday.
There are elephants at the zoo.
Kings and queens attend royal balls.
Roller coasters can be scary.
The playground is closed on Sundays.
The moon was very bright last night.
Learning to ride a bike is great.
Did the water in the cup freeze?
Is Patty afraid of the dark?
Will Becky be home for supper?
Is next Monday a holiday?
Would you like a tuna sandwich?
Could you pick up the dinosaur?
Do flamingos live at the zoo?
Did you remember the popcorn?
Test
Caramel apples are my favorite.
Silver spoons are for eating cake.
Are you visiting your parents?
Can you do the electric slide?
25
Apparatus and Procedure:
The experiment was conducted in a darkened, sound-attenuated room. Infants were
seated in their parent’s lap in the center of the room, facing a wall with a red light. One yellow
light was mounted on the wall to the left of the infant’s center line and one was mounted on the
wall to the right. The lights were approximately at the same height as the subject’s eyes.
Auditory stimuli were presented from loudspeakers mounted under the yellow side lights. A
video camera mounted below the center red light was focused on the infant and sent a continuous
signal to a monitor in a separate control room. The experimenter observed the infant from the
control room, via the video feed, and indicated the infant’s looking behavior to the computer that
controlled the experiment.
When the child was settled and facing the red light, the red light was extinguished and the
familiarization phase began. During this phase, the eight familiarization sentences associated
with the infant’s group—declarative or interrogative, Table 1—were played through both
loudspeakers. The familiarization sentences were played in 4 blocks, with the order of the
sentences randomized within each block. There were approximately 525 ms silences between
each sentence. Once the familiarization phase started, the stimuli were presented without
interruption and independent of where the infant was looking. The familiarization phase lasted
approximately 80 seconds.
In order to familiarize the infant to the contingency between its looking behavior and the
activity of the lights, which was critical for the testing phase, the lights were also activated while
the familiarization stimuli played, but the light activity was contingent on infants’ looking
behavior (Saffran, Aslin & Newport, 1996). Specifically, as soon as the familiarization phase
began, the red light was extinguished and one of the yellow side lights was randomly selected to
26
begin flashing. It continued to flash until the infant oriented towards the light, and then looked
away for at least two consecutive seconds. At that point, the side light was extinguished and the
center light started flashing again. When the infant reoriented towards the center light, the
process repeated for the duration of the familiarization period, while the auditory stimuli played
continuously.
The familiarization phase was followed by the audiovisual-contingency phase. This phase
was similar to the familiarization phase, except that the presentation of the auditory stimulus
began only when the infant was oriented towards the flashing yellow light, and only through the
loudspeaker that was on the same side as the flashing light. The auditory stimulus for each of
two trials was a synthetic pure tone with a pitch of 440 Hz and a duration of 500 ms. Once
initiated, the stimulus was repeated, with an inter-stimulus interval of 100 ms, for 15 repetitions
or until the infant looked away from the light for two consecutive seconds. When the sound
stopped playing (for either reason), the side light stopped flashing, terminating the trial, and the
center light began flashing, marking a new trial. This phase demonstrated a contingency between
the presentation of the auditory and visual stimuli, which was absent in the familiarization phase
but is critical in the test phase. The contingency phase does not occur in most experiments using
HPP, however in some experimental situations, infants’ orientation to the first one or two test
trials are exceptionally long if they do not experience the audio-visual contingency until the test
phase. Some labs use ‘warm-up trials’ as a way of mitigating against this effect; the audiovisual-
contingency phase is essentially a set of warm-up trials with non-linguistic material.
The test phase immediately followed the contingency phase. The procedure in the test
phase was identical to the audiovisual-contingency phase, except that the dependent measure of
orientation time was recorded by the controlling computer during each trial. There were eight
27
trials in total: four novel trials and four familiar trials. Familiar trials consisted of two new
sentences that matched the familiarization type, and novel trials consisted of two new sentences
that did not match the familiarization type. Critically, all test trials contained sentences that did
not appear in the familiarization phase. The order of presentation of the trials was random, with
the constraint that the first two trials were different types. The selection of the side light was
random, with the constraint that a given side could be selected in no more than three consecutive
trials. The computer recorded infants’ looking time to the flashing light during each trial.
Parents were told that if their child should become wiggly, the best time to readjust them
was when the red light was flashing, which indicated that data were not being collected.
Results
In order to make a sentence-type assessment in a given trial, an infant must, at minimum,
listen to one sentence. Since sentences were approximately 2 seconds long, listening times
under 2 seconds were removed, accounting for approximately 6% of the trials (12 out of 192).
Within-subject analyses excluded 18 additional outliers defined as listening times that were
greater than (1.5*InterQuartileRange + 3rdQuartile) or less than (1stQuartile -
1.5*InterQuartileRange) . After this process, all 24 infants still had at least five data points out of
a possible eight.
The goal of this experiment was to determine whether 7- and 9-month-olds could
distinguish between declaratives and interrogatives. A mixed effects linear regression model
(with subject as a random effect) evaluated infants’ sentence-type discrimination abilities,
comparing the mean looking time on Familiarization Type (declarative vs. interrogative), Trial
Type (novel vs. familiar) and their interaction. A mixed effects linear regression model is
similar to a repeated measures ANOVA with the advantage that a subject is not removed from
28
analysis due to a single missing value. The equality of scale was evaluated between the
declarative and polar interrogative familiarization groups to determine whether it was
appropriate to use parametric or non-parametric tests to evaluate the data. Mood’s median test is
a non-parametric way to compare the scaling of two or more samples to evaluate the
homogeneity of the population they are drawn from.
3
The Mood’s test was non-significant (p >
0.05), indicating that the scale of the two groups was not significantly different, thus parametric
tests were appropriate.
There were no significant main effect or interactions (p’s >0.1), although with a larger
sample size statistical significance may be achieved for the main effect of Sentence Type (β =
3.7, p = 0.12) and the interaction (β = -4.9, p = 0.13).
4
Seven-month-olds in the declarative
familiarization group had a mean of 12.4 s (SD = 8.0 s) for novel trials and a mean of 9.7 s (SD =
3.1 s) for familiar trials. For the interrogative familiarization group, the novel trials had a mean
of 8.7 s (SD = 5.9 s) and the familiar trials had a mean of 10.8 s (SD = 4.2 s). See Table 2.
In addition, the interrogative group looked approximately 1.6 seconds longer at
interrogative trials than infants in the declarative group. See Figure 1. Thus, contrary to Geffen
and Mintz’s (2011) results, 7-month-olds did not distinguish between declaratives and polar
interrogatives when evaluated using standard novelty/familiarity preferences.
Work by Soderstrom et al. (2010) suggests that infants may demonstrate an overall
interrogative preference when presented with the full speech signal, regardless of familiarization
3
This is similar to the Ansari-Bradley test, without the assumption of median equality.
4
A Monte Carlo simulation was conducted to provide an estimate of empirical power for each of the effects
presented. Achieved power for the main effects of Familiarization, Sentence Type, and their interaction were 27.1,
10.1, and 47.6% respectively with the current N of 24 subjects. Additional simulations were run to determine the
sample sizes need to detect each of these effects with >80% power. An N of 125 subjects was needed to provide an
empirical power of 80.8% for the main effect of Habituation. An N of 60 was needed to detect the interaction effect
at a power of 80.9%. A sample N > 375 would be needed to detect a main effect of Sentence Type at > 80% power.
29
type. This idea seems to be supported by the results presented above. To determine whether the
interrogative preference occurred at a level greater than chance, a one-sided binomial test was
conducted. Sixteen out of 24 infants listened longer to interrogative trials versus declarative
trials, though this was not a greater number than expected by chance. To assess the statistical
reliability of the listening time difference, difference scores for each subject were calculated by
subtracting the mean declarative looking time from the mean interrogative looking time (Figure
2). The mean difference score was 2.5 seconds, which was not significantly different from
chance (0) by a one sample t-test (t(23) = 1.62, p = 0.06), although there was a trend. Thus,
although subjects did not show the typical discrimination behavior — discriminating novel vs.
familiar test trials — their preference for interrogatives did approach significance, suggesting
that 7-month-olds can distinguish between interrogative and declarative trial types. To test for an
interaction of the discrimination ability with familiarization type, a Welch two sample t-test
compared difference scores for infants familiarized to declaratives (M = 2.7 s, SD = 8.5 s) and
infants familiarized to interrogatives (M= 2.2 s, SD= 6.5 s) and found no significant difference
(t(20.6) = -0.19, p = 0.85). Thus, infants demonstrated a marginal interrogative preference
regardless of familiarization type.
30
Table 2.
Beta Coefficient and Standard Error by Factor
Factor β SE p
Familiarization 3.7 2.3 0.12
Trial Type 2.2 2.2 0.33
Familiarization x Trial Type -4.9 3.1 0.13
__________ ___________________________________________________________________
31
Figure 1. Mean looking time duration for novel and familiar trials as a function of
familiarization group for 7-month-olds in Experiment 1. Error bars represent +1 standard error
of the mean.
Figure 2. Mean difference score (Interrogative-Declarative Trial) regardless of familiarization
type for 7-month-olds in Experiment 1. Positive score indicates an interrogative preference,
negative score indicates a declarative preference. Error bars represent +1 standard error of the
mean.
32
The same analyses were conducted on the 9-month-olds. All listening times under 2
seconds were removed, accounting for approximately 16% of the trials (36 out of 225). Within-
subject analyses excluded 13 additional outliers defined as listening times that were greater than
(1.5*InterQuartileRange + 3rdQuartile) or less than (1stQuartile - 1.5*InterQuartileRange) .
After this process, all 28 infants still had at least five data points out of a possible eight. The
Mood’s test was non-significant (p > 0.05), indicating that the scale of the two groups was not
significantly different, thus parametric tests were appropriate.
A mixed effects linear regression model found a main effect of Trial Type that
approached significance (β = 3.4, p = 0.09). Overall, novel trials had a mean of 9.7 seconds (SD
= 5.6 seconds) and familiar trials had a mean of 10.7 seconds (SD = 6.7 seconds). There was also
a Familiarization by Trial Type interaction that approached significance (β = -4.8, p = 0.08). See
Table 3. For the interrogative familiarization group, the novel trials had a mean of 8.15 s (SD =
5.59 s) and the familiar trials had a mean of 11.84 s (SD = 7.03 s). The declarative familiarization
group did not demonstrate a preference; the novel trials had a mean of 9.55 s (SD = 5.51 s) and
the familiar trials had a mean of 8.9 s (SD = 5.27 s). The interrogative group demonstrated
longer looking time to familiar trials (i.e. interrogatives) while the declarative group looked
longer at the novel trials (i.e. interrogatives). In addition, the interrogative group looked
approximately 2.3 seconds longer at interrogative trials than infants in the declarative group. See
Figure 3. Combined, these results suggest an overall interrogative preference. Thus, similar to
the 7-month-olds, 9-month-olds may not have shown the expected discrimination behavior
(although the results for the interrogative familiarization group were trending towards a
33
familiarity preference), but they did demonstrate discrimination in their overall preference for
interrogatives, suggesting that 9-month-olds can distinguish between declaratives and
interrogatives.
Given the trend towards an interrogative preference with the 7-month-olds, it was
important to assess whether the 9-month-olds demonstrated the same preference. Twenty out of
28 infants listened longer to interrogative trials versus declarative trials (p= 0.02, one-sided
binomial test). To assess the statistical reliability of the listening time difference, difference
scores were calculated for each subject by subtracting the mean declarative looking time from
the mean interrogative looking time (Figure 4). The mean difference score was 2.4 seconds,
which was significantly different from chance (0) by a one sample t-test (t(27) = 1.81, p = 0.04),
indicating that subjects did show an overall preference for interrogative test trials. To test for an
interaction of the discrimination ability with familiarization type, a Welch two sample t-test
compared difference scores for infants familiarized to declaratives (M= 1.5 s, SD= 7.2 s) and
infants familiarized to interrogatives (M = -3.4 s, SD = 1.9 s) and found no significant difference
(t(26) = 0.70, p = 0.49). See Figure 4. Thus, there was no difference in mean looking time to
interrogatives regardless of whether infants were familiarized to declaratives or interrogatives.
34
Table 3.
Beta Coefficient and Standard Error by Factor
Factor β SE p
Familiarization 2.6 2.3 0.28
Trial Type 3.4 1.9 0.09
Familiarization x Trial Type -4.8 2.7 0.08
__________ ___________________________________________________________________
Figure 3. Mean looking time duration for novel and familiar trials as a function of
familiarization group for 9-month-olds in Experiment 1. Error bars represent +1 standard error
of the mean.
35
Figure 4. Mean difference score (Interrogative-Declarative Trial) regardless of familiarization
type for 9-month-olds in Experiment 1. Positive score indicates an interrogative preference,
negative score indicates a declarative preference. Error bars represent +1 standard error of the
mean.
Although neither age group showed a significant interrogative preference, results
indicated trends in the same direction. Therefore, both age groups were combined for greater
power to determine whether a larger sample size (N = 52) would produce significant effects. A
mixed effects linear regression model found a significant main effect of Trial Type (β = 4.7, p
<0.01), and a significant Familiarization by Trial Type interaction (β = -5.8, p <0.01). Overall,
novel trials had a mean of 9.5 seconds (SD = 5.9 seconds) and familiar trials had a mean of 11.2
seconds (SD = 6.3 seconds). See Table 4. For the interrogative familiarization group, the novel
trials had a mean of 8.04 s (SD = 5.09 s) and the familiar trials had a mean of 12.7 s (SD = 7.38
s). For the declarative familiarization group, the novel trials had a mean of 11.0 s (SD = 6.39 s)
and the familiar trials had a mean of 9.77 s (SD = 4.73 s). The interrogative group demonstrated
longer looking time to familiar trials (i.e. interrogatives) while the declarative group looked
longer at the novel trials (i.e. interrogatives). The main effect approached significance for
36
Familiarization (β = 2.9, p = 0.09). The interrogative group looked approximately 1.7 seconds
longer at interrogative trials than infants in the declarative group. See Figure 5. Combined,
these results suggest an overall interrogative preference.
Given the trend towards an interrogative preference with each age group separately, it
was important to assess whether the combined group demonstrated the same preference. Thirty-
eight out of 52 infants listened longer to interrogative trials versus declarative trials (p <0.01,
one-sided binomial test). To assess the statistical reliability of the listening time difference,
difference scores were calculated for each subject by subtracting the mean declarative looking
time from the mean interrogative looking time (Figure 6). The mean difference score was 2.9
seconds, which was significantly different from chance (0) by a one sample t-test (t(51) = 3.32, p
<0.01), indicating that subjects did show an overall preference for interrogative test trials. To
test for an interaction of the discrimination ability with familiarization type, a Welch two sample
t-test compared difference scores for infants familiarized to declaratives (M= 1.2 s, SD= 7.3 s)
and infants familiarized to interrogatives (M = 4.7 s, SD = 4.8 s) and found a significant
difference (t(43.33) = 2.04, p = 0.05). Thus, infants in the interrogative familiarization group had
longer mean looking time to interrogatives compared to the declarative familiarization group.
37
Table 4.
Beta Coefficient and Standard Error by Factor
Factor β SE p
Familiarization 2.9 1.7 0.09
Trial Type 4.7 1.2 <0.01
Familiarization x Trial Type -5.8 1.7 <0.01
__________ ___________________________________________________________________
Figure 5. Mean looking time duration for novel and familiar trials as a function of familiarization
group for combined age groups in Experiment 1. Error bars represent +1 standard error of the
mean.
38
Figure 6. Mean difference score (Interrogative-Declarative Trial) regardless of familiarization
type for 7-month-olds in Experiment 1. Positive score indicates an interrogative preference,
negative score indicates a declarative preference. Error bars represent +1 standard error of the
mean.
The results of Experiment 1 suggest that 7- and 9-month-olds can distinguish between
declaratives and interrogatives, a preference which is manifesting as an interrogative preference.
Combining the two age groups confirmed this interrogative preference. It was important to
ascertain whether this preference truly reflected discrimination, or if some property of either set
of stimuli was responsible for the looking time differences. Thus, the prosodic properties of the
familiarization and test sentences were examined, since prosody, particularly in the final syllable,
is an important cue for distinguishing between sentence types in adult-directed speech,
To explore whether infants in this study could have used prosodic information to
discriminate sentence types, pitch peak and duration of the final syllable was measured in all 20
sentences. Max F0 for the last syllable of interrogatives (336.89 Hz, SD= 59.22 Hz) was
significantly higher than for declaratives (269.21 Hz, SD= 43.86 Hz; t(18)= -2.905, p = 0.009).
There was no significant difference in mean duration for interrogatives (396.62 ms, SD= 168.55
ms) and declaratives (380.51 ms, SD= 158.95 ms), (t(18)=- 0.22, p= 0.828). The pitch excursion
39
was also examined across the final two syllables. Pitch excursion for interrogatives (161.68 Hz,
SD= 43.38 Hz) and declaratives (153.45 Hz, SD= 65.27 Hz) was not significantly different
(t(18)= -0.332, p= 0.744). Focusing on the familiarization sentences produced the same results.
However, an examination of the test sentences yielded different results. There were no
significant differences in max F0 or pitch excursion (p >0.05). The only significant difference
was in the duration of the final syllable, with interrogatives significantly longer (M = 536.51 Hz,
SD = 29.44 Hz) then declaratives (M = 281.19 Hz, SD = 65.62 Hz).
I also examined difference scores (the change in mean pitch between the penultimate and
final syllables) for declaratives and interrogatives to ensure that the pitch contours were going in
the expected direction. As expected, the mean pitch of the final syllable in polar interrogatives
was 25 Hz higher than the penultimate syllable (final rise) and the mean pitch of the final
syllable in declaratives was 14 Hz lower than the penultimate syllable (final plateau/fall).
Examining familiarization and test items separately produced the same pattern of results. None
of these differences was significant (p > 0.05).
However, there were several exceptions. Interrogative familiarization items #1 and 2 (see
Table 1) had a higher mean pitch for the penultimate syllable than the final syllable, resulting in
a final fall, rather than the typical rise at the end of polar interrogatives.
An additional analysis was conducted at the sentence level. Although the majority of
research has focused on the final syllable(s) for distinguishing between sentence types, prosodic
contours vary across the entire utterance. Variance comparisons between sentence-type groups
were examined using a nonparametric bootstrap, which works well when populations are not
normally distributed. While there were no differences in variability in the final syllable between
sentence types, there was significant variability at the sentence level. Declaratives had
40
significantly greater variance (3019.55 Hz
2
) than polar interrogatives (203.17 Hz
2
) in the overall
Max F0 (95% CI: 711.28, 4693.28). There was also a significant difference in pitch excursion,
with declaratives demonstrating greater variance (2594.47 Hz
2
) than polar interrogatives (350.57
Hz
2
), (95% CI: 552.71, 3717.79).
Discussion
Prior research suggests that there were two possible patterns of results one could expect
for this experiment. The first possibility, based on Geffen and Mintz (2011), was that infants
would demonstrate a familiarity preference. A familiarity preference is defined by a longer
looking time to familiar test trials compared to novel test trials. The second possibility, based on
findings from Soderstrom et al. (2010) was that infants would show an interrogative preference.
Half the test trials consisted of a pair of interrogatives while the other half consisted of a pair of
declaratives. An interrogative preference is defined by a longer looking time to interrogative test
trials compared to declarative test trials, regardless of familiarization type. However, the results
demonstrated a combination of these two possible outcomes.
Addressing the first possibility, the current study aimed to replicate and expand Geffen
and Mintz’s (2011) findings that by seven months, infants can distinguish between declaratives
and polar interrogatives. However, this experiment found no significant looking time differences
to novel and familiar trials, suggesting that, contrary to Geffen and Mintz (2011), 7-month-olds
did not distinguish between declaratives and polar interrogatives. However, the combined age
group looked longer at familiar versus novel trials. Although this suggests a familiarity
preference, the effect was driven by the interrogative familiarization group, suggesting an effect
of familiarization type. Thus, neither the separate age groups, nor the combined group provide
clear support for a familiarity preference.
41
The results provide more support for an interrogative preference. Despite the lack of
significant differences between the two familiarization groups for the 7- or 9-month-olds, both of
the groups demonstrated a trend toward an interrogative preference. Nine-month-olds in
particular showed a strong trend towards discrimination, with a more than 2 second difference in
looking time to interrogatives compared to declaratives. Sixteen out of 24 7-month-olds and 20
out of 28 9-month-olds looked longer at interrogative test trials, suggesting an interrogative
preference. The combined age group provided further support, demonstrating a significant
interrogative preference, with 38 out of 52 infants looking longer at interrogatives compared to
declaratives test trials. An interrogative preference would not be surprising, given that
interrogatives occur more frequently than declaratives in daily conversation to infants and
children (44% versus 30% respectively, Newport, 1977). This frequency may lead to greater
attention to interrogatives. It is possible that interrogatives are intrinsically more interesting than
declaratives, or, this preference may be due to the greater variability in max F0 and pitch
excursion for declaratives in the current experiment. If infants were primarily attending to
prosodic information to make sentence-type distinctions, then perhaps the greater acoustic
variability made it difficult to categorize all sentences of one type into the same category,
especially for the declarative familiarization group. If it was unclear what made the declaratives
the same, perhaps the less variable nature of the interrogative stimuli was more appealing,
allowing infants to identify a commonality among the exemplars. In addition, in typical
discourse, interrogatives are often followed by a pause, while the speaker waits for a response
from the listener. Perhaps infants have begun to attend to this pause, and paid more attention to
the interrogative stimuli because the stimuli do not demonstrate this pause, and/or are not
42
followed by a declarative response. However, this suggests a level of discourse familiarity that
cannot be proved based on the current study.
The final possibility, and the possibility that this dissertation favors, is that the results
reflect a combination of discrimination (manifesting here as a familiarity preference) and
interrogative preferences. For example, both the 9-month-olds and the combined age group
demonstrated an interrogative preference in the interrogative familiarization group while the
declarative familiarization group showed a smaller difference score. This could suggest that
infants in the interrogative familiarization group are demonstrating a familiarity preference
which is boosted by an interrogative preference. It is possible that the declarative familiarization
group also initially demonstrates a familiarity preference which is being overridden by an
interrogative preference, thus the difference score that is very close to zero. It is also possible
that the interrogative preference is moderated by familiarization type, with the interrogative
group driving the results. Collectively, these findings support an interrogative preference rather
than a familiarity preference, and align with Soderstrom et al.’s (2010) findings for the full
speech signal. If results indicate a familiarity preference with the resynthesized stimuli in
Experiment 4, and an interrogative preference with word order/distributional information alone
in Experiment 5, this would suggest that the interrogative preference is related to some aspect of
word order/distributional information.
Despite, or perhaps because of, the inconsistency in results between the current study and
Geffen and Mintz (2011), it is important to evaluate the types of information infants might be
using to distinguish between declaratives and interrogatives. The full speech signal provides a
very rich source of information, containing both word order/distributional and prosodic
information, providing multiple cues and combinations of cues for infants to attend to as they
43
attempt to discriminate between these two sentence types. Investigating whether infants can use
prosodic and/or word order/distributional information to make these types of distinctions may
help make sense of the differing results between the two studies. Alternatively, perhaps the full
speech signal is too rich for all infants to attend to the same cue or combination of cues at the
same time/age. Infants may perform more consistently when they are only presented with one
cue at a time. Subsequent experiments worked to tease apart the roles of these two types of cues.
Experiment 4 investigated whether infants rely on prosodic contours while Experiment 5
evaluated the role of word order in making sentence-type distinctions.
I hypothesized that infants would use prosodic information to make initial sentence-type
distinctions, thereby allowing them to learn the correlated word order/distributional cues
associated with declaratives and polar interrogatives. Thus, the next set of experiments will
evaluate the role of prosody in making sentence-type distinctions. However, it is important to
assess the availability of prosodic information before evaluating infants’ ability to use prosodic
information to make sentence-type distinctions. Thus, Experiment 2 will evaluate whether there
is sufficient prosodic information to distinguish declaratives from polar interrogatives and wh-
questions.
44
Experiment 2: Corpus Analysis of Prosodic Characteristics of Infant-Directed Speech
In adult–directed speech, polar interrogatives are more prosodically distinct from
declaratives than wh- questions but it is unclear whether infant-directed speech shows the same
pattern. Infant-directed speech is characterized by exaggerated prosodic contours, including
higher mean pitch and expanded pitch excursion (Fernald & Kuhl, 1987). This experiment
examined typical prosodic contours in the speech of three mothers to their preverbal infants to
determine whether there is sufficient prosodic information to distinguish declaratives from wh -
questions and polar interrogatives.
I hypothesized that sentence types in infant-directed speech would demonstrate the same
prosodic contours as adult-directed speech.
Methods
Input corpora
This study evaluated input to three children from the Brent corpus from the CHILDES
database (MacWhinney, 2000) for the analysis procedure: s1, s2, and v1 (Brent & Siskind,
2001). The Brent corpus was selected because it contains speech to preverbal infants, and
provides the largest number of audio recordings of mother-child dyads of any corpora in the
database. The present study only analyzed utterances of the mother directed towards her
preverbal infant, taken from 2-3 consecutive sessions spanning a 2-4 week period depending on
the subject. Recordings were made in the subjects’ homes. Mothers were fitted with a fanny
pack and a lapel microphone located just below their mouths, and instructed to go about their
daily routines while home alone with the child (Brent & Siskind, 2001). Infants’ ages ranged
from 0;8.28 to 0;10.04 months.
45
Analysis
This study analyzed approximately 30 utterances per sentence type (declarative, wh-
question and polar interrogative) for each speaker, for a total of roughly 90 utterances per
speaker. Sentences had to meet several criteria for inclusion. Initial selection criteria were based
on structural properties. Polar interrogatives were characterized by subject and auxiliary
inversion, possibly in combination with do-support. Wh-questions were characterized by the
typical auxiliary inversion structure of polar interrogatives with a wh-word such as who, what,
where or how in the utterance initial position. Declaratives typically followed the canonical
transitive word order, although several fragments did not meet this criterion. This corpus
analysis primarily focused on the initial instance of a given utterance.
5
Selection was further
narrowed by listening to the accompanying waveform. Utterances were eliminated if they
included vocalization or background noise, making analysis difficult or impossible, for example:
laughing, crying, blowing raspberries. Unintelligible sentences were eliminated because the
mother's pronunciation helped define syllable boundaries. This corpus analysis analyzed a total
of 311 sentences (See Table 5).
5
Approximately 11% of utterances (35 out of 311) occurred two or more times in a single session, with the majority
of repetitions being questions (19 repeated utterances were wh- questions, 13 were polar interrogatives). Of those 35
utterances, repetitions of 19 utterances were analyzed. The remaining 16 utterances contained repetitions with
background vocalizations that made them difficult or impossible to analyze. The pattern of results did not change
with the addition of the duplicate utterances.
46
Table 5.
Frequency of Utterance Types
Utterance Types Examples Frequency (%)
a
A. Speaker 1 (s1)
Declarative That's your mouth. I'm gonna get it. 31 (34.4)
wh- question Where did he go? How big is he? 28 (31.1)
Polar question Are you gonna crawl? Is that your spoon? 31 (34.4)
Total 90 (28.9)
B. Speaker 2 (s2)
Declarative Now you closed it. I'll bet you can. 36 (37.1)
wh- question What did you think? Who is that? 30 (30.9)
Polar interrogative Can you find it? Is that Piglet? 31 (32.0)
Total 97 (31.2)
C. Speaker 3 (v1)
Declarative You can look at it. They're getting squished. 44 (35.5)
wh- question How do you get down? Where did it go? 39 (31.5)
Polar interrogative Do you see the kitty? Can you get that? 41 (33.1)
Total 124 (39.9)
Total Utterances 311
47
a. Frequency (%) for each sentence type is the percentage of utterances for a given speaker (e.g.
declaratives account for 34.4% of the total utterances for Speaker 1). The % total is the
percentage out of the total number of utterances analyzed (e.g. Speaker 1’s utterances account for
28.9% of the total utterances in the corpus analysis).
Separate analyses of each acoustic measure are detailed below. This study analyzed the
penultimate and final syllables given that, in many languages, the final pitch contour is a critical
(if not the most important) region in the utterance for discriminating interrogatives and
declaratives (e.g. American English – Săfárŏvá & Swerts, 2004; French – Vion & Colas, 2006).
For each utterance, coders marked the boundaries for the penultimate and final syllables for
analysis by hand. One or two coders marked the boundaries, and then I provided final
verification. Coders initially marked syllable boundaries by examining the spectrogram and
listening to the corresponding waveform. Coders also used regions of silence to mark the end
boundary of the final syllable. Disagreements were resolved by having all the coders listen to
the sentences, followed by a group discussion, and if there was still a disagreement, I decided on
the appropriate marking. Coders labeled each section with a rough English transcription of the
syllable. This was a necessary step for information extraction using the Praat script (Xu, 2012),
which carried out batch extraction of 1) mean F0, 2) pitch excursion, 3) max F0, 4) duration and
5) intensity from each labeled interval within all 311 files.
Perceptual studies with adults have demonstrated that pitch is the most important cue for
making sentence-type distinctions (e.g. Săfárŏvá and Swerts, 2004). However, there are three
48
aspects of pitch (mean F0, max F0, and pitch excursion) that could be relevant for making
sentence-type distinctions.
The first feature measured was mean F0. Infant-directed speech is characterized by
exaggerated pitch contours, including higher mean F0. This has been shown across languages
(e.g. English, Japanese, Italian) and genders (Fernald, Taeschner, Dunn, Papousek, Boysson-
Bardies & Fukui, 1989). Mean F0 is also important for calculating the pitch excursion across the
final two syllables. Higher mean F0 in the final syllable would indicate a pitch rise while higher
mean F0 in the penultimate syllable would indicate a pitch fall. Thus mean F0 could be an
important cue for distinguishing between sentence types.
The second feature measured was pitch excursion, a measure of change across the
utterance or, in this case, across the final two syllables of an utterance. A positive F0 range
indicates a final pitch rise, while a negative score indicates a pitch fall across the final two
syllables. Fisher and Tokura (1996) found that F0 range was almost twice as long for utterance-
final syllables compared to either non-final or phrase-final syllables in IDS. Thus, F0 range can
distinguish between syllable positions, although it remains to be seen if it can distinguish
between sentence types.
The third feature measured was max F0. Max F0 is an important measure, particularly
given the final rise typical of polar interrogatives, and the contrasting final flat or falling
intonation of declaratives and wh- questions in adult-directed speech (Bartels, 1999; Hedberg et
al., 2004). In fact, several languages including Russian (Makarova, 2007; Svetozarova, 1998)
and Hungarian (Gósy & Terken, 1994) rely on pitch peak alignment rather than lexical cues to
perceive distinctions between declaratives and polar interrogatives. Declaratives and
exclamations have earlier peaks than interrogatives. There is also an effect of pitch peak height,
49
where interrogatives have higher pitch peaks than declaratives. Even languages like English,
which do use lexical markings (e.g. auxiliary verbs) to distinguish sentence types, demonstrate
differences in pitch peak location between sentence types. For example, English wh- questions
show an earlier pitch rise and final F0 decline (Best et al., 1991). Thus, the location of max F0
could be a useful cue for distinguishing between sentence types. Given that adults can use this
information to distinguish between declaratives and interrogatives, it was important to determine
how early infants could make the same distinctions.
The fourth feature measured was duration. Duration is a secondary prosodic measure
used in other Germanic languages (e.g. Dutch, Orkney English; van Heuven & van Zanten,
2005) which also use pitch to distinguish between sentence types. For instance, in Dutch,
interrogatives typically have a faster speech rate than declaratives, although this difference is
more apparent in declarative questions than polar interrogatives. In addition, children sometimes
substitute durational cues for pitch cues in production studies. For example, 4-7-year-olds
produce interrogatives with increased final syllable duration compared to declaratives, indicating
young children’s sensitivity to durational differences between sentence types (Patel & Grigos,
2006). This sensitivity makes it more likely that children are using durational cues to distinguish
sentence types, provided that duration is a reliable cue. There is also evidence of durational
differences in infant-directed speech at the syllable level. A corpus analysis of infant-directed
speech (Fisher & Tokura, 1996) found evidence of both pause and vowel duration differences
between utterance-final syllables and both non-final and phrase-final syllables in infant-directed
speech. Thus, durational differences exist in infant-directed speech, and it was important to
confirm that similar durational differences were present in the current corpus analysis. In
addition, perceptual studies have demonstrated that 9-month-olds can use syllable duration
50
differences to detect syntactic boundaries, but only in conjunction with other converging cues
such as pitch changes (Jusczyk et al., 1993). Thus, these results appear to mirror the adult
findings, and suggest that infants might be sensitive to durational differences when determining
sentence type, albeit as a secondary cue.
Intensity was the final aspect of prosody assessed in this study because it correlates with
the other dimensions of prosody listed above, especially pitch, and infants appear to be sensitive
to this correlation (Fernald & Kuhl, 1987; Fisher & Tokura, 1996; Mampe, Friederici,
Christophe & Wermke, 2009). For example, in English, stressed syllables are marked by higher
pitch and intensity compared to non-stressed syllables (Hay & Saffran, 2012).
Results
The goal of this study was to determine whether there was sufficient prosodic
information in infant-directed speech to distinguish between declaratives, wh-questions and polar
interrogatives. To simplify analyses, and to highlight the similarities and differences between
sentence type, results are presented as contrasting pairs of sentence types: declaratives versus
polar interrogatives and declaratives versus wh-questions. With the exception of the pitch
excursion analysis, which only focused on differences between sentence types, not syllable
positions, data were subjected to a mixed 2 x 2 ANOVA with Sentence Type as a fixed effect
and Syllable Position (penultimate or final) as a repeated measure.
6
This study was primarily
interested in an interaction of Syllable Position with Sentence Type. If interaction is not
specifically listed in the subsequent sections, this means there was no significant interaction.
6
Adding Speaker as an additional variable did not change the pattern of results.
51
For analyzing pitch excursion, this study identified the lowest pitch value in the
penultimate and final syllables considered together, as well as their highest pitch value. If the
value came from the penultimate syllable, it was coded as negative, while values that came from
the final syllable were coded as positive. An independent-samples t-test was conducted on the
differences scores (maximum-minimum) between sentence types. Min and max F0 were not
available for 21 sentences (7%) due to incomplete pitch information extracted by the Praat script
(Xu, 2012), so they were excluded from the pitch excursion analysis. Means are listed in Table 6
and Table 8 respectively.
Declaratives versus polar interrogatives
Mean F0
Analyses revealed a significant main effect of Sentence Type F (1, 194) =16.97, p
<0.001, and a significant Sentence Type by Syllable Position interaction F (1, 194) = 12.001, p =
0.001. Polar interrogatives had significantly higher mean pitch (M=264. 87, SD= 54.23) for the
final syllable compared to declaratives (M= 221.83, SD= 53.0). There was no significant
difference between sentence types in the penultimate syllable.
In adult-directed speech, one of the primary distinctions between declaratives and polar
interrogatives is the final prosodic contour – i.e. interrogatives have a final rise and declaratives
have a final fall. Larger differences in mean pitch between penultimate and final syllables could
make it easier for infants to acoustically distinguish between sentence types. A positive score
indicates a pitch rise, while a negative score indicates a pitch fall. Declaratives demonstrated a
fall in mean pitch from the penultimate to the final syllable (M = -9.13 Hz, SD = 67.05 Hz) while
polar interrogatives showed a rise in mean pitch (M = 22.43 Hz, SD = 60.44 Hz). These results
demonstrate that mean F0 can be useful for distinguishing between sentence types.
52
Max F0
Analyses revealed a significant main effect of Sentence Type F(1, 212) = 15.830, p
<0.001, a significant main effect of Syllable Position F(1, 212) = 8.245, p <0.005, and a
significant Sentence Type by Syllable Position interaction F(1, 212) = 25.956, p <0.001.
Declaratives showed a decline in Max F0 from the penultimate syllable (M = 262.61, SD =
69.36) to the final syllable (M = 251.58, SD= 66.96), consistent with the prosodic contour of
declaratives in adult-directed speech. Polar interrogatives demonstrated a rise from the
penultimate syllable (M = 268.94, SD = 62.75) to the final syllable (M= 308.49, SD= 74.43),
consistent with the final rise pitch contour in adult-directed speech (See Table 6). This
demonstrates that max F0 provides information infants could use to make sentence-type
distinctions.
Pitch Excursion
An independent samples t-test revealed a significant difference in pitch excursion
between sentence types, t(199) = -3.36, p <0.01. Polar interrogatives had a wider pitch range
(M= 150.17 Hz, SD= 292.29 Hz) than declaratives (M= 11.79 Hz, SD= 292.19 Hz) across the
final two syllables. Specifically, polar interrogatives have higher pitch in the final syllable. The
results for declaratives are a little less clear. While declaratives have a positive difference score,
indicating that the maximum pitch occurs in the final syllable (as opposed to the penultimate
syllable), the score is closer to zero, which could suggest that this is not the case for all (or the
majority) of declaratives.
Duration
Analyses revealed a significant main effect of Syllable Position F (1, 212) = 77.411, p
<0.001, and approached significance for Sentence Type F (1, 212) = 3.623, p = 0.058. The
53
penultimate syllable was longer for declaratives (M=170.25 ms, SD= 83.62 ms) than polar
interrogatives (M=168.75 ms, SD=86.55 ms), although the difference was not significant
between sentence types, t(212) = 0.128, p= 0.898. The difference in final syllable duration
approached significance, such that declaratives (M=310.72, SD=181.73) had a marginally longer
duration than polar interrogatives (M=267.75, SD=181.73) for the final syllable, t(207.071) =
1.923, p= 0.056. This difference could be due to the fact that the last word was one syllable in
most declaratives but two syllables in at least half the polar interrogatives. These results are
consistent with adult-directed speech, where declaratives and polar interrogatives also differ in
overall duration, with polar interrogatives typically a little shorter than declaratives
7
. Larger
differences in duration, particularly in the final syllable, could make it easier for infants to
acoustically distinguish between sentence types. A positive score indicates that the final
syllable is longer, while a negative score indicates that the penultimate syllable is longer. Both
declaratives (M = 140.47, SD = 206.90) and polar interrogatives (M = 99.0, SD = 189.97) had
positive difference scores (See Table 7), indicating that the final syllable was longer for both
sentence types, although the difference scores were not significantly different between sentence
types.
Intensity
Analyses revealed a significant main effect of Sentence Type F (1, 211) = 8.162, p =
0.005, a significant main effect of Syllable Position F (1, 211) = 6.502, p = 0.011, and a
significant Sentence Type by Syllable Position interaction F (1, 211) = 4.443, p = 0.036. Polar
interrogatives had slightly higher intensity than declaratives across positions. Intensity was
7
This has been shown in a variety of languages (e.g. Manado Malay, Orkney English and Dutch; van Heuven & van
Zanten, 2005).
54
significantly higher for polar interrogatives (M= 65.05, SD= 6.62) than for declaratives (M=
61.75, SD= 8.29) in the final syllable t(205.905) = -3.225, p= 0.001, and approached significance
in the penultimate syllable for polar interrogatives (M= 63.42, SD= 6.70) compared to
declaratives (M= 61.60, SD=8.63), t(205.718) = -1.733, p= 0.085. The corresponding rise in
intensity increases across polar interrogatives as the pitch rises is consistent with findings on the
correlation between intensity and other prosodic measures (Fernald & Kuhl, 1987; Fisher &
Tokura, 1996; Mampe et al., 2009). This could indicate that adults may use increasing intensity
to signal the interrogative aspect of the utterance. Given the correlation between intensity and
other prosodic measures in infant-directed speech, intensity could be a useful tool for making
sentence-type distinctions in conjunction with other prosodic cues.
55
Table 6.
Mean and Standard Deviation by Position in Utterance
Sentence Type Penultimate Final
MEAN F0 (Hz)
Declaratives (n= 96) 230.96 (59.60) 221.83 (53.00)
Polar interrogatives(n= 100) 242.44 (57.81) 264.87 (54.23)
MAX F0 (Hz)
Declaratives (n= 111) 262.61 (69.36) 251.58 (66.96)
Polar interrogatives(n= 103) 268.94 (62.75) 308.49 (74.43)
DURATION (ms)
Declaratives (n= 111) 170.25 (83.62) 310.72 (181.73)
Polar interrogatives(n= 103) 168.75 (86.55) 267.75 (144.14)
INTENSITY (dB)
Declaratives (n= 110) 61.60 (8.67) 61.75 (8.29)
Polar interrogatives(n= 103) 63.42 (6.70) 65.05 (6.62)
56
Table 7.
Mean Difference Scores (Standard Error) and Pitch Excursion (Standard Error)
Sentence Type Final-Penultimate Highest Pitch-Lowest Pitch
MEAN F0 (Hz)
Declaratives (n= 96) -9.13 (6.84)
Polar interrogatives (n= 100) 22.43 (6.04)
DURATION (ms)
Declaratives (n= 111) 140.47 (19.64)
Polar interrogatives (n= 103) 99.00 (18.72)
PITCH EXCURSION (Hz)
Declaratives (n= 99) 11.79 (29.37)
Polar interrogatives(n= 102) 150.17 (28.94)
Declaratives versus wh- questions
Mean F0
Analyses revealed a main effect of Syllable Position, F (1, 182) = 4.095, p = 0.044. Post-
hoc analyses revealed no significant difference in mean F0 values for the penultimate or the final
syllable in wh- questions compared to declaratives (p’s > 0.1). Unlike polar interrogatives, in
adult-directed speech, wh-questions are acoustically similar to declaratives, often ending with
57
final falling or flat intonation (Bartels, 1999; Hedberg et al., 2004). Both declaratives (M = -9.13
Hz, SD = 67.05 Hz) and wh- questions (M = -9.05 Hz, SD = 53.38 Hz) demonstrated a fall in
mean pitch from the penultimate to the final syllable, consistent with adult-directed speech.
These results demonstrate that mean F0 is not useful for distinguishing between these two
sentence types.
Max F0
Analyses revealed no main effects or interactions.
Pitch Excursion
An independent samples t-test revealed no significant difference in pitch excursion
between sentence types, t(186) = -0.72, p= 0.94.
Duration
Analyses revealed a significant main effect of Syllable Position F (1, 206 ) = 106.353, p
<0.001. Final syllables (M=304.57 ms, SD= 167.29 ms) were significantly longer than
penultimate syllables (M=170.26 ms, SD= 81.44 ms). These results suggest that duration would
not be useful for distinguishing declaratives and wh- questions.
Intensity
There was a significant main effect of Sentence Type F (1, 205) = 5.089, p = 0.025.
Declaratives had higher intensity than wh- questions in the final and the penultimate syllable:
Declaratives (Final: M=61.60, SD=8.18, Penultimate: M= 61.60, SD= 8.67); wh-questions
(Final: M=59.15, SD=7.56; Penultimate: M=59.51, SD=6.78). In contrast to polar interrogatives,
which had a slightly higher intensity than declaratives, particularly in the final syllable, wh-
questions have a slightly lower intensity than declaratives.
58
Table 8.
Mean and Standard Deviation by Position in Utterance
Sentence Type Penultimate Final
MEAN F0 (Hz)
Declaratives (n= 96) 230.96 (59.60) 221.83 (53.00)
WH-questions(n= 88) 238.58 (49.75) 229.53 (57.26)
MAX F0 (Hz)
Declaratives (n= 111) 262.61 (69.36) 251.58 (66.96)
WH-questions(n= 97) 267.39 (67.44) 265.94 (75.94)
DURATION (ms)
Declaratives (n= 111) 170.25 (83.62) 310.72 (181.73)
WH-questions(n= 97) 170.27 (79.31) 297.52 (149.69)
INTENSITY (dB)
Declaratives (n= 110) 61.60 (8.67) 61.75 (8.29)
WH-questions(n= 97) 59.51 (6.78) 59.15 (7.56)
__________ ___________________________________________________________________
59
Table 9.
Mean Difference Scores (Standard Error) and Pitch Excursion (Standard Error)
Sentence Type Final-Penultimate Highest Pitch-Lowest Pitch
MEAN F0 (Hz)
Declaratives (n= 96) -9.13 (6.84)
WH-questions (n= 88) -9.05 (5.69)
DURATION (ms)
Declaratives (n= 111) 140.47 (19.64)
WH-questions (n= 97) 127.26 (16.31)
PITCH EXCURSION (Hz)
Declaratives (n= 99) 11.79 (29.37)
WH-questions(n= 89) 15.03 (34.76)
This study found significant differences between declaratives and polar interrogatives for
mean F0, max F0, pitch excursion and intensity, and a marginally significant difference between
sentence types for duration. Conversely, the only significant difference between declaratives and
wh- questions was intensity. Collectively, these results demonstrate that prosodic information is
available to distinguish between declaratives and polar interrogatives, but not between
declaratives and wh- questions. However, it remains to be seen how the statistical tendencies of
60
these acoustic cues translate into actual categorization of sentences into sentence-type categories.
Therefore, a Principal Components Analysis examined whether these acoustic measures in a
simple linear discrimination model are useful for classifying sentence types.
Multivariate Dimension Reduction Analysis and Sentence-Type Classification Accuracy
Although this study has just shown that a variety of prosodic cues provide information
that infants could use to distinguish between declaratives and polar interrogatives, it remains
unclear how useful these combinations of cues are for actual categorization. There are patterns
when considering groups of sentences, but to what degree are any of these cues individually or in
combination useful for categorizing individual sentences?
I used Principal Components Analysis (PCA) to standardize and assess 13 variables
which comprise acoustic properties of declaratives and interrogatives: Penultimate Syllable Min
F0, Final Syllable Min F0, Penultimate Syllable Max F0, Final Syllable Max F0, Penultimate
Syllable Mean F0, Final Syllable Mean F0, Penultimate Syllable Duration, Final Syllable
Duration, Penultimate Syllable Intensity, Final Syllable Intensity, Overall Minimum Pitch (over
penultimate and final syllables), Overall Maximum Pitch (over penultimate and final syllables)
and Difference Score (Final syllable mean F0 - Penultimate syllable mean F0). PCA is a
procedure analogous to factor analysis with orthogonal rotation; however devoid of theory, PCA
seeks to create latent scores that maximize the variance explained between the variables. Since
PCA creates uncorrelated (orthogonal) scores, it can play an important role in reducing the
dimensionality of many correlated variables that if modeled individually would suffer from
multicollinearity. In this analysis, PCA was utilized so as to incorporate elements of all the
variables into the eventual logistic regression model without concern for the multicollinearity of
these related variables. The number of component scores was selected based on the scree plot of
61
the eigenvalues. Four components were considered, comprising approximately 87% of the
variance in the variables. See Figure 7. Variable loadings on each component were examined to
determine the makeup of each score. Component 1 was primarily a function of Final Syllable
Duration. Component 2 was a moderate loading on all the variables excluding Difference Score,
Penultimate Syllable Duration, Penultimate and Final Syllable Intensity. Component 3 was
predominantly composed of Difference Score, and had moderate loadings on all variables
excluding Overall Minimum Pitch, Overall Maximum Pitch, Final Syllable Min F0, Final
Syllable Duration, Penultimate and Final Syllable Intensity. Component 4 was primarily a
function of Penultimate Syllable Duration and had moderate loadings on Difference Score and
Penultimate Min F0.
These scores were then used to predict sentence type in a logistic regression model within
a testing database. Half the declaratives (N = 55) and half the polar interrogatives (N = 51) were
selected randomly from all three speakers, to set the coefficients. The coefficients from this
model were then applied to the remainder of the data set, with the predicted probabilities of
sentence type compared to the observed classification. For sentence type 0 (Declaratives) versus
1 (Polar Interrogatives) discrimination, this model showed predictive accuracy for approximately
67.5% of the observations. Of the errors, approximately 57% were incorrectly classifying polar
interrogatives as statements. While 68% still leaves room for improvement, it is hard to assess
what is an acceptable level of predictive accuracy with new models such as these. Using more
components could add to predictive accuracy, although there are additional problems of
overfitting as well as multicollinearity when using the raw variables themselves.
62
For sentence type 0 (Declaratives, N=55) versus 2 (wh- questions, N=48) discrimination,
this model showed predictive accuracy for approximately 51% of the observations. Thus, this
model is at chance for categorizing declaratives and wh-questions.
Figure 7. Scree plot of the principal components of Sentence-Type categorization. Shows the
variance of the first 10 components and supports a four component analysis.
Eigen values
63
Table 10. Loadings of Individual Factors for Principal Components Analysis.
Co
mp.
1
Comp.
2
Comp.
3
Comp.
4
Comp.
5
Comp.
6
Comp.
7
Comp.
8
Comp.
9
Comp.
10
Comp.
11
Comp.
12
Comp.
13
Min 0.04 -0.23 -0.05 -0.13 -0.59 -0.10 0.04 -0.10 0.51 0.17 0.52 -0.01 0.01
Max 0.15 -0.42 0.00 0.08 0.23 0.49 -0.08 0.04 0.46 0.37 -0.39 0.07 0.00
Difference
Score 0.01 -0.05 0.52 0.30 -0.19 0.07 -0.31 -0.11 -0.43 0.53 0.14 0.01 0.01
PUMinf0 0.09 -0.23 -0.17 -0.31 -0.26 0.00 0.54 -0.40 -0.38 0.24 -0.31 -0.01 0.00
FNMinf0 0.02 -0.28 0.07 0.04 -0.58 -0.09 -0.26 0.38 -0.09 -0.35 -0.48 0.01 -0.02
PUMaxf0 0.08 -0.33 -0.38 -0.07 0.09 0.23 0.07 0.59 -0.39 0.12 0.40 -0.02 0.01
FNMaxf0 0.18 -0.41 0.27 0.12 0.09 0.35 0.06 -0.36 -0.12 -0.60 0.27 -0.06 0.00
PUDuration -0.07 0.00 -0.48 0.84 -0.14 -0.02 0.13 -0.16 0.00 -0.01 -0.03 0.00 0.01
FNDuration 0.93 0.33 -0.06 0.04 -0.09 0.01 -0.03 0.03 -0.01 0.01 -0.01 -0.01 0.00
PUPitchmean 0.11 -0.34 -0.37 -0.15 0.21 -0.40 -0.61 -0.35 -0.10 0.03 0.00 -0.02 0.01
FNPitchmean 0.19 -0.38 0.32 0.21 0.28 -0.64 0.36 0.22 0.09 0.05 -0.01 -0.03 -0.02
PUIntensity 0.01 -0.01 -0.01 0.00 0.00 -0.02 0.00 -0.02 -0.03 -0.02 0.05 0.67 -0.74
FNIntensity 0.02 -0.01 0.01 0.00 0.01 -0.03 0.02 0.00 -0.03 -0.05 0.03 0.74 0.67
64
Discussion
The results of this corpus study demonstrate that there is sufficient prosodic information
to distinguish declaratives from polar interrogatives in infant-directed speech, but not from wh-
questions. Four of the five prosodic measures – mean F0, max F0, pitch excursion, and intensity
–significantly differed between declaratives and polar interrogatives, while the durational
difference between the two sentence types approached significance. In contrast, intensity was the
only cue that distinguished between declaratives and wh- questions, with wh- questions
demonstrating lower intensity than declaratives which demonstrated lower intensity than polar
interrogatives. This provides preliminary evidence that infant-directed speech contours follow
the same patterns as adult-directed speech contours (Hedberg et al., 2004). Specifically,
declaratives and wh-questions end with a final fall, and polar interrogatives end with a final rise.
This suggests that while parents may exaggerate typical speech contours in daily speech to
infants, they do not change the shape of the contour. Instead, they simply exaggerate the features
that serve to distinguish between sentence types. Further research is needed to determine
whether infants are using these four acoustic cues (mean F0, max F0, pitch excursion and
intensity), and if so, their relative value for sentence-type discrimination.
The logistic regression model also suggests a multivariable approach to classifying
sentence types. The regression model was intended to narrow down and evaluate what variables
these components had in common, to determine which acoustic variables are best at classifying
sentence types. However, none of the variables turned out to be dominant for classification
purposes. Component 1 was predominated by Final Syllable Duration. Component 4 was
primarily a function of Penultimate Syllable Duration and had moderate loadings on two
additional variables. Component 3 was predominantly composed of Difference Score, but had
65
moderate loadings of six additional variables. Components 2 had moderate loadings of all but
four variables. There was no overlap between the predominant variables in Components 1, 3
and 4 and little overlap in moderate loadings of variables either. While these results did not limit
the number of variables necessary for classification, they still provide a model for classifying
declaratives and polar interrogatives.
These results suggest that sentence-type classification relies on varying combinations of
prosodic cues. Further research is needed to determine how well the model performs relative to
human ability. Combined with the corpus analysis, these results suggest that infants may rely on
multiple prosodic cues to distinguish between sentence types.
Although prosody is insufficient to distinguish between declaratives and wh- questions,
the ability to distinguish between declaratives and polar interrogatives provides valuable
information for making distinctions between other types of interrogatives. Polar interrogatives
and wh- questions occur in roughly equal proportions in infant-directed speech (23% and 21%
respectively; Newport, 1977), so polar interrogatives are a significant part of learners’ input.
Moreover, the characteristic inversion of auxiliaries and subjects in polar interrogatives also
occurs in wh- questions, after the wh- word (e.g. What did you have for lunch?). Using prosodic
information to differentiate polar interrogatives from declaratives could allow infants to notice
the characteristic auxiliary-subject inversion patterns. They could then detect the same patterns
in wh- questions and use the distributional similarity as a basis for distinguishing wh-questions
from declaratives. That is, infants may initially distinguish polar interrogatives from declaratives
using prosodic information, allowing them to detect the correlated auxiliary-subject inversion
pattern. Infants could then detect the same patterns in wh- questions and use this distributional
similarity as a basis for treating wh- and polar interrogatives as similar, and distinct from
66
declaratives. In this way, the learned associations between word order and prosodic patterns in
polar interrogatives may allow infants to generalize to wh- questions, given their prosodic
similarity to declaratives (Bartels, 1999; Hedberg et al., 2004).
In addition to distinguishing between sentence types, infants could also use prosodic cues
to distinguish between different types of interrogatives. For example, in languages like Russian,
which have relatively free word order and no auxiliary verbs (Rojina, 2004), pitch peak is an
important perceptual cue for distinguishing between declaratives and polar interrogatives
(Makarova, 2007; Svetozarova, 1998). But would pitch peak be equally useful in making other
sentence-type distinctions? Infant-directed wh- questions in English show an earlier pitch peak
and final F0 decline (Best et al., 1991). If wh- questions carry their emphasis, and thus their pitch
peak, on the wh- word, pitch peak location could be a useful prosodic cue for distinguishing wh-
questions and polar interrogatives. However, the current study did not analyze the location of
the pitch peak so further research is needed to substantiate this claim.
While prosody is not likely to aid discrimination of declaratives from wh-questions,
infant-directed speech provides a learner with sufficient prosodic information to distinguish
declaratives and polar interrogatives. For instance, Frota et al. (2014) found that Portuguese
learning 5- to 6-month-olds can use prosody to distinguish between declaratives and polar
interrogatives. However, it remains unclear whether English-learning infants can use this
information to make sentence-type distinctions, which will be evaluated in Experiment 4 using
resynthesized stimuli. Since few studies have looked at infants’ ability to distinguish between
resynthesized stimuli, Experiment 3 evaluated adults’ ability to make sentence-type distinctions
when this process was used before testing infants.
67
Experiment 3: Adults Can Distinguish between Resynthesized Declaratives and
Interrogatives
Experiment 2 demonstrated that there is sufficient prosodic information to distinguish
declaratives from polar interrogatives, but not wh- questions, but it is unclear whether infants can
use some or all of this prosodic information to make the same distinctions. Perceptual studies
have demonstrated that English-speaking adults rely on pitch information (e.g. Săfárŏvá and
Swerts, 2004), and to a lesser extent duration and intensity (Lieberman, 1967), to make these
sentence-type distinctions. However, these studies utilized either a full speech signal or low-pass
filtered speech. In contrast, Experiments 3 and 4 used resynthesis, an alternative to low-pass
filtering, for removing lexical and segmental information.
There are several reasons for using resynthesis instead of low-pass filtering. First,
although resynthesis does reduce naturalness, the resulting stimuli are arguably no more
unnatural sounding than sentences which have been low-pass filtered. In experiments with
infants, low-pass filtering has been employed to remove frequency components above 400 Hz
(e.g., Mehler et al., 1988 and elsewhere), which results in very unnatural spectral properties. On
the other hand, the resynthesis used in Experiments 3 and 4 removes lexical and segmental
information while leaving intact spectral properties of the human vocal tract. Ultimately, there
are always tradeoffs between naturalness and control when isolating stimulus dimensions.
Resynthesis also allows greater flexibility in eliminating or preserving prosodic cues. Thus,
resynthesis provides a better tool for evaluating whether adults and infants can distinguish
between declaratives and interrogatives when only rhythm and intonation are available. Second,
resynthesis techniques yield fewer fuss-outs from infants, leading to less attrition (Ramus &
68
Mehler, 1999). Third, it is easier to control volume levels when using resynthesized stimuli.
Low-pass filtering removes the higher frequencies, and no matter how loud the volume, it is
more difficult to hear lower frequencies.
There are very few studies examining discrimination abilities with resynthesized stimuli,
and even fewer studies evaluating infants. Ramus and Mehler (1999) tested adults’ ability to
discriminate between modified versions of English and Japanese, where all phonemes have been
replaced with /a/ and varied continuously in pitch (/a/ resynthesis speech condition). English, but
not French-speaking adults, could distinguish between these two languages. Studies by Vicenik
and Sundara (2013) tested English-speaking adults’ language and dialect discrimination ability
using the same /a/ resynthesis speech condition as Ramus and Mehler (1999). Adults could
distinguish between languages (English and German), but not between dialects of the same
language (American and Australian English). Thus, adults can make some distinctions based on
this degraded stimuli. But what about infants? Ramus (2002) tested newborns on several types
of resynthesized speech, progressively removing speech cues available for discrimination (e.g.
broad phonotactic cues), although he did not test infants on the aaa speech condition used by
Ramus and Mehler (1999) and Vicenik and Sundara (2013). This raises two questions which
will be addressed in Experiments 3 and 4: 1) Can infants distinguish between stimuli which been
modified according to the aaa speech condition and 2) If adults can distinguish between
languages, but not dialects, will they be able to distinguish between sentence types?
Experiment 3 examined American English-speaking adults' ability to distinguish
declaratives and polar interrogatives when presented with resynthesized versions of declaratives
and polar interrogatives where lexical (and segmental) information had been removed. This
experiment served as a forerunner to Experiment 4, determining whether adults could make this
69
distinction without lexical and segmental information before testing infants. If adults could not
distinguish resynthesized versions of declaratives and interrogatives, then infants, with much less
experience with language, likely could not make the distinction either.
I hypothesized that adults could make this distinction at a level above chance.
Methods
Subjects:
Sixteen native American English speakers (M = 20.0 years; 93.75 % female) were
recruited from the University of Southern California Psychology Subject Pool. Eight out of 16
subjects reported speaking one or more languages fluently in addition to English. Students
participated in the study for extra credit or to satisfy a class research participation requirement.
The study was one of many studies available.
Stimuli:
The sentences from Experiment 1 were resynthesized. The pitch contour from each
sentence was extracted and the spectral information that differentiates speech segments was
replaced with a vowel humming sound using the “Synthesize to sound (hum)” feature in Praat
(Boersma & Weenink, 2011). The resynthesis process replaces obstruents with a brief silence.
Since obstruents do not convey pitch, the resynthesis process affects the rhythmic quality of the
resynthesized speech. Thus, overall differences in the number and placement of obstruents
across sentence types could artificially introduce a confound with sentence type after resynthesis.
It is for this reason that stimuli were approximately equal in the number of obstruents. The
resulting utterances still sound speech-like and contain the spectral qualities (e.g. formant
structure) of natural speech. The design of familiarization and test trials was the same as
Experiment 1.
70
Apparatus and Procedure:
The experiment was presented in iTunes. Subjects listened to 20 sentences (10
declaratives and 10 polar interrogatives, presented in a randomized order) one at a time over
headphones. Subjects were told they would hear a number of modified sentences and were
instructed to decide whether each sentence was a declarative or a polar interrogative, and to
make their best guess if they were unsure. After listening to each sentence, subjects judged
whether it was a "Declarative" or "Interrogative." Once the subject made their choice, the
experimenter played the next sentence. Subjects heard each sentence only once
8
. Testing lasted
approximately five minutes. After the experiment, subjects were asked to explain the strategies
they used to decide if each sentence was a declarative or an interrogative.
Evaluating adults' strategies for distinguishing between declaratives and interrogatives
could provide insights into the types of cues infants rely on to make similar distinctions.
Results
The goal of this experiment was to determine whether adults could accurately identify
declaratives and interrogatives when lexical and segmental information had been removed and
only rhythm and intonation was available. , A one-sample t-test on total correct scores
investigated whether adults could correctly identify sentence types at a level above chance.
Discrimination scores were significantly above chance, t(15) = 6.63, p <0.001. Subjects
correctly identified an average of 16.06 utterances (SD = 3.66).
8
There was one exception: subjects were occasionally allowed to listen to the first sentence twice. Despite the
experimenter's warnings that speech would be unintelligible and sound strange, several subjects were still
unprepared for the unnaturalness of the resynthesized speech. If subjects seemed distracted enough that they could
not make a judgment, the subject was allowed to listen to the first sentence a second time at the experimenter's
discretion. No other repetitions were allowed.
71
Eight out of 16 subjects reported speaking one or more languages fluently in addition to
English. An independent samples t-test investigated whether there was an advantage to speaking
multiple languages fluently. Multilinguals correctly identified an average of 17.38 utterances
(SD =2.20) while monolinguals correctly identified an average of 14.75 utterances (SD = 4.46).
There was no significant difference between mono- and multilinguals, t(10)=1.49, p = 0.166.
Examining individual scores, multilinguals appear to actually be performing better than
monolinguals, with all but one multilingual correctly categorizing 16/20 or higher while
approximately half the monolinguals scored in the 13-16 out of 20 range. However, the statistics
do not support this performance difference. This might be because t-tests do not have the ability
to separate out high levels of variation like we see in the multilingual data. This variation could
be caused by the fact that the multilinguals differed in the number of additional languages (one
or two) and the languages themselves, which differ in their use of prosody. Thus, multilinguals
may be better at this task, although it is difficult to draw conclusions based on a single study.
In addition to evaluating overall accuracy, it was important to conduct an item analysis to
examine whether there were similar responses to all of the items of a specific sentence type. A
binomial test was performed for each sentence on the proportion of subjects who chose the
correct answer. Scores for individual sentence recognition ranged from 44 % to 100 % (chance
= 50%). Subjects performed significantly better than chance on 14/20 items (p's < 0.01). An
additional three sentences were identified correctly by 69% of subjects, but were not
significantly different from chance. Of the six sentences that did not meet the criteria for
statistical significance, 5/6 were interrogatives, including the three sentences yielding the worst
accuracy scores (44%, 50% and 56%, respectively #2, 3 and 8 in the interrogative familiarization
set, see Table 1). These three sentences may have had less of a perceived final rise than other
72
interrogatives, and thus more closely resembled declaratives than interrogatives. Interrogative
#2 in the familiarization sentences had a 44 Hz fall from the penultimate syllable to the final
syllable. However, interrogative #3 had a 113 Hz rise from the penultimate to the final syllable,
so it is somewhat surprising that subjects only identified this as an interrogative 50% of the time,
especially since other sentences that had less of a rise were identified as interrogatives with
higher accuracy. For example, interrogative #2 in the test sentences had a rise of only 7 Hz but
was identified as an interrogative 69% of the time. Subjects indicated that they associated rising
(final) pitch with interrogatives, and consistent pitch with declaratives.
While the t-tests assessed overall accuracy, a logistic regression adjusted for covariance
and evaluated the influence of sentence type and individual items on predicted performance.
Odds Ratios (OR) and 95% Confidence Intervals (CI) for predicted performance are reported in
Table 11, as well as the covariates across all three models. Subject effects were examined using
the null model (Model 0; subjects alone). The OR was 5.25 (95% CI =2.96, 9.31; p <0.001),
indicating that overall, subjects responded correctly at a level greater than chance (50%).
Adjusting for Sentence Type (declarative or interrogative) in Model 1, the OR remained
significant (OR = 6.93; 95% CI =3.14, 15.29; p <0.001). Subjects correctly identified
declaratives at a level greater than chance, and correctly identified declaratives 0.6 times more
often than interrogatives although this difference was not significant. A likelihood ratio chi-
square test showed no significant difference between the null model and Model 1 (See Table 12).
In Model 2, which included the individual items, the OR remained significant (OR = 9.70; 95%
CI =3.16, 29.77; p <0.001). Subjects still performed better than chance, but showed no difference
in correct identifications between sentence types. A likelihood ratio chi-square test showed that
Model 2 is a better predictor than Model 1.
73
Table 11.
Odds Ratio (OR) and 95% Confidence Intervals (CI) for Sentence Type Identification Using
Logistic Regression across Models
Model 0 Model 1 Model 2
OR (95% CI) P OR (95% CI) P OR (95% CI) P
Intercept 5.25 (2.96,9.31) <.001 6.93 (3.14,15.29) <.001 9.70 (3.16,29.77) <.001
Sentence Type 0.62 (0.32, 1.21) 0.16 0.62 (0.19, 1.99) 0.418
(Declarative= 0,
Interrogative=1)
Table 12.
Chi Square, Degrees of Freedom and p-Values.
Model Evaluation Likelihood Ratio Test
Model 1 Model 2
χ
2
2.33 16.77
df 3 1
p 0.507 <0.001
Discussion
As a group, American English-speaking adults were able to distinguish between
declaratives and interrogatives using resynthesized stimuli containing only rhythm and
intonational cues at a level greater than chance. Subjects indicated that their primary strategy for
distinguishing between sentence types consisted of listening for a final rise to indicate an
74
interrogative. However, two of the items with the lowest accuracy had average to high pitch rises
from the penultimate to the final syllable. In addition, the interrogative with the smallest pitch
rise was still identified correctly 69% of the time. Only one of the items with the lowest
accuracy in fact had a pitch fall, suggesting that there is a difference between actual pitch
changes and perceived pitch changes.
These results are consistent with the declarative bias found in previous studies (American
English – Săfárŏvá & Swerts, 2004; French – Vion & Colas, 2006) which found that subjects
were more likely to default to a declarative judgment in the absence of a final rise. For example,
Vion and Colas (2006) found a nearly 40% improvement on a gating study with French noun
phrases when final pitch information was introduced at the end of an utterance. In the absence of
this final prosodic information, subjects defaulted to a declarative judgment. This bias might
explain why subjects responded slightly more “Declarative” than “Interrogative.” However,
both of these prior studies tested participants on stimuli that contained lexical information. If
participants are truly influenced by the final rise or fall of the pitch contour, then the lack of
lexical information in the current experiment should make it easier to distinguish between
stimuli, assuming that the stimuli follow the stereotypical prosodic contours attributed to
declaratives and polar interrogatives. This suggests the possibility that subjects’
misclassification could be due to item-specific properties.
There are several other possibilities which could account for the errors that the subjects
demonstrated. One possibility is that the process of resynthesis could make it harder to
distinguish between sentence types. Resynthesizing sentences to the hum version may have
changed the sentences in some essential way that makes it harder to correctly identify several
interrogatives. For example, the number and location of obstruents could affect the overall pitch
75
contour of the utterance, since the resynthesis process essentially converts them into silence.
This could be a particular issue if the obstruents are in the final two syllables. Converting the
obstruents to silence would leave gaps in a crucial part of the prosodic contour. Although
obstruents leave gaps in the prosodic contour in typical conversation, in the absence of lexical
cues, these gaps may be more noticeable, making the sentences sound odd, which could
negatively influence identification.
Alternatively, resynthesis may emphasize some of the important differences between
stimuli, in particular the final pitch differences. The results demonstrate that, though there are
differences between declaratives and polar interrogatives, some of the interrogative stimuli may
not be the best exemplars of polar interrogatives. However, it is important to note that while
adults had some difficulty in accurately identifying sentences that served as familiarization items
in the infant behavioral experiments using HPP, they did not have difficulty in identifying the
test sentences (See Table 1).
Another possible explanation is the articulatory position of the final vowel (low, mid or
high). Different positions have different characteristic pitch patterns, and perhaps some of these
pitch patterns make the final rise less obvious. Two of the sentences consistently misidentified
have a high final vowel, one has a mid-level vowel and one has a low vowel, suggesting there is
not a relationship between the position of the final vowel and whether subjects correctly
identified it as an interrogative. However, all but one of the remaining six interrogatives has a
mid-level vowel, so perhaps being at one extreme or the other makes a difference.
The purpose of the hum study was to measure the available information in resynthesized
speech and to verify that adults could use this more limited set of information to distinguish
between declaratives and polar interrogatives before testing infants on the same stimuli. Adults’
76
ability to distinguish between declaratives and interrogatives in the absence of lexical and/or
semantic information suggests that the limited information provided in the resynthesized stimuli
is sufficient for distinguishing between sentence types. This suggests infants should be able to
do the same, although the age when they can initially make a distinction based on these distorted
stimuli remains to be determined, and is the subject of Experiment 4.
77
Experiment 4: Infants’ Ability to Distinguish between Resynthesized Declaratives and
Interrogatives
This study evaluates whether infants can distinguish between sentence types based on
prosodic information alone. As previously mentioned, infants are sensitive to prosodic
information from birth. Newborns can distinguish between languages from different rhythmic
classes (e.g. English and Japanese) even when speech has been low-pass filtered (Mehler et al.,
1988; Moon et al., 1993), but not between languages from the same rhythmic class (e.g. English
and Dutch) (Christophe & Morton, 1998; Nazzi et al., 1998). Infants’ ability to discriminate
between languages even when they are low-pass filtered (Bosch & Sebastián-Gallés, 1997;
Mehler et al., 1988; Nazzi et al., 1998), suggests that segmental information is not necessary to
distinguish between rhythmically different languages.
In addition, infants can use prosodic cues to segment words (Curtin et al., 2005; Johnson
& Jusczyk, 2001; Thiessen & Saffran, 2003) and once they become more attuned to the stress
patterns of their native language (Jusczyk et al., 1993), this allows them to segment syntactic
units (e.g. clauses and syntactic phrases; Hirsh-Pasek et al., 1987; Jusczyk et al., 1992). Thus, it
is important to evaluate the role of prosodic cues in sentence-type discrimination.
Soderstrom et al. (2010) demonstrated that infants could distinguish between declarative
statements and declarative questions when stimuli had been low-pass filtered, which suggests
that infants can discriminate between sentence types based on prosody. However, the subjects in
Soderstrom et al.’s (2010) experiment ranged from 4.5 to 24 months, so it is unclear when this
ability develops.
78
Experiment 4 evaluated whether 7-month-olds could distinguish between declaratives
and polar interrogatives, using methods and procedures identical to Experiment 1, and the
modified stimuli from Experiment 3. Based on the results of Experiments 2 and 3, I
hypothesized that 7-month-olds would be able to discriminate between sentence types based on
these resynthesized stimuli.
Methods
Subjects:
Thirty-eight typically developing English-learning 7-month-olds were recruited using the
same strategies as Experiments 1. Data from six subjects were excluded from the final analysis
due to fussiness (2), failure to complete study (1), experimenter error (1), lack of attention (1),
and four or more outlier trials (1). Of the remaining 32 infants (mean age 7.02 months, range:
6.7- 7.57 months, 18 female), 16 were randomly assigned to the declarative familiarization group
and 16 to the interrogative familiarization group (see below).
Stimuli:
This experiment used the same sentences as Experiment 3. However, there was one
difference between the stimuli in Experiment 3 and the current stimuli. Since adults had
difficulty correctly identifying several sentences in Experiment 3 (correct identification <70%),
another pilot experiment tested the same 20 stimuli recorded by a second female native English
speaker. The best versions of each sentence were selected from the two pilot studies, and a
roughly equal number of sentences were included from each speaker.
Design, Apparatus and Procedure:
Same as Experiment 1.
79
Results
This experiment followed the same analysis procedures as Experiment 1. All listening
times under 2 seconds were removed, accounting for approximately 16% of the trials (42 out of
257). Within-subject analyses excluded 19 additional outliers defined as listening times that
were greater than (1.5*InterQuartileRange + 3rdQuartile) or less than (1stQuartile -
1.5*InterQuartileRange). After this process, all 32 infants still had at least five data points out of
a possible eight.
The goal of this experiment was to determine whether infants could accurately identify
declaratives and interrogatives when lexical and segmental information had been removed and
only rhythm and intonation was available. A mixed effects linear regression model (with subject
as a random effect) evaluated infants’ sentence-type discrimination abilities, comparing the mean
looking time on Familiarization Type (declarative vs. interrogative), Trial Type (novel vs.
familiar) and their interaction. The Mood’s test was non-significant (p > 0.05), indicating that
the scale of the two groups was not significantly different.
There were no significant main effects or interactions. Novel trials had a mean of 8.2
seconds (SD = 5.5 seconds) and familiar trials had a mean of 8.6 seconds (SD = 4.5 seconds).
Infants in the declarative familiarization group had a mean of 7.7 s (SD = 5.3 s) for novel trials
and a mean of 7.9 s (SD = 4.2 s) for familiar trials. For the interrogative familiarization group,
the novel trials had a mean of 8.7 s (SD = 5.8 s) and the familiar trials had a mean of 9.4 s (SD =
4.7 s). See Table 13. Thus, contrary to my hypothesis, 7-month-olds did not distinguish between
statements and questions.
Despite the lack of significant looking time differences, and given the hint of an
interrogative preference in the earlier studies, it was important to assess whether a similar
80
preference was present in the current study. Seventeen out of 32 infants listened longer to
interrogative trials versus declarative trials, though this was not a greater number than expected
by chance. To assess the statistical reliability of the listening time difference, difference scores
were calculated for each subject by subtracting the mean declarative looking time from the mean
interrogative looking time (Figure 8). The mean difference score was 0.4 seconds, which was
not significantly different from chance (0) by a one sample t-test (t(30) = 0.45, p = 0.33),
indicating that infants did not show an overall preference for either sentence type. To test for an
interaction of the discrimination ability with familiarization type, a Welch two sample t-test
compared difference scores for infants familiarized to declaratives (M= -0.2 s, SD= 4.1 s) and
infants familiarized to interrogatives (M = -1.0 s, SD = 4.9 s) and found no significant difference
(t(27.4) = 0.74, p = 0.47). Thus, infants did not discriminate declaratives and interrogatives
regardless of familiarization type.
Table 13.
Beta Coefficient and Standard Error by Factor
Factor β SE p
Familiarization -0.8 1.8 0.65
Trial Type 0.8 1.2 0.47
Familiarization x Trial Type -0.6 1.6 0.70
__________ ___________________________________________________________________
81
Figure 8. Mean looking time duration for novel and familiar trials as a function of familiarization
group for 7-month-olds in Experiment 4. Error bars represent +1 standard error of the mean.
82
Figure 9. Mean difference score (Interrogative-Declarative Trial) regardless of familiarization
type for 7-month-olds in Experiment 4. Positive score indicates an interrogative preference,
negative score indicates a declarative preference. Error bars represent +1 standard error of the
mean.
Discussion
7-month-olds did not show reliable looking time differences to the two types of
utterances, which suggests that their discrimination abilities do not solely rely on prosodic
information. This seems to contradict the hypothesis that infants initially distinguish declaratives
and interrogatives based on prosodic information, although it is too soon to simply dismiss
infants’ ability to use prosodic information to make sentence-type distinctions.
One possibility is that while the stimuli still sounded like speech, albeit without words,
the stimuli sounded too unnatural. Perhaps infants simply did not like the stimuli, and thus did
83
not attend to the sentences the way they normally would
9
. Listening times in the current
experiment were shorter than when infants listened to sentences consisting of the full speech
signal in Experiment 1 (see Figure 1, Figure 3 and Figure 8). It is possible that, if infants were
tested on stimuli that no longer sounded like speech (e.g. sine tones), that those would be
different enough from speech that infants would focus on simply discriminating between them,
rather than comparing them to daily speech.
Another possibility is that the acoustic variability in the declarative stimuli could account
for the lack of discrimination in the current experiment. Without lexical information as an
alternative cue for making distinctions, infants would have had to rely on the inconsistent
prosodic contours amongst the familiarization items. This would have been a particular issue for
the declarative familiarization group who may have had a difficult time identifying common
characteristics amongst the exemplars in the familiarization phase. The infants in the
interrogative familiarization group looked slightly longer at both sets of test trials compared to
the declarative familiarization group, which could provide support for this idea. However, there
was a less than 1 second difference in looking times between trial types in each familiarization
group (interrogative and declarative) and there was less than a 2 second difference between the
shortest looking time (declarative novel trials) and the longest looking time (interrogative
familiar trials). Thus, it seems unlikely that the acoustic variability may have masked
discrimination in the current experiment.
Since infants are not using prosodic information to distinguish between sentence types in
this experiment, could they be using word order/distributional information instead? This seems
9
More subjects were eliminated due to fussing than in any other experiment in this dissertation.
84
somewhat unlikely given the multiple distributional patterns infants could be attending to,
including word order, co-occurring syllables or even initial words (Mintz, 2003; Mintz, Newport
& Bever, 2002), and the possibility for overly fine-grained distinctions that was mentioned
earlier. Regardless of whether infants could use word order/distributional information to make
initial sentence-type distinctions, this information is nevertheless available for distinguishing
between sentence types. Experiment 5 will evaluate whether word order information alone is
sufficient for infants to distinguish between declaratives and polar interrogatives.
85
Experiment 5: Infants’ Ability to Distinguish between Declaratives and Polar
Interrogatives Based on Word Order Information Alone
Since the results of Experiment 4 were inconclusive, it is important to evaluate the other
source of information infants could use for this task. Word order provides a good alternative
given that many interrogatives have distinct word order properties such as auxiliary inversion
and wh- fronting (wh- words occur at the beginning of an utterance). For instance, in English the
canonical ordering is subject-verb-object, as in Anna likes chocolate. Interrogatives usually
disrupt canonical word orders. In English, for example, polar interrogatives (i.e., yes/no
questions) invert the subject and auxiliary verb, and insert an utterance-initial do if no auxiliary
is present (Sam can build sand castles becomes Can Sam build sand castles? and Sam builds
sand castles becomes Does Sam build sand castles?). In addition, in cases of do-support, the
main verb no longer agrees with the subject, as it does when no auxiliary is present (Sam
builds…). Moreover, in wh-questions, a grammatical phrase that normally occurs as an
argument or adjunct in the declarative form—for example, as an object noun phrase (NP)—is
realized as a wh-word (e.g., who, what, where, how, etc.) that occurs at the beginning of the
sentence (What
t
does Sam build t
i
?)
10
. Similar kinds of word order variations occur in many
other languages. Yet the question remains when infants begin to attend to and use these
structural properties to make sentence-type distinctions.
To determine whether infants can use word order cues alone (in this case auxiliary
inversion) to make sentence-type distinctions, Experiment 5 evaluated infants on sentences with
10
The t stands for trace, and indicates how the wh-word that shares its index (i) should be interpreted within the
structure of the sentence—in this case, as the direct object of climb.
86
flattened intonation contours and equalized utterance-final vowel length. While there is little
evidence that 7-month-olds attend to sequential word order information, 12-month-olds can track
word sequences and even represent word categories (Gómez & Gerken, 1999; Mintz, 2006; Shi
& Melançon, 2010). Therefore, Experiment 5a evaluated the role of word order alone in making
sentence-type distinctions at 12-months. In addition, despite lack of evidence regarding 7-
month-olds’ attention to word order information, this very attention to word order information
may at least partially explain Geffen and Mintz’s (2011) results. Thus, Experiment 5b evaluated
7-month-olds’ ability to use word order information alone.
Based on prior work (Gómez & Gerken, 1999; Mintz, 2006; Shi & Melançon, 2010) I
hypothesized that 12-month-olds could use word order information to distinguish declaratives
and interrogatives. Given the discrepancy between Geffen and Mintz (2011) and Experiment 1
in the current dissertation, I hypothesized that 7-month-olds might be able to use word order
information to distinguish between sentence types, although it is unlikely.
Experiment 5a
Methods
Subjects:
Thirty-one typically developing English-learning 12-month-olds were recruited from
county birth records, contacted first by letter, and then by phone or e-mail to schedule an
appointment. Data from five infants were excluded from the final analysis due to fussiness
11
(1),
failure to complete study (3), and moving out of camera range (1). Of the remaining 26 infants
(mean age 12.05 months, range: 11.23 – 12.77 months, 11 female), 13 were randomly assigned
11
If an infant cried during the experiment, or was so squirmy that the experimenter judged that they were not
attending to the stimuli, then the experimenter made a note that the child was fussy, and these data were not used.
87
to the declarative familiarization group and 13 to the interrogative familiarization group (see
below).
Stimuli & Design:
This experiment used the same sentences as Experiment 1 with several modifications.
Stimuli were resynthesized to have a monotone intonation contour and equalized utterance-final
vowel lengthening. To do this, the average pitch and final vowel duration for all 20 sentences
(declaratives and interrogatives) were calculated using Praat (Boersma & Weenink, 2011). The
average pitch was 280 Hz and the average final vowel duration was 134 ms. A Praat script
(Vicenik, 2009) automatically generated a new pitch tier for each sentence, with the value set to
280 Hz for the entire sentence, and resynthesized each sentence with the new F0 contour. A
second Praat script (Plichta, 2004) adjusted the final vowel duration of each sentence to 134 ms.
Coders initially marked vowel boundaries by examining the spectrogram and listening to the
corresponding waveform. Coders then calculated the average final vowel duration across all 20
sentences, selected the region containing the final vowel in each sentence, and changed each
sentence individually to match the average final vowel duration. Thus all sentences were similar
with respect to the acoustic cues that have been reported to differ between declaratives and polar
interrogatives (Bartels, 1999; Hedberg et al., 2004). The remaining reliable difference between
interrogatives and declaratives was the modified word order (i.e., the inversion of subjects and
auxiliaries) and, in some interrogatives, the appearance of a form of the auxiliary do.
Otherwise, the design of familiarization and test trials was the same as Experiment 1.
Although discussion of prosodic differences between declaratives and interrogatives
usually focuses on differences in the final portion of utterances (and has thus far in this
dissertation), there are also potential differences in the initial portion of the utterance as well.
88
There are two related issues: 1) word frequencies and 2) categories of words (function vs.
content).
Addressing the first issue, one potential confound that is related to sentence-initial
durational differences involves the pattern of frequent and infrequent words at the beginnings of
sentences. Infants have been shown to use patterns of frequent and infrequent syllables to
organize speech (Gervain, Nespor, Mazuka, Horie, & Mehler, 2008), attesting to their sensitivity
to lexical frequencies and patterns involving frequency. Since all the interrogatives began with
high frequency function words, but only some of the declarative familiarization utterances did,
and none of the declarative test items did, there is a concern that infants could base responses on
the familiarity of word-initial frequency patterns, as opposed to sentence type. While we cannot
rule out this possibility, it is unlikely. Almost half of the declarative familiarization items began
with high-frequency functors (the and there). Infants in the declarative familiarization group
were thus exposed to almost equal numbers of frequent-first and infrequent-first items. If their
responses are driven only by this characteristic of the stimulus, one would expect differences in
responses to sentence types to be driven by infants in the interrogative familiarization group.
Thus, although it is still possible that the distribution of frequent words will play a role in
determining infants' listening behavior, it likely would not fully account for the results.
Addressing the second issue, one cannot ignore the possibility of sentence-initial
rhythmic differences as well. This is because interrogatives typically start with short function
words—auxiliaries or wh-words—whereas declaratives can begin with longer open-class words.
As a result, the initial sequence of syllables or words in interrogatives might exhibit a noticeable
short-long rhythmic pattern, whereas declaratives might be more likely to start with a relatively
long duration syllable or word. (Indeed, 5 of the 8 declaratives in my familiarization material
89
began with an open-class word.) Since, infants have been shown to be highly sensitive to this
type of rhythmic difference (Yoshida, Iversen, Patel, Mazuka, Nito, Gervain, & Werker, 2010),
and since this study did not manipulate utterance-initial durations, it was important to determine
whether these utterance-initial rhythmic properties were confounded with the experimental
manipulation. This would be an issue if, for instance, familiarization items were more similar to
same type versus cross-type test items. To test for this, for all sentences, the duration difference
was calculated from the first syllable (or word) to the second syllable (or word) by subtracting
the duration of the second from the duration of the first position. Difference scores were
calculated across the first two syllables and, separately, across the first two words, since infants
may be sensitive to rhythmic patterns at either level of analysis. Mean difference scores were
computed for the two familiarization types and the two test sentence types (Table 14). Then for
each familiarization type, the difference scores were compared for the familiarization sentences
to each of the test sentence types, using two sample t-tests; these analyses were carried out with
respect to syllables and words. See Table 15.
The main results were that, for a given familiarization type, test sentence types did not
differ with respect to their similarity to the familiarization type. Specifically, neither declarative
nor interrogative test sentences differed from declarative familiarization sentences in a
statistically reliable way, and this was true at the syllable- and word-levels. In addition, both sets
of test sentences had duration difference scores which differed from interrogative familiarization
sentences in the same way; this pattern occurred at the word- and syllable-level. Thus, the
sentence-initial rhythmic properties of both types of test sentences differed from the interrogative
familiarization sentences in the same way, and neither test sentence type differed from
declarative familiarization sentences in sentence-initial rhythmic properties. Taken together,
90
these analyses indicate that, across familiarization conditions, rhythmic (dis)similarity of test
items and familiarization items was not confounded with sentence type familiarity—my critical
independent variable.
Table 14.
Duration Difference Scores Averaged Over Sentence Type and Familiarization Type, Calculated
at the Word Level and Syllable Level. Difference Scores Were Computed by Subtracting the
Duration of the Second Position from the Duration of the First Position.
____________________________________________________________________________
Level Sentence Type Familiarization/Test Mean Duration Diff (SE)
Word
Interrogative Familiarization 173 (69)
Declarative Familiarization -2 (88)
Interrogative Test -12 (0.20)
Declarative Test -27 (18)
Syllable
Interrogative Familiarization 71 (31)
Declarative Familiarization 54 (66)
Interrogative Test -12 (0.20)
Declarative Test -151 (32)
______________________________________________________________________________
91
Table 15.
T Statistics and Associated p Values for Comparisons of Duration Difference Scores between
Familiarization and Test Sentence. Comparisons Are Reported for Analyses at the Word and
Syllable Level.
___________________________________________________________________________
Analysis Level Familiarization Type Test Type t (df) p
Word
Interrogative Interrogative 2.68 (7) .03
Interrogative Declarative 2.81 (8) .02
Declarative Interrogative 0.11 (7) .92
Declarative Declarative 0.27 (7.5) .79
Syllable
Interrogative Interrogative 2.66 (7) .03
Interrogative Declarative 5.01 (3) .01
Declarative Interrogative -0.64 (7) .54
Declarative Declarative 1.33 (7.7) .22
Apparatus and Procedure:
Same as Experiment 1
MacArthur-Bates Communicative Developmental Inventories (MCDI)
I used the MCDI to roughly evaluate infants’ receptive and expressive language abilities,
a widely used measure of early communication development (MCDI; Fenson, Pethick, Renda,
92
Cox, Dale, & Reznick, 2000). Parents completed the MCDI short form: Level I (for ages 8 to 18
months), an 89-word vocabulary checklist, indicating words their children understood and words
their children produced.
Results
This experiment followed the same data exclusion procedures as Experiment 1. All
listening times under 2 seconds were removed, accounting for approximately 7 % of the trials
(15 out of 216). Within-subject analyses excluded five additional outliers defined as listening
times that were greater than (1.5*InterQuartileRange + 3rdQuartile) or less than (1stQuartile -
1.5*InterQuartileRange). After this process, all 26 infants still had at least six data points out of
a possible eight.
Shapiro-Wilk tests for normality indicated that infants’ responses to novel and familiar
trials were not normally distributed (W=.74, p<.001 for novel trials, W=.82, p<.001 for familiar
trials). My analyses thus compared median listening times to novel versus familiar trial types
(Gómez & Lakusta, 2004). Overall, infants listened longer to novel (Mdn = 9.0 s) versus
familiar sentence types (Mdn = 7.9 s). Twenty out of 26 infants listened longer to novel versus
familiar trials (p<.005, one-sided binomial test). To assess the statistical reliability of the
listening time difference, we calculated difference scores for each subject by subtracting the
median novel looking time by median familiar looking time (Figure 1). The median difference
score was 1.8 seconds, which was significantly different from chance (0) by a Wilcoxon signed
rank test (V = 285, p <.005), indicating that 12-month-olds discriminated the sentence types. To
test for an interaction of the discrimination ability with familiarization type, we used a Wilcoxon
rank sum test to compare median difference scores—subtracting familiar from novel listening
times for each infant—for infants familiarized to declaratives (M=1.9s) and infants familiarized
93
to interrogatives (M=3.0s) and found no significant difference (W = 90, p = 0.801). Thus, infants
discriminated declaratives and interrogatives equally whether familiarized to declaratives or
interrogatives.
Figure 10. Boxplot depicting difference scores: Median listening times to novel stimuli
subtracting out median listening times to familiar stimuli calculated by subject, including both
familiarization groups.
Given the hint of an interrogative preference in the earlier studies, it was important to
assess whether a similar preference was present in the current study (in addition to the novelty
preference reported above). Eleven out of 26 infants listened longer to interrogative trials versus
declarative trials, though this was not a greater number than expected by chance. To assess the
94
statistical reliability of the listening time difference, difference scores were calculated for each
subject by subtracting the mean declarative looking time from the mean interrogative looking
time. The median difference score was -0.2 seconds, which was not significantly different from
chance (0) by a Wilcoxon signed rank test (V = 160, p = 0.53), indicating that 12-month-olds did
not discriminate between declarative and polar interrogative trials. To test for an interaction of
the discrimination ability with familiarization type, a Wilcoxon rank sum test compared median
difference scores for infants familiarized to declaratives (M= 1.2 s) and infants familiarized to
interrogatives (M= -1.4) and found a significant difference (W = 31, p <0.01).
To evaluate the relationship between infants’ vocabulary (receptive and expressive) and
their sentence-type discrimination ability, three linear regression analyses explored the
relationship between each individual’s difference score (familiar listening time subtracted from
novel listening times) and 1) their receptive language score, 2) their expressive language score
and 3) their composite language score (receptive + expressive) as evaluated by the MCDI
(Fenson et al., 2000). None of the correlations was significant (p’s > 0.5).
Discussion
These results show that 12-month-olds can discriminate interrogatives from declaratives
solely on the basis of word-order patterns, when the utterances contain no distinguishing
prosodic information. This is the first demonstration, to my knowledge, that infants can rely on
word-order alone to make distinctions between sentence types, and demonstrates that by the time
they are a year old, infants represent something about the distributional differences between
declaratives and interrogatives. However, these results raise two further questions: 1) What is
the nature of the distributional properties that infants use to differentiate declaratives and
interrogatives, and 2) How do they initially learn to link those properties to sentence categories?
95
What Distributional Properties do Infants Associate with Interrogatives?
Regarding the first question, one possibility is that infants have heightened attention to
initial words. With the exception of Experiment 2, all of the interrogatives in this dissertation
were polar interrogatives, characterized by a sentence-initial auxiliary or do-support in the
absence of an auxiliary, followed by the subject. Children pay more attention to the beginning
(Newport, Gleitman & Gleitman, 1977) and end of sentences (Seidl & Johnson, 2006; Slobin,
1973) than the middle. Therefore, it is reasonable to consider that infants might use utterance-
initial words as a salient cue for forming initial sentence-type categories. Prior to coming into
the lab, infants may have learned to associate the initial occurrence of auxiliaries with
interrogatives
12
. However, in the current experiment, infants would have had to have made
generalizations beyond simply matching particular auxiliaries across familiarization and test
utterances, since interrogatives in the test set contained different auxiliaries from interrogatives
in the familiarization set. This means that either infants had previously learned to associate each
utterance-initial auxiliary with the interrogative sentence type, or they treated auxiliaries as a
category, and associated the utterance-initial occurrence of the auxiliary category with the
interrogative sentence type. However, while there is cross-linguistic evidence that 14-month-olds
have some representation of the functional categories (Höhle, Weissenborn, Kiefer, Schulz, &
Schmitz, 2004; Shi & Melançon, 2010), there is little evidence that infants under 12 months can
represent categories (Höhle et al., 2004; although see Shi & Lepage, 2008). The latter do,
nevertheless, have some expectations about the correct placement of determiners within noun-
12
When referring to the question sentence type in the context of infants’ knowledge, the term is intended only to
designate a formal differentiation from the canonical statement type. It is not intended to imply any semantic or
functional knowledge associated with the formal distinction.
96
phrases (Hallé, Durand & de Boysson-Bardies, 2008). If infants’ representation of auxiliary
verbs is similarly limited, it is unlikely that 12-month-olds could have begun to distinguish polar
interrogatives based on the category auxiliary occurring in sentence initial position.
A related possibility is that infants based their sentence differentiation on the inversion of
the auxiliary and the subject. That is, sequences like will she be, would you like, can you do, etc.,
constitute disruptions to the canonical sequencing of these words in declarative sentences, and
infants may have learned to associate these distributional patterns with the interrogative sentence
type. Accordingly, the critical information is not that the utterance begins with an auxiliary—a
category we just saw 12-month-olds probably do not represent—but rather that a word that often
occurs in first position {she, you, the, …} is now preceded by one of a small set of words, for
example, can you instead of you, or does the instead of the, and so on. Even without
representing categories, it is possible that infants have learned to associate these displacement
patterns with interrogatives on an item-specific level. In principle, these relational patterns might
be sufficient for infants to mark the interrogatives as formally non-canonical.
The findings do not provide evidence regarding the specific distributional patterns infants
used, and whether the representations involved grammatical categories, or were more item-
specific. However, they do show that whatever the representations are, they generalize over
numerous exemplars of polar interrogatives.
Regardless of what specific words or classes of words infants are attending to, the child’s
individual vocabulary (as assessed by the MCDI) does not seem to be related to their ability to
distinguish between statements and questions. However, the majority of words included on the
MCDI Short Form: Level I consisted of frequent nouns and action verbs. There were few
auxiliary verbs or other words commonly found in an utterance-initial position (e.g. determiners,
97
pronouns). Thus, perhaps it is not the general vocabulary which is important, but the more
specific vocabulary. In other words, if a child recognizes more words typically found in
utterance-initial position (e.g. auxiliaries, determiners, etc.), perhaps this would show a stronger
relationship to their ability to distinguish between sentence types, especially when only presented
with word order/distributional information.
How do Infants Learn to Associate Distributional Patterns with Interrogatives?
The second question concerns how infants could learn to associate distributional patterns
with different sentence types. One possibility is that infants use distributional information to
make initial sentence-type distinctions. As discussed in the previous section, this would require
that the representations involve grammatical categories, not individual lexical patterns. For
example, infants might differentiate utterances based on whether they begin with an auxiliary or
a noun phrase as opposed to whether they begin with can, is, are or do versus we or the. While
the distributional information necessary to make distinctions based on individual lexical items is
likely to exist (Mintz et al., 2002; Redington, Chater, & Finch, 1998), it is not immediately
obvious how such a mechanism could avoid making linguistically irrelevant distinctions as well.
For example, a mechanism that classifies sentences based on the initial sequence of words might
correctly determine that sentences that begin with {can, does, would, will} followed by {my, the,
his, a} are distinct from those that begin with {the, my, his, a} followed by {can, does, would
will}, but would the same mechanism also then erroneously determine that sentences of the latter
type are distinct from those beginning with {he, she, it} before {can, does, would, will}? While
12-month-olds are able to carry out some kinds of distributional analyses (e.g., Gómez &
Gerken, 1999; Mintz, 2006), other kinds of analyses, including this one, require a greater degree
of constraint specificity than has been reported in infants. The analyses required here to make
98
just the right kinds of distinctions and not erroneous ones would seem to require very specific
constraints, and it is not evident that infants are constrained in this way. On the other hand, if
infants differentiated sentence types based on some other source of information, then they could
later learn to associate the initial word-order patterns that correlate with each specific sentence-
type.
Prosody might be just such a source of information. As previously discussed, the final
pitch contour is an important perceptual cue for sentence-type discrimination (American English-
Săfárŏvá and Swerts, 2004; Australian English- Wales & Taylor, 1987; French- Vion & Colas,
2006). There is a wealth of evidence demonstrating infants’ sensitivity to prosodic information
from early on (e.g., Mehler et al., 1988; Moon et al., 1993; Nazzi et al., 1998), and some
evidence that infants exploit prosodic information to discriminate declaratives and interrogatives
(Frota et al, 2014; Geffen & Mintz, 2011; Soderstrom et al., 2010). Thus, while the prior results
suggest that prosody does play a role, it is also possible that infants in the prior experiments were
responding to non-prosodic properties of the material, a theory that is supported by the results of
the current experiment.
While infants may initially distinguish between declaratives and polar interrogatives,
another alternative to consider is that infants could first learn to differentiate wh-questions on
distributional grounds, and then apply their knowledge to polar interrogatives. Wh- questions
have several constraints which may make it easier for infants to make distinctions between wh-
questions and other sentence types. These limitations include 1) limited number of wh- words, 2)
limited initial phonemes (beginning with /w/ or /h/ in English, with comparable phonological
similarities in question operators in other languages), 3) limited number of positions within an
utterance (generally restricted to a sentence initial position, with the exception of in situ wh-
99
words) and 4) limited set of immediately following words (e.g. auxiliaries). This combination of
factors could make it easier for infants to distinguish wh- questions from other sentence types,
especially declaratives. A mechanism that responded to such a strictly regulated set of cues
would likely make a different set of mistakes than one relying on the distributional properties of
polar interrogatives (in the absence of prosodic information). The correlated distributional
properties—e.g., subject-auxiliary inversion and do-support—could then be detected and used as
criteria for classifying polar interrogatives together with wh-questions. Interestingly, if infants
generalize from wh- questions to polar interrogatives, the prosodic correlates of polar
interrogatives may not be necessary in early sentence-type discrimination. This could be useful
given the prosodic variability found in declaratives. Of course, in a modified version of this
account, the ability to discriminate polar interrogatives could develop independently of the
ability to discriminate wh-questions, and the former could rely on prosodic information.
The current experiment demonstrates that infants are sensitive to word
order/distributional information for making sentence-type distinctions. This may explain the
discrepancy in results between Geffen and Mintz (2011) and Experiment 1 in the current
dissertation. If 7-month-olds are at least partially relying on word order/distributional
information to distinguish between declaratives and polar interrogatives, the limited number of
initial words in Geffen and Mintz (2011) may have made this distinction a little easier. To test
this hypothesis, Experiment 5b evaluated 7-month-olds on resynthesized versions of the stimuli
from Geffen and Mintz (2011) using the same modifications as Experiment 5a.
Experiment 5b
To further examine infants’ sensitivity to word order information, Experiment 5b
evaluated 7-month-olds’ ability to distinguish flattened versions of the sentences from Geffen
100
and Mintz (2011). I hypothesized that seven-month-olds would not be able to distinguish
between sentence types.
Methods
Subjects:
Twenty-nine typically developing English-learning 7-month-olds were recruited from
county birth records, contacted first by letter, and then by phone or e-mail to schedule an
appointment. Data from three subjects were excluded from the final analysis due to fussiness
(1), four or fewer data points of a possible eight (1), and less than 50% daily English input (1).
Of the remaining 26 infants (mean age 7.07 months, range: 6.73 – 7.5 months, 9 female), 13
were randomly assigned to the declarative familiarization group and 13 to the interrogative
familiarization group (see below).
Stimuli & Design:
Stimuli consisted of nine declaratives and nine polar interrogatives. Familiarization
consisted of seven exemplars of one type of sentence (e.g. declaratives) with the remaining two
sentences of each type in the test phase. Critically, all test trials contained sentences that did not
appear in the familiarization phase. Stimuli were matched on number of syllables and duration. A
full list of stimuli can be found in Table 16.
Stimuli were recorded by a female native English speaker in an infant-directed register,
which exaggerates prosodic dimensions and is intrinsically appealing to infants (Fernald, 1984).
This experiment made the same modifications to stimuli as Experiment 5a, resulting in a
monotone intonation contour and equalized utterance-final vowel lengthening. The average
pitch was 260 Hz and the average final vowel duration was 183 ms.
Otherwise, the design of familiarization and test trials was the same as Experiment 1.
101
Table 16.
Familiarization and Test Sentences for Both Groups.
Statement Familiarization Question Familiarization
Familiarization
We are having French toast for
breakfast.
We're going to the park today.
The cat is playing with the ball.
Your sweater is very pretty.
Let's go outside and blow bubbles.
We just bought a pair of new shoes.
We saw elephants at the zoo.
Did you see a funny movie?
Can we go to the beach today?
Do you want some watermelon?
Can I have a sugar cookie?
Do you want to read a story?
Do you want to go down the slide?
Can I play with the cute brown
puppy?
Test
We had turkey and grapes for lunch.
The pirates hid the treasure chest.
Can you bring me the teddy bear?
Would you like a tuna sandwich?
Apparatus and Procedure:
Same as Experiment 1
Results
This experiment followed the same analysis procedures as Experiment 1. All listening
times under 2 seconds were removed, accounting for approximately 7% of the trials (15 out of
208). Within-subject analyses excluded 17 additional outliers defined as listening times that
were greater than (1.5*InterQuartileRange + 3rdQuartile) or less than (1stQuartile -
102
1.5*InterQuartileRange). After this process, all 26 infants still had at least six data points out of
a possible eight.
The goal of this experiment was to determine whether Geffen and Mintz’s (2011) results
were due to 7-month-olds’ attention to word order/distributional information.
A mixed effects linear regression model (with subject as a random effect) evaluated
infants’ sentence-type discrimination abilities, comparing the mean looking time on
Familiarization Type (declarative vs. interrogative), Trial Type (novel vs. familiar) and their
interaction. The Mood’s test was non-significant (p > 0.05), indicating that the scale of the two
groups was not significantly different.
There were no significant main effect (p’s >0.1), although with a larger sample size
13
statistical significance may be achieved for the main effect of trial type (β = -5.5, p = 0.12). See
Table 17. Regardless of familiarization group, infants looked longer at novel (M= 13.0, SD =
11.9) versus familiar trials (M = 9.1, SD = 3.8). There was also no significant interaction (p >
0.1). Infants in the declarative familiarization group had a mean of 11.9 s (SD = 9.4 s) for novel
trials and a mean of 9.6 s (SD = 3.3 s) for familiar trials. For the interrogative familiarization
group, the novel trials had a mean of 14.0 s (SD = 14.3 s) and the familiar trials had a mean of
8.5 s (SD = 4.3 s). Thus, 7-month-olds did not distinguish between declaratives and polar
interrogatives using word order information alone. See Figure 11.
13
A Monte Carlo simulation was conducted to provide an estimate of empirical power for each of the effects
presented. Achieved power for the main effects of Familiarization, Sentence Type, and their interaction were 13.1,
73.4, and 24.5 % respectively with the current N of 26 subjects. Additional simulations were run to determine the
sample sizes need to detect each of these effects with >80% power. An N of 350 subjects was needed to provide an
empirical power of 80.8% for the main effect of Habituation. An N of 33 was needed to detect the interaction effect
at a power of 80.9%. A sample N of 135 would be needed to detect a main effect of Sentence Type at > 80% power.
103
Given the hint of an interrogative preference in the earlier studies, it was important to
assess whether a similar preference was present in the current study. Sixteen out of 25 infants
listened longer to interrogative trials versus declarative trials, though this was not a greater
number than expected by chance. To assess the statistical reliability of the listening time
difference, difference scores were calculated for each subject by subtracting the mean declarative
looking time from the mean interrogative looking time (Figure 12). The mean difference score
was -1.6 seconds, which was not significantly different from chance (0) by a one sample t-test
(t(25) = -0.64, p = 0.74), indicating that 7-month-olds did not discriminate between sentence
types. To test for an interaction of the discrimination ability with familiarization type, a Welch
two sample t-test compared median difference scores for infants familiarized to declaratives (M
= 2.3 s, SD = 9.2 s) and infants familiarized to interrogatives (M= -5.5 s, SD= 14.7 s) and found
no significant difference (t(20.1) = -1.62, p = 0.12). Thus, infants did not discriminate between
declaratives and interrogatives regardless of whether they were familiarized to declaratives or
interrogatives.
104
Table 17.
Beta Coefficient and Standard Error by Factor
Factor β SE p
Familiarization -2.1 3.5 0.56
Trial Type -5.5 3.4 0.12
Familiarization x Trial Type 3.2 4.8 0.52
__________ ___________________________________________________________________
Figure 11. Mean looking time duration for novel and familiar trials as a function of
familiarization group for 7-month-olds in Experiment 5b. Error bars represent +1 standard error
of the mean.
105
Figure 12. Mean difference score (Interrogative-Declarative Trial) regardless of familiarization
type for 7-month-olds in Experiment 5b. Positive score indicates an interrogative preference,
negative score indicates a declarative preference. Error bars represent +1 standard error of the
mean.
Discussion
As previously discussed, the primary limitation of Geffen and Mintz (2011) was the
confound in stimuli similarity between familiarization and test items of the same sentence type.
One suggestion was that the results were due to infants creating one category for sentences that
began with can and did and another for sentences that began with we.
However, infants did not show reliable looking time differences between test trial types
in the current experiment. This suggests several possibilities: 1) infants may require further
experience in order to learn the specific word-order patterns, 2) at seven months, some infants
may still require a combination of distributional and prosodic information to make sentence-type
distinctions or 3) some 7-month-olds may be attending to distributional information while others
primarily rely on prosodic information.
106
Addressing the first possibility, Experiment 5a seems to support this idea, with 12-month-
olds using word order information where 7-month-olds did not. It remains to be seen whether
infants can use word order alone to make distinctions at a slightly younger age (e.g. nine
months).
Addressing the second possibility, while infants may rely on a combination of
distributional and prosodic cues, Experiment 4 seems to argue against the necessity of prosodic
information alone. Infants did not show reliable looking time differences between novel and
familiar trials, suggesting that they did not distinguish between sentence types when only
prosodic information was available. However, Experiment 1 demonstrated a trend towards
discrimination, for 7- and 9-month-olds individually, and significant discrimination (both a
familiarity and an interrogative preference) for the combined age group. These results suggest
that seven-month-olds may be able to use prosody, at least in combination with distributional
information.
The third possibility may explain the results throughout the dissertation. If some infants
are preferentially attending to distributional information, while others are preferentially attending
to prosodic information, these two groups could essentially cancel out any discrimination we
might otherwise observe. Future research could evaluate this question using individual analyses
on each subject.
While it is still possible that infants in Geffen and Mintz (2011) were using word
order/distributional information to distinguish between sentence types, the current study does not
provide support for this claim.
107
General Discussion
This dissertation provides evidence that infants can distinguish between declaratives and
(polar) interrogatives and evaluates when and how they begin to make these distinctions.
The clear interrogative preference demonstrated by the combined age group (combined
with the trend demonstrated by the individual age groups) provides evidence that infants can
make sentence-type discrimination when presented with the full speech signal. This dissertation
also provides evidence that this interrogative preference is not due to a single set of
circumstances. These results replicated Soderstrom, Ko & Nevzorova’s (2011) findings, and
extended them with tightly controlled age groups, a different method (HPP) and a different
interrogative type (polar interrogatives).
One potential explanation for this interrogative preference is the greater frequency of
interrogatives in infants’ daily input. Newport (1977) found that infants hear more interrogatives
than declaratives (44% and 30% respectively) in daily speech and thus, could be more sensitive
to interrogatives, which could bias infants’ perception. Unfortunately, information about daily
input to the subjects in the current studies was not collected, making it difficult to draw
conclusions about the validity of this hypothesis.
The interrogative preference could also result from the combination of prosodic and word
order/distributional information available in the full speech signal. Infants demonstrated an
interrogative preference when tested on the full speech signal (Experiment 1), but not when cues
were presented in isolation: prosody alone (Experiment 4), or word order alone in 7-month-olds
(Experiment 5b). Soderstrom et al. (2010) similarly found an interrogative preference only when
infants were exposed to the full speech signal. These results suggest that both prosody and word
order may be necessary for infants to demonstrate an interrogative preference.
108
When Do Infants Begin to Distinguish between Declaratives and Polar Interrogatives?
Experiment 1 replicated Geffen and Mintz (2011) while addressing the lexical confound
between familiarization and test items of the same sentence type versus cross type. Although
infants did not show the expected discrimination pattern – novel versus familiar – there was a
trend towards an interrogative preference for 7- and 9-month-olds. Nine-month-olds in
particular showed a strong trend towards discrimination, with a more than 2 second difference in
looking time to interrogatives compared to declaratives. In addition, the combined age group
demonstrated both a familiarity and an interrogative preference, which seem to be driven by the
interrogative familiarization group. Collectively, these results suggest that infants may be able to
distinguish between declaratives and polar interrogatives by seven months, which is in line with
Geffen and Mintz’s (2011) findings.
What Types or Combinations of Cues Do Infants Use to Distinguish between Sentence Types?
It was also important to evaluate how infants distinguish between declaratives and polar
interrogatives. One hypothesis was that infants would make initial sentence type distinctions
using prosody, allowing them to learn the correlated word order/distributional properties
characteristic of declaratives and polar interrogatives. However, despite the availability of
prosodic information to distinguish declaratives from polar interrogatives (although not from wh-
questions; Experiment 2), seven-month-olds did not demonstrate discrimination based on
prosody alone (Experiment 4). These results do not provide support for the hypothesis that
infants use prosody to make initial sentence-type distinctions.
However there is so much variability in declaratives, it is hard to conceive of prosody as
useful for classifying declaratives in English. For example, the experimental stimuli in this
dissertation demonstrated greater acoustic variability in max F0 and pitch excursion for the
109
declarative stimuli versus the interrogative stimuli. It is possible that the distinguishing
properties seen in Experiment 2 might be too subtle to detect given the overall variability in the
declarative stimuli. Declaratives may be perceived only as not polar interrogatives. This may
explain why so many labs (e.g. Frota et al., 2014; Soderstrom et al., 2010) find distinctions
between polar interrogatives and everything else. Infants would then have to rely on an alternate
source of information for making initial sentence-type distinctions.
Word order provides a likely alternative. Experiment 5a demonstrated that by 12 months,
infants can use word order information alone to distinguish between declaratives and polar
interrogatives, although they did not demonstrate discrimination at seven months (Experiment
5b). These results could suggest maturational differences in attention to distributional
information. However, it is also possible that a conclusive finding of discrimination may have
been masked by individual differences within the data. For example, the interrogative
familiarization group had a mean looking time of 14.0 s (SD = 14.3 s) for novel trials and 8.5 s
(SD = 4.3 s) for familiar trials. Such a large standard deviation is usually a sign of wide variation
in performance, in this case, looking times. This could suggest individual differences in
attention to cues for distinguishing between sentence types, and word order/distributional
information more specifically – or simply a lack of attention. Unfortunately the results do not
provide a clear answer. The latter option seems somewhat unlikely given the large looking time
demonstrated by 7-month-olds in the interrogative familiarization group. Perhaps the answer lies
somewhere in between, where some infants listened for a very long time, and some infants paid
almost no attention at all. Combined, these results suggest age-related changes in sentence-type
discrimination and support the hypothesis that older infants can distinguish between sentence
types based on word order alone (Experiment 5a).
110
The results of Experiments 5a and 5b raise several questions: 1) How well do these
results generalize to other types of interrogatives and 2) Do infants actually use word
order/distributional information to make initial distinctions, allowing them to learn the correlated
prosodic properties that characterize each sentence type?
Addressing the first question, polar interrogatives and wh- questions occur in roughly
equal proportions in infant-directed speech (Newport, 1977), so it is important to determine
whether infants can distinguish equally well between declaratives and both types of
interrogatives. Interrogatives provide a rich and potentially confusing source of information
about sentence types given the variety (e.g. polar interrogatives, wh- questions) and the lack of
consistency in cues (Bartels, 1999; Gunlogson, 2001). Wh- questions are more prosodically
similar to declaratives in both adult- (Bartels, 1999; Hedberg et al., 2004) and infant-directed
speech (Experiment 2), but have similar underlying structural properties to polar interrogatives.
Thus, a clear follow-up would be a replication of Experiment 5a with declaratives and wh -
questions.
Addressing the second question, as previously discussed, if infants do make initial
sentence-type distinctions using distributional information, it is possible they begin by
distinguishing between declaratives and wh- questions. Although this may seem like a surprising
alternative, wh- questions have a number of constraints (including fewer possible initial
phonemes and fewer possible initial words than a polar interrogative) which may be easier to
learn than polar interrogatives. This combination of factors could make it easier for infants to
distinguish wh- questions from other sentence types, especially declaratives. It is still possible
that the ability to discriminate polar interrogatives could develop independently of the ability to
discriminate wh-questions, and the former could rely on prosodic information.
111
Theoretical implications
The outcomes of these experiments could make several important contributions.
Broadly, it would contribute to our understanding of infants' early linguistic representations, the
cues infants attend to in order to construct these representations, and whether these
representations change within the first year. This project could add significant new knowledge
to the types of distinctions pre-verbal infants can make about grammatically different sentence
types, which could have broader impacts in theories of child language acquisition. These
findings would be of particular importance to theories of early syntactic development: As
previously mentioned, distinguishing between declaratives and interrogatives is an important
ability for English-learners. Identifying an utterance as an interrogative is critical for learning
the syntax of interrogatives, but more broadly, distinguishing sentence types could be important
for avoiding errors in syntax acquisition that might occur if interrogatives and declaratives were
analyzed as the same type of utterance (Pinker, 1984). Thus, distinguishing declaratives and
interrogatives early in language development from non-syntactic information could facilitate the
acquisition of syntax. Additionally, by mapping out these abilities and sensitivities in typically
developing children, these experiments could impact our understanding of, and lead to the
development of a method for assessing individual differences in early language perception for
individuals at risk for language impairments.
Limitations
One potential limitation of this dissertation was the greater acoustic variability in
declaratives compared to interrogatives.
14
Experiment 1 aimed to replicate and expand Geffen
14
This does not apply to the stimuli used in Experiment 5b.
112
and Mintz’s (2011) findings using stimuli with greater initial-word variability. While these
experiments controlled for a number of factors (number of syllables, number of obstruents, no
more than two sentences beginning with the same initial word), they did not control for the
acoustic variability between sentence types. As previously discussed, there was no acoustic
variability between declaratives and polar interrogatives in the final syllable but a sentence-level
analysis demonstrated significantly greater variability in max F0 and pitch excursion for
declaratives. In addition, the difference in overall duration approached significance, with
declaratives not only longer than interrogatives, but also showing greater variability between
exemplars. This greater acoustic variability for declaratives could partially explain the
interrogative preference demonstrated in the majority of experiments, although there are several
other possible explanations which were detailed above. However, when presented with only
prosodic information, which could have served to highlight this acoustic variability in the
absence of lexical cues, infants did not demonstrate an interrogative preference (or
discrimination at all). This suggests that this variability does not solely account for the
interrogative preference. Future work is needed to replicate these studies with more constrained
stimuli.
Another potential limitation is that while HPP is a popular method used in infant studies,
it may not be the best method for assessing discrimination. It is difficult to tell if infants are
distinguishing between two sets of stimuli when they may not have become familiar with one set
of stimuli in the first place. Future work could replicate these studies using another paradigm
like habituation, to confirm that the results are not due to an inadequately sensitive method.
Future Directions
113
To my knowledge, these are among the first studies of their kind to examine a variety of
factors infants may be using to distinguish between declaratives and interrogatives. Soderstrom
et al. (2010) provided some preliminary evidence that infants may use prosodic information to
distinguish between declarative statements and declarative questions. In addition, Frota et al.
(2014) demonstrated that European Portuguese-learning infants can distinguish between
declaratives and interrogatives using only prosodic information.
While Frota et al.’s (2014) work expands the infant sentence-type discrimination
literature to an additional language, more research is needed examining infants’ capabilities
cross-linguistically. While this dissertation and Soderstrom et al. (2011) suggest that, in English
at least, younger infants may benefit from the combination of prosodic and word
order/distributional information, work remains to evaluate whether other languages rely on the
same or different combinations of cues for making these distinctions. For example, how would
these distinctions work in languages like Japanese, which places question markers in a sentence-
final position, or tonal languages like Chinese, where prosodic information may play a stronger
role given the lexical function of tones for distinguishing grammatical meaning?
The results of Experiment 4 in the current dissertation were inconclusive as to the role of
prosody in making sentence-type distinctions. Therefore further research is needed to address
this issue. One possibility for evaluating infants’ ability to use prosodic information to make
sentence-type distinctions is to test infants on a foreign language (e.g. Russian). While it would
be necessary to ensure that none of the subjects had prior exposure to the foreign language, it
could provide a good test of infants’ sensitivity to prosodic information without having to
remove lexical content.
114
A related issue is that while these studies tested infants’ ability to categorize utterances
based on their prosodic characteristics in a way that is relevant for distinguishing declaratives
from interrogatives, they did not test whether 7-month-olds, outside of the experimental settings,
treat these prosodic patterns in any special way. That is, they did not test whether 7-month-old
English learners have a representation, based on experience, of a prosodic pattern characteristic
of polar interrogatives (e.g., final pitch rise). Future work could evaluate whether the results of
these experiments are experiment-internal or generalizable by replicating Experiments 1 and 4
and overlaying an unnatural prosodic contour, one that is not associated with any sentence type
(in English). If infants have already begun to recognize some of the prosodic contours that
characterize sentence types such as declaratives and interrogatives in English, one might expect
that infants would have trouble learning the unnatural contours compared to the familiar
contours.
Another area of future research could evaluate infants’ response to different types of
interrogatives. Although most interrogatives are characterized by AUX-inversion and the final
prosodic rise, there are exceptions to both of these rules (e.g. wh- questions end with a final fall
and declarative questions do not demonstrate inversion). Future research could evaluate two
questions: 1) How do polar interrogatives help infants learn to recognize wh- questions as
belonging to the same category? and 2) Do infants show the same pattern of response to all types
of interrogatives? In other words, can infants distinguish equally well between declaratives and
various interrogative types (e.g. polar interrogatives, wh- questions, declarative questions), and if
so, does this ability develop at the same time?
With regard to the first question, as previously discussed, wh- questions often include the
same word order sequences as polar interrogatives (e.g. What did you have for lunch?). Polar
115
interrogatives may help infants recognize the word order patterns that signal an interrogative, and
help infants recognize that polar interrogatives and wh- questions are part of the same category.
Newport et al. (1977) found that children whose mothers use more auxiliaries in polar
interrogatives also used more auxiliaries in other types of sentences. Infants pay special
attention to utterance-initial words, so polar interrogatives provide a good opportunity for infants
to begin attending to auxiliaries. Once learners begin to attend to auxiliaries in polar
interrogatives, they can begin using them in other types of interrogatives, such as wh- questions,
where the auxiliary appears in an utterance-medial position. Our lab is currently in the process
of replicating Experiment 5a comparing declaratives and wh-questions.
The second question addresses the inconsistency of cues between different types of
interrogatives. For instance, adults often produce wh- questions with final falling or flat
intonation, similar to declaratives (Bartels, 1999; Hedberg et al., 2004). Experiment 2
demonstrated that infant-directed speech follows the same prosodic patterns as adult-directed
speech for declaratives, polar interrogatives and wh- questions. While prosodic characteristics
may unite declaratives and wh-questions, lexical and distributional information serves to unite
polar interrogatives and wh-questions. Thus, infants cannot rely solely on one type of cue for
distinguishing between declaratives and all types of questions. If infants are in fact relying on
the final pitch of an utterance to categorize sentences as either declaratives or interrogatives, this
could lead them to incorrectly categorize wh- questions as declaratives rather than interrogatives.
Thus, infants may show different levels of discrimination depending on the type of interrogative
being tested. This is supported by the Soderstrom et al. (2010) studies, which demonstrated that
infants do show different response patterns for different types of interrogatives (declarative
questions and wh- questions). Infants were able to distinguish between declarative statements
116
and declarative questions in both full speech and low-pass filtered conditions. Infants did not
show a significant difference in looking times between novel and familiar trials when tested on
wh- questions, although infants familiarized with interrogatives showed a trend towards a
novelty preference, suggesting they may be able to distinguish between declaratives and wh-
questions. However, given the wide age range, it is unclear if all subjects would be able to make
this distinction or only some (perhaps the older infants). Future research could evaluate infants’
ability to distinguish between declaratives and wh- questions, using a narrower age range of
infants, perhaps 7- or 9-month-olds. If it turns out that infants are incorrectly grouping wh-
questions with declaratives when only presented with prosodic information that may indicate that
infants rely on both prosodic and word order cues to accurately categorize various types of
interrogatives and declaratives, as was suggested by Experiment 1 in the current dissertation.
All of the current experiments are aimed at examining discrimination abilities in infants,
yet none of them address the question of whether infants understand the distinctions between
declaratives and interrogatives. A future experiment could use a functional paradigm to
determine when infants (or toddlers) begin to understand this distinction between sentence types,
although it is not immediately clear how this can be done. The best approach might be to
evaluate interrogative types individually, rather than studying all types of interrogatives
simultaneously given the structural and prosodic differences between interrogative types (e.g.
polar interrogatives vs. wh- questions vs. declarative questions). For example, polar
interrogatives generally require a response from the listener (with the exception of rhetorical
questions), whereas the related declarative does not. One could present infants with scenarios in
which an individual responds with yes or no to an interrogative versus another scenario where an
117
individual responds to a declarative and evaluate whether infants look longer at one or the other.
This would get at some aspects of the pragmatics of interrogative use.
Conclusion
The studies in this dissertation were designed to evaluate when and how infants
distinguish between declaratives and interrogatives. The results of this dissertation provide a
hint that infants may be able to distinguish between declaratives and polar interrogatives by
seven months when presented with the full speech signal. In addition, by one year, infants are
sensitive to structural differences between declaratives and polar interrogatives. However,
further studies are needed to evaluate the generalizability of these results to other interrogative
types (e.g. wh- questions) as well as the time course of this ability. While there is no evidence
suggesting that infants can use word order information earlier than a year, there is evidence that
infants are sensitive to this information at two months (in combination with prosodic
information; Mandel et al., 1996). It is important to determine how infants hone their sensitivity
to word order information alone
In summary, the current experiments yield insights into the time course and mechanisms
underlying infants’ ability to differentiate sentence types, a critical ability in language
acquisition. Despite the lack of significant differences between the two familiarization groups
for the 7- or 9-month-olds when presented with the full speech signal, both age groups
demonstrated a trend toward an interrogative preference. In addition, the combined age group
demonstrated a significant interrogative preference, suggesting that infants can distinguish
between declaratives and polar interrogatives. However, seven months may be too young for
infants to distinguish between declaratives and polar interrogatives when presented with isolated
cues. Seven-month-olds did not demonstrate discrimination when presented with prosodic
118
contours alone (Experiment 4) or word order information alone (Experiment 5b). It remains to be
seen whether a change in methodology or stimuli would demonstrate discrimination at seven
months. In addition, infants are sensitive to structural differences between declaratives and polar
interrogatives by 12 months. These results suggest a shift in discrimination ability between
seven and twelve months. Future research is needed to evaluate the generalizability of these
results not only to novel utterances of the same types (e.g. declaratives and polar interrogatives)
but to other types of interrogatives (e.g. declarative questions, wh- questions) as well.
119
References
Abercrombie, D. (1967). Elements of general phonetics. Chicago: Aldine.
Bartels, C. (1999). The intonation of English statements and questions: a compositional
interpretation. New York & London, Garland Publishing Inc.
Best, C., Levitt, A., & McRoberts, G. (1991). Examination of language-specific influences in
infants’ discrimination of prosodic categories. In Proceedings of the 12th International
Congress of Phonetic Sciences (Vol. 4, pp. 162-165).
Boersma, P. & Weenink, D. (2011). Praat: Doing phonetics by computer (version 5.2.23)
[Computer program]. Available from www.praat.org.
Bosch, L., & Sebastián-Gallés, N. (1997). Native-language recognition abilities in 4-month-old
infants from monolingual and bilingual environments. Cognition,65(1), 33-69.
Brent, M. & Siskind, J. (2001). The role of exposure to isolated words in early vocabulary
development. Cognition, 81, 31-44.
Christophe, A., & Morton, J. (1998). Is Dutch native English? Linguistic analysis by 2-month-
olds. Developmental Science, 1, 215-219.
Curtin, S., Mintz, T. & Christiansen, M. (2005). Stress changes the representational landscape:
Evidence from word segmentation. Cognition, 96, 233-262.
Fenson, L., Pethick, S., Renda, C., Cox, J., Dale, P., & Reznick, J. (2000) Short form of
versions of the MacArthur Communicative Development Inventories Applied
Psycholinguistics, 21, 95-115.
Fernald, A. (1984). The perceptual and affective salience of mothers’ speech to infants.
In L. Feagans, C. Garvey, & R. Golinkoff (Eds.), The origins and growth of
communication. New Brunswick, NJ: Ablex.
120
Fernald, A., & Kuhl, P. (1987). Acoustic determinants of infant preference for motherese
speech. Infant Behavior and Development, 10(3), 279-293.
Fernald, A., Taeschner, T., Dunn, J., Papousek, M., de Boysson-Bardies, B., & Fukui, I. (1989).
A cross-language study of prosodic modifications in mothers’ and fathers’ speech to
preverbal infants. Journal of child language, 16(3), 477-501.
Fisher, C. & Tokura, H. (1996). Acoustic cues to grammatical structure in infant directed
speech: Cross linguistic evidence. Child development, 67, 3192-3218.
Frota, S., Butler, J., & Vigário, M. (2014). Infants' Perception of Intonation: Is It a Statement or a
Question?. Infancy, 19(2), 194-213.
Geffen, S. (2010). Can infants discriminate between declaratives and interrogatives?
(Unpublished Master’s thesis). University of Southern California, Los Angeles.
Geffen, S., & Mintz, T. (2011). Seven-month-olds' Discrimination of Statements and Questions.
In 36th Boston University Conference on Language Development (BUCLD), Boston
University.
Gerken, L., Jusczyk, P., & Mandel, D. (1994). When prosody fails to cue syntactic
structure: 9-month-olds’ sensitivity to phonological versus syntactic phrases. Cognition,
51, 237-265.
Gervain, J., Nespor, M., Mazuka, R., Horie, R., & Mehler, J. (2008). Bootstrapping word order
in prelexical infants: A Japanese-Italian cross-linguistic study. Cognitive Psychology,
57(1), 56–74. doi:10.1016/j.cogpsych.2007.12.001.
Gómez, R. & Gerken, L. (1999). Artificial grammar learning by one-year-olds leads to specific
and abstract knowledge. Cognition, 70, 109-135.
Gómez, R. & Lakusta, L. (2004). A first step in form ‐based category abstraction by 12 ‐
121
month ‐old infants. Developmental science, 7(5), 567-580.
Gósy, M. & Terken, J. (1994). Question marking in Hungarian: Timing and height of pitch
peaks. Journal of phonetics, 22, 269 – 281.
Gunlogson, C. (2001). True to form: Rising and falling declaratives as questions in
English. Ph.D. Dissertation, University of California, Santa Cruz, United States --
California. Retrieved September 10, 2011, from Dissertations & Theses: Full Text.
(Publication No. AAT 3032261).
Hallé, P., Durand, C., & de Boysson-Bardies, B. (2008). Do 11-month-old French infants
process articles? Language and Speech, 51(Pt 1-2), 23–44.
Hay, J. & Saffran, J. (2012). Rhythmic grouping biases constrain infant statistical
learning. Infancy, 17(6), 610-641.
Hedberg, N., Sosa, J. & Fadden, L. (2004). Meanings and configurations of questions in
English. Proceedings of speech prosody 2004, Nara (Japan), 309 – 312.
Hirsh-Pasek, K., Kemler Nelson, D., Jusczyk, P., Cassidy, K., Druss, B. & Kennedy, L.
(1987). Clauses are perceptual units for young infants. Cognition, 26(3), 269-286.
Höhle, B., Weissenborn, J., Kiefer, D., Schulz, A., & Schmitz, M. (2004). Functional elements in
infants' speech processing: The role of determiners in the syntactic categorization of
lexical elements. Infancy, 5(3), 341-353.
Houston, D., Horn, D., Qi, R., Ting, J., & Gao, S. (2007). Assessing speech discrimination in
individual infants. Infancy, 12, 119-145.
Johnson, E., & Jusczyk, P. (2001).Word segmentation by 8-month-olds: When
speech cues count more than statistics. Journal of Memory & Language, 44(4), 548–
567.
122
Jusczyk, P., & Aslin, R. N. (1995). Infants′ detection of the sound patterns of words in fluent
speech. Cognitive Psychology, 29(1), 1-23.
Jusczyk, P. Cutler, A. & Redanz, N. (1993). Infants preference for the predominant
stress patterns of English words. Child Development, 64, 675-687.
Jusczyk, P., Hirsh-Pasek, K., Kemler Nelson, D., Kennedy, L., Woodward, A. & Piwoz, J.
(1992). Perception of acoustic correlates of major phrasal units by young infants.
Cognitive Psychology, 24, 252 – 293.
Kemler Nelson, D., Hirsh-Pasek, K., Jusczyk, P. W., & Wright Cassidy, K. (1989). How the
prosodic cues in motherese might assist language learning. Journal of Child Language,
16, 55-68.
Kemler Nelson, D., Jusczyk, P., Mandel, D., Myers. J., Turk, A., & Gerken, L. (1995).
The Head-Turn Preference Procedure for Testing Auditory Perception. Infant Behavior
and Development, 18, 111–116.
Lieberman, P. (1967). Intonation, perception, and language. Cambridge, MA: MIT Press.
MacWhinney, B. (2000). The CHILDES Project: Tools for analyzing talk. Third Edition.
Mahwah, NJ: Lawrence Erlbaum Associates.
Makarova, V. (2007). The effect of pitch peak alignment on sentence type identification in
Russian. Language and speech, 50, 385-422.
Mampe, B., Friederici, A. D., Christophe, A., & Wermke, K. (2009). Newborns' cry melody is
shaped by their native language. Current biology, 19(23), 1994-1997.
Mandel, D, Kemler Nelson, D, & Jusczyk, P. (1996). Infants remember the order of words in a
spoken sentence. Cognitive Development,11(2), 181-196.
Mehler, J., Jusczyk, P., Lambertz, G., Halsted, N., Bertoncini, J., & Amiel-Tison, C. (1988). A
123
precursor of language acquisition in young infants. Cognition, 29, 143-178.
Mintz, T. (2003). Frequent frames as a cue for grammatical categories in child directed speech.
Cognition, 90, 91-117.
Mintz, T. (2006). Finding the verbs: distributional cues to categories available to young
learners. In K. Hirsh-Pasek & R. M. Golinkoff (Eds.), Action Meets Word: How
Children Learn Verbs (pp. 31-63). New York: Oxford University Press.
Mintz, T., Newport, E. & Bever, T. (2002). The distributional structure of grammatical
categories in speech to young children. Cognitive Science, 26, 393-424.
Moon, C., Cooper, R., & Fifer, W. (1993). Two-day-olds prefer their native
language. Infant Behavior and Development, 16, 495-500.
Nazzi, T., Bertoncini, J., & Mehler, J. (1998). Language discrimination by newborns:
Toward an understanding of the role of rhythm. Journal of Experimental Psychology, 24,
756-766.
Newport, E. (1977). Motherese: the speech of mothers to young children. In N. J.
Castellan, D. B. Pisoni & G. R. Potts (Eds.), Cognitive theory: Vol. 2.
Hillsdale, New Jersey: Lawrence Eribaum Associates.
Newport, E., Gleitman, H., & Gleitman, L. (1977). Mother, I'd rather do it myself: Some
effects and non-effects of maternal speech style. In C. E. Snow & C. A. Ferguson (Eds.),
Talking to children (pp. 109-149). Cambridge, UK: Cambridge University Press.
Patel, R., & Grigos, M. (2006). Acoustic characterization of question-statement
contrast and 4, 7, and 11 year old children. Speech communication, 48, 1308-1318.
Pike, K. (1945). The intonation of American English. Ann Arbor: University of Michigan
Press.
124
Pinker, S (1984). Language learnability and language development. Cambridge, MA: Harvard
University Press.
Plichta, B. (2004). Akustyk for PRAAT. User Manual. Department of Linguistics, Germanic,
Slavic, Asian and African Languages: Michigan State University.
http://bartus.org/akustyk/tutorials/adj_dur.php
Ramus, F. (2002). Language discrimination by newborns: Teasing apart phonotactic, rhythmic,
and intonational cues. Annual Review of Language Acquisition, 2(1), 85-115.
Ramus, F., & Mehler, J. (1999). Language identification with suprasegmental cues: A study
based on speech resynthesis. The Journal of the Acoustical Society of America, 105(1),
512-521.
Redington, M., Chater, N., & Finch, S. (1998). Distributional information: A powerful cue for
acquiring syntactic categories. Cognitive Science, 22(4), 425-469
Rojina, N. (2004). The acquisition of wh-questions in Russian. Nordlyd, 32 (1), 68–87.
Săfárŏvá, M. & Swerts, M. (2004). On recognition of declarative questions in English. In
Proceedings of Speech Prosody. Nara, Japan.
Saffran, J., Aslin, R., & Newport, E. (1996). Statistical learning by 8-month-old
infants. Science, 274(5294), 1926-1928.
Seidl, A. & Johnson, E. (2006). Infant word segmentation revisited: Edge alignment facilitates
target extraction. Developmental Science, 9, 565-573.
Shi, R., & Lepage, M. (2008). The effect of functional morphemes on word segmentation in
preverbal infants. Developmental Science, 11(3), 407-413.
Shi, R., & Melançon, A. (2010). Syntactic Categorization in French-Learning Infants. Infancy,
15(5), 517–533.
125
Slobin, D. I. (1973). Cognitive prerequisites for grammar. In C. Ferguson & D. I. Slobin, eds.,
Studies in child language development. New York: Holt, Rinehart and Winston.
Soderstrom, M., Ko, E., & Nevzorova, U. (2010, March) Infants Discriminate Yes/No
Questions from Declaratives? Poster presented at the 17th Biennial Conference on Infant
Studies, Baltimore, MD.
Soderstrom, M., Ko., E. & Nevzorova, U. (2011). It's a question? Infants attend
differently to polar questions and declaratives. Infant Behavior and Development, 34,
107-110.
Svetozarova, N.D. (1998). Intonation in Russian. In D. Hirst & A. Di Cristo, eds., Intonation
systems: A survey of 20 languages. Cambridge: University Press.
Thiessen, E., & Saffran, J. (2003). When cues collide: Use of stress and statistical
cues to word boundaries by 7- to 9-month-old infants. Developmental Psychology, 39,
706–716.
Thiessen, E., & Saffran, J. (2007). Learning to learn: Infants' acquisition of stress-based
strategies for word segmentation. Language Learning and Development, 3, 73-100.
van Heuven, V., & Van Zanten, E. (2005). Speech rate as a secondary prosodic characteristic
of polarity questions in three languages. Speech Communication, 47(1), 87-99.
Vicenik, C. (2009). Flat-intonation-resynthesizer.
http://www.linguistics.ucla.edu/faciliti/facilities/acoustic/praat.html
Vicenik, C., & Sundara, M. (2013). The role of intonation in language and dialect discrimination
by adults. Journal of Phonetics, 41(5), 297-306.
Vion, M. & Colas, A. (2006). Pitch cues for the recognition of yes – no questions in French.
Journal of Psycholinguistic Research, 35, 427 – 445.
126
Wales, R. & Taylor, S. (1987). Intonation cues to questions and statements: How
are they perceived? Language and Speech, 30, 199-211.
Warner, N., & Arai, T. (2001). The role of the mora in the timing of spontaneous Japanese
speech. Journal of the Acoustical Society of America, 109 (3), 1144–1156.
Xu, Y. (2005-2012). ProsodyPro.praat. Available from:
http://www.phon.ucl.ac.uk/home/yi/ProsodyPro/
Yoshida, K., Iversen, J., Patel, A., Mazuka, R., Nito, H., Gervain, J., & Werker, J.
(2010). The development of perceptual grouping biases in infancy: a Japanese-English
cross-linguistic study. Cognition, 115(2), 356–361. doi:10.1016/j.cognition.2010.01.005
Abstract (if available)
Abstract
To determine when and how infants begin to distinguish between declaratives and interrogatives, five experiments were carried out with typically developing American English‐learning 7‐ to 12‐month‐olds. In Experiment 1, although there were no significant results for discrimination for either age group, there was a trend towards discrimination between declaratives and polar interrogatives. Additionally, the combined age group demonstrated a significant interrogative preference, suggesting infants can distinguish between sentence types. Experiments 2-5 evaluated the contribution of prosodic and word‐order information in infants’ ability to distinguish between sentence types. Given infants’ early sensitivity to prosodic information, I hypothesized that infants could use prosody to make initial sentence‐type distinctions. Accordingly, Experiment 2 analyzed prosodic measures across penultimate and final syllables in infant‐directed wh-questions, polar interrogatives, and declaratives. The results showed that declaratives and polar interrogatives differed on several dimensions, but declaratives and wh-questions did not, similar to adult‐directed speech. Thus, while prosody is not likely to aid discrimination of declaratives from wh-questions, infant‐directed speech provides a learner with sufficient prosodic information to distinguish declaratives and polar interrogatives. ❧ Experiment 3 tested adults’ ability to make sentence‐type distinctions in the absence of lexical information, before testing infants. The results showed that adults were able to distinguish between declaratives and polar interrogatives when only prosody was available and suggests that the limited information provided in the resynthesized stimuli is sufficient for distinguishing between sentence types. This suggests infants should be able to do the same, although age of initial distinction remains to be determined. ❧ Experiment 4 evaluated 7‐month‐olds’ ability to make sentence‐type distinctions in the absence of lexical information. The results showed that despite the availability of prosodic information, infants did not use prosody alone to distinguish between sentence types. This suggests that infants may require other types of information to distinguish between sentence types. Given the syntactic differences between declaratives and polar interrogatives (e.g. AUX‐inversion), word order is a plausible alternative. ❧ Experiment 5 assessed word‐order cues by presenting 7‐ and 12‐month‐olds with sentences that had flattened intonation and different word order. Experiment 5a tested 12‐month‐olds’ ability to distinguish between sentence types when only word‐order information was available. The results showed that 12‐month‐olds discriminated sentence types from word order alone. These results suggest that word order may be sufficient for distinguishing between declaratives and polar interrogatives. ❧ Experiment 5b tested 7‐month‐olds’ ability to distinguish between sentence types based on word order information alone. Infants did not show a reliable looking time difference, suggesting that they did not distinguish between declaratives and polar interrogatives based on word order alone. This dissertation provides evidence of infants’ ability to distinguish between declaratives and polar interrogatives, manifesting as an interrogative preference. Seven‐month‐olds did not distinguish between sentence types based on prosodic or word order information alone, but 12‐month‐olds did make distinctions based on word order alone. One limitation was the exclusion of wh- questions from the infant experiments. Future studies should evaluate the generalizability of these findings to wh- questions, as well as specific distributional features infants are attending to.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Can infants discriminate between declaratives and interrogatives?
PDF
Prosody and informativity: a cross-linguistic investigation
PDF
Interaction between prosody and information structure: experimental evidence from Hindi and Bangla
PDF
Mechanisms underlying acquisition of non-adjacent dependencies
PDF
Learning about word learning in ambiguous situations
PDF
Signs of skilled adaptation in the co-speech ticking of adults with Tourette's
PDF
Building blocks of action and goal comprehension in a newborn visual system
PDF
Considering the effects of disfluent speech on children’s sentence processing capabilities and language development
PDF
Syntax-prosody interactions in the clausal domain: head movement and coalescence
PDF
Detecting joint interactions between sets of variables in the context of studies with a dichotomous phenotype, with applications to asthma susceptibility involving epigenetics and epistasis
Asset Metadata
Creator
Geffen, Susan (author)
Core Title
When and how infants discriminate between declaratives and interrogatives
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Psychology
Publication Date
07/03/2014
Defense Date
05/14/2014
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
declarative,Infant,interrogative,OAI-PMH Harvest,prosody,word order
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Mintz, Toben H. (
committee chair
), Farver, Jo Ann M. (
committee member
), Iskarous, Khalil (
committee member
), Sundara, Megha (
committee member
), Wood, Justin N. (
committee member
)
Creator Email
segeffen@gmail.com,swingchicla@aol.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-430349
Unique identifier
UC11287041
Identifier
etd-GeffenSusa-2612.pdf (filename),usctheses-c3-430349 (legacy record id)
Legacy Identifier
etd-GeffenSusa-2612.pdf
Dmrecord
430349
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Geffen, Susan
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
declarative
interrogative
prosody
word order