FUNDAMENTAL FREQUENCY CHARACTERISTICS AFFECT JUDGMENTS OF
SEGMENTAL, SUPRASEGMENTAL, AND INDEXICAL PROPERTIES OF CHILDREN
WHO STUTTER
By
Danielle Gage

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Communicative Sciences and Disorders - Master of Arts
2015

ABSTRACT
FUNDAMENTAL FREQUENCY CHARACTERISTICS AFFECT JUDGMENTS OF
SEGMENTAL, SUPRASEGMENTAL, AND INDEXICAL PROPERTIES OF CHILDREN
WHO STUTTER
By
Danielle Gage
Stuttering is a speech and language disorder characterized by disruptions in the
production of speech sounds. Stuttered speech often includes repetitions of words or parts of
words, prolongations of speech sounds, blocks or pauses, interjections such as “um”, or “like”, or
delays of initiating sounds. A critical question from a theoretical standpoint is whether the
perception of prosody within a speech signal is independent of the perception of other
components of the speech signal or interdependent with other components. Prior findings have
indicated that the speech of typically developing child talkers, when modified to a lower
fundamental frequency (F0) level, was judged to be slower, less fluent, and less intelligible than
speech with a higher F0. The present study tested the generalizability of these findings to speech
of children who stutter. F0 of speech of children who stutter was raised, lowered, or kept at its
original level. Listeners assessed anxiety, cognitive abilities, fluency, gender, likeability, speech
rate, speech and language abilities based on an 8 point scale. In addition, listeners estimated the
age of the speaker, and the percentage of intelligible words. Results are discussed in terms of
theories of speech perception and the practical implications for stuttering therapy.

ACKNOWLEDGMENTS

I would like to thank my thesis committee member, Dr. Laura Dilley, Ph.D, Dr. Soo-Eun
Chang, Ph.D, CCC-SLP, and Dr. Eric Hunter, Ph.D, for all of their support with this research.
Additionally, I thank Elizabeth Wieland, Amanda Depalsmaker, Kayla Tillman, Stephanie
Schmidt, and other members of the MSU Speech Perception-Production Laboratory and MSU
Speech-Neurophysiology Laboratory for their assistance with this project.

iii

TABLE OF CONTENTS

LIST OF TABLES………………………………………………………………………………...v
LIST OF FIGURES……………………………………………………………………………....vi
KEY TO ABBREVIATIONS……………………………………………………………………vii
INTRODUCTION………………………………………………………………………...............1
METHODS………………………………………………………………………………………13
Participants……………………………………………………………………………….13
Stimuli…………………………………………………………………………………....13
Spectral Modifications of Stimuli………………………………………………………..14
Stimulus Lists……………………………………………………………………………16
Task……………………………………………………………………………………....17
Procedure…………………………………………………………………………….......20
Apparatus………………………………………………………………………………...20
RESULTS………………………………………………………………………………………..21
Age……………………………………………………………………………………….21
Anxiety…………………………………………………………………………………...23
Cognitive Abilities……………………………………………………………………….24
Fluency…………………………………………………………………………………...26
Gender……………………………………………………………………………………27
Intelligibility……………………………………………………………………………..29
Likeability………………………………………………………………………………..31
Speech Rate………………………………………………………………………………32
Speech and Language Abilities…………………………………………………………..33
DISCUSSION……………………………………………………………………………………36
Effects of F0 modifications………………………………………………………………36
Effects of severity level and interactions of severity…………………………………….39
Theoretical implications………………………………………………………………….41
Clinical and practical implications……………………………………………………….42
Directions of Future Research…………………………………………………………...44
APPENDICES.……………….. ………………………………………………………………...50
Appendix A: Participant Instructions and Reference Sheet……………………………...51
Appendix B: Participant Post-Experiment Questionnaire..…...…………………………54
REFERENCES…………………………………………………………………………………..56

iv

LIST OF TABLES

Table 1: Stimulus Lists…………………………………………………………………………..47
Table 2: Attribute Averages for Modification and Severity Conditions…………………………49

v

LIST OF FIGURES

Figure 1. Age ratings as a function of modification condition (Lowered, Original, Raised), and severity
level (Mild, Moderate, Severe). Participants were asked to estimate the speaker’s age in years and
months. The average age in months is represented in the graph above. Error bars indicate ± 1 SEM…..21
Figure 2: Anxiety ratings as a function of modification condition (Lowered, Original, Raised), and
severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating listener was positive the
speaker was non-anxious and comfortable in the speaking situation, and 8 indicating the listener was
positive the speaker was anxious and uncomfortable in the speaking situation. Error bars indicate ± 1
SEM………………………………………………………….....................................................................23
Figure 3: Cognitive ability ratings as a function of modification condition (Lowered, Original, Raised),
and severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating listener was positive
the speaker had no cognitive impairment, and 8 indicating the listener was positive the speaker did have a
cognitive impairment. Error bars indicate ± 1 SEM……………...............................................................24
Figure 4. Fluency ratings as a function of modification condition (Lowered, Original, Raised), and
severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating listener perceived the
speaker to be completely fluent, and 8 indicating the listener perceived the speaker to be completely
disfluent. Error bars indicate ± 1 SEM…………………………………………………………………....26
Figure 5: Gender ratings as a function of modification condition (Lowered, Original, Raised), and severity
level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating listener perceived the speaker to
be positively male, and 8 indicating the listener perceived the speaker to be positively female. Error bars
indicate ± 1 SEM……………………………………………......................................................................27
Figure 6: Percentage of intelligible words as a function of modification condition (Lowered, Original,
Raised), and severity level (Mild, Moderate, Severe). This dependent variable was assessed by listeners
providing a perceived percentage of words understood between 0%, indicating none of the words were
understood, and 100%, indicating that all of the words were understood. Error bars indicate ± 1
SEM…………………………………………………………………………………….............................29
Figure 7: Likeability ratings as a function of modification condition (Lowered, Original, Raised), and
severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating listener perceived the
speaker to be strongly likeable, and 8 indicating the listener perceived the speaker to be strongly
unlikable. Error bars indicate ± 1 SEM………………………………………………………..…….........31
Figure 8: Rate of speech ratings as a function of modification condition (Lowered, Original, Raised), and
severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating listener perceived the
speaker to be speaking fast, and 8 indicating the listener perceived the speaker to be speaking slowly.
Error bars indicate ± 1 SEM……………………………………………………........................................32
Figure 9: Figure 8. Speech and language ability ratings as a function of modification condition (Lowered,
Original, Raised), and severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating
listener was positive the speaker had no speech and/ or language delay and/or disorder, and 8 indicating
the listener was positive the speaker did have a speech and/or language delay and/or disorder. Error bars
indicate ± 1 SEM………………………………………………..................................................................33

vi

KEY TO ABBREVIATIONS

Children who stutter …………………………………………………………………………CWS
Persons who stutter …………………………………………………………………………..PWS

vii

INTRODUCTION
Stuttering is a speech disorder characterized by disruptions in the production of fluent
speech, otherwise known as disfluencies (ASHA, 2014). These disfluencies can include
repetitions of words or parts of words, prolongations of audible or inaudible (i.e., block) speech
sounds, interjections such as “um” or “like”, or unintentional delays of initiating sound (ASHA,
2014). Stuttering is often described as a timing disorder, disintegration or discoordination of
subglottic and glottal processes, and of articulation with phonation and respiration (Adams,
1975; Bergmann, 1986; Perkins, 1981; Perkins, Rudas, Johnson, & Bell, 1976; Van Riper, 1982).
These disruptions may arise from a number of possible mechanisms related to aberrant speech
motor control and/or its interactions with linguistic, cognitive and emotional processes (Kleinow
& Smith, 2000; Smith & Kelly, 1997; Tasko, McClean, & Runyan, 2007). Stuttering affects
approximately 5% of children and 1% of the general population; however, the etiology of
stuttering is yet unresolved.
Stuttering plays a significant role in the perceived effectiveness of an individual’s
communication. There is reasonable evidence that indicates that children who stutter are likely to
have concomitant language, articulation, and phonological concerns (Bloodstein, 1995; St. Louis
& Hinzman, 1988; Wolk, Conture, & Edwards, 1990). In fact, 30-40% of children who stutter
also exhibit phonological and/or articulatory disorders (Wolk, Conture, & Edwards, 1990;
Yaruss & Conture, 1996). Anderson, Wagovich, and Hall (2006) found that children who stutter
display an asynchrony in speech-language processing systems (i.e., speech-language
dissociation), which may imply slower phonological processing, resulting in increased phoneme
errors. Increased phonemic errors, as seen in PWS, may be perceived as a breakdown in
communication. A study by Von Tiling (2011) found that PWS are afraid of being viewed as

1

“communicatively incompetent” due to perceived difficulties in making precise points and
arrangements in speech, as well as avoiding and correcting for misunderstandings. The ability to
adapt messages effectively and appropriately to communicative interaction contexts is generally
perceived to be hindered in PWS, which may further indicate a deficit in overall effective
communication (Spitzberg & Cupach, 1984; Rickheit, Strohner, & Vorwerg, 2008).
Moreover, stuttering often leads to negative psychosocial consequences such as lack of
confidence, self-assertion and general self-advocacy skills, as well as lack of vitality, social,
emotional, and mental functioning, which may commonly result in elevated levels of anxiety
(Bricker-Katz, Lincoln, McCabe, & 2010; Craig, Blumgart, Tran, 2009; Craig & Tran, 2006;
Craig, Tran, Wikesuriya, & Boord, 2006). Individuals with any kind of overt disorder, such as
stuttering, are often perceived with more negative connotations. In addition to negative
perceptions, PWS also face negative sanctions as well (i.e pity, condescension, exclusion)
(Klassen, 2001). Research by Gabel, Blood, Tellis, and Althouse (2004) reported that stuttering
negatively affected the perceptions of appropriate career choices for people who stutter,
demonstrating that people who stutter (PWS) were advised against becoming an attorney, a
judge, a minister, or a psychologist. These researchers also found that PWS often remain
constrained by the impact that stuttering has on his/her communication, and that they frequently
struggle with a fear of speaking, as well as a fear of negative evaluation by others. Therefore, it
is common for PWS to apply learned and compensatory techniques to assist in fluent speech
production, which can then decrease fear of speaking situations and increase overall confidence
(Bricker-Katz, Lincoln, & McCabe, 2010).
Stuttering can be assessed in a variety of ways either quantitatively or qualitatively (i.e.,
impressionistically or perceptually). Quantitative assessment of stuttering can be accomplished

2

by trained Speech-Language Pathologists using the Stuttering Severity Instrument (SSI), by
counting the frequency of disfluencies per hundred words or syllables, and the duration of
disfluencies. Quantitative scales have been shown to produce highly reliable results, and often
indicate that fluent speakers stereotype PWS in similarly consistent ways (Hughes, Gabel, Irani,
& Schlagheck, 2010). However, qualitative judgments are necessary to some degree even when
administering quantitative assessments, as clinical judgments are incorporated. Therefore,
qualitative judgements may often provide for a more functional and holistic assessment, as such
depends jointly upon overlapping speech components, primarily in children who stutter (Ferrer,
L., Scheffer, & Shriberg, 2010). Often, the foundation of many stuttering treatment approaches
include techniques whose replicability and implementation depends largely upon clinician
insight and judgment (Ingham, 1975; Ingham & Lewis, 1978; Ingham, Montgomery, & Ulliana,
1983). Purely qualitative studies have been rare, but may have more validity and are generally
able to provide more information regarding listener perceptions and attitudes toward PWS
(Hughes, Gabel, Irani, & Schlagheck, 2010).
The effective use of overlapping speech cues are critical to understanding the intentions
of speech throughout communicative interactions, particularly when working with clinical
populations such as children who stutter (Beilby & Byrnes, 2012). Much research has noted a
relationship between the segmental, suprasegmental, and indexical components of speech, as
they all simultaneously interact during speech. First, segmental aspects of speech refer to the
organization of sequences of phonetic or syllabic components, relying on the lexical elements of
the speech signal, otherwise referred to the gross communicative message (Kent, 1983). Second,
suprasegmental, or prosodic, aspects of speech, are commonly characterized by the rhythm and
intonation patterns of speech which are used to modulate and enhance the meaning of a spoken
3

speech signal, direct attention to, and identify personal aspects of the speaker (Grossman, Bemis,
Skwerer, & Tager-Flusberg, 2010; Darwin, 1975; Pallier, 1996). Third, indexical aspects of a
speech signal refer to properties of an individual’s speech variability and sociolinguistic identity
(e.g., gender, age, race, class, speaking style), and group membership (e.g., normal talker, or
talker with a speech-language disorder). Indexical aspects of speech can also refer to differences
in vocal tracts (Peterson & Barney, 1952), glottal waves (Monsen & Engebreston, 1977),
articulatory dynamics (Ladefoged, 1980), and native dialects (cited in Goldinger, 1998). In
combination, these three aspects of speech (segmental, suprasegmental, indexical) are the
constitutive components of speech signals. Therefore, it is of interest as to how the properties of
segmental, suprasegmental, and indexical attributes overlap in order to consequently impact the
perceptions of listeners in everyday situations.
The perception of prosody within a speech signal has been viewed by many researchers
as independent of other components of the speech signal. Many theories of speech processing
implicitly or explicitly claim independence of segmental, suprasegmental, and indexical
properties, thus implying that it is possible to extract one aspect in order to accurately assess
overall communicative competence. In fact, Halle (1985) and Joos (1948) refer to this process as
“speaker normalization”, which proposes that all suprasegmental and indexical aspects are
filtered out of the segmental stream. Remez, Fellowes, and Rubin (1997) characterized this
position by saying that in order to identify a word, the listener discards acoustic aspects in order
to form an abstract representation of the linguistic message. Likewise, to identify speakerspecific information, the listener discards acoustic information specific to phonemes to form a
representation of voice quality. Joos (1948) further suggested that voice details can be
considered “noise”, and are therefore irrelevant to phonetic perception. Therefore,

4

suprasegmental and indexical aspects of the signal such as voice quality, utterance speed, and
other circumstantial information are irrelevant when determining lexical and semantic content,
and that it is rare for listeners to truly recall the salient acoustic properties of a speech signal
(Halle, 1985). This idea was further supported by Studdert-Kennedy (1976) who proposed that
since spoken word perception is seemingly unaffected by subtle variations in the speech signal,
lexical entries are more abstract, and therefore unrelated. Pisoni (1993) refuted this idea
suggesting that abstract representations enable lexical interpretation, providing robust, contextinsensitive perception. Such theories assume that surface information, such as suprasegmental
and indexical, aspects are filtered in speech perception.
By contrast, a growing body of literature supports the notion of interdependencies among
suprasegmental, indexical, and segmental components in the perception of a speech signal
(Pierrehumbert, 2003, Psioni, 1993). Such research states that perceived difficulties in segmental
aspects of communication, such as those suggested for PWS, can extend into the effective use of
prosodic characteristics and interact with indexical components of speech. Recent research has
speculated that PWS have “unstable speech systems” that may be interrupted by variability
within the speech signal. Packman, Onslow, Richard, and Doorn (1996) explored whether
altering the prosodic aspect of syllable stress could reduce segmental stuttering behaviors. It was
found that when PWS decreased the variability within the speech signal, stuttering behaviors
decreased as well. This notion supports the relationship between speech motor execution
(segmental), speech variability (indexical) and syllabic stress (suprasegmental) in providing a
comprehensive speech signal for PWS. Bergmann (1986) suggested that suprasegmental effects
of speech may be closely related to timing (segmental) within a speech signal, especially in
PWS. This was supported by Riley-Graham (2011) who demonstrated that PWS have increased

5

difficulty producing the intended phonemes with the proper suprasegmental properties of timing,
rhythm, and pitch. Furthermore, Ladd, Silverman, Tolkmitt, Bergmann, and Scherer (1985)
supported this idea of overlapping components stating that emotional and attitude dependent
(indexical) factors affect overall intonation patterns and F0 values. Furthermore, changes in F0
could be potentially anchored by segmental components and structure within the overall speech
signal, suggesting that segmental and suprasegmental information align when conveying speech
signals (Arvaniti, Ladd, & Mennen, 1998; Ladd, Faulkner, Faulkner, & Schepman,1999, Bruce,
1977). In general, such studies have indicated a dependence between suprasegmental
components and speech rate. Bergmann (1986) further speculated regarding the relationship of
indexical components stating that ‘person-dependent’ (indexical) and ‘communication-related’
(i.e., communication demands, expectations) psychological factors also significantly impact
motor execution. An additional study by Sommers and Barcroft (2006), hypothesized that
differences among indexical components of speech (i.e, speaking style), would interact with
lexical components of speech. Considered with previous findings, the results largely indicated
that variability in indexical, suprasegmental, and segmental components of speech do
significantly impact identification of spoken words (Mullennix, Pisoni, & Martin, 1989;
Sommers, Nygaard, & Pisoni, 1994).
In general, aspects of speaker variation (indexical) interact with presentation rate
(segmental), giving rise to the potential of affecting listener’s perceptual identification of the
speech context and of the speaker themselves (Goldinger, Pisoni, & Logan, 1991; Klatt & Klatt
(1990). Because of such difficulties, many stuttering therapies focus on desensitizing PWS to
reactions and attitudes of listeners, therefore suggesting that listener perceptions are pertinent for
enhancing fluency (Johnson, 1959; Woods, 1978; Van Riper, 1982). A study by Ingham, Martin,

6

and Haroldson (1985) investigated whether speech naturalness could be used to not only assess
speech quality, but to modify speech quality. These researchers stated that speech naturalness
scores based on perceptual ratings from naïve listeners could be used as functional contingencies,
and could potentially lead to new approaches to stuttering therapy. Therapy approaches also
often attempt to produce stutter-free speech by introducing new fluency-enhancing or
modification patterns of speech. Such strategies include, but are not limited to whispering,
choral speech and singing, paced/metronomic speech (Wingate, 1969, 1970; Stager, Jeffries, &
Braun, 2004; Ingham, Bothe, Wang, Purkhiser, & New, 2012). However, therapy approaches
consequently introduce speech patterns that are audibly different from the speech of individuals
who do not stutter (Franken, Boves, Peters, & Webster, 1992). These therapy approaches
correlate with Wingate’s (1969, 1970) previously proposed a “vocalization hypothesis”, which
states that various forms of novel speech (e.g., whispering, singing, paced speech), create
enhanced fluency for PWS due to alterations in the manner of vocalization. Wingate further
hypothesized that fundamental frequency is one such alteration of vocalization, and such a
change could lead to increased fluency patterns in PWS. A study by Healey (1982) sought to
justify Wingate’s hypothesis and identify certain parameters of speaking fundamental frequency
(F0) associated with people who do and do not stutter. Results indicated that the F0 for PWS was
slightly lower and yielded more standard deviations than normally fluent speakers. The F0 range,
rate, and mean were significantly greater in PWS than in non-stuttering groups (i.e.
suprasegmental aspects). The reduction of F0 frequency variability across studies is one speech
aspect that can be objectively measured when comparing speech-production differences between
PWS and fluent speakers.

7

Since specific segmental properties of a speech signal (e.g., articulation rate) are often
targeted in speech therapy for stuttering and other disorders, it is of interest to determine the
extent to which one informational component of the signal (e.g., the segmental lexical content)
might be influenced by other informational components of the acoustic signal (e.g., prosody
(suprasegmental), and speaker identity (indexical) content). The interdependency of these three
components raises the possibility that judgments of one component could be influenced through
an indirect route of manipulating another component. For example, the segmental component of
a persons’ speech could be made to sound more or less intelligible by altering the
suprasegmental content. Previous studies conducted by Henry, McAuley, and Zaleha (2009), and
Melara and Marks (1990) suggest this is a possibility, demonstrating that speech rate is not
merely a function of speed within the speech signal, but also is dependent on other interrelated
factors. Bergmann (1986) also stated that often segmental irregularities are perceived rather as
prosodic (suprasegmental) irregularities and vice versa due to the overlap in components. These
findings are consistent with interdependency theories that depict an interdependence of
perceptual judgments of pitch, timing, and/or personal profile. More specifically, manipulations
of prosodic characteristics, such as fundamental frequency, have been shown to increase
perceived fluency in PWS (Wingate, 1969, 1970).
Given that perceptual judgments of prosody may be subject to interdependencies of
acoustic factors, a suprasegmental variable worthy of study is the effect of altered F0. In
particular, findings of interdependency among components raise the possibility that targeting a
property such as overall F0 might influence the perceptions of components that seem to relate
more to the segmental component, such as intelligibility or fluency.

8

The contrasting hypotheses between independent and interdependent speech processing
theories were investigated by Goldinger (1998), who put forward an influential episodic model
of language processing. Goldinger’s episodic model proposed that segmental information, as
well as aspects of indexical information shown to be reflected in the speaker’s voice, mood,
speaking rate, etc., should be explored in combination. The crux of episodic theories is that
perceptual details are stored in memory and are integral to later perception (Goldinger,1998,
Pierrehumbert, 2003, Pisoni, 1993, Mullennix, Pisoni, & Martin, 1989, Mullennix & Pisoni,
1990). This idea ties into the pre-existing theory of perceptual normalization being an integral
component of spoken-word recognition (Mullennix, Pisoni, & Martin, 1989; Mullennix &
Pisoni,1990). Although the idea of speaker normalization, as proposed by Halle (1985), and Joos
(1948) may be true to some extent, it can be generally stated that normalization is necessary for
speech perception to some degree. Many theories of speech perception assume that special
processes match stimuli to canonical representations in memory, which is done via speaker
normalization. Speaker normalization also states that abstract lexicon can be assumed to be
motivated by signal variation and is considered a perceptual problem to be solved by recognition
systems of listeners (McClelland & Elman, 1986; Morton, 1969; Studdert Kennedy, 1976,
Gertsman, 1968). Additional theories state that automatic normalization plays a significant role,
primarily when no effects of speaker variation are evident, implying that an “automatic
normalization process occurs”(Goldinger, 1998). Regardless, of differing theories,
normalization is reflected in speech perception, but more as a hypothesis, rather than an
assumption. However, assumptions can be made regarding memory for stored “surface” details
of communication as a whole, including nonlinguistic stimuli (i.e faces, pictures, pitch and
tempo, social interactions) (as cited in Goldinger 1998). Additionally, linguistic processes often

9

create lasting, detailed memories (Goldinger, 1998). Semon, 1909, 1923 (as cited in Goldinger,
1998) proposed that words are stored in memory as smaller pieces of a larger context, and
without any context, the retrieval seems abstract. Voice information is a critical component of
retrieval of context of spoken words Garner (1974).Such episodic and perceptual normalization
theories can account for interdependency among segmental, suprasegmental, indexical
properties. Recent developments in linguistic theories also support this model, proposing that
listeners do in fact retain memory traces of the fine-grained acoustic information
(suprasegmental) in the speech signal along with speaker information (indexical) (Goldinger,
1996) This further refutes theories of independence and indicates interdependencies between the
various components of segmental, suprasegmental, and indexical aspects of speech. Such a
theory could have significant consequences for identification of spoken words, the effects of rate
and talker variability, and/or establishing the importance of perceptual normalization (Sommers,
Nygaard, & Pisoni, 1994). However, the applicability of such theories as they relate to speech
processing issues in clinical speech-language pathology has largely not been investigated.
Due to the suspected interdependencies between speech characteristics, and the limited
amount of recent literature, it is critical to continue to investigate the carryover between streams
of segmental, suprasegmental, and indexical information within a speech signal. Dilley, Wieland,
Gamache, McAuley, and Redford (2013) explored this carryover by manipulating the overall
formants and fundamental frequency (F0) of speech for typically-developing five-year old
children and adults in order to mimic the indexical, age-related changes in F0 and formants
which occur during normal development. Listeners heard speech files in which both the formants
and F0 were unmodified (i.e., children sounded like children, and adults sounded like adults), as
well as conditions where the formants and F0 had been modified (i.e., the children sounded like

10

adults, and adults sounded like children). Listeners then made judgments of segmental,
suprasegmental, and indexical properties of the talker. If segmental and/or suprasegmental
properties of speech are processed independently of indexical properties (Joos, 1948; Halle,
1985), then the F0 and formant modification should not have affected the judgments of these
properties. Contrary to this view, the results from this study indicated an overlap between the
three streams, where the indexical/age-related manipulation affected judgments of segmental
and/or suprasegmental properties. Further, these findings indicated that the speech of child
talkers, when modified to a lower pitch contour, was judged to be slower, less fluent, and less
intelligible. The same child talkers with modified speech were judged to be less likeable, more
anxious, and more likely to have cognitive and/or speech-language impairments. The results
supported the hypothesis that an interdependence exists between various aspects of the speech
signal that can affect a wide range of impressionistic judgments of attributes that are commonly
assumed to be orthogonal or quasi-orthogonal to prosodic, segmental, and talker attributes. These
findings may be critical to understanding how speech is perceived by a listener, as well as how
various factors/streams overlap in order to effectively communicate.
This thesis study builds on the work of Dilley and colleagues (2013) which identified
interdependencies in perception among segmental, suprasegmental, and indexical aspects of
speech in typically-developing children. Specifically, we investigated whether the
interdependencies demonstrated by Dilley et al. (2013) can be generalized to clinical populations
such as children who stutter. If this interdependency proves true, the perception of speech would
depend on influential factors or patterns of one component by other components (i.e., segmental,
suprasegmental, indexical). Therefore, one research question this thesis study will be
investigating is whether an F0-only manipulation would provide support for the hypothesis of

11

interdependencies among speech components (segmental, suprasegmental, and indexical) in the
clinical population of children who stutter? Another research question that was explored was will
manipulating overall F0 to be higher or lower affect impressionistic judgments of segmental
and/or suprasegmental properties relating to anxiety, cognitive abilities, fluency, gender,
intelligibility, likeability, speech rate, and/or speech-language abilities of children who stutter?
This thesis study investigated the effect of one prosodic characteristic (i.e., F0) on other
impressionistic judgments of prosodic, segmental and indexical aspects of speech in children
who stutter. We also explored whether changing the indexical properties of a speech sample will
affect judgments of segmental and/or suprasegmental properties of children who stutter. To test
if indexical and/or suprasegmental information would be found to be interdependent of
segmental information in speech, we altered the suprasegmental F0 to be higher or lower in
pitch, which we expected might be more likely to enhance the segmental perceptions of fluency
in a speech sample. Additionally, an F0-only manipulation revealing similar results would
demonstrate interdependence among components in perceptual judgments. These findings give
rise to the possibility of new clinical approaches aimed at altering volitional F0 in a manner
which might result in more positive perceptual evaluations of segmental, suprasegmental, or
indexical aspects of speech of PWS.

12

METHODS
Participants
Thirty three students (F = 25, M = 8) from Michigan State University participated in the
experiment in return for partial credit in a university course. Participants were recruited via
SONA network, and were native speakers of American English who were at least 18 years of age
(M = 21.3) with self-reported normal hearing. The design of the experiment was a 3
(modification condition: Lowered, Original, Raised) x 3 (Severity level: Mild, Moderate, Severe)
mixed factorial. Modification and Severity measures were both within-subjects variables.

Stimuli
The stimuli were obtained via recordings of spontaneous speech produced in a childfriendly, quiet, experimental room in the Speech Neurophysiology Laboratory at Michigan State
University. Speech samples were collected from 15 children, three to ten-year-old (6 female; 9
male, with a mean age of: 6 months, 2 years, or 74.6 months) who stutter. The 15 children were
chosen based on previously established severity of his/her fluency disorder into three groups of
Mild, Moderate, and Severe. This was arranged so that five different children, both males and
females, were in each of the three severity groups. Severity ratings were determined based on
previous diagnoses from speech language pathologists. None of the children had any
concomitant speech, hearing, or developmental disabilities, besides fluency.
Speech samples selected for the present study included speech segments extracted from a
ten minute narrative task from the wordless picture book “Where are you Frog” by Mercer
Meyer (1967). Segments of speech were extracted from the first time the child was exposed to
this specific reading stimulus, so that the child had never read/seen the wordless picture book

13

“Where are you Frog?” prior to this task. First, each full 10-minute recording of the wordless
picture book was segmented into three speech segments from each CWS so that each segment
contained various disfluencies. Each speech segment ranged from 4-10 seconds long, with the
mean duration across all 15 stimuli being 4.2 seconds. This resulted in a total of 45 speech
samples (3 segments x 15 children = 45), and approximately 30 seconds of speech from each
child (6-10 seconds x 3 segments = approximately 30 seconds) in each severity group prior to
manipulations. Disfluencies were characterized by syllable repetitions, whole or part word
repetitions, prolongations of audible or inaudible speech sounds, and/or unintentional delays in
initiating sound (ASHA, 2004). Other disfluencies (OD’s) that were included in the
segmentation included interjections, revisions, and phrase repetitions.

Spectral Modifications of Stimuli
Spectral modifications to alter F0 of each of the three speech segments from each CWS
was conducted in Praat using spectral modification tools (Boersma & Weenink, 2002). Specific
values for global settings for the pitch measurement factors and modification parameters were
restricted to particular ranges for natural-sounding child speech. The first step to completing
spectral modifications using Praat was to change global spectral parameters for a talker using
pitch-synchronous overlap-add algorithm (Moulines & Charpentier, 1990). This involved setting
a global value for each talker for two pitch measurement factors (i.e., pitch floor, and pitch
ceiling), and two modification parameters (i.e., formant shift ration and new pitch median). The
standard values for child speech were as follows:
A. Time step (s) = 0.01
B. Minimum pitch (Hz) = 120 Hz for children

14

C. Maximum Pitch (Hz) = 700 Hz for children
Two other modification parameters (i.e., pitch range factor and duration factor) were left at the
default value of 1.0 for all talkers. Next, the three extracted speech segments from the 15 CWS,
each underwent separate modifications of F0 values using Praat by shifting the frequencies of the
F0 from the original sound file. Values for shifting frequencies were selected for which the talker
seemed to be most convincingly transformed in pitch, yet still natural sounding for the child’s
chronological age and gender. For each segment, the frequencies were shifted by -50.0, 0.0, and
+50.0 Hz to produce the modified Lowered, Original, and Raised speech segments, respectively.
Manipulation for the Original level assured that artifacts created by manipulation itself existed in
all three levels, not just the speech samples where the F0 frequency was altered. Fifty hertz was
chosen as the amount of modification because it corresponded to 1.32 standard deviations for
children ages 5-10 as reported in Lee et al., 1999, JASA paper on children’s speech qualities.
These modifications allowed for holding constant the prosodic attributes, such as speech rate, F0
contour, pause timing, and duration. The modifications resulted in nine speech files per CWS (3
speech segments (1, 2, 3) x 3 modification levels (Lowered, Original, Raised). Therefore, there
were 135 total speech segments across all CWS, so that all 3 speech segments, and all 3
modification conditions were represented for all 15 CWS (3 modification types x 3 modification
levels x 15 =135). The Lowered, Original, and Raised modification conditions were expected to
correspond to speakers’ sounding older, of the same age, or younger, respectively.
Following the modifications, the speech sounded very natural for the age of the children,
with little if any perceptual indication of modification. To further minimize possible artifacts of
modification, the speech was carefully checked auditorily by researchers and piloted.
Additionally, participants were asked to complete a post-experiment questionnaire in which they

15

identified any patterns or abnormalities in the speech sample in order to further asses any
perceptions of oddities observed in the speech samples. Of 33 participants, only 5 participants
(15%) made any mention to observable oddities throughout the experiment. Some of the
oddities described included poor sound quality, background noise, and computerized voices.

Stimulus Lists
After the modifications were completed, three stimulus lists were created consisting of 5
practice trials and 45 experimental trials. The 5 practice trials oriented participants to the task
and allowed them the opportunity to ask questions. Speech segments used in the practice trials
came from different CWS who were not represented in the experimental trials. E ach of the three
conditions and severity levels were represented in the practice block. The 45 experimental trials
consisted of all three speech segments from each child (15 children who stutter), with each
segment represented in a different modification condition (Lowered, Original, Raised) (15 x 3 =
45). The experimental speech segments were then divided into 3 blocks of 15 samples with a
mandatory short break of 45 seconds between each block. Three blocks were utilized in order to
break up the speech samples for two main reasons. The first reason was in order to give the
listener a short break. The second reason was in order to present a new speech segment from
each of the CWS in a new condition (Original, Raised, Lowered).
To create lists for the experiment, stimuli were first blocked by child speaker, so all
listeners heard and judged one fragment from each of the 15 talkers before hearing additional
fragments from a speaker. The order of presentation of child speakers within blocks was
pseudorandomized, with the constraint that CWS within the same severity level were not heard
consecutively. The same sequence of child talkers was used for all blocks. Three stimulus lists

16

were then created using a Latin Square experimental design by counterbalancing the pairing of
modification condition (Raised, Original, Lowered) with speech fragments. A Latin Square
design ensured that each listener heard each of the three speech samples from each of the 15
CWS only once throughout the experiment while being exposed to all Modification conditions.
The pairing of Modification conditions with speech fragments in each list was further
constrained within a block such that listeners heard a specific sequence of Modification
conditions that repeated every three stimuli within a list (e.g., Original, Raised, Lowered,
Original, Raised, Lowered, etc. for the first block).
The stimulus lists are represented in Appendix A: Table 1: Stimulus Lists.

Task
Participants listened to 50 short speech fragments over headphones and answered 9
questions immediately following each speech fragment. The questions that followed asked the
participant to provide impressionistic judgements of the speaker on nine attributes of the speaker
based upon an eight point scale, which was displayed on the screen.
The first rating that the participants were asked to provide impressionistic judgments on
was in regards to the age of the speaker (0-100 years). Participants estimated the speaker’s age
in years and months using the format Age. Month (age separated from month by a period). For
example, 1 year and 3 months was depicted as 1.3. Participants were told in instructions that
typically developing children begin speaking by one year of age and begin speaking in phrases
by three years of age.
The second rating that the participant was asked to provide impressionistic judgments on
was in regards to the gender of the speaker. Participants assessed his/her confidence in regards

17

to the gender of the speaker, with 1 indicating that he/she was positive the speaker was a male,
and 8 indicating that he/she was positive the speaker was a female.
The third rating that the participants were asked to provide was in regards to the speech
rate of the speaker. Participants assessed his/her perception of how fast the speaker was
speaking, with 1 indicating that the speaker was speaking fast, and 8 indicating that the speaker
was speaking slow.
The fourth rating that the participants were asked to provide was in regards to the
intelligibility of the speaker. Participants assessed his/her perception of what percentage of
words were understood, with 0 indicating 0%, or none of the words were able to be understood,
and 100 indicating that 100%, or all of the words were able to be understood. Participants were
informed that intelligibility refers specifically to how much you could understand from what they
were actually saying.
The fifth rating the participants were asked to provide was in regards to the speaker’s
cognitive abilities. Participants assessed his/her confidence in regards to the overall cognitive,
or mental abilities, of the speaker, with 1 indicating that he/she was positive that the speaker had
no cognitive impairment, and 8 indicating that he/she was positive the speaker did have a
cognitive impairment. Participant were informed that cognitive abilities refers specifically to the
underlying mental abilities of the speaker, and they were asked to rate the speaker on whether or
not you believe there could be some sort of mental delay or disability.
The sixth rating the participants were asked to provide was in regards to the speaker’s
speech and language abilities. Participants assessed his/her confidence in regards to the overall
speech and language abilities of the speaker, with 1 indicating that he/she was positive that the
speaker had no speech-language delay or disorder, and 8 indicating that he/she was positive that

18

the speaker did have a speech-language delay or disorder. Participants were informed that the
speech and language delay/disorder rating differs from the cognitive abilities rating in that the
speech and language rating refers specifically to the output of speech, rather than the underlying
abilities. Participants were reminded to remember to include his/her perception of both disorders
and delays in the speaker.
The seventh rating the participants were asked to provide was in regards to the speaker’s
anxiety. Participants assessed his/her confidence in regards to the overall anxiety of the speaker,
with 1 indicating that the speaker was completely non-anxious/comfortable, and 8 indicating that
the speaker was completely anxious/uncomfortable. Participants were informed that anxiety
could be evidenced by how timid or shy the speaker was, or how excitable the speaker was in the
speaking situation.
The eighth rating the participants were asked to provide was in regards to the speaker’s
likeability. Participants assessed his/her confidence in regards to the overall likeability of the
speaker, or how enjoyable he/she would be as a conversational partner, with 1 indicating that the
speaker was strongly likeable, and 8 indicating that the speaker was strongly unlikeable.
The ninth, and final rating that the participants were asked to provide was in regards to
the speaker’s fluency. Participant assessed the overall degree of fluency of the speaker when
speaking, with 1 indicating that the speaker was completely fluent, and 8 indicating that the
speaker was completely disfluent. Participants were informed that the term fluency in this case
referred to the broad spectrum of fluent speech characteristics, whether by normal talkers or
people who stutter, and that disfluency referred to the broad spectrum of disfluent speech
characteristics. Participants were directed to pay attention to the difference between fluency and
intelligibility ratings, Intelligibility is again referring to how much you can understand, whereas

19

fluency is referring to the overall fluidity of the speech. Therefore, it is possible for an
individual to be completely intelligible, while simultaneously being completely disfluent.
Participants were given a reference sheet hat outlined each of the attributes and the rating scales
for reference throughout the experiment.

Procedure
The 33 participants were assigned randomly to each of the three stimulus lists (n = 11 per
list). Before each experiment, participants completed 5 practice trials, and then given a chance to
ask questions before beginning the actual experiment. Participants were not made aware of the
speech modifications. Throughout the experiment, the scale associated with each question was
displayed on the screen in order to reinforce consistent use of scale endpoint referents, as well as
to encourage use of intermediate variables.

Apparatus
Participants listened to the speech fragments over Sennheiser HD 280 headphones. The
experiment was presented to participants using E-Prime Version 2.0 Professional (Psychology
Software Tools, Inc.) running on a Lenovo Intel Core2 Duo CPU E8500 computer with a 19-in.
monitor.

20

RESULTS
A 3x3 repeated measures ANOVA with factors of modification (Lowered, Original,
Raised conditions), and severity level (Mild, Moderate, Severe) was conducted for each
dependent variable. Results for these dependent variables are discussed in turn below.

Age

Figure 1. Age ratings as a function of modification condition (Lowered, Original, Raised), and
severity level (Mild, Moderate, Severe). Participants were asked to estimate the speaker’s age in
years and months. The average age in months is represented in the graph above. Error bars
indicate ± 1 SEM.

Due to a coding glitch, the deliminator that was used to separate years from months was
not recorded in the program throughout the experiment, yielding a concatenation of years and
months in digits in which the initial zeros were omitted. Data was therefore hand coded to
separate digit strings into years and months. In doing so, experimenter assumptions had to be
made. Reasonable assumptions were based upon the fact that participants were informed upon
initial instruction that children typically begin speaking first words at approximately 1 year of
21

age, and begin speaking in phrases at about 3 years of age. Single digits 1-9 were thus converted
to represent a round number of years, assuming that participants would not estimate a child
speaking in phrases to be less than 12 months. For responses that indicated 10-16, the initial two
numbers were converted to represent years, with the final digit assumed to represent months. .
For example, if a response was noted as 111, it was assumed that the participant intended the
string of digits to represent 11.1, eleven years, one month. Throughout the analyzing of this
data, intent assumptions were made consistently across all modifications and severity levels
alike, therefore limiting possible confounds for the variables in questions.
Figure 1 shows the average rating for perceived age in months for each modification and
severity condition. There was a significant main effect of modification on ratings of age
(F(2,64) = 63.47, p < .001, ηp2 = .672). Moreover, there was a significant main effect of severity
(F(2,64) = 25.316, p < .001, ηp2 =.450). Finally, there was no significant interaction between
modification and severity (F(4,124) = .248, p = .910, ηp2 =.008).
Next, a one-way ANOVA was conducted on modification. There was a significant effect
of modification (F(2,64) = 63.47, p < .001, ηp2 = .672). Post hoc, Bonferroni-corrected pairwise
t-test comparisons demonstrated that CWS in the Original condition (M = 73.51 months) were
perceived as being older than CWS in the Raised condition (M = 67.89 months, p < .001), and as
being younger than CWS in the Lowered (M = 78.69, p < .001) modification conditions.
Additionally, CWS in the Lowered modification condition were perceived as being significantly
older than CWS in the Raised modification condition (p <.001).
Finally, a one-way ANOVA was conducted on severity. There was a significant effect of
severity (F(2,64) = 25.316, p < .001, ηp2 =.450). Post hoc, Bonferroni-corrected pairwise t-test
comparisons demonstrated that children in the Mild severity condition (M = 77.32 months) were

22

perceived as being older than children in both the Moderate (M = 67.69 months, p < .001), and
CWS in the Severe (M = 75.08, p <.001). Additionally, CWS in the Severe severity group were
perceived as being older than CWS in the Moderate (p < .001) severity conditions.
There was no significant interaction between modification and severity (F(4,124) = .248,
p = .910, ηp2 =.008).

Anxiety

Figure 2. Anxiety ratings as a function of modification condition (Lowered, Original, Raised),
and severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating listener
was positive the speaker was non-anxious and comfortable in the speaking situation, and 8
indicating the listener was positive the speaker was anxious and uncomfortable in the speaking
situation. Error bars indicate ± 1 SEM.

Figure 2 shows the average rating for anxiety for each modification and severity
condition. There was a significant main effect of modification on ratings of anxiety (F(2,64) =
5.029, p = .009, ηp2 = .136). Moreover, there was a significant main effect of severity (F(2,64) =

23

24.058, p < .001, ηp2 = .429). Finally, there was a significant interaction between modification
and severity (F(4,128) = 2.622, p = .038, ηp2 = .076).
In order to tease apart the significant interaction, a one-way ANOVA was conducted at
each level of severity to examine the simple effects of modification. The only level of severity
that showed a significant effect of modification was the Moderate level (F(2,64) = 6.853, p =
.003, ηp2 = .171). For the Moderate level, post-hoc, Bonferroni-corrected, t-tests showed the
Original modification (M = 3.56) was perceived as significantly lower in anxiety than the Raised
(M = 4.15, p = .005).

Cognitive Abilities:

Figure 3. Cognitive ability ratings as a function of modification condition (Lowered, Original,
Raised), and severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating
listener was positive the speaker had no cognitive impairment, and 8 indicating the listener was
positive the speaker did have a cognitive impairment. Error bars indicate ± 1 SEM.

Figure 3 shows the average rating for perceived cognitive abilities for each modification
and severity condition. There was a significant main effect of modification on ratings of
cognitive ability (F(2,64) = 12.420, p < .001, ηp2 = .280). Moreover, there was a significant main
24

effect of severity (F(2,64) = 24.058, p < .001, ηp2 =.657). Finally, there was no significant
interaction between modification and severity (F(4,128) = 2.622, p = .038, ηp2 =.076).
Next, a one-way ANOVA was conducted on modification. There was a significant effect
of modification (F(2,64) = 12.420, p < .001, ηp2 = .280). Post hoc, Bonferroni-corrected pairwise
t-test comparisons demonstrated that speakers in the Original condition (M = 3.34) were
perceived as having significantly better cognitive abilities than in the Raised condition (M =
3.53, p < .014), and Lowered (M = 3.85, p < .001) modification conditions.
Finally, a one-way ANOVA was conducted on severity. There was a significant effect of
severity (F(2,64) = 61.210, p < .001, np2 = .657). Post hoc, Bonferroni-corrected pairwise t-test
comparisons demonstrated that children in the Mild severity condition (M = 3.111) were
perceived as having significantly better cognitive abilities than children in the Moderate (M =
3.352, p = .033) and in the Severe (M = 4.251, p < .001) severity conditions. Additionally,
children in the Moderate severity condition were perceived as having significantly better
cognitive abilities than in the Severe severity condition (p < .001).

25

Fluency

Figure 4. Fluency ratings as a function of modification condition (Lowered, Original, Raised),
and severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating listener
perceived the speaker to be completely fluent, and 8 indicating the listener perceived the speaker
to be completely disfluent. Error bars indicate ± 1 SEM.

Figure 4 shows the average rating for fluency for each modification and severity
condition. There was a significant main effect of modification on ratings of fluency (F(2,64) =
5.701, p = .005; ηp2 =.151). Moreover, there was a significant main effect of severity (F(2,64) =
55.249, p < .001; ηp2 <.001). Finally, there was no significant interaction found between
modification and severity (F(4,128) = 0.955, p = .435; ηp2 =.029).
Next, a one-way ANOVA was conducted on modification. There was a significant effect
of modification (F(2,64) = 5.701, p = .005; ηp2 =.151). Post hoc Bonferroini corrected pairwise
comparisons demonstrated that the Original modification (M = 5.170) was significantly more
fluent than the Raised (M = 5.463, p = .016) modification condition.

26

Finally, a one way ANOVA was conducted on severity. There was a significant main
effect of severity observed (F(2,64) = 55.249, p < .001; ηp2 <.001). Post hoc Bonferroini
corrected pairwise comparisons demonstrated that the Mild (M = 4.875) was perceived as
significantly more fluent than the Moderate (M = 5.147, p = .014), and the Severe (M = 5.956, p
< .001) severity levels. Additionally, the Moderate severity level was perceived as significantly
more fluent than the Severe (p < .001) severity level.
No significant interaction was found between modification and severity (F(4,128) =
0.955, p = .435; ηp2 =.029).

Gender

Figure 5. Gender ratings as a function of modification condition (Lowered, Original, Raised),
and severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating listener
perceived the speaker to be positively male, and 8 indicating the listener perceived the speaker to
be positively female. Error bars indicate ± 1 SEM.

27

Figure 5 shows the average rating for gender for each modification and severity
condition. A significant main effect of modification was observed when collapsing across all
severity levels (F(2,64) = 11.173; p < .001; ηp2 =.259). Moreover, there was a significant main
effect of severity observed (F(2,64) = 66.315, p < .001, ηp2 = .675). Finally, there was a
significant interaction found between modification and severity (F(4,128) = 4.319, p = .003, ηp2
= .119).
A one-way ANOVA was conducted on modification. There was a significant main effect
of modification (F(2,64) = 11.173; p < .001; ηp2 =.259). Post-hoc Bonferroni-corrected pairwise
comparisons demonstrated that the Original modification (M = 3.95) was perceived as
significantly less female than the Raised (M = 4.56, p < .001) and trended towards less female
for the Lowered (M = 4.24; p =.077) modifications.
Finally, a one-way ANOVA was conducted on severity. There was significant main
effect of severity was observed when collapsing across all modification levels (F(2,64) = 66.315,
p < .001, ηp2 = .675). Post-hoc Bonferroni-corrected pairwise comparisons demonstrated that
the Severe severity (M = 3.37) was perceived as significantly less female than the Moderate (M
= 4.57, p < .001) and Mild (M = 4.81, p < .001) severity levels.
A significant interaction was found between modification and severity (F(4,128) = 4.319,
p = .003, ηp2 = .119). Therefore, a one-way ANOVA was conducted to look at the simple effects
of modification at each level of severity. The Mild severity showed a significant effect of
modification (F(2,64) = 13.200, p < .001, ηp2 = .292), and Bonferroini-corrected post-hoc tests
showed the Raised modification (M = 5.41) was perceived as significantly more female than the
Original (M = 4.34, p < .001), and the Lowered (M = 4.69, p = .005) modification conditions.
The Moderate severity showed a significant effect of modification (F(2,64) = 4.267, p = .018, ηp2

28

= .118), and Bonferroini-corrected post-hoc tests showed that the only significant difference was
where Original (M = 4.30) was perceived as significantly less female Raised (M =4.90, p = .031)
modification conditions. The Severe severity had no significant main effect of modification.

Intelligibility

Figure 6. Percentage of intelligible words as a function of modification condition (Lowered,
Original, Raised), and severity level (Mild, Moderate, Severe). This dependent variable was
assessed by listeners providing a perceived percentage of words understood between 0%,
indicating none of the words were understood, and 100%, indicating that all of the words were
understood. Error bars indicate ± 1 SEM.

Figure 6 shows the average percentage of perceived intelligible words for each
modification and severity condition. A significant main effect of modification was observed
when collapsing across all severity levels (F(2,64) = 11.530, p < .001, ηp2 = .265). Moreover,
there was a significant main effect of severity (F(2,64) = 17.481, p < .001, ηp2 = .353). Finally,
29

there was no significant interaction was found between modification and severity with (F(4,128)
=.612, p = .655; ηp2 = .019).
A one-way ANOVA was conducted on modification. There was a significant effect of
modification (F(2,64) = 11.530, p < .001, ηp2 = .265). Post-hoc Bonferroni-corrected pairwise
comparisons demonstrated that the Original modification (M = 62.33) was perceived as
significantly more intelligible than the Lowered (M = 54.75; p = .003) and the Raised (M =
53.53, p < .001) modification conditions.
Finally, a one-way ANOVA was conducted on severity. There was a significant main
effect of severity was observed when collapsing across all modification levels (F(2,64) = 17.481,
p < .001, ηp2 = .353). Post-hoc Bonferroni-corrected pairwise comparisons demonstrated that the
Mild severity (M = 60.98) was perceived as significantly more intelligible than the moderate (M
= 56.91, p = .007) and severe (M = 52.71, p < .001) severity levels. Additionally the Moderate
severity level was perceived as significantly more intelligible than the Severe (p = .024) severity
level.
No significant interaction was found between modification and severity (F(4,128) =.612,
p = .655; ηp2 = .019).

30

Likeability

Figure 7. Likeability ratings as a function of modification condition (Lowered, Original, Raised),
and severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating listener
perceived the speaker to be strongly likeable, and 8 indicating the listener perceived the speaker
to be strongly unlikable. Error bars indicate ± 1 SEM.

Figure 7 shows the average rating for each modification and severity condition. A
significant main effect of modification was observed when collapsing across all severity levels
(F(2,64) = 9.434; p < .001; ηp2 =.228). Moreover, a significant main effect of severity was
observed (F(2,64) = 13.450; p < .001; ηp2 =.296). Finally, there was no significant interaction
was found between modification and severity (F(4,128) =.827, p = .510; ηp2 =.025).
A one-way ANOVA was conducted on modification. There was a significant effect of
modification (F(2,64) = 9.434; p < .001; ηp2 =.228). Post hoc Bonferroini corrected pairwise
comparisons demonstrated that the Original modification (M = 3.729) was perceived as
significantly more likable than the Lowered (M = 4.065; p = .001), and the Raised (M =3.923, p
=.046) modification conditions.

31

A one-way ANOVA was conducted on severity. There was significant main effect of
severity was observed (F(2,64) = 13.450; p < .001; ηp2 =.296). Post hoc Bonferroini corrected
pairwise comparisons demonstrated that the Severe severity condition (M = 4.248) was perceived
as significantly less likeable than the Mild (M = 3.758, p < .001) and the Moderate (M = 3.711; p
< .001) severity levels.
No significant interaction was found between modification and severity (F(4,128) =.827,
p = .510; ηp2 =.025).

SpeechRate

Figure 8. Rate of speech ratings as a function of modification condition (Lowered, Original,
Raised), and severity level (Mild, Moderate, Severe). Ratings were assessed with 1 indicating
listener perceived the speaker to be speaking fast, and 8 indicating the listener perceived the
speaker to be speaking slowly. Error bars indicate ± 1 SEM.

Figure 8 shows the average rating for speech rate for each modification and severity
condition. No significant main effect of modification was observed when collapsing across all
severity levels (F(2,64) = 2.794; p = .069; ηp2 =.080). However, a significant main effect of
32

severity was observed (F(2,64) = 37.754; p < .001; ηp2 =.541). There was no significant
interaction found between modification and severity (F(4,128) = 0.062, p = .993; ηp2 =.002).
A significant main effect of severity was observed when collapsing across all
modification levels (F(2,64) = 37.754; p < .001; ηp2 =.541). Post hoc Bonferroini corrected
pairwise comparisons demonstrated that the Mild (M = 4.422) was perceived as significantly
lower than the Severe (M = 5.156; p < .001) modification condition, and that Moderate severity
condition (M = 4.501) was significantly lower than the Severe severity condition (p <.001).
No significant interaction was found between modification and severity (F(4,128) =
0.062, p = .993; ηp2 =.002).

Speech and Language Abilities

Figure 9. Speech and language ability ratings as a function of modification condition (Lowered,
Original, Raised), and severity level (Mild, Moderate, Severe). Ratings were assessed with 1
indicating listener was positive the speaker had no speech and/ or language delay and/or
disorder, and 8 indicating the listener was positive the speaker did have a speech and/or language
delay and/or disorder. Error bars indicate ± 1 SEM.

33

Figure 9 shows the mean rating for speech and language abilities for each modification
and severity condition. A significant main effect of modification was observed (F(2,64) = 5.936,
p = .004; ηp2 =.156). Moreover, a significant main effect of severity was observed (F(2,64)
=86.447, p < .001, ηp2 =.730). Finally, a significant interaction was found between modification
and severity (F(4,128) = 3.536, p = .009; ηp2 =.100).
Next, a one-way ANOVA was conducted on modification. There was a significant main
effect of modification (F(2,64) = 5.936, p = .004; ηp2 =.156). Post hoc Bonferroini corrected
pairwise comparisons demonstrated that the Original modification (M = 4.919) was perceived as
significantly lower than the Lowered modification (M = 5.222; p = .014), and the Raised
modification (M = 5.042, p < .001) levels.
A one-way ANOVA was conducted on severity. There was a significant main effect of
severity was observed (F(2,64) =86.447, p < .001, ηp2 =.730). Post hoc Bonferroini corrected
pairwise comparisons demonstrated that the Mild severity (M = 4.374) was perceived as having
significantly better speech and language skills than the Moderate (M = 4.691, p = .055) and the
Severe (M =6.119, p < .001) severity conditions. Additionally, the Moderate severity was
perceived as having significantly better speech and language skills than the Severe severity level
(p < .001).
A significant interaction was found between modification and severity (F(4,128) = 3.536,
p = .009; ηp2 =.100). Therefore, a one-way ANOVA was conducted to look at the simple effects
of modification at each level of severity. The only level of severity that showed a significant
effect of modification was the Mild severity level (F(2,64) = 6.853, p = .003, ηp2 = .171), and
Bonferroini-corrected post-hoc tests showed the Original modification (M = 3.99) was perceived

34

as having significantly better speech and language abilities than the Raised (M = 4.59, p = .001)
and trended towards being better for the Lowered (M = 4.54, p = .043) severity level.

35

DISCUSSION
In this experiment, spectral modification was used to alter fundamental frequency (F0)
characteristics for young children who stutter (CWS). This was done in order to determine the
possible effect that such modifications may have on the impressionistic judgments of
suprasegmental, segmental, and indexical characteristics of a speaker, particularly regarding the
clinical population of children who stutter with varying stuttering severity levels. Naïve listeners
judged modified speech segments and provided impressionistic judgements on perceived age,
anxiety levels, cognitive abilities, gender, intelligibility, likeability, speech rate, and speech and
language abilities. Spectral modification was used to allow for targeted manipulation of F0 to
examine its effects on impressionistic judgments while controlling for other acoustic variables,
including duration, speech rate, overall F0 contour, and pausing, as well as factors such as
linguistic content and voice quality.
A main finding of this research was that modifications of F0 influenced naive listener
perceptions on all measures examined. The selected measures related to a variety of
suprasegmental, segmental, and indexical characteristics for this targeted clinical population of
CWS, including speech quality (i.e., intelligibility and fluency). In the following, we will recap
the main findings, describing in turn how F0 modifications and severity influenced
impressionistic judgments for CWS. We then describe the theoretical and clinical implications of
this work and identify some future directions for the research.

Effects of F0 modifications
Segmental and suprasegmental attributes for which F0 significantly affected judgments
included fluency, intelligibility, and speech rate. CWS were perceived overall as being

36

significantly more fluent in the Original modification condition than in the raised modification
condition. Additionally, CWS were perceived to be significantly more intelligible in the
Original modification condition than in the Lowered or the Raised modification. Contrary to
ratings of fluency and intelligibility, there was no significant interaction of modification on F0
regarding the suprasegmental attribute of speech rate. However, the p-value (p = .069) may
indicate a possibility of a Type Two error, thereby implying an effect that could not be detected
reliably in the present study design. Overall, these results imply that F0, a primarily
suprasegmental characteristic, can influence judgements of other suprasegmental and segmental
attributes of CWS. The fact that many of the impressionistic judgements were often more
positive in the Original modification condition may suggest that listeners may have been
sensitive to whether F0 levels were age-appropriate for the CWS, such that having an ageappropriate F0 may be important for optimal perceptual evaluations of fluency and intelligibility.
F0 modification was also shown to significantly influence perceptual evaluations of
indexical attributes for CWS. Recall that indexical attributes refer to specific properties of an
individual’s speech variability and sociolinguistic identity, such as gender, age, race, class,
speaking style, or group membership (e.g., normal talker, or talker with a speech-language
disorders). The results of this study indicated that a number of indexical attributes were affected
by F0 manipulation, including personal identity (gender and age), clinical variables (cognitive
and speech and language abilities), and personal presentation variables (anxiety and likeability).
Age was one such indexical attribute that was affected by F0 modifications that can be
said to influence listener’s perception of a key identity feature of the speaker. Often, F0 is
correlated to the indexical attribute of age, so that it can be expected that younger children have a
more naturally raised F0, and older children have a more naturally lower F0. If atypical F0

37

ranges are observed, it can be said that listeners will naturally judge and perceive the speaker
more harshly. The findings reported here reflect this notion in that the Original modification and
the Mild severity level reflect the most positive and appropriate ratings based upon the CWSs’
chronological age. The trends observed in this study were expected, being that the Lowered
modification resulted in perceived older children, and the Raised modification resulted in
perceived younger children. This trend is also observed throughout the severity groups, where
CWS in the Mild and Severe severity groups were perceived as older than CWS in the Moderate
severity group, implying that CWS in the Moderate severity group were perceived as closer to
the child’s chronological age. Therefore, it can be stated again that the further F0 is from a
typical range expected for the speaker, regardless of severity level, the less positive listener
perceptions will be associated to that CWS.
Gender was another indexical attribute that was affected by F0 modifications that can be
considered a key personal identity feature of a speaker. Similar to age related findings, F0 is
oftencorrelated to the indexical attribute of gender and can even affect perceptions of speaker
likeability (Perry, Ohde, Ashmead, 2001; Re, O’Connor, Bennett, Feinberg, 2012). The findings
reported here showed significant effects of F0 modification on gender judgments; however, F0
modification interacted with severity and will be discussed more below.
Additional indexical attributes that were affected by F0 modification included cognitive
abilities and speech-language abilities of CWS. Results indicated that the Original condition led
to more positive impressions of CWSs’ cognitive abilities (i.e., judgments of less likelihood of
impairment) than both the Lowered and the Raised modification conditions. Similarly, the
Original modification level was associated with more positive impressions of CWSs’ speechlanguage abilities (i.e., judgments of less likelihood of speech-language impairment) than both

38

the Lowered and the Raised modification conditions. This is consistent with the hypothesis that
having an age-appropriate F0 level results in more positive perceptual evaluations of cognitive
and speech-language abilities.
Further indexical attributes regarding presentation variables that were affected by F0
modification included anxiety and likeability. Our results demonstrated that F0 only affected
judgments of anxiety for CWS at the Moderate severity level, with children in the Original
modification condition being perceived as significantly less anxious than the Raised modification
condition. Impressionistic judgments of likeability indicated that overall, CWS in the Original
modification were perceived as significantly more likeable than both the Lowered and the Raised
modification conditions. These results show that modifying F0 played a significant role on
impressionistic judgements of speaker attributes of cognitive abilities, and speech and language
abilities. The fact that the impressionistic judgements trended towards more positive for the
Original modification conditions is consistent with the hypothesis that having an age-appropriate
F0 is important for optimal evaluations of a speaker’s likeability and degree of perceived
anxiety.

Effects of severity level and interactions with severity
Another main finding of this research was that stuttering severity levels also were
associated with significant differences in listener perceptions for all measures examined. The
segmental and suprasegmental attributes for which significant differences of severity were
demonstrated included fluency, intelligibility, and speech rate. Listeners gave impressionistic
ratings of degree of perceived fluency that mirrored the diagnostic evaluations that had been
provided by speech language pathologists (mild, moderate, or severe). Thus, CWS in the Mild

39

severity group were perceived overall as being more fluent than CWS in the Moderate and
Severe groups, and CWS in the Moderate severity group were perceived overall as more fluent
than CWS in the Severe group. The severity levels also affected perceived intelligibility, with
CWS in the Mild severity group being perceived as more intelligible than those in the Moderate
group, and CSD in the Moderate group in turn being perceived as more intelligible than those in
the Severe group. Another suprasegmental variable that was affected by severity was speech rate
results indicated that CWS in both the Mild and Moderate severity groups were perceived as
speaking faster than CWS in the Severe group. Not surprisingly, these results indicated that the
less severe the stuttering severity, the more positive the impressionistic judgements. The findings
that perceptual evaluations mirror diagnostic categories provided by speech-language
pathologists further serves to validate the methodology used in this study.
F0 modification and severity interacted to influence gender judgments. For mild and
moderate severity levels, a Raised F0 yielded more “girl” judgments than an Original F0 level.
Significant effects of severity, and interactions with severity, must be taken with a grain of salt,
because the distribution of males and females was not controlled across severity levels. In
particular, 4/5 (80%) in the Severe group were male, a fact which would be expected to cause
speakers in the Severe group to be judged as sounding, on average, more like boys than speakers
in the other two severity groups. Thus, effects of severity level probably mask inequalities in the
distribution of the gender across severity levels.
Clinically related indexical variables regarding cognitive and speech and language
abilities of CWS were affected by stuttering severity levels as well. CWS in the Mild severity
group were perceived as having better cognitive abilities than CWS in the Moderate group, who
in turn were perceived as having better cognitive abilities than CWS in the Severe groups.

40

Similarly, CWS in the Mild severity condition were perceived as having better speech and
language abilities than CWS in the Moderate group, who in turn were perceived as having better
speech and language abilities than CWS in the Severe group. These results again indicate that
the lower the severity, the more positive impressionistic judgments and further serve to validate
the methodology used in the present study. Although CWS in the lower severity groups overall
were perceived more positively, this study further demonstrated that modifications of F0 affected
ratings for all CWS, regardless of severity level. Thus, F0 modifications did affect perceptual
ratings from listeners overall. Therefore, CWS who are more constrained by the impact of
his/her fluency, such as those in the more severe severity groups, could work on altering
volitional F0 in order to increase perceptual ratings of speech and self.
Personal presentation indexical attributes were also affected by stuttering severity levels,
including impressionistic judgments of anxiety and likeability of CWS. Specifically, we found
that CWS in the Moderate severity group only were perceived as having lower anxiety in the
Original modification condition than in the Raised or Lowered modification conditions.
However, for likeability ratings, CWS were perceived as less likable if they were in the Severe
severity group rather than the Mild or Moderate groups. Overall, the results of severity levels
suggests that the lower the stuttering severity, the more positive the evaluations in
impressionistic judgments of a variety of attributes.

Theoretical implications
These results have significant theoretical implications regarding the hypothesis of
interdependencies between various aspects of speech. In particular, the findings support the
hypothesis that ratings of suprasegmental, segmental, and indexical attributes are strongly

41

interdependent on each other, rather than being characteristics that can be isolated perceptually
from the speech of a talker. This is reflected in patterns of perceptual judgments by listeners. In
particular, F0, which is often described as a suprasegmental attribute, had effects on attributes
that are variously considered segmental, suprasegmental, and/or indexical. The present findings
are the first to reveal this pattern of interdependence among segmental, suprasegmental, and
indexical components in perception of speech of a clinical population, namely, CWS. These
findings suggest that perceptual bias in judgements of suprasegmental, segmental, and indexical
speaker characteristics may generally be automatic. It appears that listeners unintentionally and
unconsciously perceive and judge speakers based on a variety of characteristics, with little to no
control over such bias (cf. Dilley et al., 2013).

Clinical and practical implications
These findings have significant implications for developmental prosody research, as well
as for clinical practice for targeted clinical populations such as CWS. Impressionistic
judgements are common in both clinical and real world evaluation of communication, providing
for qualitative analysis of communication and social interactions at a more functional level. The
present research showed that modifying the F0 of CWS significantly affected the impressionistic
judgments and perceptions of listeners. Recall that the spectral manipulation method used here
allowed for control of a variety of other acoustic variables across conditions, including duration,
speech rate, overall F0 contour, and pausing, as well as factors such as linguistic content and
voice quality. Therefore, it can be stated that F0 manipulations alone, significantly influence
perceptual ratings for a clinical population of CWS. Our initial hypothesis anticipated a more
non-monotonic trend, where increases in F0 would yield consistently better (or worse)

42

impressionistic judgments. However, a number of dependent variables showed non monotonic
trends, with more positive evaluations for the Original modification condition on a number of
impressionistic dimensions. This may be due to a tendency of subjects to make evaluations
partly based on the appropriateness of F0 for a child’s judged chronological age. The Original
modification condition had age-appropriate F0 levels, while the Lowered and Raised levels were
associated with having lower than or greater than age-appropriate F0 levels, respectively. If this
hypothesis is correct, then children who naturally have an F0 level which is unusually high or
unusually low for his/her chronological age would be expected to be judged more negatively on
a variety of dimensions than a child with a typical F0 level for his/her chronological age. This
hypothesis further raises the possibility that therapies could be developed which were targeted at
altering the F0 level of children, or perhaps adults, when it was atypical. This could involve
lowering F0 which was unusually high, or raising F0 which was unusually low. Additional work
is needed to test the hypothesis that the typicality of F0 levels relative to chronological age
influences impressionistic judgments, and to assess the clinical benefits of therapeutic strategies
aimed at F0 modification for enhancing impressionistic judgments of talkers.
Given the generally small effects of F0 on dependent variables in the present study, it
seems possible that the greatest clinical significance may obtain for CWS with borderline
severities. Some CWS are often overlooked or wait-listed for speech therapy due to being on the
border between two diagnostic categories (e.g., normal vs. mild stuttering). For these children,
there may be particularly great clinical significance of F0 level, in that changing F0 level might
yield different classifications of e.g., the degree of fluency for that child in a way that alters their
clinical classification.

43

In summary, the results of this experiment reveal that naïve listeners are influenced by
changes in F0, with the data generally consistent with the hypothesis that listeners have more
positive judgments when the F0 level of CWS is age and gender appropriate. Therefore, it can
be stated that F0 manipulations alone significantly influence perceptual ratings of clinical
populations, and that these effects obtain not only for typically developing children and adults
(Dilley et al., 2013) but also for a clinical population of CWS. More research is warranted in this
area to determine the clinical significance of F0 modifications in other populations. These
findings also support the hypothesis that ratings of suprasegmental, segmental, and indexical
attributes are interdependent of each and depend not only on isolated characteristics of a talker’s
speech, but also on the listener’s overall assessment of an individual.

Directions of Future Research
This study has implications for future research and findings in the field of speech
language pathology. Both theoretical and clinical implications were observed regarding the
interdependencies of suprasegmental, segmental, and indexical characteristics, as well as in the
influence of F0 specifically in speech and listener perceptions. This study would be worth
investigating more using the clinical population of adults who stutter. The results may reveal
more significant differences, as adult ratings from listeners would be more concrete. Often,
many participants stated difficulty honestly rating the CWS, primarily with the likeability
attribute, due to the young age of the children, it was difficult to assess negatively. It would also
be worth investigating further whether aligning F0 values to more age appropriate values
specifically would significantly affect impressionistic judgments. More research would need to
be done regarding natural F0 for children. Additionally, it would be interesting to utilize video,

44

rather than just speech audio for listener perceptions in order to analyze how secondary
characteristics of fluency disorders affect listener perceptions as well. This study could be
designed so that participants first hear only audio and provide ratings, and then immediately
watch the same speech fragment over video and provide ratings. This may mean that rating
scales would have to be tweaked slightly for the video ratings as fluency disorders will be much
more observable. Furthermore, parsing apart gender perceptions regarding age appropriate F0,
and perceived gender roles in modified speech with more balanced ratios of male: female
speakers would also be interesting to explore. An additional area to explore in order to assess the
generalization of the results from this thesis, would be to consider altering volitional F0 in
natural speaking contexts, rather than in spectrally modified contexts. This would require that
CWS be trained to adequately alter F0 within natural ranges for typical speaking situations.
Listeners would then be asked to rate CWS on the same nine attributes to provide perceptual
ratings on segmental, suprasegmental, and indexical attributes of the speaker. This kind study
would be interesting in order to determine whether the results of this thesis study generalize
when all other aspects are not held constant such as in natural speaking situations, as compared
to spectral modifications where all other factors are uninfluenced. Additional research and
design would be necessary. The effects of F0 on perceptions of age would also be a variable for
further in-depth investigation in order to assess whether modifications in F0 would significantly
alter the perceived age of speakers regardless of segmental content of the speech signal. Overall,
the present results demonstrate the there is significant potential to develop clinical interventions
based on F0 in order to manage listener impressions related to a variety of segmental,
suprasegmental, and indexical properties of speech and talkers. Furthermore, altering the
anchors of the rating scales and making the scale smaller, as well as eliminating the extreme

45

attributes associated with the end variables (1 and 8) would be an area of change, in order to
eliminate any additional bias that may unintentionally affect listener perceptions.

46

Table 1: Stimulus Lists
LIST ONE
SF_1_O
SM_1_R
SM__L
SM_1_O
SF__R
SM_1_L
SM_1_O
SM_1_R
SF_1_L
SM_1_O
SF_1_R
SM_1_L
SF_1_O
SM_1_R
SF_1_L

LIST TWO
SF_2_L
SM_2_O
SM_2_R
SM_2_L
SF_2_O
SM_2_R
SM_2_L
SM_2_O
SF_2_R
SM_2_L
SF_2_O
SM_2_R
SF_2_L
SM_2_O
SF_2_R

LIST THREE
SF_3_R
SM_3_L
SM_3_O
SM_3_R
SF_3_L
SM_3_O
SM_3_R
SM_3_L
SF_3_O
SM_3_R
SF_3_L
SM_3_O
SF_3_R
SM_3_L
SF_3_O

SF_2_R
SM_2_L
SM_2_O
SM_2_R
SF_2_L
SM_2_O
SM_2_R
SM_2_L
SF_2_O
SM_2_R
SF_2_L
SM_2_O
SF_2_R
SM_2_L
SF_2_O

SF_3_O
SM_3_R
SM_3_L
SM_3_O
SF_3_R
SM_3_L
SM_3_O
SM_3_R
SF_3_L
SM_3_O
SF_3_R
SM_3_L
SF_3_O
SM_3_R
SF_3_L

SF_1_L
SM_1_O
SM_1_R
SM_1_L
SF_1_O
SM_1_R
SM_1_L
SM_1_O
SF_1_R
SM_1_L
SF_1_O
SM_1_R
SF_1_L
SM_1_O
SF_1_R

SF_3_L
SM_3_O
SM_3_R
SM_3_L
SF_3_O
SM_3_R
SM_3_L
SM_3_O
SF_3_R

SF_1_R
SM_1_L
SM_1_O
SM_1_R
SF_1_L
SM_1_O
SM_1_R
SM_1_L
SF_1_O

SF_2_O
SM_2_R
SM_2_L
SM_2_O
SF_2_R
SM_2_L
SM_2_O
SM_2_R
SF_2_L
47

Table 1 (cont’d)
SM_3_L
SF_3_O
SM_3_R
SF_3_L
SM_3_O
SF_3_R

SM_1_R
SF_1_L
SM_1_O
SF_1_R
SM_1_L
SF_1_O

SM_2_O
SF_2_R
SM_2_L
SF_2_O
SM_2_R
SF_2_L

This table shows the ordering of the three stimulus lists, created using a Latin-Squares experimental
design to counterbalance across all measures. Each stimulus list is divided into 3 blocks of 15
samples each (15 CWS).

48

Table 2: Attribute Averages for Modification and Severity Conditions
Attribute and Severity
Age-Mild (in years)
Age-Moderate (in years)
Age-Severe (in years)

Lowered
6.72
5.04
5.21

Original
6.99
4.27
4.53

Raised
6.23
4.22
4.77

Anxiety-Mild
Anxiety-Moderate
Anxiety-Severe

3.92
3.89
4.67

3.75
3.56
4.42

4.10
4.15
4.37

Cognitive Abilities-Mild
Cognitive Abilities-Moderate
Cognitive Abilities-Severe

3.39
3.73
4.42

2.79
3.10
4.13

3.16
3.22
4.21

Fluency-Mild
Fluency-Moderate
Fluency-Severe

4.93
5.13
5.98

4.64
4.95
5.92

5.06
5.36
5.97

Gender-Mild (larger = more female)
Gender-Moderate (larger = more female)
Gender-Severe (larger = more female)

4.69
4.48
3.55

4.34
4.30
3.20

5.41
4.90
3.37

Intelligibility-Mild (percentage)
Intelligibility-Moderate (percentage)
Intelligibility-Severe (percentage)

60.99
53.29
50.18

65.65
62.87
58.50

56.30
54.57
49.72

Likeability-Mild
Likeability-Moderate
Likeability-Severe

3.95
3.88
4.36

3.50
3.50
4.19

3.82
3.75
4.19

Speech Rate-Mild
Speech Rate-Moderate
Speech Rate-Severe

4.56
4.63
5.33

4.41
4.50
4.38

4.30
4.38
5.03

Speech and Language Abilities-Mild
Speech and Language Abilities-Moderate
Speech and Language Abilities-Severe

4.54
4.85
6.27

3.99
4.58
6.19

4.59
4.64
5.89

Table 2 shows the average ratings for each attribute examined based upon severity level and
modification condition. Recall that higher ratings represent more negative impressionistic
judgments, with the exception of age (represented in years and months), intelligibility
(represented as a percentage), and gender (higher ratings represent more female-like).
49

APPENDICES

50

Appendix A: Participant Instructions and Reference Sheet

In this study, you will listen to 50 short speech fragments over headphones and then answer a
few questions immediately following each speech fragment. The questions that follow will ask
you to rate the speaker on 9 aspects of the speaker based on an eight point scale.

There will be

one practice block of 5 speech fragments, followed by a chance to ask questions before the rest
of the experiment continues. Please use that time to clarify any questions that you may have.

The first rating that you will provide will be in regards to the age of the speaker. You
should estimate the speaker’s age in years and months using the format Age.Month (age
separated from month by a period). For example: 1 year and 3 months, depicted as 1.3. Recall
that children typically begin speaking first words at approximately one year of age, and begin
speaking in phrases at approximately 3 years of age.
The second rating that you will provide will be in regards to the gender of the speaker.
You should assess your confidence in regards to the gender of the speaker, with 1 indicating that
you are positive the speaker is a male, and 8 indicating that you are positive the speaker is a
female.
The third rating that you will provide will be in regards to the speech rate of the speaker.
You should assess your perception of how fast the speaker is speaking, with 1 indicating that the
speaker is speaking fast, and 8 indicating that the speaker is speaking slowly.
The fourth rating that you will provide will be in regards to the intelligibility of the
speaker. You should assess your perception of what percentage of words were understood, with

51

0 indicating 0%, or none of the words were able to be understood, and 100 indicating that 100%,
or all of the words were able to be understood.
The fifth rating you will provide will be in regards to the speaker’s cognitive abilities.
You should assess your confidence in regards to the overall cognitive, or mental abilities, of the
speaker, with 1 indicating that you are positive that the speaker has no cognitive impairment, and
8 indicating that you are positive the speaker does have a cognitive impairment.
The sixth rating you will provide will be in regards to the speaker’s speech and language
abilities. You should assess your confidence in regards to the overall speech and language
abilities of the speaker, with 1 indicating that you are positive that the speaker has no speechlanguage delay or disorder, and 8 indicating that you are positive that the speaker does have a
speech-language delay or disorder.
The seventh rating you will provide will be in regards to the speaker’s anxiety. You
should assess your confidence in regards to the overall anxiety of the speaker, with 1 indicating
that the speaker is completely non-anxious/comfortable, and 8 indicating that the speaker is
completely anxious/uncomfortable.
The eighth rating you will provide will be in regards to the speaker’s likeability. You
should assess your confidence in regards to the overall likeability of the speaker, or how
enjoyable he/she would be as a conversational partner, with 1 indicating that the speaker is
strongly likeable, and 8 indicating that the speaker is strongly unlikeable.
The ninth, and final rating you will provide will be in regards to the speaker’s fluency.
You should assess the overall degree of fluency of the speaker when speaking, with 1 indicating
that the speaker is completely fluent, and 8 indicating that the speaker is completely disfluent.

52

1. Age: Format Age.Month (age separated from month by a period). Example: 1 year and 3
months, depicted as 1.3.
2. Gender: Assess the gender of the speaker as male or female, with 1 indicating positively
male, and 8 indicating positively female.
3. Speech rate: Assess the rate of speech, or how fast the speaker is speaking, with 1
indicating positively fast, and 8 indicating positively slow.
4. Intelligibility: Assess what percentage of the phrase was able to be understood, with 0
indicating that 0%, or none of the words were able to be understood and 100 indicating
that 100%, or all of the words were able to be understood.
5. Cognitive abilities: Assess the overall cognitive, or mental abilities of the speaker, with
1 indicating that you are positive the speaker has NO cognitive impairment, and 8
indicating that you are positive the speaker HAS a cognitive impairment.
6. Speech and Language abilities: Assess the overall speech and language abilities of the
speaker, with 1 indicating that you are positive the speaker has NO speech and/or
language delay/disorder, and 8 indicating that you are positive the speaker HAS a speech
and/or language delay/disorder.
7. Anxiety: Assess the overall anxiety/comfort of the speaker, with 1 indicating that the
speaker is positively non-anxious and comfortable, and 8 indicating that the speaker is
positively anxious and uncomfortable.
8. Likeability: Assess the overall likeability of the speaker, or how enjoyable he/she would
be as a conversational partner, with 1 indicating that the speaker is strongly likeable
(would go out of your way to speak to him/her), and 8 indicating that the speaker is
strongly dislikable (would not choose to intentionally speak to him/her)
9. Degree of Fluency: Assess the overall degree of fluency of the speaker when speaking,
with 1 indicating that the speaker is completely fluent, and 8 indicating that the speaker is
completely disfluent.

53

Appendix B: Participant Post-Experiment Questionnaire

1. What is your age?
2. What is your gender?
3. What is your major/area of study?
4. How many years of college have you completed?
5. Describe how much and what kind of interaction you have with children on a regular
basis?

6. Describe how much and what kind of interaction you have with children with disabilities
(cognitive, speech-language, developmental, or otherwise).

7. Did you notice any patterns throughout the experiment? If so, please describe the types
of patterns you may have noticed, and how you figured them out.

8. What were your general impressions regarding the sound quality of the speech files?

9. Did you notice any oddities within the speech samples themselves throughout the
experiment? If so, please describe the types of oddities that you may have noticed.

54

10. Do you think the speech played over the headphones was modified? If yes, please
describe.

11. What do you think was the purpose of this study?

Please leave additional feedback and comments on the back of this paper. Thank you for your
time.

55

REFERENCES

56

REFERENCES

Adams, M. R. (1975). A physiologic and aerodynamic interpretation of fluent and stuttered
speech. Journal of Fluency Disorders, 1, 35-47.
Anderson, J. D., Wagovich, S. A., & Hall, N. E. (2006). Nonword repetition skills in young
children who do and do not stutter. Journal of Fluency Disorders, 31(3), 177-199.
ASHA (2014). Stuttering. American Speech Language Hearing Association. Retrieved
from HYPERLINK "http://www.asha.org/public/speech/disorders/stuttering/"
www.asha.org/public/speech/disorders/stuttering/ on February 2, 2014.
Arvaniti, A., Ladd, D. R., & Mennen, I. (1998). Stability of tonal alignment: the case of Greek
prenuclear accents. Journal of Phonetics, 26(1), 3-25.
Beilby, J. M., Byrnes, M. L., & Young, K. N. (2012). The experiences of living with a sibling
who stutters: a preliminary study. Journal of Fluency Disorders, 37(2), 135-148.
Bergmann, G. (1986). Studies in stuttering as a prosodic disturbance. Journal of Speech and
Hearing Research, 29, 290-300.
Bloodstein, 0. (1995). A handbook on stuttering (5th ed.). San Diego, CA: Singular Publishing
Group, Inc.
Bricker‐Katz, G., Lincoln, M., & McCabe, P. (2010). Older people who stutter: barriers to
communication and perceptions of treatment needs. International Journal of Language &
Communication Disorders, 45(1), 15-30.
Bruce, G. (1977) Swedish word accents in sentence perspective. Lund: Gleerup
Craig, A., Blumgart, E., Tran, Y. (2009). The impact of stuttering on the quality of life in adults
who stutter. Journal of Fluency Disorders, 34, 61-71.
Craig, A., Tran, Y., Wijesuriya, N., & Boord, P. (2006). A controlled investigation into the
psychological determinants of fatigue. Biological Psychology, 72, 78–87.
Craig, A., & Tran, Y. (2006). Chronic and social anxiety in people who stutter. Advances in
Psychiatric Treatment, 12, 63–68.
Cutler, A., & Norris, D. (1988). The role of strong syllables in segmentation for lexical access.
Journal of Experimental Psychology: Human Perception and Performance, 14(1), 113.
Darwin, C. J. (1975). The dynamic use of prosody in speech perception. In A. Cohen &
Nooteboom (Eds.), Structure and process in speech perception New York:
Springer-Verlag.
57

S.

Dilley, L. C., Wieland, E. A., Gamache, J. L., McAuley, J. D., & Redford, M. A. (2013).
Age-related changes to spectral voice characteristics affect judgments of prosodic,
segmental, and talker attributes for child and adult speech. Journal of Speech, Language,
and Hearing Research, 56(1), 159-177.
Ferrer, L., Scheffer, N., & Shriberg, E. (2010, March). A comparison of approaches for modeling
prosodic features in speaker recognition. In Acoustics Speech and Signal Processing
(ICASSP), 2010 IEEE International Conference on (pp. 4414-4417). IEEE.
Franken, M. C., Boves, L., Peters, H. F., & Webster, R. L. (1992). Perceptual evaluation of the
speech before and after fluency shaping stuttering therapy. Journal of Fluency
Disorders, 17(4), 223-241.
Garner, W. (1974). The processing of information and structure. Potomac, MD: Erlbaum
Gabel, R. M., Blood, G. W., Tellis, G. M., & Althouse, M. T. (2004). Measuring role entrapment
of people who stutter. Journal of Fluency Disorders, 29(1), 27-49.
Gerstman, L. H. (1968). Classification of self-normalized vowels. IEEE Transactions on Audio
and Electroacoustics, AU-16, 78—80.
Goldinger, S. D. (1996). Words and voices: episodic traces in spoken word identification and
recognition memory. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 22(5), 1166.
Goldinger, S. D. (1998). Echoes of echoes? An episodic theory of lexical access. Psychological
Review, 105(2), 251.
Goldinger, S. D., & Azuma, T. (2004). Episodic memory reflected in printed word naming.
Psychonomic Bulletin & Review, 11(4), 716-722.
Goldinger, S. D., Pisoni, D. B., & Logan, J. S. (1991). On the nature of talker variability effects
on recall of spoken word lists. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 17(1), 152.
Grossman, R. B., Bemis, R. H., Skwerer, D. P., & Tager-Flusberg, H. (2010). Lexical and
affective prosody in children with high functioning autism. Journal of Speech Language
Pathology, and Hearing Research, 53(30), 778-793.
Halle, M. (1985). Speculation about the representations of words in memory. In V. A. Fromkin
(ed.), Phonetic Linguistics (pp. 101-104). New York, NY: Academic Press.
Healey, E. C. (1982). Speaking fundamental frequency characteristics of stutterers and
nonstutterers. Journal of Communication Disorders, 15(1), 21-29.

58

Henry, M. J., McAuley, J. D., & Zaleha, M. (2009). Evaluation of an imputed pitch velocity
model of the auditory tau effect. Attention, Perception, & Psychophysics, 71(6), 1399
-1413.
Houston, D. M., & Jusczyk, P. W. (2000). The role of talker-specific information in word
segmentation by infants. Journal of Experimental Psychology: Human Perception and
Performance, 26(5), 1570.
Hughes, S., Gabel, R., Irani, F., Schlagheck, A. (2010). University students’ explanations for
their descriptions of people who stutter: An exploratory mixed model study. Journal of
Fluency Disorders, 35 (2010), 280-298.
Ingham, R.J. (1975). Operant methodology in, stuttering therapy. Eisenson (Ed.), Stuttering: A
second symposium. New York: Harper & Row.
Ingham, R.J., & Lewis, J. I. (1978). Behavior therapy and stuttering: And the story grows.
Human Communication, 1978, 3, 125-152.
Ingham, R. J., Martin, R. R., Haroldson, S. K., Onslow, M., & Leney, M. (1985). Modification of
listener-judged naturalness in the speech of stutterers. Journal of Speech, Language, and
Hearing Research, 28(4), 495-504.
Ingham, R. J., Montgomery, J., Ulliana, J. (1983). The effect of manipulating phonation duration
on stuttering. Journal of Speech and Hearing Research, 26, 579-587.
Ingham, R. J., Bothe, A. K., Wang, Y., Purkhiser, K., & New, A. (2012). Phonation interval
modification and speech performance quality during fluency-inducing conditions by
adults who stutter. Journal of Communication Disorders,45(3), 198-211.
Johnson, W. (1959). The onset of stuttering: research findings and implications. Minneapolis,
MN University of Minnesota Press.
Joos, M. (1948). Acoustic phonetics. Language, 24(2), 5-136.
Kent, R. D. (1983). The segmental organization of speech. The production of speech (pp. 57-89).
Springer New York.
Klatt, D. H., & Klatt, L. C. (1990). Analysis, synthesis, and perception of voice quality variations
among female and male talkers. The Journal of the Acoustical Society of America, 87(2),
820-857.
Klassen, T. R. (2001). Perceptions of people who stutter: Re-assessing the negative
stereotype. Perceptual and Motor Skills, 92(2), 551-559.

59

Kleinow, J., & Smith, A. (2000). Influences of length and syntactic complexity on the speech
motor stability of the fluent speech of adults who stutter. Journal of Speech, Language,
and Hearing Research, 43, 548–559.
Kuhl, P. (1991). Human adults and human infants show a “perceptual magnet effect” for the
prototypes of speech categories, monkeys do not. Perception & Psychophysics, 50,
93-107.
Ladd, D. R., Faulkner, D., Faulkner, H., & Schepman, A. (1999). Constant" segmental
anchoring" of F0 movements under changes in speech rate. The Journal of the Acoustical
Society of America, 106(3 Pt 1), 1543-1554
Ladd, D. R., Silverman, K. E., Tolkmitt, F., Bergmann, G., & Scherer, K. R. (1985). Evidence
for the independent function of intonation contour type, voice quality, and F0 range in
signaling speaker affect. The Journal of the Acoustical Society of America, 78(2), 435
444.
McClelland, J. L., & Rumelhart, D. (1985). Distributed memory and the representation of
general and specific information. Journal of Experimental Psychology: General, 114,
159-188.
Morton, J. (1969). Interaction of information in word recognition. Psychological Review, 76,
165—178.
Melara, R. D., & Marks, L. E. (1990). Interaction among auditory dimensions: Timbre, pitch,
and loudness. Perception & Psychophysics, 48(2), 169-178.
Mullennix, J. W., & Pisoni, D. B. (1990). Stimulus variability and processing dependencies in
speech perception. Perception & Psychophysics, 47(4), 379-390.
Mullennix, J. W., Pisoni, D. B., & Martin, C. S. (1989). Some effects of talker variability on
spoken word recognition. The Journal of the Acoustical Society of America, 85(1), 365
378.
Packman, A., Onslow, M., Richard, F., & Doorn, J. V. (1996). Syllabic stress and variability: A
model of stuttering. Clinical Linguistics & Phonetics, 10(3), 235-263.
Pallier, C. (1996). The role of suprasegmentals in speech perception and acquisition.
Phonological structure and language processing: Cross-linguistic studies, 12, 145.
Perkins, W. H. (1981). Implications of scientific research for treatment of stuttering--A lecture.
Journal of Fluency Disorders, 6, 155--162.
Perkins, W. H., Rudas, J., Johnson, L., & Bell, J. (1976). Stuttering: Discoordination of
phonation with articulation and respiration. Journal of Speech and Hearing Research, 19,
509-522.
60

Perry, T. L., Ohde, R. N., & Ashmead, D. H. (2001). The acoustic bases for gender identification
from children’s voices. The Journal of the Acoustical Society of America, 109(6), 2988
2998.
Pierrehumbert, J. (2003). Phonetic diversity, statistical learning and acquisition of phonology.
Language and Speech, 46(2-3), 115-154.
Pisoni, D. B. (1993). Long-term memory in speech perception: Some new findings on talker
variability, speaking rate and perceptual learning. Speech Communication, 13(1),
109-125.
Porfert, A. R., & Rosenfield, D. B. (1978). Prevalence of stuttering. Journal of Neurology,
Neurosurgery & Psychiatry, 41(10), 954-956.
Prosek, R. A., & Runyan, C. M. (1982). Temporal characteristics related to the discrimination of
stutterers' and nonstutterers' speech samples. Journal of Speech and Hearing Research,
25(1), 29.
Re, D. E., O'Connor, J. J., Bennett, P. J., & Feinberg, D. R. (2012). Preferences for very low and
very high voice pitch in humans. PloS one, 7(3), e32719.
Remez, R. E., Fellowes, J. M., & Rubin, P. E. (1997). Talker identification based on phonetic
information. Journal of Experimental Psychology: Human Perception and
Performance, 23(3), 651.
Rickheit, G., Strohner, H., & Vorwerg, C. (2008). The concept of communicative competence.
Handbook of communication competence, 15-62.
Riley-Graham, J. (2011). Self-Monitoring and Feedback in Disordered Speech Production.
Runyan, C. M., & Adams, M. R. (1979). Unsophisticated judges' perceptual evaluations of the
speech of “successfully treated” stutterers. Journal of Fluency Disorders, 4(1), 29-38.
Smith, A., & Kelly, E. (1997). Stuttering: A dynamic, multifactorial model. In R. F. Curlee &
G.M. Siegel (Eds.), Nature and treatment of stuttering: New directions (pp. 204–217).
Needham Heights, MD: Allyn and Bacon.
Sommers, M. S., Nygaard, L. C., & Pisoni, D. B. (1994). Stimulus variability and spoken word
recognition. I. Effects of variability in speaking rate and overall amplitude. The Journal
of the Acoustical Society of America, 96(3), 1314-1324.
Sommers, M. S., & Barcroft, J. (2006). Stimulus variability and the phonetic relevance
hypothesis: Effects of variability in speaking style, fundamental frequency, and speaking
rate on spoken word identification. The Journal of the Acoustical Society of
America, 119(4), 2406-2416.
61

Spitzberg, B. H., & Cupach, W. R. (1984). Interpersonal communication competence.
Stager, S. V., Jeffries, K. J., & Braun, A. R. (2004). Common features of fluency-evoking
conditions studied in stuttering subjects and controls: an H215O PET study. Journal of
fluency disorders, 28(4), 319-336.
St. Louis, K., & Hinzman, A. (1988). A descriptive study of speech, language, and hearing
characteristics of school-aged stutterers. Journal of Fluency Disorders, 13, 331-355.
Studdert-Kennedy M. Speech perception. In: Lass NJ, editor.Contemporary Issues in
Experimental Phonetics. New York: Academic Press; 1976. pp. 243–293.
Tasko, S. M., McClean, M. D., & Runyan, C. R. (2007). Speech motor changes coincident with
stuttering treatment. Journal of Communication Disorders, 40, 42–65.
Van Riper, C. (1982). The nature of stuttering (2nd ed.). Englewood Cliffs, NJ: Prentice-Hall.
Von Tiling, J. (2011). Listener perceptions of stuttering, prolonged speech, and verbal avoidance
behaviors. Journal of communication disorders, 44(2), 161-172.
Wingate, M. E. (1969). Sound and pattern in “artificial” fluency. Journal of Speech, Language,
and Hearing Research, 12(4), 677-686.
Wingate, M. E. (1970). Effect on stuttering of changes in audition. Journal of Speech and
Hearing Research, 13(4), 861.
Wolk, L., Conture, E. G., & Edwards, M. L. (1990). Coexistence of stuttering and disordered
phonology in young children. South African Journal of Communication Disorders, 37,
15-20.
Woods, C. L. (1978). Does stigma shape the stutterer? Journal of Communication Disorders, 11,
483-487.
Van Riper, C. (1982). The nature of stuttering. Engleulood Cliffs, NJ: Prentice-Hall.
Yaruss, J. S., & Conture, E. G. (1996). Stuttering and phonological disorders in
children.Examination of the Covert Repair Hypothesis. Journal of Speech, Language,
and Hearing Research, 39(2), 349-364.

62