Chapter II

Review of Literature

The review of literature is covered under the following headings.

1. Disfluency and Dysfluency

2. Major studies on normal fluency and abnormal fluency

3. Frequency of speech disfluency

4. Type of speech disfluency

5. Duration of speech disfluency

6. Factors affecting disfluency

1. Disfluency and Dysfluency

“The prefix “dis” denotes separation, negation or reversal; the core reference is

“apart,” the simplest equivalent is not. The prefix signifies a contrast with

whatever the suffix may be. Words containing “dis” as a prefix, thus are very

general referents. In case of “disfluency” the referent is “all kinds of speech

features which contrast with fluency”” (Wingate, 1984). According to Hegde

(1978), fluency is speech that is free of all forms of disfluency. A complete

understanding of the disorder of stuttering is largely dependent on the

understanding of the characteristics of fluency and the possible variations or

disruptions in fluency which can be put under the general category of

“disfluencies”

The prefix “Dys” refers to something that is abnormal. This prefix has been used

widely in medicine to refer to the anomalies in organic structure and function.

“Dysfluency” signifies abnormality of fluency; it includes but is not limited to

13
 stuttering” (Wingate, 1984).

According to Wingate (1984) “Fluency” is an illusion, a fact borne out by a

considerable amount of research on normal speech. Speech perceived as normal

fluent speech typically contains a variety of “disfluencies”; in fact it is

characterized by “disfluencies” other than those appropriately referred to as

instances of “disfluency.”

2. Major studies on normal fluency and abnormal fluency

The focus in disfluency and stuttering research is to investigate relationship, if

any, between stuttering and normal disfluency in young children. The

fundamental question addressed is whether any “natural lines of demarcation”

exist between the disfluency of those who stutter and those who are considered

normal speakers. This would depend on the definition of abnormal fluency and

fluent speech. If “abnormal” is viewed as qualitatively distinct from normal, then

stuttering cannot be defined in terms of any feature of a child‟s speech, and

therefore has to be defined at least in part on the basis of the evaluative

judgments of listeners. However if a relative, quantitative, or dimensional

concept of abnormality is accepted, stuttering in young children can be

considered as a relatively severe degree of normal disfluency. The first option is

embodied in Johnson‟s (1959) avoidance or interaction hypothesis. The second

forms the basis of continuity hypothesis advocated by Bloodstein (1970).

The Avoidance Hypothesis was advocated by Johnson (1959) which played a

pivotal role in his diagnosogenic theory of stuttering. His theory stated that

certain parents or other adults are unusually anxious or perfectionistic about the

14
child‟s speech fluency. Such adults tend to react to the child‟s normal

disfluencies as speech impairment. The child may then develop anticipatory

avoidance reactions through anxiety about the speech hesitancy. Before its

diagnosis as stuttering, this hesitancy may be of greater or lesser degree

depending upon factors of environment, personality or heredity, but in most

cases according to this theory, it does not differ markedly in amount or quality

from that of other children who do not come to be regarded by their parents as

stutterers. Thus according to Johnson, stuttering is largely a disorder in the

evaluations placed upon a child‟s fluency. The avoidance theory says that if one

child‟s speech repetitions are normal, there is little justification for calling

another child‟s speech repetitions anything else just because there are more of

them. An important implication of this theory was that a relatively large amount

of disfluency was to be found in the speech of normally developing children.

According to the Continuity Hypothesis, if parents of stutterers and non

stutterers tended to describe the same disfluencies in their children‟s speech, this

might be because stuttering and certain aspects of normal disfluency in young

children are not categorically different. This differed from Johnson‟s advocation

that stuttering is something that can be sharply differentiated from normal.

According to this view, Bloodstein (1960a, 1970, 1975) discussed that mild

tensions and fragmentations are an ordinary feature of the speech of young

children as a result of communicative pressures and difficulties. When the

tensions and fragmentations become intensified by communicative pressures or

failures that are severe and chronic, they tend to be identified as episodes of

stuttering. This view emphasizes that there is an overlap between stuttering and

15
normal nonfluency and that “there are no “stutter-like” disfluencies that are not

to be found in the speech of many nonstutterers” (Bloodstein, 1975). Shames

and Sherrick (1963) advanced a continuity hypothesis in the context of their

analysis of stuttering and normal disfluency as operant behavior. They suggested

that stuttering may be normal disfluency that has increased in frequency due to

inadvertent reinforcement by individuals in the child‟s environment. The

continuity hypothesis says that if one child‟s speech repetitions are stuttering,

there is little reason to call another child‟s repetitions anything else merely

because there is relatively normal amount of them.

The two hypotheses have a vast distinction between them in their ultimate

implications for theory and therapy. First of all, the source of the child‟s problem

lies in the perceptual and evaluative problems of a listener as a result of which

the child comes to regard disfluencies as a matter for concern and to struggle to

avoid them. Second, the causes of the problem lies largely in the nature of the

stuttering that are latent in almost all children‟s speech and in the variety of

possible factors that might tend to increase them.

One of the pioneering studies on disfluencies seen in children with stuttering and

their normally fluent counterparts comes from the research carried out by

Johnson and Associates (1959). In the third phase of their study, the authors

investigated the disfluencies produced by 89 children with stuttering in the age

group of 2.5 to 8 years and an equal number of genders and age matched normal

speaking children. The data were analyzed separately for the two genders and for

the experimental and control groups. Their findings showed no statistically

significant difference in both groups with respect to gender. The authors

16
suggested that their data may be used as normative reference in evaluating

appropriately derived measures of nonfluency of children within the indicated

sex, age and socioeconomic groups. They emphasize that “nonfluencies occur in

the speech of children generally, and that their distribution is dimensional rather

than categorical. There are no “natural” lines of demarcation between “normal”

and “abnormal” degrees of nonfluency.” (Johnson and associates, 1959)

Another finding was with respect to differential frequencies of types of

nonfluency. The major portion of nonfluencies of the experimental group male

subjects, were sound and syllable repetitions, word repetitions and interjections.

Prolonged sounds, revisions and phrase repetitions occurred to a lesser degree.

In, females of the experimental group, although findings were similar, greater

proportions of interjections than sound and syllable or word repetitions were

found. Slightly greater than three fourths of nonfluencies of both male and

female subjects in the control group were interjections, revisions and word

repetitions, with sound and syllable repetitions occurring with much lesser

frequency.

Group differences revealed significant difference (p< 0.01) for the frequency

index measures of male experimental and control groups for types such as

sound and syllable repetition, word repetition, phrase repetition, prolonged

sounds and for all the categories combined and broken words (p< 0.05). In all

these, means were higher in the experimental group compared to the control

group. No significant differences were found for types of interjections, revisions

and incomplete phrases. In case of females, significant differences were observed

for sound and syllable repetitions, word repetitions and all categories combined

17
(p< 0.01) and prolongations (p< 0.05). Again in all these experimental group

means were higher than the control group. No significant differences were

obtained for interjections, phrase repetitions, revisions, incomplete phrases and

broken words.

Although Johnson‟s study has provided very useful and relevant information, it is

criticized on the fact that (1) there are large differences between two and eight

year old children in almost all areas of development which makes analysis of

group data from children in this age range invalid and (2) the prolonged interval

from onset reported for the stuttering in their study (approx. 18 months) makes it

hard to consider the features of their disfluent speech to be that of early stuttering

(Zebrowski, 1995). Hence in later studies (e.g., Yairi and Lewis, 1984, Yairi and

Ambrose, 1992, Yairi, Ambrose and Niermann, 1993) several researchers

attempted to objectively examine the speech of children within a much narrower

age range and within a small interval from the reported stuttering onset.

Yairi (1981) studied the spontaneous speech outputs of 33 children, 18 girls and

15 boys ranging between 24 and 33 months. Speech samples (of 500 words

length) were transcribed verbatim. Various types of disfluencies were identified

and classified as given by Johnson (1961) and Williams, Silverman and Kools

(1968). However, word repetitions were further differentiated as single syllable

words and poly syllable words. The frequencies of occurrence of each disfluency

type and of the total number of disfluencies were tallied to provide individual

and group data. The author found that the total number of disfluencies varied

widely across subjects, from 0 to 25.6 disfluencies per 100 words spoken. Four

types of disfluencies - part word repetitions, single-syllable word repetition,

18
interjection and revision were observed. The standard deviations were large,

exceeding group mean values in several instances, which indicate that 2 year

olds are extremely heterogeneous in respect to disfluency. Also, boys were found

to be more disfluent than the girls. The largest differences occurred in part word

repetitions and interjections. However, using two-tailed t tests (p= .05; df=31) no

statistically significant gender differences were identified in specific disfluency

types or in the total number of disfluencies. The author also studied the

percentages of subjects who uttered each disfluency type within certain

frequency ranges. The data showed that although all types of disfluency were

represented in the speech of 2 year olds, not all the subjects uttered all the types.

Over 80% of the subjects did not exhibit even one instance of polysyllabic word

repetition. Also, for 29 children, the percentage of single-syllable words in the

total words which they repeated exceeded the percentage of single-syllable words

which they uttered in their speech samples. The authors made three important

observations. First, more than half of the subjects exhibited two types of

disfluencies - Disrhythmic Phonation and Tense Pause - which were not

included in the previous studies of 2 year olds. Second, relatively high

percentage of subjects did not use revisions in their speech. Third, higher

frequency of part-word repetition was exhibited by boys compared to girls. In

addition to the frequency, the author also examined the extent of each instance of

disfluency, i.e., how many times did subjects repeat a particular disfluency?

Repetition units were defined as the number of utterances in excess of the first

production in case of part word, word or phrase repetition. In case of interjection,

the number also included the first production. It was found that although there

were isolated instances of up to 5 repetition units, on an average, instances of

19
part-word, word or phrase repetition were short, consisting of the original

uttering of a segment followed by one extra production. More repetition units

occurred more often on single syllable words, interjections were also almost

exclusively events of single uttering of sounds. However, interjections were

sometimes prolonged by uttering a succession of different sounds.

Yairi (1982) analyzed spontaneous verbal output of 500 words, of 33, 2 year old

children (18 girls and 15 boys). The subjects were divided into two groups.

Subgroup I consisted of 13 younger children in the group (24-26 months).

Subgroup II comprised of the older children (10 boys and 10 girls), in the age

between 29 to 33 months. Four spontaneous speech samples were taken for each

in subgroup I and three samples for each in subgroup II were recorded at 4-

month interval. The classification of disfluencies used by Yairi (1981) was used

in this study. Data was calculated to compute frequency of disfluency per 100

words. Since no significant gender difference was reported in the previous study,

data for boys and girls were combined. Results showed that for the entire group,

there was a reduction in disfluency, throughout the 8- month span during which

three speech samples were recorded. The average total number of disfluencies,

declined from 6.5 disfluencies per 100 words at the beginning of the study, (29

months) to 5.10 (33 months) to 4.1 (37 months). Subgroup I revealed a general

trend of increase in the number of disfluencies from the beginning to the end of

the third year of their life. However, the subjects in Subgroup II showed a sharp

decline in disfluencies during the 8 month period. To summarize, results showed

that the number of disfluencies at the beginning of third year, is low, but

gradually increases with time. The normal disfluency peaks in the latter part of

20
the year. Data also showed that increase in disfluency for Subgroup I was mainly

due to two disfluencies – Revision and Phrase Repetition. In subgroup II, there

was consistent decline across all disfluency types but noticed more for Part word

repetitions and Interjections. Hence “taken together, the overall picture obtained

from these observations, is that the year between the age 2 and 3 is a volatile

period in speech development as far as the parameter of disfluency is

concerned.” (Yairi, 1982). This period is marked by instability in the fluency of

ongoing speech and brief episodes of fluency disruption, lasting up to several

weeks may occur in some children, which may be regarded as parents or other

observers as stuttering (Yairi, 1982). The author concluded that two year old

children should not be treated as a homogenous group in studies of disfluency.

The above studies included English speaking children in the age range of 24 to

33 months. Wexler and Mysak (1982) studied the disfluency characteristics in 36

non stuttering males aged 2, 4 and 6 yrs (12 in each age group) (types of

disfluency and relationships among the different types). Speech samples of 100

utterances were transcribed from tape recordings of these children. They

identified and classified a total of 7 disfluency types according to revision of

(Johnson‟s (1961) classification system. Total frequency of various disfluency

types was investigated. In addition, two kinds of compound disfluencies-

clustering, a term used by Silverman (1969) to describe the occurrence of more

than one disfluency on the same word or consecutive words or both, and

oscillation, a term used by Mysak (1978) to describe the number of repetitions

per instance of disfluency - were also investigated. Frequency of disfluency was

measured as percent for 100 words. Within each age group the frequency of

21
individual disfluency types were compared. The authors reported that the two

most commonly occurring disfluency types at each age levels were revision-

incomplete phrases and interjections. This was followed by frequency of Phrase

repetitions and word repetitions. The least occurring type was part-word

repetitions for 2 and 4 year olds. Disrhythmic phonations, part-word repetitions,

phrase repetitions, and word repetitions occurred least frequently for 4 and 6 year

olds. Further, a comparison of the disfluency types was made across the different

age groups. The results indicated that 2 year olds had significantly higher

disfluency frequencies for word repetitions and phrase repetitions compared to 4

or 6 year olds. Also 2 year old children showed a higher frequency of

disrhythmic phonations compared to 6 year old children. The authors reported a

developmental decrease with respect to clustering and oscillation, but the

differences between the age groups were not statistically significant (supporting

studies by Wexler, 1978, 1982). Also, high variability between subjects was

noticed in 2 year old children compared to the older age groups on most

measures. The authors concluded that (1) certain disfluency types occur more

frequently than others within each age group, (2) Patterns of disfluency types had

almost similar distribution across these ages, except for greater frequencies for

certain types in the speech of 2 year olds and (3) Significant differences were

observed among the age levels studied in the frequency of occurrence of

particular disfluency types, which supported the concept of developmental

changes in disfluency.

Results of Pearson product moment correlations indicated highest correlations

between revision-incomplete phrase and interjections, between revision--

22
incomplete phrase and word repetitions and between phrase repetitions and

disrhythmic phonations. Also, low correlations were found between interjections

and phrase repetitions, between tense pause and word repetitions, and between

tense pause and interjections. Significant correlations between pair wise

combinations of part word repetitions, word repetitions and disrhythmic

phonations, between part word repetitions and phrase repetitions, although part

word repetitions correlated with word repetitions and word repetitions with

phrase repetitions were also noted. The authors concluded that “there is a range

in the relationships between individual disfluency types. Certain clusters of

disfluency types exist inviting speculation about different sources for the

different clusters of disfluency types. A linguistic factor could be responsible

with certain disfluency types, as opposed to a motor factor in other types. The

moderately high correlations between types that appear, intuitively, to have a

linguistic component i.e., between revision-incomplete phrase (RIPs) and

interjections and between RIPs and word repetitions tend to support the linguistic

hypothesis. The relatively lower disfluency frequencies observed for the types

that may appear, to have a motor component, i.e., part-word repetitions and

disrhythmic phonations, especially in the speech of the youngest age group, tend

to support the motor hypothesis (Wexler and Mysak, 1982).

DeJoy and Gregory (1985) identified and studied nine types of disfluency, in 60

non stuttering males aged 3.5 and 5 years of age. The speech samples were

transcribed verbatim and the following disfluency types were identified (1) part

word repetitions, (2) word repetitions, (3) phrase repetitions, (4) revision, (5)

interjections, (6) incomplete phrases and (7) disrhythmic phonations (Williams,

23
Silverman and Kools, 1968). Two types of pauses grammatical and

ungrammatical of one second or longer in duration were identified using a stop

watch. The frequencies of disfluencies were computed and expressed as the

number of occurrences per 100 words. The results indicated that certain

disfluencies commonly associated with young children‟s speech (i.e., repetitions,

incomplete phrases and disrhythmic phonations) declined significantly. Two

disfluency types that discriminated significantly between the older and younger

children were part word repetitions and disrhythmic phonations, which the

authors hypothesize to reflect “motor factor” (supporting study by Wexler and

Mysak, 1982). These two disfluency types correlated significantly in the older

group, suggesting that they vary independently in younger children but together

may reflect a less mature speech motor system in the later preschool years. They

also found that interjections and ungrammatical pauses did not decline

significantly across the ages, suggesting that these are disfluencies that may

characterize more of adult like speech.

Disfluencies in 10 male and 10 female black first grade children (5.6 to 7.3

years) was examined by Brutten and Miller (1988). They found that frequencies

of eight types of measured disfluencies varied considerably. The mean number

of disfluencies per 100 words ranged from 2.74 to 0.365. Interjections and

broken words were the most and the least frequently seen disfluencies. Whole

word repetitions were more than both part-word and phrase repetitions. The

main effect for sex was not found to be statistically significant. Girls evidenced

slightly more of all types of disfluencies except prolongations. However, the

overall amount of disfluency was quite similar. Whole word repetitions and

24
revisions were noticeably seen more in girls than boys. The interaction between

sex and type of disfluency was not statistically significant. The authors imply

that their findings may put to rest the belief that greater disfluency among boys is

one of the causative factors of stuttering.

The above major studies were carried out to provide information about

disfluency characteristics of normally speaking children. There were more

studies which employed a comparison of speech disfluencies of children with

stuttering and normally speaking children to help a speech clinician determine

whether a given child‟s speech disfluency is within normal limits. According to

Adams (1977) from a clinical view point, “data on speech characteristics of

young children, just beginning to stutter are needed to develop meaningful

norms, that can be used in refining and validating differential diagnostic schemes

of early childhood stuttering”.

Yairi and Lewis (1984) collected speech samples from 10 2-year old and 10 3-

year old children, (5 boys and 5 girls) with stuttering and 10 matched normally

speaking children. They analyzed 8 disfluencies, in their subjects‟ speech of

approximately 500 syllables long. They found that overall stutterers were three

and a half times more disfluent than the control subjects. (M=21.54 for

stutterers, versus, M= 6.16 for the non stuttering group). According to them, in

spite of large variability within each group, stutterers and non stutterers could be

recognized as belonging to two distinct population samples. Their data showed

that normally speaking 2 and 3 year olds, exhibited almost all types of

disfluencies. Mostly they exhibited interjections, revisions and single unit part

word repetitions. They found an overall increase in the number of disfluencies in

25
the stuttering group. This increase was most significant for part word repetition

and disrhythmic phonation. Also, stutterers repeated part words twice or more

per instance, while non stutterers rarely repeated a segment more than once.

Meyers (1986) studied the disfluencies produced by 24 young stuttering and

nonstuttering boys (4 to 5 years old) interacting in three dyadic sessions. Twelve

subjects were stutterers and 12 were non stutterers. The three sessions included

each child interacting with his own mother, an unfamiliar mother of a stutterer,

and an unfamiliar mother of a nonstutterer. Exact conversations (10 minutes)

were transcribed from the videotaped play periods. Nonfluent words were

counted and categorized as stuttering types or normal types of disfluency.

Stuttering types included part-word repetitions, prolongations, broken words,

and tense pauses. Normal disfluency types included whole-word and phrase

repetitions, revisions, incomplete phrases, and interjections. Percentages of

disfluent words were calculated by dividing the number of disfluencies (of a

given type) by the total number of words. Results of the study showed that,

averaged across sessions, stutterers exhibited a higher percentage of stuttering

behaviour (M = 13.5%, SD = 6.4%) than did nonstutterers (M = 0.2%, SD =

0.4%). Stutterers were quantitatively and qualitatively different in their

stuttering types of disfluency. Stuttering children produced significantly more

part-word repetitions, prolongations, and tense pauses than did nonstuttering

children. The most common stuttering type of disfluency for the stutterers was

part-word repetitions (M = 8.7%, SD = 4.0%). Prolongations (M = 3%, SD =

3%), tense pauses (M = 1.6%, SD = l.0 %), and broken words (M = 0.l%, SD =

0.3%) occurred less frequently. Nonstuttering children had few part word-

26
repetitions (M = 0.2%, SD = 0.4%) and emitted no prolongations, tense pauses,

or broken words. In terms of the normal disfluency types, the nonstutterers had

significantly more whole-word repetitions (M = 1.1%, SD = 1.1%) than did

stutterers (M = 0.4%, SD = 0.5 %,) and more revisions (M = 0.7%, SD = 0.6%)

than did stutterers (M = 0.2%, SD = 0.4%). Stutterers (1.8%) and nonstutterers

(1.1%) did not differ in the emission of other normal types of disfluency (i.e.,

phrase repetitions, incomplete phrases, and intejections). The author also found

that there were no differences between the children interacting with the different

mothers on the speech and fluency measures. Meyers concluded that in terms of

stuttering behaviours, her results support the hypothesis that stuttering children

(4 to 5 years) are quantitatively and qualitatively different from nonstutterers.

Stuttering children with severe disfluency problems differed quantitatively from

those with moderate problems. Total frequency of disfluency was highly

consistent for both groups of children in all the three sessions except for two

disfluency types (tense pauses and whole-word repetitions) which showed

significant variability across sessions. This was the first report on stuttering type

and normal type disfluencies.

Zebrowski (1991) studied the frequency and duration characteristics in a 300

word conversational sample of 10 stuttering and 10 non stuttering children (mean

age-4 yrs). The time since onset of stuttering, was 1 year or less as reported by

the mothers. They found stuttering children to be more disfluent than their non

stuttering peers. The average frequency of all speech disfluency (M=13%), per

100 words, was significantly higher than that of non stuttering children. (M=5%).

The mean duration of stutterers and non stutterers sound/syllable repetitions

27
was 556 ms and 520 ms respectively. The mean duration of sound

prolongations of stutterers and non stutterers was 435 ms and 404 ms. Also,

with respect, to sound/syllable repetitions, they found stuttering and non

stuttering children averaged 2.4 and 2.1 repeated units per instance of

repetition. In word repetitions, it was 2.3 and 2.2 for stuttering and non

stuttering children, respectively. They found no significant difference between

the two groups in either the average duration of sound/ syllable repetitions and

sound prolongations or the average number of repeated units. The author

examined frequency, duration, and repeated units. However, type of disfluency

was not examined.

The above studies have several limitations. The number of children included was

small, children were within wide age ranges and the size of the speech samples

was small. To explain, Johnson and associates used a wide age range from 2 to 8

years, Yairi (1981, 1982) used 33 children, Wexler and Mysak (1982) used only

12 children in the age group of 2, 4 and 6 years and De Joy and Gregory (1985)

studied children in the range 3.5 to 5 years. Wexler and Mysak used only 100

utterances while Yairi and Lewis used only 500 syllables for disfluency analysis.

Meyers (1986) used samples of only 10 min length. These diversities among

studies prompted Ambrose and Yairi (1999) to provide a normative reference for

early stuttering which would provide a basis for the differential diagnosis of

stuttering from normal disfluency. They also attempted to discuss the possible

differences in gender and age with respect to disfluencies in the preschool

children. Speech samples were recorded from 90 children with stuttering in the

age range of two to five years within six months of stuttering onset and from 54

28
age matched normally fluent children. They used a six category classification

system for the disfluencies. These included (1) part word repetitions, (2) single

syllable word repetitions, (3) disrhythmic phonation (comprising of

prolongations, blocks and broken words), (4) interjections (5) revision or

abandoned utterances and (6) multisyllable or phrase repetitions. The first three

types were combined to form Stuttering like Disfluencies (SLD) and the second

three categories formed Other Disfluencies (OD). For part word and single

syllable word repetitions, if a segment was repeated more number of times, the

number of extra times it was repeated was computed as Repetition Units (RU).

The authors found that the main difference between the groups was determined

by the proportion of SLD. These are thus the types of fluency disruptions that are

most typical in persons with stuttering. For the group with stuttering, SLD

constituted 66% of the total disfluency whereas in the normal group it was 34%.

Other disfluencies constituted 34% of the total disfluency for the stuttering group

and 76% for the control group. Part word repetitions occurred with maximum

frequency in the stuttering group while in the control group, interjections and

revisions provided major contribution to the disfluency count. No significant

differences on gender differences within and between the experimental and

control groups were found. Also, no statistically significant age difference was

found for both the groups. However, part word repetitions and repetition units

tended to significantly decrease with age. Other disfluencies tended to increase

with age, but statistical significance could not be obtained.

Ambrose and Yairi (1999) also developed a weighted measure for SLD which

was the weighted sum of part and one syllable word repetitions and disrhythmic

29
phonation per 100 syllables (repetitive disfluencies * mean number of iterations

+ disrhythmic phonations * 2). Using this measure, with a cut off of 4%, all

children were correctly assigned to their respective fluency groups i.e., weighted

SLD scores were found to be continuous across age groups and lie below 4.00

for all the control group and above 4.00 for the experimental group.

Pellowski and Conture (2002) investigated disfluencies in 72 native English

speaking children ranging from 3 to 4 years of age of which 36 children had

stuttering and the remaining 36 were age and gender matched normal speakers.

They measured SLD, OD, total disfluencies (TD) and the number of repetition

units (RU). They also calculated the weighted SLD measure as described by

Ambrose and Yairi (1999). Results showed a statistically significant difference

between the groups for stuttering like disfluencies and total disfluencies. 81%

and 42% of the total disfluencies comprised of SLDs in the experimental and

control groups respectively while, OD occupied 58% of the total disfluencies in

the controls and 19 % in the group with stuttering. For the weighted SLD

measure, results showed significant overall difference between the two groups.

The mean weighted SLD measure was 20.1 and 1.2 in children with stuttering

and normals, respectively. Hence in this study, a weighted SLD value of 4% and

unweighted SLD of 3% could almost correctly classify 97% of the children as

stuttering or normal.

All the above studies are in English. However, stuttering is a fluency disorder

observed across languages and cultures (Bloodstein, 1995, Cooper & Cooper,

1998; Shapiro, 1999, Van Borsel, Maes & Foulon, 2001; Van Riper, 1971). One

must be cautious in generalizing findings of English – speaking children to other

30
linguistic and cultural groups. In this context, Carlo and Watson (2003) studied

the disfluent speech of 32 normally fluent monolingual, Spanish speaking

children from Puerto Rico, The total frequencies and types of speech disfluencies

were examined in 15 children (8 girls and 7 boys) aged 3.5 to 4 years and 17

children (8 girls and 9 boys) aged 5.0 to 5.5 years. The study examined the

influence of gender and age on the speech disfluencies, the amount and type of

and comparing the results of their study with those using English speaking

children. Spontaneous speech samples were elicited during a free play interactive

activity which yielded a sample size between 500 and 800 syllables. The first 50

syllables were omitted from analysis. Elicited samples were transcribed

verbatim. Disfluencies were identified in each sample using adaptations of

classification systems described by De Joy and Gregory (1985), Yairi (1981) and

Campbell and Hill (1987). For the inter observer reliability, another qualified

speech language pathologist reviewed 20% of the samples. Means, standard

deviations, and ranges of total disfluencies per 100 syllables were computed for

boys and girls in the younger and older groups. Findings suggested that means

of total disfluencies ranged from 5.21 for the younger boys to 8.26 for the older

girls. With age and gender groups combined, the frequency range of total speech

disfluencies was between 3.0 and 5.99. Results showed no statistically

significant difference in the total speech disfluencies exhibited by boys and

girls or by the younger and older age groups. Interaction effect between age

and gender was also not seen. This differed from the earlier findings of English

studies that there is a decrease in the total disfluencies with an increase in

chronological age (DeJoy & Gregory, 1985; Wexler, 1982; Wexler and Mysak,

1982; Yairi, 1997). The absence of gender effect was in congruence with earlier

31
reports of English speaking children (Ambrose and Yairi, 1999; Haynes and

Hood, 1977; Kools and Berryman, 1971; Yairi, 1981, 1982; Yairi and Lewis,

1984). High amounts of variability were also observed in the types of speech

disfluencies within each age group (represented by the high standard

deviations). They also found no statistically significant difference in the

proportions of disfluency types exhibited by the two age groups. Rank orders of

the disfluency types revealed that for both groups, single-syllable word

repetitions, revisions, and interjections were the most frequent disfluency types.

Broken words, blocks and repetitions of more than one syllable were less

frequent types. They also did not find any significant difference in frequency of

total speech disfluencies between boys and girls. The authors thus suggest a

possibility that the relationship between age and disfluency may be different for

Spanish and English speaking children.

Natke, Sandrieser, Pietrowsky and Kalveram (2006) studied the speech

disfluencies of 24 German-speaking preschool children who stutter and 24

gender and age matched comparison children who do not stutter. They wanted to

describe how German-speaking preschool children who stutter and do not stutter

display stuttering like and normal disfluencies including number of iterations, to

explain how powerful classification measures for the diagnosis of stuttering are

for German-speaking pre-school children and to see how disfluency patterns of

native English and German speaking children close to the onset of stuttering

differ. Speech samples were both audio and video recorded. Speech samples

were transcribed orthographically and analyzed using the computer program

CLAN (MacWhinney, 2000), where a special post-coding system for disfluencies

32
was added. To avoid warming up effects, at the beginning of the session, the first

100 syllables of each speech sample were excluded from disfuency analysis.

Speech samples collected over two sessions constituted a minimum of 1000

syllables. Five types of disfluencies: prolongation, blocks and repetitions of

sounds, syllables and one syllable words were defined and grouped as Stuttering-

like disfluencies (SLD), (Yairi, 1997). Multi-syllable word repetitions, phrase

repetitions, interjections, revisions and interrupted utterances were defined and

combined as other disfluencies (OD). For repetitive disfluency types, the number

of iterations were coded and counted separately. For analysis, a syllable based

metric was used. Percentages for each disfluency type and the number of

iterations were calculated for each participant and were used to derive group

means. Wilcoxon signed rank tests revealed that differences between mean

frequencies of SLD and OD were not significant.

Disfluency data showed that CWS and CWNS showed all types of SLD. All

types of SLD, as well as grouped SLD, were shown more frequently by CWS

than by CWNS and this was statistically significant (p<0.001). For OD, however,

there were no differences between the groups, with the exception of multi-

syllable word repetitions, which were shown more frequently by CWS than by

CWNS. With regard to repetition units, CWS showed a significantly higher

number than CWNS in SLD repetitive disfluencies (sound, syllable and one

syllable repetitions) with p-values < .009. OD repetitive disfluencies (phrase and

multi syllable word repetitions) did not differ significantly in the number of

iterations (p> .28) between CWS and CWNS. A weighted SLD measure was

additionally calculated according to Ambrose and Yairi (1999). With a cut off

33
of 4% of the weighted SLD measure, three CWS are below the cut off and no

CWNS above.

Their results were consistent with reports for English speaking children, of this

age group (Ambrose & Yairi, 1999; Pellowski and Conture, 2002) in that specific

types of disfluencies for CWS and CWNS differ significantly at a very early age

and all disfluency types classified as SLD are produced significantly more often

by children who stutter than by children who do not stutter. However, the

difference between part word repetitions and other disfluency types was larger

for English samples than in German samples. The overall frequency of SLDs

from the English samples was found to be higher (11%, range 3-40%) than those

produced by German speaking children (9.2%, range 2-25%). A cut off of 3%

SLD was shown to be a powerful measure for the diagnosis of stuttering in

German speaking children. This was the same cut off used by Pellowski and

Conture (2002) for English speaking children. The overall results were consistent

with the findings of Ambrose and Yairi (1999) that “very early stuttering is

distinct from normal disfluency” and “therefore it cannot be said that all children

go through at least a brief period of “stuttering””.

Boey, Wuyts, Heyning, Bodt and Heylen (2007) conducted a study to compare

the characteristics of stuttering-like disfluencies in a group of Dutch speaking

children who stutter (n= 693) with a group of normally speaking children (n=79).

Their findings suggested that the frequency of SLD in the stuttering group (M=

15. 71) was significantly higher than in the control group (M= 0.42); 3% SLD

criterion, for distinguishing children who do from children who do not stutter

resulted in high sensitivity (0.95) (i.e., the proportion of individuals who

34
stutter) and high specificity (0.98) (i.e., the proportion of individuals who do not

stutter); approximately 77% of the children who stutter exhibited some form of

prolongation or blocks but none of the normally fluent children produced

prolongation or blocks; duration of SLD was longer for stuttering children than

non stuttering children and physical tension during SLD was estimated higher for

stuttering than non-stuttering children.

Characteristics of stuttering like disfluencies in 8 monolingual Portuguese school

age children (mean age 10 years) was studied by Valente and Jesus (2011). They

also compared it with an equal number of normally fluent children. An inclusion

criterion of 3 or more SLD per 100 words was used for the group of children

who stutter. Speech samples were obtained through the reading task. The authors

have developed an assessment tool (Jesus and Valente, 2010), incorporating the

evaluation of factors that characterize SLD i.e. frequency and types of SLD,

duration and physical tension of concomitant symptoms (Guitar, 2006; Riley,

2009). The frequency and types of SLD during reading in both groups were

analyzed. The authors identified a total of 75 SLD in the group of children with

stuttering and 8 in the normally fluent group. They found that SLD were more

frequent in children who stutter than in the normally fluent children. The groups

total SLD was statistically significant. However, the results for monosyllabic

whole word repetitions and sound/syllable repetitions (SLDs observed in both the

groups) were not. The authors also found that the mean number of repetition

units for the group with stuttering was higher than for the normal group, a

finding in congruence with studies by (Ambrose and Yairi, 1999) and Boey et

al. (2007). The mean duration of different types of SLD was higher in first

35
group (stuttering) than the controls. However, statistically significant

differences could not be obtained. In their study, the authors also make a

reference to the physical tension associated with SLD. They found some amount

of physical tension associated with SLD in the stuttering group and no tension

was seen in the normally fluent group. The authors inferred from their results that

type, frequency and physical tension of SLD are parameters which differentiate

children who stutter from matched normally fluent children. Duration and

repetition units do not differentiate the two groups.

In the Indian context, Indu (1990), Nagapoornima (1990), Yamini (1990) and

Rajendraswamy (1991) proposed a fluency test each in different age groups in

Kannada, a south Indian Dravidian Language. This was based on disfluency data

of 12 children in the age groups of 3-4 (Nagapoornima), 4-5 (Indu), 5-6 (Yamini)

and 6-7 (Rajendraswamy) years. Simple pictures, cartoons and pictures depicting

Pancatantra stories were used in these tests. A total disfluency of greater than 25

– 30 % was considered to be abnormal. The percent disfluency is high because

unlike in English where several iterations of sound/syllable are considered as one

instance of repetition, the authors calculated each iteration as one repetition.

Geetha, Karanth, Rao & Ravindra (2000) aimed at classifying and sub grouping

young disfluent children based on family history of stuttering, speech and

nonspeech characteristics, and concomitant speech and language problems;

developing a comprehensive procedure to facilitate early, easy, and objective

differential diagnosis of disfluent children. Artificial Neural Network (ANN), a

computer program, was used for this purpose on two groups (of 25 and 21

children each) of disfluent children below the age of 6 years. The authors

36
developed Disfluency Assessment Procedure for Children (DAPC) in Kannada.

This consisted of historical indicators, attitudinal indicators, behavioral (speech)

indicators, articulation assessment and language assessment to differentiate

between NNF and stuttering. Results of their comparison were - Types of

disfluencies, especially the prolongations and articulatory fixations were present

only in stuttering children. Sound/syllable repetitions were present in both the

groups but their frequency, number of unit repetitions, presence of schwa, and

rate of repetitions were useful in differentiating stuttering children. Pauses, both

audible and inaudible, were present in both NNF and stuttering groups, specially

the audible ones. There was slight over-lap in the frequency of disfluencies

among the NNF and stuttering children when sound syllable repetitions and

prolongations were considered. NNF children had 2–3% disfluencies while it

ranged from 3–30% in stuttering children, with majority showing around 10%

disfluencies; all NNF children had fleeting disfluencies while the stuttering

children exhibited disfluencies ranging from half a second to as long as 60

seconds with majority having disfluencies of 2–9 seconds duration; Physical

concomitants or secondary behaviors were observed only in stuttering children

(in about 60–70%), the most common ones being eye blinking, nose flaring,

grimaces, or frowning; the rate of speech was faster in stuttering children (60–

70%), compared with NNF children; with respect to severity of the problem

based on Stuttering Severity Instrument, there was slight overlap among the NNF

and stuttering children. On the historical and attitudinal factors there was overlap

in scores in about 10% of NNF and very mild stuttering children. In general,

NNF children had scores below 5 whereas most stuttering children‟s range was

between 5–15. “Attitudinal factors (of self and significant others) are important

37
in determining stuttering disfluencies as negative attitudes have important

repercussions” (Geetha et al., 2000); behavioral factors clearly differentiated

stuttering and NNF groups- NNF children had zero scores whereas it ranged

from 3–20 for stuttering children; Also, articulation and language problems were

the most common associated problems found in stuttering children. ANN could

predict the sub classifications of normal nonfluency and stuttering with 92%

accuracy although the size of the training sample (25) was comparatively small.

The authors concluded ANN to be a useful clinical tool for objectifying

diagnostic procedures, with regard to stuttering.

Rathika, Kanaka, Sunila and Rajashekhar (2012) analysed the disfluencies in 48

Typically Developing Tamil Speaking Children between 4 - 8 Years (with 12

children in each age group. A total of seven types of disfluencies were identified.

They included filled pauses (pauses filled with sounds like “mm”…”um” etc);

unfilled pauses (silent pauses having duration greater than 300 ms); repetition

(repetition of sounds, syllable repetition (repetition of syllable), part word

repetition, whole word repetition, phrase repetition., parenthetical remark, false

starts, audible inspiration, and prolongation. Total percentage of disfluencies

and percentage of individual disfluency types were calculated. The results

indicated pauses to have highest percentage of disfluencies in all the age groups;

among the repetitions, whole word repetitions occurred most frequently,

followed by part word repetition; syllable and phrase repetition showed an

increasing trend till 7 years with age and reduced by 8 years; prolongation was

the only disfluency that was not seen in any of the children between 4-8 years

(only one child exhibited between 5-6 years) and the total percentage of

38
 disfluencies ranged from 17.1 (7-8 years) to 30 (5-6 years). However, the

findings of the study were not in congruence with the Western studies.

From the earlier studies, it becomes evident that various characteristics of

disfluencies have been studied both in the speech of children with no stuttering

(CWNS) and children with stuttering (CWS). Analysis of these features may also

aid in the differential diagnosis of normal disfluency from stuttering. These

specific features include -frequency of speech disfluency; type of speech

disfluency and proportion of type and duration of instances of disfluency in terms

of number of repeated units and other temporal aspects of sound, syllable or

word repetitions.

3. Frequency of speech disfluency

“Frequency of disfluency refers to how often a child produces disfluent speech

units (e.g., words or syllables) within a sample of predetermined size or duration

(e.g., the number of disfluent words in 100 words; the number of disfluent

syllables in the total number of syllables produced; the number of disfluent

syllables or words in one minute of conversational speech)” (Zebrowski, 1995).

When the frequencies of all of the child‟s disfluencies are measured, we need to

know how many of them are normal. One of the earliest databases in this regard

was compiled by Johnson and associates (1959). They examined the evidence for

the “diagnosogenic” theory of stuttering. Johnson hypothesized that at the time

the child is first “diagnosed” a stutterer by his parents, the child‟s disfluencies are

no different from those of non stuttering children. These authors compared the

disfluencies produced by 89 male children with stuttering between the ages of

39
2.5 to 8 years of age and a like number of age and gender matched non stuttering

children (68 boys and 21 girls). Results indicated that although there was some

overlap, the children who stuttered were more than twice as disfluent overall

than their normally disfluent counterparts (an average of 18 disfluencies per

100 words compared with 7 disfluencies per 100 words) and produced more of

almost every type of speech disfluency.

Haynes and Hood (1977) included 30 non stuttering elementary school children

where 5 males and 5 females were considered in 3 distinct chronological age

groups- 4, 6 and 8 years. They found that the groups did not differ significantly

on the total frequency of disfluency per 100 words.

Yairi (1981) analyzed the frequencies of occurrence of disfluencies in the verbal

output of 33 2-year old normally speaking children. They found that less than 6

disfluencies per 100 words in about half of the subjects while the other half

uttered between 6 and 13 disfluencies. Also there was maximum concentration of

subjects close to the lower end i.e., less than two disfluencies per 100 words.

In Wexler and Mysak‟s (1982) study, frequency of disfluency was calculated per

100 words spoken for non stuttering male children in the age group of 2, 4 and 6

years. They found that the mean frequency of total disfluency was greater for

two year olds (M=14.6) when compared to 4 and 6 year olds (M=9.1) for both

groups. Also 2 year olds had significantly higher frequencies for disfluencies

such as word repetitions and phrase repetitions than the 4 and 6 year olds. There

was more variability between the 2 year olds (represented by high standard

deviation in scores) than for the four and six year olds.

40
Yairi and Lewis (1984) analyzed the disfluencies in the speech of a group of 10

two and 10 three year old children within two months of stuttering onset, and

those produced by their normally fluent peers. Their results revealed that on an

average, the stuttering children were three and a half times more disfluent

overall than their control counterparts (M=21.54 vs. M=6.16 disfluencies per

100 syllables). Thus, a proportionate increase was found when compared to

Johnson et al.‟s between group comparison measures.

DeJoy and Gregory (1985) found that 3.5 year olds were more disfluent

(M=11.40) than the 5 year olds (M=9.30) in the frequency count of total

disfluencies.

Carlo and Watson (2003) in their study of 15 children (8 girls and 7 boys) aged

3.5 to 4 years and 17 children (8 girls and 9 boys) aged 5.0 to 5.5 years, found

that that means of total disfluencies ranged from 5.21 for the younger boys to

8.26 for the older girls. With age and gender groups combined, the frequency

range of total speech disfluencies was between 3.0 and 5.99. Results were not

significantly different for frequency of disfluencies, across age and gender.

In the Indian studies, a total disfluency of greater than 25 – 30 % was considered

to be abnormal. The percent disfluency is high compared to the reports of

English studies, where several iterations of sound/syllable are considered as one

instance of repetition, the authors calculated each iteration as one repetition.

Although many researchers have measured disfluencies per number of words

spoken, there exists a good argument for measuring disfluencies per number of

syllables spoken. Andrews and Ingham (1971) first recommended the practice of

41
assessing frequency of stuttering in relation to the syllables spoken. Following

their logic, Yairi (1997) noted that as children get older, they are more likely to

use multisyllable words, and it is possible to have disfluencies on more than one

syllable in a word, as in “di-dinosa-sa-saur”. Thus in several studies Yairi et

al.have assessed disfluencies in children as number per 100 syllables (Hubbard &

Yairi, 1988; Yairi & Lewis, 1984; Yairi and Ambrose, 1996). The results of all

these studies indicate that the average, normally disfluent preschool child has

about seven disfluencies for every 100 words spoken. If measured in syllables,

this would be closer to six disfluencies per 100 syllables. This figure might be a

little high throughout the preschool period (Yairi, 1997a), but many children go

through a period of increased disfluency at age 2 or 3 years that will reach this

level. The range in frequency of normal disfluency is important, especially if

frequency of disfluency is used to make a clinical decision. Johnson and

associates (1959) and Yairi (1981) found that although many non stuttering

children have only one or two disfluencies per 100 words, at least one child in

their samples had slightly more than 25 disfluencies per 100 words. Hence,

frequency of disfluencies may not, by itself be a definitive clinical measure.

Additional studies of stuttering frequency relatively close to onset have shown

that in conversation with their mothers, young stuttering children were between

two and half and three and half times more disfluent than age and gender

matched non stuttering peers (Meyers, 1986). Zebrowski, (1991) in his

comparison of 10 stuttering and 10 non stuttering children, found stuttering

children to be almost 2 and half times more disfluent (M=13%) when compared

to normals (M=5%). Pellowski and Conture (2002) found that among 3 and 4

42
 year olds, CWNS showed total disfluencies more than 4 times greater than CWS.

The results of all these studies provide interesting findings. First, reports of the

studies that examined normally fluent children, found frequencies to be different

for different age groups. With younger children (around 2-3 years) showing

higher frequencies compared to the older group (4 – 6 year olds). However, some

report no effect of age or gender on disfluencies. Second, studies that compared

disfluencies of stuttering and non stuttering children, reported that children who

stuttered were at least twice as disfluent as non stuttering children, with average

frequencies of disfluency consistently above 15% of the total number of words or

syllables spoken. Third, with each in each groups studied, there is lots of

variability with respect to the frequency, suggested by the high Standard

Deviation obtained in the scores. SDs are noted to be approximating the mean.

4. Type of speech Disfluency

Frequency of speech disfluency is highly variable both amongst and between the

different ages in preschool children. Hence, on its own it cannot reliably

differentiate between stuttering and normally disfluent children. Therefore it is

necessary to take into account the types of disfluencies that children produce, as

well as the proportions of these different types in order to categorize a child as

stuttering or normally disfluent. There have been several studies (Zebrowski and

Conture, 1989; Conture, 1990 among others) of interest reporting several types

of disfluencies produced by young stuttering and non stuttering children, and

their distribution. Most of these studies report that irrespective of age, or the time

of interval between data collection and onset, there is a substantial overlap

43
between children who stutter and their non stuttering counterparts, in the types of

disfluencies they produce. It has been consistently found that children who

stutter, even very young, ones relatively close to the onset of the problem,

produce more within word speech disfluencies (Zebrowski and Conture, 1989;

Conture, 1990,) than their normally disfluent peers. Within-word speech

disfluencies consist of sound/syllable or part word repetitions and sound

prolongations, and are those in which the child disrupts a connection or

“transition” between sounds within a word (Zebrowski, 1994b).

As per the results of Johnson and associates (1959), 89 stuttering and 89 non

stuttering children in their study differed in mean overall frequency of speech

disfluency (18 disfluencies per 100 words, versus 7 disfluencies per 100 words).

Also, they found that both groups of children essentially produced the same types

of disfluencies. As mentioned earlier it was this finding, along with his analysis

of responses obtained by parents to detailed interview questions, which led to his

formulation of the “diagnosogenic” theory of stuttering onset. He gave a

justification that because there was considerable overlap in the disfluencies

produced by the two groups, and there was no specifically distinctive type of

disfluency produced by either group of children, analysis of child‟s speech would

not lead to a definite diagnosis of stuttering in children. He believed that it is the

parent‟s response and reaction to their child‟s speech disfluencies that give rise to

the “stuttering” label. Based on these interpretations of the data, Johnson

concluded that, “most, and possibly all, very young children speak with sufficient

nonfluency or fluency to be classified as „stutterers‟ or as „normal speakers‟ by

appropriately motivated parents, or other listeners”.

44
Johnson opined that there is similarity in types of disfluencies produced by the

stuttering and non stuttering children. However, a re- analysis of the data

revealed between group differences in the proportions of the different disfluency

types produced. The children who stuttered produced significantly more sound

and syllable, whole word and phrase repetitions, broken words and prolonged

sounds than the non stuttering children. Also, among the nonstuttering children,

interjections, revisions and word repetitions were the most common disfluency

types.

Kools and Berryman (1971), in their study of 92 first grade children, found that

non stutterers showed more interjections than any other type of disfluency.

Word repetitions, revisions and incomplete phrases were also the prevalent types.

The authors opine that “the tendency for non stutterers to exhibit marked

quantities of interjections, word repetitions, revisions, and incomplete phrases,

but to present relatively little disfluency in those categories of disfluency

commonly associated with stuttering behavior, suggests that stuttering behavior

may be qualitatively and quantitatively different from the disfluent speaking

behavior of non stuttering children.”

Yairi (1981) reported, in his 33 2- year old normal subjects, two clusters of

common disfluency types - repetitions of speech segments of one syllable or less

(one syllable words or parts of words were repeated) and interjections and

revisions. The most common disfluency type appears to change as the child

grows older. In a follow-up of his earlier study, Yairi (1982) found that children

between 2 and 3 and 1/2 years showed an increase in revisions and phrase

repetitions, but a decrease in part word repetitions and interjections. Results of

45
Yairi‟s study indicated that as non stuttering child matures, part-word repetitions

decline, even if other disfluency types increase. Further, Yairi suggests that

increase in part word repetitions longitudinally is a sign that may warrant

concern.

Yairi and Lewis (1984) used a modified classification scheme based on

Johnson‟s (1959) eight types of speech disfluency for identifying disfluencies in

ten two and three year old children and their matched controls. They found that

both the groups produced “all recognized types of disfluency”, but in varying

amounts. The most frequently occurring disfluencies in the speech of stutterers,

in ranked order, were part-word repetitions, disrhythmic phonations and single

syllable word repetitions. For the normally speaking counterparts, most

frequently occurring disfluencies in ranked order were interjections, part-word

repetitions and revision-incomplete phrase. Further, the proportion of with-in

word disfluencies differentiated the two groups; with the children who stuttered

producing significantly more part word- repetitions and sound prolongations than

the non stuttering children.

Similarly, Meyers (1986) reported that four – and five- year old boys who

stuttered produced significantly more part word repetitions, sound prolongations

and tense pauses (sometimes referred to as inaudible sound prolongations) than

did normally disfluent boys of the same age. Meyers observed that the non

stuttering children in her study, “emitted no prolongations, tense pauses or

broken words”, an observation which may be related to the age of these children.

According to Zebrowski (1991), “there was no particular type of speech

46
disfluency that was exhibited by either the stuttering or non stuttering children”

when describing the disfluency types produced by young children within 12

months of stuttering onset and a group of age and gender matched normally

disfluent children. However, the findings were concordant with earlier studies, in

that the stuttering children produced more within word disfluecies, overall than

their non stuttering counterparts. Their most frequently produced disfluency

types were sound prolongations and sound/ syllable repetitions, which

significantly differentiated the two talker groups.

Yairi and Ambrose (1999) found that the main difference between the stuttering

and control groups was determined by the proportion of SLD (types of fluency

disruptions that are most typical in persons with stuttering). For the group with

stuttering, SLD constituted 66% of the total disfluency whereas in the normal

group it was 34%. Other disfluencies constituted 34% of the total disfluency for

the stuttering group and 76% for the control group. Part word repetitions

occurred with maximum frequency in the stuttering group while in the control

group, interjections and revisions provided major contribution to the disfluency

count. Weighted SLD scores were found to be continuous across age groups and

score below 4.00 for used to assign subjects in the control group and above

4.00 for the experimental group.

Pellowski and Conture (2002), in their study on 36 stuttering and 36 non

stuttering children, also calculated the weighted SLD measure as described by

Ambrose and Yairi (1999). Results showed a statistically significant difference

between the groups for stuttering like disfluencies and total disfluencies. 81%

and 42% of the total disfluencies comprised of SLDs in the experimental and

47
 control groups respectively while, OD occupied 58% of the total disfluencies in

the controls and 19 % in the group with stuttering. According to them, weighted

SLD value of 4% and unweighted SLD of 3% could almost correctly classify

97% of the children as stuttering or normal.

In summary, variability in the amounts and types of disfluencies exhibited by

young children has been reported by a number of investigators (e.g., De Joy and

Gregory, 1985; Haynes and Hood, 1977; Wexler and Mysak, 1982; Yairi, 1981,

1997). Also, the scheme of stuttering like and other disfluencies, helps to better

understand the overlap of disfluencies seen between normally fluent and

stuttering children. SLDs have significantly found to be more in children with

stuttering than children with no stuttering. OD seems to overlap between the two

groups.

5. Duration of speech Disfluency

Absolute Duration of repetitions and prolongations: Duration of

sound/syllable repetition and sound prolongations is measured from the onset of

the acoustic energy associated with the disfluent initial sound in the word to the

cessation of acoustic energy for the audibly prolonged or final iteration of the

repeated sound/syllable (Kelly and Conture, 1988; Zebrowski and Conture,

1989). In a study of young children within 12 months of stuttering onset,

Zebrowski (1991) reported that the mean duration of these children‟s sound/

syllable repetitions was 556 ms while the instances of sound prolongation

averaged 435 ms in duration. No significant differences were obtained between

these measures and those obtained from a group of matched non stuttering

48
children. Zebrowski (1994) suggested that “the clinical utility of duration lies not

in its use as a distinguishing characteristic of stuttering and normally fluent

children, but its relationship to other aspects of stuttered and fluent speech of

children”. He reported that children producing a high proportion of sound

prolongations also tended to produce longer prolongations and slower (fluent)

articulatory speech rate and therefore these children may be more likely to

require direct intervention for stuttering compared to their normally fluent peers.

Zebrowski (1994) suggested using both stuttering duration and frequency,

when rating the severity of stuttering.

Gorden and Luper (1992a, b), suggested that in several widely used protocols for

identifying early stuttering measure or estimation of stuttering duration has been

considered. Many of these protocols directly or indirectly specify that instances

of stuttering (particularly sound prolongations) of one second or longer

distinguish stuttering from normal disfluency (e.g., Cooper and Cooper, 1985;

Curlee, 1980; Pindzola, 1987; Van Riper, 1982). The important findings from

these studies include (1) children who stutter produce instances of stuttering

which are typically shorter than one second in duration; (2) sound/syllable

repetitions tend to be longer in duration than sound prolongations. However,

Conture and Kelly (1991), Kelly and Conture (1992), Louko, Edwards and

Conture (1990), Niermann, Yairi and Peterson (1993) and Yaruss and Conture

(1993) have all found that for a large part young children who stutter produce

stuttering that are one second or shorter in duration.

Number of repeated units within sound, syllable and word repetitions: The

duration of sound/ syllable part word and whole word repetitions is expressed as

49
the number of repeated units (RU). The number of extra times a segment is

repeated for part and single syllable word repetition is tabulated as repetition

units. This measure of RU within instances of sound, syllable and word

repetitions has been identified as distinguishing characteristics of early stuttering.

According to Sander (1963), the parameter of extent can have a substantial

impact on the perception of stuttering. Hence greater the number of repetition

units, more is the probability of correctly identifying as “stuttering”. However,

support for this notion has been inconsistent.

Branscom, Hughes and Oxtoby (1955) were one of the first to use the concept of

repetition units in the speech of normally nonfluent children. Among the 25 3-

year olds, 79% of word and syllable repetitions contained just one extra unit.

17% contained 2 extra units, 2 % contained 3 units, 1% contained 4 units and 1%

contained 5 units. In Johnson et al.‟s study, combining part and whole word

repetitions for males and females, the average number of syllable and word

repetition units was 1.46 for children who stutter, and 1.09 for the normally

disfluent children. Statistically significant difference was obtained at .01 level.

Adams (1980), Curlee (1980), and Pindzola (1987) all support that

sound/syllable or part word repetitions consisting of more than two repeated

units reflects beginning stuttering, as opposed to normal disfluency.

In Yairi (1981)‟s study on 33 2-year olds, for boys, the mean number of units

ranged from 1.12 (part word repetition), to 1.28 (monosyllabic word repetition).

Girls had a score of 1.08 (part word repetitions) and 1.07 (single-syllable word

repetitions). Yairi and Lewis (1984) reported that the young , beginning

50
stuttering children in their study, produced significantly more repeated units per

instance of part word repetition than did their non stuttering peers (M=1.72

repeated units for the stuttering children and M=1.12 repeated units for the non

stuttering children). The authors also found almost no overlap between the two

groups in terms of the number of repetition units (for part word repetitions). The

range was 1 to 11 and 1 to 2 in children with stuttering and normal controls,

respectively. Also, instances of repeating part words more than once were

infrequent and instances of repeating more than twice was rare in normally

disfluent children. In contrast, Zebrowski (1991) found no significant difference

in the mean number of iterations produced by young children representing the

same two talker groups (M= 2.4 repeated units for stuttering children versus M=

2.1 for the non stuttering children. in her study, the final production of the target

sound was also counted in the repetition units. However, Johnson et al. (1959),

and Yairi and Lewis, (1984) have counted only those segments in excess of the

intended segment.

Ambrose and Yairi (1995) found that the maximum number of repetition units

per instance of disfluency was 17 for children who stutter and 4 among the

normally disfluent children. The mean number of repetition units per instance of

repetition was calculated for each subject. The individual scores were used to

derive group means for males and females in the two groups. The range of

individual means in the stuttering group was nearly 5 units per instance with a

total group mean of 1.70. For the normally disfluent group, the range of

individual means was 1.44 units per instance with a total group mean of 1.16.

The authors reported that the group of children who stuttered had significantly

51
more repetition units per instance than the control group. But, no significant

difference was obtained between genders on repetition units. The ratio of single-

unit to multiple unit repetition was 2.06 in the stuttering group and 6.58 in the

control group.

Pellowski and Conture (2002) found a mean number of repetition units (based on

all productions of part word and monosyllabic whole-word repetitions) to be 2.0

(SD=1) with individual subject mean ranging from 1.0 to 5.5 units of repetition

in CWS, while for CWNS, it was 1.1 (SD= 0.48), with a range of 1.0 to 1.6.

Natke et al., (2006) in their study of 24 German children (5 years or below) who

stutter and 24 controls, reported one iteration in CWNS. CWS showed

significantly higher number of iterations than CWNS in SLD repetitive

disfluencies (sound, syllable and one syllable repetitions), while OD repetitive

disfluencies (phrase and multi syllable word repetitions) were not significantly

different between CWS and CWNS.

Counting Total disfluencies: This simply means counting all the disfluencies

present in a speech sample with no distinction being made as stuttering or

normal. Yairi (1997) stated that “disfluency counts have been the classic metric

of the disorder [stuttering] as well as the dependent variable of interest in both

clinical and experimental studies”. This metric is archetypal as it has not changed

substantially since it was first introduced by Johnson in the late 1950s (Johnson,

1961; Johnson et al., 1959). Here, the clinician / researcher transcribes a speech

sample and then classifies each transcribed disfluency to a particular type. Thus,

in this method, no differentiation is made between normal disfluencies and

52
stuttering. Adams (1977) suggested that a child should be designated as

stuttering, when greater than 10 % of the words spoken are disfluent. Ambrose

and Yairi (1999) emphasized on the proportions of SLDs in relation to other

disfluencies in the diagnosis of stuttering in children. Several later studies

considered variations on Johnson‟s method of classification of disfluencies, as

useful for differentiating between normally disfluent and stuttered speech

(Ambrose and Yairi, 1999; Pellowski and Conture, 2002). Johnson‟s system was

modified by later investigators, thus restricting comparisons among studies that

used its many variants. For example, Gregory and Hill (1993) and Zebrowski

(1991) used 10 disfluency characteristics; Carlo and Watson (2003) used 15

disfluency types. Yairi and Ambrose (1992) used 8, Yairi, Ambrose and

Niermann (1993) used 7, Yairi and Ambrose (1999) used only 6 disfluency

types.

Stuttering Like Disfluencies: The SLDs measure was introduced by Yairi and

Ambrose (1992). According to Yairi (1995), ““Stuttering like Disfluency”

reduces the semantic confusion of labeling disfluencies of non stuttering children

as “stuttering”, and recognizes that not all disfluencies of persons who stutter are

necessarily “stuttering””.

Ambrose and Yairi (1999) reported a normative reference for early stuttering.

They examined speech samples of 90 stuttering children aged 2 to 5 years within

6 months of stuttering onset and from 54 age matched normally fluent children.

Disfluencies in the speech samples were identified and classified and modified as

per the system given by Johnson et al. (1959). Prolongations, blocks and broken

words were combined into a category called disrhythmic phonation. Yairi and

53
Lewis (1984) had modified the system by dividing the category of word

repetition into single syllable word and multisyllable word repetition. Yairi and

Ambrose further eliminated the between word category of tense pause because of

the unusual difficulty in identification and the frequent confusion with external

noises generated by toys as well as speech sounds made by children or adults. As

multi-syllable word repetition was rare in English speaking children‟s usage, and

was similar to phrase repetition (i.e., both contained more than one syllable),

these two categories were combined.

The six remaining categories included (a) part-word repetition, (b) single-syllable

word repetition, (c) disrhythmic phonation- prolongations, blocks and broken

words, (d) interjection, (e) revision /abandoned utterances and (f) multi-

syllable/phrase repetition. The first three categories were combined to get the

measure stuttering-like disfluencies (SLD); the second three categories were

also combined and labeled as other disfluencies (OD). “Other disfluencies” was

used to designate specific types of disfluencies, whereas the wording “normal

disfluencies” referred to the disfluencies produced by normally fluent children. If

more than one disfluency type occurred in a word, each was counted to provide

accurate representation of the child‟s disfluent speech. The number of extra times

a segment was repeated for part and single-syllable word repetition was also

tabulated as repetition units.

The disfluency types that constitute an SLD have varied to a great extent. Yairi

(1997) reported SLD as constituting: part word and mono syllabic word

repetition, disrhythmic phonation and tense pauses. Justification for using these

terms to define SLDs comes from earlier study by Yairi and Lewis (1984), who

54
used parent judgments to classify their children as having a stuttering problem or

as being normally fluent. The children‟s speech samples were analyzed, and the

disfluencies characterizing those with a stuttering problem were extracted. It was

reported that only two disfluency types distinguished between the stuttering and

normal groups: part word repetitions and disrhythmic phonation. The

frequency of single- syllable word repetitions did not differentiate between the

two groups, however, the number of iterations per instance of repetition did.

However, monosyllabic repetitions were included as SLDs. Yairi (1996)

highlighted the lack of consistency among researchers in classifying a mono

syllabic word repetition as an occasion of stuttering. Wingate (2001) remarked

that monosyllabic word repetitions are usually considered to be common in

normally fluent speech. However, Ambrose and Yairi (1999) and Yairi, Watkins,

Ambrose and Paden (2001) opined that monosyllabic repetitions are distinctive

in young stuttering speakers and therefore should be included as an SLD.

Williams, Silverman and Kools (1968) labeled disrhythmic phonation and tense

pause as SLDs emerged from modification of Johnson‟s (1961) system. A

disrhythmic phonation was defined as occurring only within words; similar

behaviours that occurred between words (or between part words or non words)

were identified as tense pause (tension). Disrhythmic phonation ultimately

emerged as a label for a variety of disfluency subtypes.

According to Yairi and Ambrose (1999) disrhythmic phonation “includes sound

prolongations, silent blocks, and also broken words and other within word

interruptions (but not repetitions) that disturb the continuity of words.” Finally,

tense pause was omitted from the SLD metric because of the “difficulty in

55
reliable identification” (Ambrose and Yairi, 1999). Pellowski and Conture

(2002), and Carlo and Watson (2005) use different sets of disfluencies within

their “SLD” and “other disfluency” measures. In summary, the SLD unit was

described as linked to a word; it did not refer to events that occurred between

words (Ambrose and Yairi, 1999; Yairi and Ambrose, 1999).

Several studies have used SLDs to classify children who stutter or to

differentially diagnose from normal disfluency. Yairi and colleagues opine that a

child must display at least three SLDs per 100 syllables in a speech sample in

order to diagnose a child to have stuttering. The classification by Yairi (1997),

Yairi and Ambrose (1999), Pellowski and Conture (2002), and Zakheim and

Conture (2003) included three or more SLDs per 100 words (not syllables) in a

speech sample. However, not all SLDs are stuttering events. For e.g. single part

word repetitions, single syllable word repetitions or short prolongations typically

occur within the speech of non stuttering children and listeners have no difficulty

in classifying their speech as normally fluent. Based on this, Yairi (1997)

emphasized the fact that SLD should not be considered as a synonym for

stuttering. Yairi and Ambrose (1999) opined that the three SLD threshold need

not be surpassed in order for a speech sample to be judged as displaying

stuttering. However, they also admit that “few young normally fluent children,

especially 2 –to 3- year olds, may exhibit up to 5 SLDs. To increase the

diagnostic strength of the SLD measure, Ambrose and Yairi (1999) proposed a

weighted SLD score which requires considerable data analysis.

Weighted Measure: The weighted SLD score is a metric that considers the

frequency of repetitions, the extent of repetitions (number of iterations), and the

56
 presence and duration of disrhythmic phonation. It is calculated by using the

following formula:

The authors justify the multiplication of disrhythmic phonation as these are

infrequent in early stuttering and rare in normally fluent children (Johnson et al.,

1959; Yairi & Lewis, 1984). They have also traditionally been considered an

advanced symptom. In addition, they make a considerable contribution to the

perception of severity (Costello & Ingham, 1984; Zebrowski & Conture, 1989).

Brief disrhythmic phonation might be weighted by a factor of 1; noticeable, more

tense by a factor of 2 and obvious or distracting, by a factor of 3. To avoid the

necessity of actual time measurements, the middle factor of 2 was chosen.

Ambrose and Yairi (1999) claim that a weighted SLD score will “provide more

powerful information about the presence and severity of stuttering”. They also

claim that a “weighted SLD score of 4 or above should be suspected of stuttering

regardless of the number of SLD”. However, these claims are not well supported.

According to Ambrose and Yairi (2001), the weighted score “is a mathematical

method of differentiating stuttering from disfluency. However, the authors stated

that the resulting score “is not to be taken as an absolute line of demarcation”.

6. Factors affecting fluency

Age: Review of several studies on disfluency has shown that disfluency of

normal speaking children tends to decrease with age. Colburn and Mysak (1982),

and Yairi (1982) observed variations in disfluency at 2 years of age in

57
longitudinal studies. Evidently, some children become more disfluent as

language emerges. Several investigators (Haynes and Hood, 1977; Wexler, 1982;

De Joy and Gregory (1985) report decrease in the frequency of disfluency with

advance in age (4 to 8 years). Haynes and Hood (1977) also observed

interjections to significantly increase between ages 4 and 8. Word repetitions,

however exhibited a noticeable downward shift. Shifting was also observed for

revisions and incomplete phrases. “The implication of this shifting is that

researchers and clinicians should be aware of possible changes in the specific

types of disfluency, as a function of increasing chronological age when

attempting to form or to interpret normative data on disfluency”. (Haynes and

Hood, 1977). De Joy and Gregory (1985) observed a decrease in word, part-word

and phrase repetitions, incomplete phrases and “disrhythmic phonations”

(prolongations, broken words or hard attacks) with increase in chronological age

from 3.5 to 5.5 years. Ambrose and Yairi (1999) studied did not find any

significant difference for the age factor, or for the interaction between age and

group in children from 2 to 5 years of age. However, they found that among the

SLDs, part- word repetitions, single syllable word repetitions and repetition units

were highest in the 3 year old range which decreased later. Thus the literature is

equivocal on the effect of age on disfluencies.

Gender: Glasner and Rosenthal (1957) found sex ratio was only 1.4 to 1 among

153 children who were reported by their parents to have stuttered at some time

before entering the first grade. Andrews and Harris (1964) found that the sex

ratios based on incidence figures tend to be lower than ratios based on the

prevalence of stuttering at a given time, which may be due to shorter episodes of

58
stuttering in girls than in boys. They reported that the sex ratio averaged from 2.6

to 1. Johnson and Associates (1959), Kools and Berryman (1971), Haynes and

Hood (1977), or Ratusnik, Kiriluk and Ratusnik (1979), Ambrose and Yairi

(1999) did not find any sex differences.

Kools and Berryman (1971) analyzed speech samples from 46 male and 46

female first-grade children. They found no significant difference between

genders. Only incomplete phrases were significantly different, with more males

than females exhibiting it.

Ratusnik et al., (1979) studied the rate of disfluency in 4 groups of 36

preschoolers (mean age- 4 years 10 months) each, equally divided by sex.

Subjects included lower status black, lower status white, middle social status

black and middle social status white children. They identified 6 disfluency types

– interjections, word repetitions, part word repetitions, phrase repetitions,

revisions and incomplete phrases - in a 300 word sample. They found disfluency

on 7% of the words. They reported that “sex was not associated with rate of

disfluency, and there were no interactions between sex and social status

background or sex and racial background”.

Language and speech development: With respect to normal disfluency, there

have been somewhat inconsistent findings. In 62 2- to 5- year old children, Davis

(1940) found no relationship between amount of speech repetition and measures

of language maturity such as mean length of response, amount of verbal output,

vocabulary, number of correct speech sounds, intelligibility, and percentage of

functionally correct responses. Berryman and Kools (1975) found that among 92

59
first graders, disfluency was not related to reading ability, intelligence or judges‟

ratings of language level. Haynes and Hood (1977) found that among children

aged 4, 6 and 8 years, there was no relationship between disfluency and syntactic

proficiency as measured by Lee‟s Developmental Sentence Score. Also, De Joy

and Gregory (1985) selected 10 3-5 year old children with the highest number

and 10 with the lowest number of disfluencies of various types and compared the

two groups of samples with respect to the Developmental Sentence Score. The

disfluent children scored lower in syntactic maturity than did the fluent ones in

case of word repetitions, pauses, incomplete phrases, and for the 5- year olds in

case of part-word repetitions. Yaruss, Newman and Flora (1999) studied

syntactic complexity, utterance length and disfluency in the spontaneous speech

of 12 normally fluent children (6 boys and 6 girls, ages, 44-64 months). Only 50

– utterance spontaneous speech samples were analyzed for the presence of more

typical (produced by normally fluent speakers; consisting of repetitions of

phrases, revisions and interjections) and less typical disfluencies ( seen more in

individuals who stutter, including, repetitions of words, sounds or syllables,

prolongations or block. Utterance length and syntactic complexity (measured

using Developmental Sentence Scoring Procedures, given by Lee, 1974) was

calculated. The authors found that disfluent utterances were longer and more

syntactically complex than fluent utterances in normally fluent children‟s

conversational speech.

Westby (1979) compared stutterers, highly disfluent normal speakers and

typically disfluent normal speakers in kindergarten and first grade on several

measures of grammar, vocabulary, and semantic aspects of language. The highly

60
 disfluent normals scored significantly lower than the typical normals on most of

the measures. There was no difference between the highly disfluent normal

speakers and the stutterers. Twenty four children (3.1 – 5.11 years - 12 were

CWS and 12 CWNS) were studied for the influence of utterance length and

complexity relative to the children‟s mean length of utterance (MLU) on

stuttering-like disfluencies (SLDs) for CWS and nonstuttering-like disfluencies

(nonSLDs) for CWNS. Four measures were obtained - utterance length (MLU),

utterance complexity; mean % of stuttering like disfluencies (i.e., sound/syllable

repetitions, sound prolongations, broken words and monosyllable whole-word

repetitions) per 100 words; mean % of nonstuttering-like disfluencies (e.g.,

phrase repetitions, interjections, revisions) per 100 words; also number of SLDs

or nonSLDs per utterance. Results showed that utterances above children‟s MLU

were more stuttered or disfluent and that both stuttering-like as well as

nonstuttering-like disfluencies occurred on utterances that were both long and

complex. The authors concluded that that “the relative “match” or “mismatch”

between linguistic components of an utterance (i.e., utterance length and

complexity) and a child‟s language proficiency (i.e., MLU) influences the

frequency of the child‟s stuttering/speech disfluency” (Zacheim and Conture,

2003).

Additional factors that are known to affect disfluency include situational factors

and familiality. These are however, not discussed here as they were not

considered for the present study.

The review of literature indicated that (1) the frequency of speech disfluencies

reportedly decreases after age 3. Also, researchers have observed decrease in certain

61
types of disfluencies (whole and part word repetitions, phrase repetitions, incomplete

phrases, revisions and disrhythmic phonations) with age. Some disfluency types like

interjections have been reported to increase with age; (2) Some studies have shown

boys to be more disfluent than girls. However, most of the studies show no

statistically significant gender differences; (3) Both repetition units and absolute

duration of speech disfluencies are used in disfluency analyses. Results of most

studies reveal that CWNS produce a maximum of 1-2 iterations per instance of

repetitive disfluency, while more than 2 iterations is more characteristic of CWS.

Also, duration of disfluencies is not significantly different in CWNS than CWS,

although CWS may produce disfluencies that are slightly longer. However, on an

average CWS and CWNS produce disfluencies that are within 1 sec; (4) Disfluency is

found to be related to syntactic complexity i.e. disfluencies occur on longer and more

complex syntactic utterances; and (5) Few studies described the fluency behavior of

preschool children in languages other than English, like Spanish, German, Dutch and

Kannada. With respect to Indian languages, normative data in the preschool period are

not available.

In this context, the present study investigated speech disfluencies in 2.1 to 6 year old

typically developing Kannada speaking children. The specific objectives of the study

were to investigate (a) number, type, and duration of speech disfluencies exhibited by

2.1 ≤ 3, 3 ≤ 4, 4 ≤ 5, 5 ≤ 6 year old typically developing Kannada speaking children,

(b) the effect of gender (Boys and Girls only) and age on speech disfluencies in these

children, and (c) to examine the clinical relevance of the measures derived from the

study.

62