Journal of Undergraduate Research
Volume 4, Issue 4 - December 2002

Temporal Characteristics of Disfluencies in Young Children

Anne Schaedler

ABSTRACT

This study examined the temporal characteristics of speech disfluencies produced by young children during a sentence imitation task. Participants were 11 children who stutter (CWS) and 11 age- and sex-matched children who do not stutter (CWNS). Participants repeated 10 sentence pairs in which the phonetic similarity of the syllable onsets differed. Resultant speech samples were analyzed to determine whether the groups differed in (a) the duration of within-word disfluencies (b) the speed at which they terminated overt speech errors, and (c) the speed at which they initiated speech error repairs. Results indicated that the within-word disfluencies of CWS were significantly longer than those of CWNS, and that the speed at which CWS initiated speech error repairs was significantly shorter than that for CWNS. There was, however, no significant difference between groups for the duration of speech errors. Effects of phonetic context on the temporal characteristics of children’s speech disfluencies are discussed.

INTRODUCTION

Current diagnostic methods for children who stutter call for determining the number and type of within-word disfluencies that a child produces as well as the length of time (duration) that these disfluencies last (Conture, 2001). The duration of within-word disfluencies has also been used as an indicator of stuttering severity (Conture, 2001; Riley, 1980), and as a means for evaluating stutterers’ progress in fluency therapy (Johnson et al., 1959). Some authors (e.g., Bloodstein, 1987) have questioned the diagnostic validity and usefulness of disfluency duration measures. Such questions primarily have stemmed from the lack of objective data concerning the duration of within-word disfluencies produced by children who stutter (CWS) and children who do not stutter (CWNS). Accordingly, several recent studies have investigated the temporal characteristics of children’s within-word disfluencies.

Both Zebrowski (1991) and Kelly and Conture (1992) measured the duration of sound/syllable repetitions (SSRs) and sound prolongations (SPs) produced during conversation by preschool-aged CWS and CWNS and found no significant differences between the groups. Throneburg and Yairi (1994) measured the overall duration and the duration of silent intervals within children’s whole- and part-word repetitions. They found that the CWS exhibited shorter silent intervals between repetition units than CWNS, which suggested that the tempo or pace of a repetition might be a useful measure in diagnosis of stuttering.

Although these studies offer insight into the usefulness of disfluency duration measures, they present important three limitations. First, in previous studies, conclusions about the duration of within-word disfluencies were based upon a small number of observations. For example, in Zebrowski’s (1994) study, the 10 CWNS produced a total of only 21 within-word disfluencies. Second, in previous studies, researchers have not controlled for the length or linguistic context of disfluent words. It is unclear how such variables might affect the duration of within-word disfluencies. Third, previous studies only address stutter-like disfluencies. They have not explored other disfluency types, such as linguistic revisions. The study of such disfluencies might provide insight into whether CWS differ from CWNS in the speed at which they detect and repair overt errors in spoken speech.

The purpose of this study was to compare the duration of SSRs, SPs, speech errors, and revisions produced by preschool-age CWS and CWNS. To address limitations associated with previous studies, disfluencies were elicited via a sentence repetition task that controlled for potentially confounding linguistic and phonetic variables. Data were examined to determine whether (a) within-word disfluencies differ in duration between groups, b) the groups differed in the speed at which they terminate overt speech errors, and c) the groups differed in the speed at which they initiate speech error repair. Such research should provide additional information about usefulness of temporal measures in the assessment of CWS.

METHOD

Participants

Participants were 11 CWS and 11 age- (+/- 3 months) and sex- matched CWNS. The mean age for both groups was 5.6 years (SD = 1.9 years). All participants were monolingual, Standard American English speakers with no concomitant speech-language problems. Among the CWS, two presented severe stuttering, 6 presented moderate stuttering, and 3 presented mild stuttering.

Data Collection

Each child repeated 20 3-syllable-long phrases and sentences, some of which resembled “tongue twisters,” in that all syllable-initial and some of the syllable-final consonants were phonetically similar. Consequently, these stimuli were likely to elicit disfluencies and/or nonsystematic speech errors. The stimuli were presented in random order to each child. Younger participants (< 7 years) repeated each target, five times (i.e., 100 total). Older participants (> 7 years) produced three trials of five repetitions for each target (i.e., 300 total). All productions were videotape recorded.

Data Analysis

Following data collection, verbatim transcripts of all productions were prepared. Each of the 1-syllable words was coded for phonetic accuracy and fluency. Words featuring articulation errors (e.g. “Snue’s snake soup”) were coded as “inaccurate”. Inaccurate words were then coded as “revised” if the speech error was repaired (e.g. “Snue’s- Sue’s snake soup.”). Syllables featuring within-word disfluencies were coded to reflect the type of disfluency they contained (i.e., SSR, SP, or SSR+SP).

All revisions that occurred on non-stuttered words and all stuttered disfluencies that occurred on phonetically accurate words were digitized using Computerized Speech Lab (CSL), so that they could be viewed as amplitude waveforms and sound spectrograms. Pertinent segments of speech were measured (in milliseconds) using on-screen cursors and associated time codes.

Within-word disfluencies

The duration of within-word disfluencies was defined as time elapsed from the onset of the syllable-initial consonant associated with the disfluency to the onset of the nucleus for the syllable. (e.g. for the prolongation “sssnake,” the duration of the acoustic segment “sssn-” was measured.

Revisions

The duration of each speech errors was defined as time elapsed from the onset of the inaccurate speech segment to the termination of acoustic energy associated with the inaccurate speech segment. The duration of speech error revisions was defined as the time elapsed from the onset of the speech error to the initiation of the speech error revision. The duration of the silent interval between the error cut-off and the initiation of the speech error revision was derived by subtracting the error duration from the revision duration.

RESULTS

Within-Word Disfluencies

Of the 11 CWS, 9 produced within-word disfluencies while completing the experimental task. The participants who did not produce within-word disfluencies and their pairs were excluded from the within-word disfluency duration portion of the study. The nine CWS produced 88 total measurable disfluencies (i.e. 52 SPs, 23 SSRs, and 13 SSR+SPs disfluencies). The nine CWNS also produced 88 measurable disfluencies (i.e. 24 SPs, 49 SSRs, and 15 SSR+SPs disfluencies). On average, the children in each group produced a total of 9.78 within-word disfluencies.

Table 1 shows the mean duration of within-word disfluencies for both groups. Statistical analysis indicated that the mean duration of within-word disfluencies was significantly longer in CWS than in CWNS (U = 11; p < .01). Table 1 also shows the mean duration of within-word disfluency by type. As can be seen, CWS and CWNS produced SSRs of comparable length; however, the mean duration of SP’s of CWS was 114.5 ms longer than those of CWNS, and the mean duration of SSRs+SPs of CWS was1029 ms longer than those of CWNS. Due to the small number of disfluencies produced for each type, group differences could not be assessed using inferential statistics.

Table 1 Mean Disfluency Durations and Standard Deviations (in milliseconds) for Each Type of Within-word Disfluency and for Within-words Disfluencies, Overall
Group	Within-word disfluencies	Within-word disfluency type
		Sound/Syllable repetitions	Sound prolongations	SSR + SP disfluencies
CWS
M	906.9	1044.8	664.2	2088.0
SD	280	230	310	800
CWNS
M	913.2	730.3	549.7	1059.0
SD	540	250	350	220 Revisions Of the 11 CWNS, 10 produced measurable revisions. Thus, the subject who did not produce any revisions and her matched pair were excluded from this analysis. The remaining 10 CWS produced 71 measurable revisions. The 10 CWNS produced 80 measurable revisions. Table 2 shows the mean duration of measurable error terms within speech errors for CWS and CWNS. Statistical analysis indicated that there was no significant difference between CWS and CWNS in the duration of error terms within speech errors (U = 72; p > .05). Table 2 also shows the mean duration of the (silent) interval between the speech error cut-off and the initiation of the speech error revision for CWS and CWNS. Results of the statistical analysis indicated that the silent interval of speech error revisions was significantly shorter in CWS than in CWNS (U = 78; p < .05). Group Table 2 Mean Durations and Standard Deviations (in milliseconds) for Speech Errors and Silent Intervals of Speech Error Revisions for CWS and CWNS Speech error Silent Interval CWS M 389.3 288.7 SD 170 150 CWNS M 410.0 368.0 SD 120 60 Context effects The durations of within word disfluencies in phonetically similar and control utterances were informally compared. The nine CWS included in the within-word disfluency analysis produced 56 within-word disfluencies on phonetically similar utterances and 32 within-word disfluencies on control utterances. The nine CWNS included produced 64 within-word disfluencies on phonetically similar utterances and 24 within-word disfluencies on control utterances. Among CWS, the mean duration of within-word disfluencies on phonetically similar utterances was 1053 ms (SD = 392 ms), and on dissimilar utterances it was 951.3 ms (SD = 442 ms). Among CWNS, the mean duration of within-word disfluencies on phonetically similar utterances was 717 msec (SD = 242 ms), and on dissimilar utterances it was 776.5 ms (SD = 420 ms). DISCUSSION There were two principal findings of this study. The first was that the within-word disfluencies of CWS were significantly longer than those of CWNS. This finding is contrary to results reported by others (e.g., Throneburg & Yairi, 1994; Yairi & Hall, 1993; Zebrowski, 1991). This discrepancy may relate to methodological differences between the present study and the others listed, such as the nature of the speaking task and the ages of the participants. It is also important to consider that Yairi and Hall only considered the duration of sound/syllable repetitions while the present study measured the duration of all types of stutter-like disfluencies. Although there were not enough data to assess differences in disfluency duration across within-word disfluency types, it seems that the SSR+SP disfluencies contributed heavily to the between group difference on this measure. That is, there was a small difference between mean duration of prolongations of CWS and CWNS and almost no difference between the mean duration sound/syllable repetitions of CWS and CWNS. However, the duration of SSR+SP disfluencies in CWS were two times longer than those in CWNS. This may explain why findings of the present study are not consistent with previous studies, as SSR+SP disfluencies were not considered in previous studies. The second finding was that the silent interval between speech error cut-offs and the initiation of speech error revisions was significantly shorter in CWS than in CWNS. This finding is consistent with previous studies (Throneburg & Yairi, 1994; Yairi & Hall, 1993), which suggested that the duration of silent intervals between iterations within whole- and part-word repetitions were significantly shorter in CWS than in CWNS. The findings of the present study suggest that the speech error revisions of CWS feature similar temporal patterning. It was also found that the duration of the error terms within speech errors did not significantly differ between groups, suggesting CWS and CWNS take a similar length of time to detect speech errors and initiate speech error repairs. Although context effects could not be assessed with inferential statistics, it is interesting to note that the mean duration within-word disfluencies of CWS was 100 ms longer in the phonetically similar context than in the control context, suggesting that phonetic context effects on disfluency duration may warrant additional research. This was a preliminary study. Although, the speech task was successful, it did not elicit enough specific types of within-word disfluencies or enough disfluencies in similar and dissimilar contexts to assess the statistical significance of the differences between the groups in these categories. Further research should be conducted with more participants and a larger set of data, which could be elicited from an increased number of trials per participant. An additional improvement would be to include participant pairs closer in age than those in the present study. Finally, further research should include analyses of repetition unit duration and silent interval duration during within-word disfluencies. Figure 1. Mean duration (in milliseconds) for each type of within-word disfluency and for within-word disfluencies, overall. WWD = overall within word disfluencies; SP = sound prolongations; SSR = sound/syllable repetitions; SP+SSR = sound/syllable repetition and sound prolongation. Figure 2. Mean duration (in milliseconds) for silent intervals of speech error revisions for CWS and CWNS. REFERENCES Bloodstein, O. (1995). A handbook on stuttering (5th ed.). San Diego: Singular Publishing. Conture, E. (2001). Stuttering: It’s nature, diagnosis, and treatment. Needham Heights, MA: Allyn & Bacon. Curlee, R. (1980). A case selection strategy for young disfluent children. Seminars in Speech, Language, and Hearing, 1, 277-287. Johnson, W. and Associates. (1959). The onset of stuttering. Minneapolis: University of Minnesota Press. Kelly, E. & Conture, E. (1992). Speaking rates, response time latencies, and interrupting behaviors of young stutterers, nonstutterers, and their mothers. Journal of Speech and Hearing Research, 35, 256-1267. Riley, G. (1980). Stuttering severity instrument for children and adults. Austin, TX: Pro-Ed. Throneburg, R., & Yairi, E. (1994). Temporal dynamics of repetitions during the early stage of childhood stuttering: An acoustic study. Journal of Speech and Hearing, Research 37, 1067-1075. Yairi, E., & Hall, K. (1993). Temporal relations within repetition of preschool children near the onset of stuttering: A preliminary report. Journal of Communication Disorders, 26, 231-44. Zebrowski, P. (1991). Duration of speech disfluencies of beginning stutterers. Journal of Speech and Hearing Research, 34, 483-491. Zebrowski, P. (1994). Duration of sound prolongation and sound/syllable repetition in children who stutter: Preliminary observations. Journal of Speech & Hearing Research, 37, 254-263. --top-- Back to the Journal of Undergraduate Research College of Liberal Arts and Sciences | University Scholars Program | University of Florida | © University of Florida, Gainesville, FL 32611; (352) 846-2032.