A longitudinal investigation of duration and temporal variability in

1 downloads 0 Views 131KB Size Report
approximately 10%, and temporal variability decreased by about 40%. For the individual children .... 2 years apart, which subsequently will be referred to as time.
A longitudinal investigation of duration and temporal variability in children’s speech production Bruce L. Smith,a) Mary Kay Kenney, and Sarmad Hussain Department of Communication Sciences and Disorders, Northwestern University, 2299 North Campus Drive, Evanston, Illinois 60208-3570

~Received 27 July 1995; accepted for publication 19 December 1995! A number of cross-sectional, acoustic studies have found that young children’s speech segments tend to be longer and more variable than those of older children and adults. However, very little longitudinal information of this nature is available that considers changes across time for individual children. The present investigation is a longitudinal analysis of several temporal characteristics of the speech of 12 children of various ages who were each seen twice, approximately 1 21 years apart. For the group, durations decreased on average from the initial to the follow-up recordings by approximately 10%, and temporal variability decreased by about 40%. For the individual children, however, it was found that some of them showed few, if any, changes in some of the temporal measurements made at the two different times, whereas others showed substantial differences. Younger children also did not necessarily show longer durations or greater variability than older children, nor did younger children always show greater changes across time than older children. Thus, although cross-sectional studies indicate that there is a general tendency when comparing groups for increased age to be associated with shorter durations and reduced variability, individual children may not evidence such patterns or changes across time. © 1996 Acoustical Society of America. PACS numbers: 43.70.Ep, 43.70.Fq

INTRODUCTION

In recent years, a number of acoustic studies have observed that children’s speech segments, syllables, words, etc., tend to decrease in duration and also become less variable from younger to older groups of subjects ~e.g., Chermak and Schneiderman, 1986; Kent and Forner, 1980; Smith, 1978, 1992, 1994; Smith et al., 1983; Tingley and Allen, 1975!. Although both duration and variability have commonly been observed to continue decreasing until as late as 10–12 years of age before reaching adultlike levels ~Kent and Forner, 1980; Smith et al., 1983!, the basis for such changes is not especially clear. They are typically assumed to be due to ‘‘neuromuscular maturation,’’ to greater experience with the process of speech production, or to a combination of both maturation and experience. While both maturation and experience are likely to be involved in the duration and variability changes that occur during children’s speech production development, additional research is needed to help clarify the particular effects of these factors. A related issue associated with these commonly observed decreases in duration and temporal variability in children’s speech pertains to what the nature of any relationship between these two measures might be. Although it has been suggested that decreases in variability with increased age may be merely a consequence of segment durations getting shorter ~e.g., Chermak and Schneiderman, 1986; Kent and Forner, 1980!, more recent research has determined that there is typically only a rather modest correlation between these two temporal measures ~Smith, 1992, 1994!. An exa!

Corresponding author. E-mail: [email protected]

2344

J. Acoust. Soc. Am. 99 (4), Pt. 1, April 1996

ample of the reasonably independent nature of children’s speech segment durations and their temporal variability is that duration appears to reach adultlike levels considerably earlier in the process of development than does variability ~Smith, 1992, 1994!. Furthermore, to the extent that these two measures are correlated, it is possible that they could be correlated only as a result of their individual relationships to other, more global aspects of children’s development. Although a variety of acoustic studies have provided considerable information about these and other aspects of children’s speech production development, a number of issues remain unresolved. One reason for this is that most acoustic investigations of children’s speech production have employed cross-sectional designs, which involve studying several groups of children at different age levels. A significant weakness of utilizing a cross-sectional approach for examining speech production development is that such studies tend to utilize group comparisons rather than focusing on patterns that individual subjects show across time. Even when cross-sectional studies consider the performance of individual children ~e.g., Smith, 1992, 1994!, they are still limited in that they are not able to investigate patterns of development that occur for individual children over a period of time. Despite the fact that cross-sectional analyses are useful ~and are clearly a more time-efficient means of studying ‘‘development’’!, it is essential to conduct longitudinal analyses of speech production to determine what types of changes occur for individual children during various periods of their development. Such studies should allow one to separate out putative ‘‘developmental’’ effects that could be partly due to different groups of subjects being used in cross-sectional

0001-4966/96/99(4)/2344/6/$6.00

© 1996 Acoustical Society of America

2344

studies, and they may help clarify, for example, questions pertaining to the contributions of experience versus maturation in the process of speech production development ~Schwartz, 1995; Smith, 1995!. As noted by von Hofsten ~1989!, there are certain aspects of development that cannot be understood as well, if at all, without obtaining longitudinal data. If one relies only upon pooled data based on groups of individuals, this will ‘‘smear the developmental function, hide important transitions, and make it look smooth and uneventful’’ ~p. 952!. A recent study by Robb and Tyler ~1995! provides a good example of some of the benefits that result from conducting longitudinal research about children’s speech production and the types of information that simply cannot be obtained from a cross-sectional perspective. Robb and Tyler investigated ~monthly! changes in duration in the premeaningful and meaningful consonant–vowel ~CV! productions of seven young children from the time they were about 8 –14 months of age until they were about 20–26 months of age. Although they observed very little change in the durations of what were judged to be nonmeaningful CV vocalizations, they did find that meaningful CV productions, on average, tended to decrease somewhat in duration across the approximately one year that they followed their subjects. However, the variation in CV durations both within and among subjects was quite extensive. For example, some subjects showed relatively little change over the entire year, whereas others manifested considerable reductions in duration. In addition, some subjects evidenced decreases in duration across a period of time, followed by subsequent increases to levels at or above their average durations at younger ages. Although the children Robb and Tyler ~1995! followed were in an age range during which one would expect many interesting changes to be occurring, the observations they were able to make were limited to some extent due to the fact that it is extremely difficult to exercise any degree of experimental control with 1- to 2-year-old children. For example, all CV syllables produced by each child were combined in determining the median duration for that child at any given age, with no control as to the place, manner or voicing characteristics of the consonants, or the vowel heights involved in such productions. Therefore, although there was probably some degree of similarity in the segments produced across ages for a given child, intrinsic consonant and vowel duration differences likely resulted in considerable variability across age comparisons for individual children. Furthermore, such factors are likely to have led to even greater differences when comparing among the seven children who were studied. In addition, at any particular age individual children often produced a total of only five to ten CV syllables, which is also likely to have contributed to the considerable variability that was sometimes observed in CV durations both within and across children. Thus some of the benefits of a reasonably ‘‘naturalistic’’ study such as this one are, at the same time, potential limitations that can only be overcome by studying children who are older and able to participate in more controlled, experimental research. Aside from Robb and Tyler’s ~1995! investigation, there have been relatively few longitudinal, acoustic studies of 2345

J. Acoust. Soc. Am., Vol. 99, No. 4, Pt. 1, April 1996

children’s speech production development, and those that have been conducted have typically tended to focus on rather specific acoustic characteristics such as voice onset time or final syllable lengthening ~e.g., Port and Preston, 1974; Robb and Saxman, 1990; Goodell and Studdert-Kennedy, 1993; Snow, 1994!. Although such details are clearly useful in helping to understand children’s speech production development more fully, somewhat more general issues concerning temporal development in children’s speech have received little attention from a longitudinal perspective. For example, it is not known whether all children show decreases in duration and/or variability over time, as might be assumed on the basis of group data. It is also not clear whether temporal changes across time occur more rapidly for younger versus older children, as one might expect as a result of crosssectional research ~i.e., reductions in duration and variability do not necessarily decrease linearly across age groups!. In addition, it is not known whether children who show longer ~or shorter! durations at one time in their development ~when compared to other subjects! maintain similar, relative levels of performance when compared to those same children at other sampling periods. Thus there are a variety of important and interesting issues concerning general, temporal characteristics of children’s speech production that can be addressed only by considering longitudinal data. The present investigation was an attempt to obtain information about some of the above issues regarding duration and variability by studying a group of children who were seen two times, approximately 1 21 years apart. In contrast to Robb and Tyler’s ~1995! study, the 12 children who were studied in the present investigation were older than might initially seem desirable for a longitudinal study of speech production development, in that they ranged from about 7 to 11 years of age when they were first recorded. The rationale for studying children of these ages was that temporal variability was one of the issues being considered, and it was, therefore, important to exercise careful experimental control across subjects and data collection sessions. In addition, the particular speech tasks that were involved were more demanding than children younger than about 7 or 8 years of age could perform. One reason that a more complicated experimental task was involved in the present investigation is that the subjects were initially seen as part of a different study that was designed to assess children’s and adults’ abilities to consciously/actively control variability in their speech production ~Smith and Kenney, 1994!. Approximately 1 21 years after that study was conducted, it was determined that all 12 of the children from that project still lived in the Chicago area and could be seen for a follow-up analysis, based on a portion of the original investigation. Thus they were recorded a second time to determine what temporal changes, if any, had occurred in certain aspects of their speech since the original data were collected. I. METHODS

Twelve normally developing children ~4 males and 8 females! were recorded at two different times approximately 1 12 years apart, which subsequently will be referred to as time 1 ~initial recording! and time 2 ~second recording!. At time 1 Smith et al.: Duration and temporal variability

2345

they ranged from about 7 to 11 21 years of age, including one 7-year-old, five 8-year-olds, three 9-year-olds, two 10-yearolds, and one 11-year-old ~M 58 years, 10 months!. At time 2 they ranged from about 8 21 to 13 years of age ~M 510 years, 5 months!. All subjects were monolingual speakers of General American English with no reported history of speech, language, hearing, or neurological problems. At both times 1 and 2, the target words ‘‘saucer’’ and ‘‘sissy’’ were each produced 25 times ~in succession! in isolation and then 25 times each in the carrier phrase, ‘‘Say__again,’’ as cued by a visual prompt from the experimenter at approximately 1- to 2-s intervals. ~Whether ‘‘saucer’’ or ‘‘sissy’’ was produced first or second within the isolated and the sentence contexts was randomly determined.! This resulted in 100 productions by each of the 12 children at both times 1 and 2 ~i.e., 2 words and 2 sentences 3 25 repetitions each!. The productions of all subjects for both time periods were audio-taped using a Sony TC-D5M cassette recorder, and they were subsequently digitized with a Macintosh IIci computer using SOUNDEDIT software ~sampling rate511 kHz; low-pass filtered at 5.5 kHz!. Because temporal variability of the subjects’ productions was one of the parameters of interest ~in addition to duration!, the stimuli that were utilized were chosen because of the ease with which they could be segmented. This minimized measurement variability as much as possible in evaluating speakers’ production variability. Aspects of the speech signal that were selected for analysis involved several intervals of different lengths, which included both the stressed vowel @Ä# ~or @Å#! and the medial consonant @2# of ‘‘saucer’’ from the single-word productions. Segmentation was based on the substantial changes in waveform shape and amplitude that occur between vowels and voiceless fricatives. For example, the medial @2# interval of ‘‘saucer’’ was measured from the termination of voicing in the preceding @Ä# vowel to the onset of periodicity in the following @Ñ# vowel. Two larger speech intervals were also analyzed according to similar criteria. One was the VC sequence @I´S# formed by the two medial segments from ‘‘sissy.’’ The other was the VCVC sequence @|t sÄr s# from the phrase, ‘‘Say saucer again,’’ i.e., from the onset of periodicity for the vowel @|# in ‘‘say’’ to the onset of periodicity for @Ñ# ~following the medial @2# of ‘‘saucer’’!. For each subject, 100 acoustic measurements were thus made at both times 1 and 2 ~with standard deviations computed from the 25 repetitions of each of the four segments/intervals that were measured!. To determine intrajudge measurement reliability, approximately 10% of the data were remeasured by the same person about one month after they were initially measured. When comparing absolute values between times 1 and 2, subjects’ average durations were found to be within 2–3 ms of one another, and standard deviation values differed by less than 1 ms. II. RESULTS

Before looking at the findings for individual subjects, it is helpful to consider some group data as a frame of reference. As shown in Fig. 1 for all 12 children and for all the segments and VC~VC! intervals combined, durations of the 2346

J. Acoust. Soc. Am., Vol. 99, No. 4, Pt. 1, April 1996

FIG. 1. Group comparisons ~N512 children! of time 1 versus time 2 measurements for duration, standard deviation, and coefficient of variation. Findings are also compared to averaged adult data (N512) for the same production tasks, shown as a value of 1.

segments and the longer intervals decreased from time 1 to time 2 by an average of approximately 9%–10%. Average decreases were slightly greater for @I´S# and @|t sÄr s# ~12% and 9%, respectively: t52.90, df511, p,0.05; t52.45, df511, p,0.05! than for the individual segments @Ä# and @2# ~7% and 6%, respectively: t51.55, df511, p.0.05; t51.30, df 511, p.0.05!. It can also be seen in Fig. 1 that the decrease observed from time 1 to time 2 for temporal variability ~as measured by standard deviation, based on the square root of the average variance! was considerably greater than the average decrease of about 10% that occurred for duration. For all the children and measurements combined, variability was found to decrease by an average of approximately 40% from time 1 to time 2. Decreases ranged from 29% for @2# ~t52.70, df 511, p,0.05! to 50% for @|t sÄr s# ~t52.77, df511, p,0.05!, with changes of 38% and 43% for @I´S# ~t53.89, df511, p,0.01! and @Ä# ~t53.86, df511, p,0.01!, respectively. In addition to the fact that the relative changes observed from time 1 to time 2 were considerably greater for temporal variability than for duration, it is also important to note in Fig. 1 that variability was much less adultlike than duration at both time periods. @The horizontal dashed line at ‘‘1’’ in Fig. 1 represents the average values for these measures for the same stimuli produced by 12 adults from the initial investigation, i.e., when the children’s time 1 data were collected ~Smith and Kenney, 1994!.# Figure 1 also shows coefficient of variation data ~standard deviation divided by average duration, times 100!, indicating that this commonly used method of representing variability also shows a substantial decrease of approximately 35% between time 1 and time 2. In terms of general duration and variability changes, these group findings comparing the children at two different points in their development are thus quite consistent with what very likely would have been observed if two separate sets of children that differed in age by approximately 1 21 years had been studied in a cross-sectional design. As previSmith et al.: Duration and temporal variability

2346

FIG. 2. Average duration ~upper panel! and average standard deviation ~lower panel! at time 1 versus time 2 for each of the 12 children. The horizontal dashed lines indicate the range of averages for 12 adults performing the same tasks.

ously indicated, however, an important benefit of longitudinal research is to be able to compare individual subjects with themselves at different times. In order to focus more on general patterns pertaining to duration and variability and thereby avoid dealing with differences that sometimes occur among various, specific segments ~van den Heuvel et al., 1994!, the top portion of Fig. 2 shows duration changes for each child’s ‘‘average segment’’ ~i.e., the mean obtained for @Ä#, @2#, @(,2#, and @|,2,Ä,2#!1 across the 1- 21-year time span, with each data point based on 100 productions by each child. As can be seen, 11 of the 12 children evidenced duration decreases from time 1 to time 2 ~i.e., the black square is above the gray diamond!, although several of the changes were not very large. ~In this figure, the children increase in age within the group as a whole from left to right.! In contrast, one subject ~No. 4, an 8-year-old! showed an increase in average segment duration of approximately 13% from time 1 to time 2. The top portion of Fig. 2 also indicates the range of average durations for 12 adults who produced these same stimuli ~i.e., between the two horizontal dashed lines!.2 It can be seen that at time 1 six of the children ~1, 4, 5, 6, 7, 2347

J. Acoust. Soc. Am., Vol. 99, No. 4, Pt. 1, April 1996

and 10! had average durations on the order of about 10% or more above the upper end of the adult range, and six of them showed durations within the adult range. By time 2 only two subjects ~1 and 4! continued to have average durations more than 10% above the adult range. The other four subjects that had been above the adult range at time 1 ~5, 6, 7, and 10! all showed decreases in average duration by time 2 that put them at or very near the upper end of the adult range. The lower portion of Fig. 2 shows changes in the average variability of these same segments for the 12 individual children. As can be seen, ten of the subjects showed decreases in standard deviation ~based on the square root of the average variance! from time 1 to time 2 ~i.e., the black square is above the gray diamond!, although a couple decreases were very small. In contrast, two subjects showed small increases in average variability from time 1 to time 2. The adult range of variability is again shown by the two horizontal dashed lines. It can also be seen that 10 of the 12 children ~1–10! had standard deviations that were 10% or more above the upper end of the adult range at time 1. Only subjects 11 and 12 ~the two oldest children! were at or within 10% of the upper end of the adult range at time 1. By time 2 about half the subjects still had average values more than 10% above the adult range. Substantial decreases in variability occurred for about half of the children, however, such that they were within or at least substantially closer to the upper end of the adult range by time 2, compared to the considerably greater variability they had shown at time 1. In addition, when thinking in terms of averages across groups of children of different ages ~i.e., as in a crosssectional study!, one might have expected reasonably monotonic decreases in both duration and variability to be evident with increasing age among the individual children. However, it can be seen in Fig. 2 that values for at least some of the younger children were not greater than those of several older children. For example, subjects 2 and 3, two of the youngest children ~both 8-year-olds at time 1!, show durations ~upper panel! within the adult range at time 1, whereas several older children ~e.g., 5, 6, 7, and 10; at time 1: two older 8-yearolds, a 9-year-old, and a 10-year-old! were well above the adult range at time 1, and were only at the upper end of the adult range by time 2. The situation regarding variability ~lower panel! is somewhat more in line with what might have been expected ~i.e., greater variability for younger children!, but there are nonetheless relatively young children such as 3 and 5 who showed considerably lower standard deviation values at time 1 than several older children ~e.g., 6, 7, and 9!. A related observation of interest is that although one might reasonably have predicted that the younger children would show a tendency toward larger decreases than the older children for both duration and variability between times 1 and 2, this was not routinely the case. As can also be seen in Fig. 2, the youngest subject ~#1! showed essentially no change in average duration from time 1 ~age 7,1! to time 2 ~age 8,6!. Yet, subjects 5, 6, 7, and 10, all of whom were 1–3 years older than subject 1, showed average duration decreases for time 1 versus time 2 of 10%–20% ~25–50 ms!. Very similar findings can be observed in the lower portion of Smith et al.: Duration and temporal variability

2347

Fig. 2 for the average variability data. Despite the fact that the pattern seen here may be somewhat more closely associated with age, there are still notable, individual differences. For example, subject 3, one of the younger children, showed a reasonably low amount of variability at time 1 ~age 8,2! and no decrease in variability by time 2 ~age 9,5!, whereas older subjects such as 6, 7, and 9 showed substantial decreases from time 1 to time 2. Thus younger children did not necessarily show greater changes in duration or variability across the approximately 1 21-year time span when compared with various older subjects that were also above the adult range at time 1. When considering changes for both duration and standard deviation for each child from time 1 to time 2 for all four of the temporal intervals ~@Ä#, @2#, @ I`S#, and @|t 2Är s#! combined, it can be seen in Fig. 2 that all 12 subjects showed at least a small decrease in one or the other of these temporal measures, and 9 of the 12 subjects showed decreases in both. As an example of the fact that duration and variability are not particularly closely related in the performance of individual subjects, it can be seen, for example, that although subject 1 ~the youngest child at ages 7,1 and 8,6! showed virtually no change from time 1 to time 2 for average duration, she showed a decrease of nearly 50% for variability. Similarly, subject 4 showed an increase of approximately 15% in average duration from time 1 to time 2, but a decrease of about 45% in variability. Of the 11 subjects who showed decreases in duration from time 1 to time 2 when averaging across all of the temporal intervals combined, none of them showed reductions of more than about 20%. In contrast, of the ten subjects who showed decreases in variability, eight of them showed reductions of 30% or more. Finally, it was also of interest to determine whether those children who showed longer durations and/or greater variability at time 1 were the same subjects manifesting greater durations or variability at time 2. For each of the eight individual temporal intervals that were measured ~i.e., four segments/intervals for both duration and variability!, correlations between time 1 and time 2 were found to be positive and on the order of r50.4– 0.6. For three of the four duration comparisons and two of the four variability comparisons, the correlations approached or reached significance (p,0.05), which indicates that there was a tendency for children to maintain at least somewhat similar levels of temporal performance relative to the other children in the group when comparing times 1 and 2. As an example of this, the average segment durations for each of the 12 children for all stimuli combined at time 1 versus time 2 were compared and found to result in a correlation of 0.68 ~df510; p,0.05!. In this regard, it can be seen in Fig. 2 that the same six subjects ~2, 3, 8, 9, 11, 12! had the shortest average segment durations at both time 1 and time 2. III. DISCUSSION

This longitudinal analysis of certain acoustic characteristics of the speech of 12 children over a period of approximately 1 21 years resulted in some interesting observations that could not have been obtained utilizing a cross-sectional de2348

J. Acoust. Soc. Am., Vol. 99, No. 4, Pt. 1, April 1996

sign. Although cross-sectional studies have typically found that duration and variability tend to decrease when comparing groups of younger children with groups of older children and adults ~e.g., Kent and Forner, 1980; Smith, 1992, 1994; Smith et al., 1983; Tingley and Allen, 1975!, the longitudinal data of the present study indicated that several younger children evidenced shorter durations and/or less variability than some of the older children. Such a finding is not entirely unique to a longitudinal investigation, as such patterns can sometimes be seen in figures from various cross-sectional studies that present data for individual subjects in different age groups ~e.g., Kent and Forner, 1980; Smith, 1992, 1994!. However, what was determined in the present study that has not ~and could not have! been observed in cross-sectional investigations is that despite changes from time 1 to time 2 for the group as a whole, not all the individual subjects showed decreases in duration or variability over the approximately 1 21 years between recordings. Several children demonstrated little, if any, change in duration and/or variability across the two sampling periods. These findings suggest that although some amount of neuromuscular maturation and additional experience with speech production can be assumed to have occurred for all the children during the 1 21 years between samples, such factors did not necessarily result in decreases in the temporal parameters of interest for each of the individual subjects. Thus the generalization often made on the basis of cross-sectional data that duration and variability decrease with increased age must be made somewhat cautiously given the longitudinal patterns shown by individual children. In addition, although one might reasonably have expected the younger children to show greater temporal ‘‘development’’ than the older children ~e.g., Thelen, 1991! between times 1 and 2 ~i.e., greater reductions in duration and/or variability!, that was not necessarily the case either ~as seen in Fig. 2!. In at least certain instances, some of the younger children showed little difference between the two time periods for one or both of the temporal parameters of interest, whereas some older children showed rather substantial decreases across time compared to certain younger children. This might be an example of von Hofsten’s ~1989! suggestion that ‘‘the rate of development is different for different subjects. Some develop quickly, whereas others develop slowly. One and the same child may develop quickly at certain ages and slowly at others’’ ~p. 953!. That is, those younger children who showed few changes in one or the other of these temporal parameters may have been on somewhat of a developmental ‘‘plateau’’ for a period of time, while some of the older children were evidencing periods of change during the time of this study. Another interesting finding from this longitudinal investigation that could not have been made utilizing a crosssectional design is that there was at least somewhat of a tendency for children whose durations were longer or whose variability was greater at time 1 to also show longer durations or greater variability at time 2. This suggests that there may be a certain amount of learned and/or inherent aspects of children’s speech production that do not change substantially over time relative to other children within a group. Smith et al.: Duration and temporal variability

2348

Although merely speculation at this point, it could even be the case that such factors persist over longer periods of time and are to some extent responsible for the fact that some adults are observed to be faster talkers, whereas others are found to be slower talkers ~Crystal and House, 1982, 1988a, 1988b, 1990!. That is, such adults might be showing patterns that have been carried over, at least in part, from speaking characteristics and/or capabilities they developed as children. As with any study, there were limitations associated with the present investigation that provide directions for subsequent research. For example, because the present study was not specifically designed to obtain information about effects of experience versus maturation on temporal aspects of children’s speech production, it would be useful to develop a method that could allow careful consideration of such effects ~e.g., Siren and Wilcox, 1995!. As alluded to previously, it would also generally be preferable to study younger children, who would presumably be developing more rapidly. If 2- to 4-year-old children could be studied with a simpler task, but still in a carefully controlled manner, one might observe changes occurring at a more rapid rate than is likely to have been the case with the older children in this investigation ~although younger children might also show periods of little or no change!. As a result, one would not only be able to obtain information about temporal parameters at a more critical time in the process of speech production development, but more samples ~perhaps at 4- to 6-month intervals! could also be obtained within a reasonable period of time. In conclusion, several observations were made from this longitudinal investigation that could not have been obtained on the basis of a cross-sectional study involving two or more groups of children of different ages. For this group of 12 children collectively, segment and syllable durations decreased from the initial to the follow-up recordings by approximately 10%, and temporal variability decreased by about 40%. In looking at the data for individual children, however, it was found that some of them showed few, if any, decreases in one or even both of these temporal measures during the 1 21-year interval between samples. In addition, younger children also did not necessarily show longer durations or greater variability than older children, nor did they always manifest greater changes in these temporal measures across time than older children. Thus, although there is a general tendency in cross-sectional studies of groups of children for increased age to be associated with shorter durations and/or less variability, individual children may or may not show such changes over rather substantial time periods. This suggests that neither maturation nor additional experience in producing speech unequivocally results in temporal decreases for all children across all time periods, despite the tendency for group data to show such changes. 1

Because temporal intervals of different lengths were measured to avoid possible effects that can occur if only single segments are studied, the individual segments of the VC and VCVC intervals were not actually measured for this comparison because those stimuli had been measured as VC and VCVC ‘‘units.’’ However, in order to be able to average the individual segment durations for @Ä# and @2# with the durations of the VC and VCVC intervals, the total durations of the VC and VCVC intervals were divided by the number of segments each contained to provide an estimate of the

2349

J. Acoust. Soc. Am., Vol. 99, No. 4, Pt. 1, April 1996

‘‘average segment duration’’ within each of these intervals. It is important to note, however, that dividing the total duration of, for example, a VCVC interval by 4 results in the same ‘‘average segment duration’’ as if the four individual segments of such an interval had each been measured, and were then added together and divided by 4. That is, as long as the same ‘‘end points’’ for such an interval are utilized, the average duration for the individual segments will be the same regardless of how the ‘‘internal’’ segmentation is accomplished. 2 These adult data were also taken from the previous study ~Smith and Kenney, 1994!. Only 1 of the 12 adults originally recorded was still available to study at time 2, so that it was not possible to rerecord them and determine what amount of variation might have occurred if they had also produced the 1 same stimuli 1 2 years later. The data collected from the one adult subject who was available at time 2, however, showed only small differences compared to her time 1 data, with all of her time 2 values for both duration and variability falling within the adult ranges shown in Fig. 2 for the 12 adult subjects at time 1. Chermak, G. D., and Schneiderman, C. R. ~1986!. ‘‘Speech timing variability of children and adults,’’ J. Phon. 13, 477– 480. Crystal, T. H., and House, A. S. ~1982!. ‘‘Segmental durations in connected speech signals: Preliminary results,’’ J. Acoust. Soc. Am. 72, 705–716. Crystal, T. H., and House, A. S. ~1988a!. ‘‘A note on the variability of timing control,’’ J. Speech Hear. Res. 31, 497–502. Crystal, T. H., and House, A. S. ~1988b!. ‘‘Segmental durations in connected-speech signals: Current results,’’ J. Acoust. Soc. Am. 83, 1553– 1573. Crystal, T. H., and House, A. S. ~1990!. ‘‘Articulation rate and the duration of syllables and stress groups in connected speech,’’ J. Acoust. Soc. Am. 88, 101–112. Goodell, E. W., and Studdert-Kennedy, M. ~1993!. ‘‘Acoustic evidence for the development of gestural coordination in the speech of 2-year-olds: A longitudinal study,’’ J. Speech Hear. Res. 36, 707–727. Kent, R. D., and Forner, L. L. ~1980!. ‘‘Speech segment durations in sentence recitations by children and adults,’’ J. Phon. 8, 157–168. Port, D. K., and Preston, M. S. ~1974!. ‘‘Early apical stop production: A voice onset time analysis,’’ J. Phon. 2, 195–210. Robb, M. P., and Saxman, J. H. ~1990!. ‘‘Syllable durations of preword and early word vocalizations,’’ J. Speech Hear. Res. 33, 583–593. Robb, M. P., and Tyler, A. A. ~1995!. ‘‘Durations of young children’s word and nonword, vocalizations,’’ J. Acoust. Soc. Am. 98, 1348 –1354. Schwartz, R. G. ~1995!. ‘‘Effect of familiarity of word duration in children’s speech: A preliminary investigation,’’ J. Speech Hear. Res. 38, 76 – 84. Siren, K. A., and Wilcox, K. A. ~1995!. ‘‘Effects of lexical meaning and practiced productions on coarticulation in children’s and adults’ speech,’’ J. Speech Hear. Res. 38, 351–359. Smith, B. L. ~1978!. ‘‘Temporal aspects of English speech production: a developmental perspective,’’ J. Phon. 6, 37– 68. Smith, B. L. ~1992!. ‘‘Relationships between duration and temporal variability in children’s speech,’’ J. Acoust. Soc. Am. 91, 2165–2174. Smith, B. L. ~1994!. ‘‘Effects of experimental manipulations and intrinsic contrasts on relationships between duration and temporal variability in children’s and adults’ speech,’’ J. Phon. 22, 155–175. Smith, B. L. ~1995!. ‘‘Variability of lip and jaw movements in the speech of children and adults,’’ Phonetica 52, 307–316. Smith, B. L., and Kenney, M. K. ~1994!. ‘‘Assessing duration and variability in speech and non-speech tasks performed by adults,’’ J. Acoust. Soc. Am. 96, 699–705. Smith, B. L., Sugarman, M. D., and Long, S. H. ~1983!. ‘‘Experimental manipulation of speaking rate for studying temporal variability in children’s speech,’’ J. Acoust. Soc. Am. 74, 744 –749. Snow, D. ~1994!. ‘‘Phrase-final syllable lengthening and intonation in early child speech,’’ J. Speech Hear. Res. 37, 831– 840. Thelen, E. ~1991!. ‘‘Motor aspects of emergent speech: A dynamic approach,’’ in Biobehavioral Foundations of Language, edited by N. Krasnegor ~Erlbaum, Hillsdale, NJ, 1991!, Chap. 15, pp. 339–362. Tingley, B. M., and Allen, G. D. ~1975!. ‘‘Development of speech timing control in children,’’ Child Dev. 46, 186 –194. van den Heuvel, H., Rietveld, T., and Cranen, B. ~1994!. ‘‘Methodological aspects of segment-and speaker-related variability: A study of segmental durations in Dutch,’’ J. Phon. 22, 389– 406. von Hofsten, C. ~1989!. ‘‘Motor development as the development of systems: Comments on the special section,’’ Dev. Psychol. 25, 950–953.

Smith et al.: Duration and temporal variability

2349