Aspects of speech production in dyslexics

The durational aspects of quantity production has not yet been studied in dyslexics. In general, the temporal aspects of the production side of communication has been studied to a much lesser extent than that of the perception side. Accordingly, the information gained in a number of perceptual studies indicating differences between dyslexics and controls is clearly out of balance with the scant information provided by a handful of studies on temporal aspects of dyslexics' speech production skills. This is clearly a defect in light of one of the prevalent views among linguists that the two sides of communication may reflect each other (e.g., Ingram 1989), and, thus, deviations in the perceptual side of communication may have consequences in production or vice versa. In order to gain a perspective on the speech production side of dyslexics a brief review of the studies on production that are connected with temporality is provided here.

The only study up to present day which deals purely with the durational aspects of speech production in dyslexics is the master's thesis of Scanlon (1994). As well as studying the accuracy, she also studied the articulation rate of syllables and the overall duration of utterances in the speech of an 8 year old girl in comparison to that of a 8 year old boy. The children were studied in pseudoword repetition task, in a disyllabic tongue twister¹ repetition task, and in a sentence² repetitions task. Scanlon (1994) found that the speech of the 1

2

In the tongue twister task the subjects were required to repeat 11 times four disyllabic words.

In the sentence repetition task proper English words were used in four short sentences.

dyslexic subject was significantly faster than that of the control subject. As to the accuracy measure, Scanlon found that the dyslexic subject made significantly more errors than the control subject in the pseudoword repetition task. Furthermore, the performance of the dyslexic subject was significantly more variable between the two testing sessions in comparison to that of the control subject. Scanlon concluded that the dyslexic subject had considerable more difficulties with accuracy in comparisons to the control subject, especially with the pseudoword repetition task. Also the dyslexics' performance was more variable between the two testing sessions than was that of the control subject. She suggested that the dyslexic subject's tendency to produce the stimuli at a faster rate may have had a compromising effect on her phonological accuracy performance, and that consequently, or in addition, she was unable to sufficiently monitor her performance. When examining the evidence provided by Scanlon's study several aspects need to be emphasised.

First, because it is restricted to the relatively small data of one dyslexic subject the results of this study are not easily generalisable. Secondly, some methodological aspects, for instance the fact that the material to be repeated was produced by the experimenter live in the testing situations, should have been more controlled in order to facilitate the interpretation of the results.

Further evidence is needed in order to draw reliable conclusions on the temporal aspects of dyslexics speech. Furthermore, simply by the virtue of the fact that Scanlon's study focused on the rate of articulation her results are not directly relevant to the present study.

Brady, Shankweiler and Mann (1983) also studied dyslexic children's production of speech in a word repetition task. They employed among other things tasks in which 8 to 9 year-old children were required to repeat monosyllabic words either with or without masking noise. Whereas they found no difference in the performance of the subject groups in the condition without masking noise, in the masking noise condition the dyslexic children missed the stop consonants significantly more often than did the control children (Brady et al. 1983, 360). Brady et al. (1983, 360) suggested that this finding could be an indicator of the particular difficulties of dyslexics in processing stops but they considered more likely the possibility that the employed masking noise¹ was of the type that particularly affected the perception of stops. Clearly the former speculation cannot be valid on the grounds of the evidence of their study, since the two subject groups performed equally well in the condition with no masking noise with exactly the same stimuli. An important aspect of the results to be taken in consideration is the fact that the performance levels were at ceiling levels with both of the subject groups, and thus the test may not have been successful in assessing group difference in ideal listening conditions.

However, the fact that the temporal features of the signal were retained in the masking noise whereas the spectrum was not, may have had an effect on the perception of the target stimuli. Since stop sounds include rapid transitions the conflicting information of the noise signal and the target signal may have had a 1 The words were masked by a noise signal which was formed by computing it directly from each digitized speech signal to be masked. The masked noise had the same time-varying characteristics as the speech signal but the spectrum was flattened by a long-term frequency spectrum. (Brady et al. 1983, 356-357).

complicating effect in the perception of the temporal features of the target signal. If dyslexics have problems with the temporal dimensions of speech this factor could partly explain the poorer performance of the dyslexics in this task.

In fact, this speculation is further warranted by their other experiment (Brady et al. 1983), in which they found that there was no difference between the subject groups in identifying environmental sounds with masking noise. The idea that the dyslexics may have had problems with temporal cues in the perception of inherently rapid stop sounds which were reflected in the production of the sounds makes the study of Brady and her colleagues relevant to the present study which has similar h_ypothesis.

Snowling, Goulandris, Bowlby and Howell (1986) experiments were similar to that of Brady et al. The children, which were divided into groups of 9 to 12 year old dyslexic children and their age matched and reading age matched controls, were required to repeat monosyllabic stimuli in both noise masking condition^l and without masking signal condition. The stimuli consisted of both words (divided into high and low frequency classes) and pseudowords (which were constructed from real words by changing an initial feature of the real word used in the study). Their data showed that dyslexics did not differ from the controls in repeating high frequency words with or without noise but that their performance differed significantly from the controls when repeating low frequency words and pseudowords in both of the listening conditions². The performance in the lexical decision tasks was also poorer in the dyslexia group than in the control groups. Snowling and her colleagues (1986) concluded that their results suggest that dyslexics seemed to have difficulties with segmentation.

Before drawing conclusions regarding the evidence provided by Snowling et al., however, one should consider the fact that the pseudowords employed in their study included words like "gake" (derived from "cake") and, thus, the lexically conflicting information in the stimuli may have lead the performance of the dyslexics astray. It is conceivable that the dyslexics relied more heavily on contextual information than did the control subjects and thus they were not able to analyze the stimuli to the extent the control group subjects did. Another aspect of the choice of stimuli in the experiments of Snowling et al. is that by changing the initial sounds of the real words when making the pseudowords, in half of the words this meant a change in the voicing of stop sounds. In English (the language used in their study) this entails that the temporal characteristics were changed by means of VOT values between the real words and pseudowords. Thus, the dyslexics' difficulties with the temporal dimension may have had an effect at least partly on Snowling et al.'s results. An unfortunate detail about their study is that, while claiming to be an extension of Brady et al.'s study, Snowling et al. failed to do the same error analysis as Brady et al. did and thus were not able to provide further 1

2

The noise masking condition used in the study of Snowling et al. (1986) was constructed using exactly the same method as Brady et al. (1983) had employed.

Interestingly, Snowling et al's results did not agree with the results of Brady' et al which indicated that dyslexics were more affected by the masking noise than the control children. Snowling et al. speculate that the fixed order of presntation (noise condition always first) in the Brady et al.' study may have affected their results.

evidence on the possible problems of dyslexics in the processing of stop consonants. In conclusion, it seems that there was a difference in the dyslexics and the controls in the study of Snowling et al. in processing phonological information, with the dyslexics either failing to perceive the crucial characteristics of sounds or relying more heavily on the previously acquired lexicon or both. Furthermore, since Snowling et al., also found differences between the subject groups in the lexical decision tasks, it may be tempting to consider the possibility that the results reflected the dyslexics possible difficulties with producing the required sounds (e.g. gake) with significant temporal cues. This explanation is not plausible here, however, since the subjects did not need to repeat the words but rather indicated their responses by yes or no answers.

In addition to studying accuracy, Brady, Poggie and Rapala (1989) studied children's reaction times in repetition tasks with monosyllabic, multisyllabic and pseudoword stimuli. They found no group differences in the reaction times between the subjects with different reading abilities. In the accuracy measure, the poor readers, who were tested to be at least 9 months behind the average reading level of their age children (8-9 year olds), were significantly less accurate at repeating the multisyllabic and pseudoword stimuli than the above average readers. Brady et al. (1989, 120) speculated that the reason why there were no reaction time differences could have been that the subjects did not have severe problems with reading ability and that the children were too old to demonstrate any basic differences in the speed of processing. They (1989) concluded that their evidence suggests that poor readers have encoding problems but that they are not slower than average readers. The fact that the dyslexics in their study were not slower than the controls is interesting in light of the findings of Experiment 3. Here the dyslexic adults were significantly slower than the controls in their atta-responses. But since the two studies dealt with the two different sides of the communication, i.e., perception and production, these findings are not directly comparable.

In a more recent study, Stone and Brady (1995) extended the findings of Brady et al.'s study (1989) by studying the same age children with the similar subject sampling screening criteria. Their evidence gave further affirmative evidence on poor readers difficulties with accuracy in the pseudoword repetitions tasks among other tasks, but also that their performance was not significantly slower than that of their agemates and younger reading level matched children. As the researchers themselves noted (1989, 120; 1995, 112) their evidence may be limited since the subjects were not necessarily severely dyslexics, and, therefore, their evidence is not directly applicable to severely dyslexic populations. However, there exists ample evidence, some of it not reviewed here (Brady et al. 1989, Stone & Brady 1995, Catts 1986, Kamhi and Catts 1986, Kamhi, Catts, Mauer, Apel and Gentry 1988, Catts 1989, 1993, Hansen & Bowey 1994, Apthorp 1995), that dyslexics have difficulties with accuracy scores in speech repetition tasks especially with pseudowords. Thus, it seems that performance in the pseudoword repetition tasks reflects onto the output phonological level the possible difficulties dyslexics have with phonological processing. If it is assumed that perception abilities are reflected in production abilities then the difficulties observed in perception of duration in quantity categorisation in Finnish may also be reflected in production. On

the basis of this idea, the durational dimensions of speech production in connection with dyslexia is studied in the following two experiments.