Procedure

The assessment of the categorisation abilities of the infant subjects was made using a conditioned head-turn procedure. The technique used here was a modified version of that of Kuhl and her colleagues (e.g., Kuhl 1985a, Kuhl, Williams, Lacerda, Stevens & Lindblom 1992.). The head-turn procedure is based on infants' instinct to turn toward a new sound source.

In the procedure used here the aim was to condition infants to turn their heads towards the loudspeaker whenever they perceived a change in the quantity category within an auditory sequence. Correct head turns were visually reinforced. (Some of the details of the procedure are described in the Procedure section of Experiment 3 on pages 94-96).

The procedure involved three consecutive experimental phases as well as two practice phases. Initially 30 background stimuli (atal) were presented in order to familiarise the infants with the sounds and the experimental situation. The stimuli were presented with a constant 1000 msec interstimulus interval. The stimuli were played at a comfortable listening level, at approximately 70 dB.

rhythmicity to prevent any rhythmic movements of the parents from influencing the reactions of the infants.

00 PICTURE 1

--

Picture taken in the test situation. An infant sat on his parent's lap. An assistant was seated m the right hand side. At an angle of 90^° to the left of the infa:it and at eye level two boxes were situated. The windows of the. boxes were shaded in order to conceal the bunny and the bear, which were lit when the toys started to drum. The loudspeaker from which the stimuli were presented was situated on the top of the boxes. The video camera was placed at a 135^° angle in front and to the left of the infant. The researcher and the parent both wore headphones and heard music while the infant was tested.

In the second practice phase the infants were presented with 8 change trials (see the type of trials in Fig. 10, page 95). The stimulus ata8 served as a target stimuli. In order to facilitate the orienting response towards the sound change the target stimuli were presented at a 10 dB higher level than the background stimuli during the first 4 trials. In the subsequent practice trials, the intensity of the target was reduced in two 3-dB steps until the target and background were equated at 70 dB. The presentations of the target stimuli were automatically accompanied with the visual reinforcement^l, i.e., the mechanical toys moved.

In the first experimental phase, the same · background and target stimuli were presented. This time all the stimuli were presented at the same sound level (70 dB) and the system was calibrated regularly using a Brtiel and Kjaer precision sound-level meter (Type 2235). In this phase only correct headturn responses were reinforced. A headturn response was recorded by the experimenter if the infants turned at least approximately 30^° toward the loudspeaker during the observation period (the observation period for scoring the infants' head-turn response began with the onset of the first target stimulus and ended 1 sec after the fourth target repetition making a total response interval of approximately 6 sees see Fig. 10). The computer program kept a running tally of the infants' correct responses and when a predetermined criterion of 3 consecutive correct head turn responses was met the conditioning phase was terminated and the criteria testing phase was initiated.

In the criteria testing phase the same stimuli (atal and ata8) were used as in the previous phases. This time, control trials were also introduced.

These were added to measure if there was spontaneous turning in the direction of the loudspeaker during the experiment. The computer software presented the trials in a pseudorandom order with the stipulation that only three trials of one type could occur consecutively. Throughout the experiment the experimenter and the assistant were naive as to the type of trial presented. If a head-turn button was pressed on a control trial, the computer did not present the visual reinforcer. The proceeding criterion in this phase was 6 out of 8 correct responses (hits and correct rejections).

The categorisation phase began immediately after the previous phase.

In this phase all the stimuli from the continuum were presented in four trials (a total of 32 trials). The order of the trials was randomised with the constraint that all the stimuli were presented once before they could be presented again, and that only three trials with stimuli from one of the categories could be presented in a row. Only headturn responses in the trials with stimuli ataS to ata8 were reinforced. In both the criterion testing phase and categorisation phase, probe trials were also included. These "wake-up"

trials occurred whenever the infants failed to respond correctly on three consecutive trials. These trials comprised three change trials in which the first target stimuli were presented 7 dB louder than the background stimuli,

The presentation of the visual reinforcers was timed so that they were shown immediately after the second target stimulus was presented. In this way the infant also had a chance to respond to the change in stimuli prior the reinforcement.

and the second 4 dB louder, and the third at the same intensity level as the background stimuli. These same trials were also utilised if the testing was reinitiatedl after it had been interrupted for a longer period of time. Neither the probe trials or the refresh trials were accounted for in the final data.

The whole experiment took 11 minutes to complete, without interruptions and with the subject reaching the criteria for conditioning (initially 3 consecutive correct head turns when stimulus number 8 was presented, eventually 6 out 8 correct responses) with the lowest possible number of stimuli presentations. On average, the test took the young infants 25 minutes to complete (the time ranging from 17 to 40 minutes).

All the data were rescored from the video tapes for head turns by a second judge. The interscorer agreement was extremely high as assessed by correlations. The correlation between the original and reliability scorings for trials averaged .98 across the means ranging from .95 to .99. Also the experimenter who had scored the headturns online and reported after the test that she had forgotten to press the button once during the test or that she had pressed the response button accidentally, pointed out these incorrect cases from the data. These incidences constituted approximately 2% of all the responses.

3.4.3 Results and discussion

Conditioning phases

The data from the conditioning phases of Experiment 4 were analysed in order to see how many trials the six-month-old infants needed in order to be conditioned to turn their heads towards the ataB stimulus which had substantially long duration of the stop closure in comparison to the standard stimulus atal. The results demonstrated that the majority of the infants were able to consistently discriminate the difference between the stimuli atal and ata8. In fact, 83 percent of all the infants met the criterion for proceeding on to the practice phase (three correct headturns out of three successive trials) with the minimum possible score. The rest of the infants proceeded in the experiment with the following scores: four infants after four trials, five after five trials, two after two trials, three after seven trials, and one after ten trials. According to ANOV A there was no statistical difference between the two subject groups in the performance in this initial conditioning phase (F .3213, P= .5723; the mean amount of trials was 3,5 in the GR+ group and 3,4 in the GR- group). Neither was there a difference between the two genders in this respect (ANOV A, F 3.3016, P= .0727). In general, the results reveal that the infants as young as 6 months old can discriminate the difference in the speech sounds according to duration.

These trials were used to refresh the task in the infant.

In the practice phase the mean average score for proceeding in the experiment was 12, (the scores ranged between 9 to 29) for all the infants (the lowest possible score according to the proceeding criterion was 6). There was no statistically significant difference between the two subject groups in this phase of the experiment. Neither was there a difference between girl and boy infants in responding to the stimuli. The mean reaction times were approximately 2568 msec in GR+ group and 2654 msec in GR- group. The mean reaction times were not statistically different either between the subject groups or the two genders.

In the practice phase so called wake-up trials were also utilised in cases where the infants failed to turn their heads on three consecutive change trials. The mean of the wake-up trials was of the order of ,2 and ,3 in the GR+ and GR- groups respectively. The results on these data showed that the amount of wake-up trials between the subject groups as well as the two gender groups was relatively even. The amount of wake-up trials, as well as the attained scores for preceding in the experiment indicate that these young infants were able to perform in an adult-like fashion when the stimuli were as distinct as they were here. For reasons of possible fatigue or inattentiveness or for other similar reasons the scores were lower in the practice phase in comparison to those of adults in Experiment 3. All in all, the results of the conditioning phases show that the six-month-old infants from the Finnish speaking environment were able to categorise the two stimuli with considerably different word medial consonant durations. In addition, the categorisation abilities did not differ between GR+ and GR­

infants or between girls and boys. Thus, there is not deviancy in the risk infants in the basic categorisation abilities. This means that there cannot be any cross differences between the high genetic risk for dyslexia infants and those with no such risk in the processing of large durational changes within speech sounds.

Categorisation phase

The data on the categorisation phase were first analysed in order to discover whether six-month-old infants were able to categorise stimuli with several durational variations of a single sound. In order to do so, the data of the control infants were investigated. The total-trial time data showed that there were similar tendencies to the adult data in the categorisation function: the stimuli from the shorter end of the durational continuum were perceived as distinct from the stimuli from the longer end of the continuum. The difference between the adult and infant data is in the degree of distinctiveness in the two categories, the adult results revealing clearly more distinct differences between the two quantity categories (see Fig.

21). However, the steepness of the category functions between the infant and adult data is relatively similar. It appears that the main categorisation function shows similar tendencies in six-month old infants to that of adults.

The reasons for the less steep categories of the infant data can probably be

attributed to some intervening factors such as inattentiveness and fatigue during the testing. Obviously the results may also demonstrate that these young infants do not categorise the stimuli with variable sound duration into two distinct categories as consistently and systematically as adults.

(I) ::: curves are clearly different but the directional tendencies are similar between the adult and infant data.

The possible categorical nature of the infant speech perception with durational parameters gains additional support from the fact that these results showed a distinct place in the continuum which divided the responses into two categories. In addition, the locations of the . category boundary between infants and adults were very close to each other: The category boundary in the control infant data was located at approximately 142 msec whereas that of the adult data was at 136 msec. Similar tendencies have been demonstrated earlier in categorisation studies using another durational parameter, VOT. For example, Kuijpers (1993, 98) studied categorisation abilities of Dutch children and adults and discovered that the category boundary was more elevated in young children than in old children and adults. It therefore seems that there is a developmental trend in the location of the category boundary according to durational parameters.

The question whether this category boundary is learned or whether it is based on some natural durational boundary cannot be solved with the present experiment. In order to do so experiments with infants with different language backgrounds would have to be conducted. Most likely the category boundary revealed here is located in a natural auditory

psychophysical area but there may be other similar areas in a durational continuum which speakers of other languages utilise in categorising sounds.

It should be noted that the tasks in the adult and infant data are not directly comparable. The reason for this is that in the instructions already the categorical nature of the task was pointed out to the adults. The task of the young infants was more like a discrimination task on the basis of which the categorical assumptions can be inferred.

· The infant data were analysed next in terms of the possible differences between the high genetic risk for dyslexia infants and those with no such risk. The total-time trial data in percentages of atta-categorisations are graphically illustrated in Fig. 22. These results indicate that both subject groups' responses show similar categorical tendencies. However, the results differ remarkably in one respect. The control infants categorised the stimulus ata4 more often as atta in comparison to the risk infants. In fact, the difference between the two groups in this respect reached a highly significant difference when tested with a chi-square test (Pearson x² = 23.32418, P = .0000). There were statistically significant differences in the responses to the other stimuli: the stimulus atal was perceived less often as atta by GR+ infants, (Pearson x^{2 =} 7.58245, P ⁼ .00589), and the stimulus ata5 was perceived by the GR- infants more often as atta (Pearson x² = 8.45933, P

= .00363). In addition, the difference in responding to stimulus ata7 reached almost a significant difference level with the GR- infants categorising the stimulus more often as atta in comparison to the GR+ infants (Pearson x² = 3.75102, P = .05278).

In terms of the category boundary the data revealed that it was located approximately at 142 msec in the GR- infants and around 180 msec in the GR+ infants. This means that the GR+ infant's categorisations were significantly further away from the adult categorisations. Furthermore, the degree of steepness in the category function indicates that the GR+ infants had less distinctive durational categories in comparison to the GR- infants and the adults. The results are strongly suggestive that reason for the less sharp categories is a perceptual temporal deficiency of the GR+ infants.

The temporal processing ability has been shown to improve with age (Eilers, Bull, Oller & Lewis 1984, Morrongiello & Trehub 1987, Elfenbein, Small &

Davis 1993) which entails that these GR+ infants may be developmentally behind the GR- infants. It is possible that they would catch up with the GR­

infants later on in the development. Further research is needed in order to elucidate this possibility. Although developmental catching up seems feasible it is highly unlikely that all of the GR+ infants would do so.

Therefore, the interpretation is favoured here according to which dyslexia may be connected to a permanent temporal processing deficiency.

--FIGURE22

L-.r'

_�

_...-

�

-V

2 3

V -v"

../

-4 5 ATA

,.,,v-,/v ...

-6 7

-l..,.a

8

- GR+group ..,.._ GR-group

The mean percentage of Atta-categorisations in the high genetic risk for dyslexia infants (GR+) and those with no such risk (GR-). There is a highly significant difference in the responses between the two subject groups to the stimulus ata4 (Pearson x² = 23.32418, P = .0000) with the GR+ infants categorising the stimulus more often as atta.

In order to discover whether the observed group differences were caused by differences in the response biases of the subject groups, the d'score analyses were also conducted on the infant data. The data were analysed for statistical significance by t-tests for independent samples. The t-tests for equality of means indicated that there was no statistical difference between the responses of the subject groups in the responses to any of the stimuli between the two subject groups. The d'score curves show (Fig. 23) that the difference in the categorisation functions according to the response percentages cannot be due to the fact that the GR- infants were more inclined to turn their heads during the observational period than the GR+

infants. In conclusion, it appears that the infants with high risk for dyslexia needed remarkably longer duration in order to categorise the stimuli as atta in comparison to the control infants. These are similar tendencies as in the adult group data in Experiment 3. Also previous research has shown similar trends in adult' dyslexic data (Steffens et al. 1992). Also previous research has shown evidence that dyslexics may have a perceptual temporal processing deficit specifically with brief acoustic signals (e.g., Tallal 1980, McCroskey & Kidder 1980, Watson 1992). It also appears that the GR+

infants have less distinct representations of the categories according to the durational features of speech.

2

The discrimination data of Fig. 22 expressed in terms of d'scores. The 8 stimuli are shown as numbers in the abscissa and the group d'scores are shown in the ordinate. The d'scores of the two subject groups show that in terms of the statistical differences in the percentage data the difference in responding to ata4, ataS, ata7 ^and ata2 is not due to the fact that GR- infants were more inclined to tum their heads without a change in the stimuli.

The head-tum data were also analysed by investigating the reaction times of the head turn responses in the two subject groups. It should be remembered, though, that this measure is not as objective as it was in the adult data since here the times were recorded from the point at which the experimenter pressed the response button in identifying the head-turns of the infants. The average reaction times pooled across the stimuli revealed that both of the subject groups indicated a change in the category after two stimuli presentations in the observation period. This result means that the infants were remarkably slower in their responses than the adults. This feature of the results could be expected on the basis of previous research in reaction times with different age subject groups. Young infants and children are slower in their reactions than older children and adults. Here the slowness can partly be explained in terms of the motoric abilities related to head turns which are not very advanced in six-month-old infants. In addition, it cannot be ruled out that the reaction times of the scorerers and their decisional strategies may have influenced these results taking account the fact that the task of the online scorerer was not always simple since the head turns of the infants were not always obvious. The reaction time data pooled across the stimuli revealed that there was a slight difference between the subject groups with the GR- infants being faster in their responses than the GR+ infants (ANOV A F 4.5143 P= .0338). This result shows similar tendencies to those found in the adult data in which the dyslexic adults

The head-tum data were also analysed by investigating the reaction times of the head turn responses in the two subject groups. It should be remembered, though, that this measure is not as objective as it was in the adult data since here the times were recorded from the point at which the experimenter pressed the response button in identifying the head-turns of the infants. The average reaction times pooled across the stimuli revealed that both of the subject groups indicated a change in the category after two stimuli presentations in the observation period. This result means that the infants were remarkably slower in their responses than the adults. This feature of the results could be expected on the basis of previous research in reaction times with different age subject groups. Young infants and children are slower in their reactions than older children and adults. Here the slowness can partly be explained in terms of the motoric abilities related to head turns which are not very advanced in six-month-old infants. In addition, it cannot be ruled out that the reaction times of the scorerers and their decisional strategies may have influenced these results taking account the fact that the task of the online scorerer was not always simple since the head turns of the infants were not always obvious. The reaction time data pooled across the stimuli revealed that there was a slight difference between the subject groups with the GR- infants being faster in their responses than the GR+ infants (ANOV A F 4.5143 P= .0338). This result shows similar tendencies to those found in the adult data in which the dyslexic adults

In document Familial dyslexia and sound duration in the quantity distinctions of Finnish infants and adults (sivua 120-142)