Auditory Discrimination Tasks
3.6. Ü Differences in performance on auditory discrimination tasks Performance on each auditory discrimination task was significantly
correlated with performance on each other task. This is in contrast to studies by Bountress and Lauderberg (1981), Bountress (1984) and Cracknell (1987) which failed to find significant correlations between performance on different measures o f auditory discrimination. However, it has already been suggested that some outlying scores may have affected the analysis in the Bountress (1984) study. The finding o f significant correlations in the data reported here may reflect, to some extent, the use o f matched stimuli. The auditory discrimination picture task was more highly correlated with the ABX task which used matched stimuli, than with the same / different task which used different stimuli.
However, there were some differences in task performance. When performance was compared across the 3 tasks, using percentage o f items correct, there were significant differences between task performance. Differences in task performance may be the result o f cognitive factors, or o f the different levels o f speech input processing involved. Performance on the ABX task was poorer than on the auditory discrimination picture task and this may reflect differences between speech processing levels. The auditory discrimination picture task required the matching o f spoken stimuli to existing phonological representations, and the ABX task required the comparison between two non-words for which there are no lexical representations. But performance on the ABX task could have been constrained by cognitive factors, such as the level o f conceptual understanding required for the task, and / or attentional and STM demands in retaining two stimuli for comparison with a third.
For the same / different task, there was no effect on performance when words were presented rather than non-words. This suggests that for the stimuli and task used here, access to lexical representations did not aid discrimination. It is possible that the children tended to use bottom-up processing to discriminate both words and non-words.
It is difficult to make any further assumptions about the relationships between performance on different tasks because ceiling effects were present in the auditory discrimination picture task and the same / different task data, for the older children.
3.6.::! Differences in performance on speech production tasks.
Although performance was strongly correlated across the 3 speech production tasks, a significant main effect o f task was revealed, with a significant interaction between age and task, suggesting a complex pattern o f differences in performance for different ages o f children.
For the 3 year-old children there was no significant difference in performance between word and NWR performance, however scores on both repetition tasks were significantly better than naming scores. The children were able to produce the target words in naming, with prompting where necessary, so poorer scores were not the result o f children not knowing some o f the words, or not recognising some o f the pictures. As naming required the child to access their own lexical representations for production, this finding is interpreted to suggest that there were some inaccuracies / immaturities within these lexical representations for the youngest children. For repetition tasks, where bottom-up processing could support performance, the availability o f the adult’s models led to better performance. In repeating stimuli, there was no effect on production o f existing lexical representations o f the real words, as opposed to the non words.
At the age o f 4 years, word repetition had become significantly more accurate than NWR, suggesting that by this age the use o f existing lexical representation had an advantageous effect on speech production. Word repetition was still significantly better than naming, suggesting that any inaccuracy / immaturity of lexical representations is still affecting naming performance. Whereas in the word repetition task the child hears an adult
model o f accurate pronunciation and has an opportunity to update an immature motor program.
For 5 year-old children there is no longer any significant difference in performance between word repetition and naming, suggesting that lexical representations have matured and become more accurate. NW R continues to be significantly less accurate than the speech production tasks that use words, suggesting that there is a beneficial effect o f lexical representation on output speech processing. This pattern was maintained in the 6 and 7
year-old children.
In the youngest children there was no significant difference between word and non-word repetition accuracy, however performance on these two tasks becomes increasingly divergent with age, with a greater rate o f increase in accuracy for word repetition rather than non-word. Young
children are frequently exposed to unfamiliar words and phonological forms and seem to have a facility to quickly leam how to say them (Rice, 1990). Their usual response on hearing new words is to repeat them (Speidal, 1989, cited in Gathercole et al., 1994). So the pattern o f performance observed may reflect some facility with NW R at this younger age that aids new word learning, but that is subsequently weakened. Another possibility is that the 3 year-old children process both words and non-words presented for repetition in a similar way, favouring a bottom-up approach to both in which no use is made o f lexical representations. It might be possible to illustrate this by analysing errors in word and non
word repetition. I f bottom-up processing is being used, one would expect to find that non-words would not be lexicalised and that words would sometimes be produced as non-words. If top-down processing is occurring.
one would expect to find some lexicalisation o f non-words. It would be interesting to carry out such an analysis and compare errors o f lexicalisation and non-lexicalisation across age groups, to see if the younger children are tending to use bottom-up processing to do these tasks more than older children are.
The finding that in the older age groups, children are less accurate in NWR than in naming or word repetition could be interpreted in a number of ways. One possibility is that the STM requirement o f holding a novel
string o f phonemes, as opposed to a known word, affects performance. This interpretation would be in accord with Gathercole et al. (1994b), who view NWR performance as being reliant on PSTM capacity. For the 3 year-old children, who showed no difference in word and non-word repetition performance, this did not seem to be a limitation. But one interpretation given to this finding above, was that both words and non words presented for repetition were being processed bottom-up, in which case, if PSTM is influencing performance, it will have an effect on both sets o f stimuli. However, it is equally possible that the speech processing requirements o f either discriminating all the phonemes o f the non-word without top-down support, or creating a new motor program could be affecting performance.
The effect o f stimulus length in the speech production tasks was significant, with the children’s responses being less accurate for longer words. There is a changing pattern o f relationships between task and stimulus length across the age groups. For the 3 year-old children, word repetition is better than NWR, which is better than naming. The interaction between task and length approached significance, with naming affected
more greatly by word length than the repetition tasks. Lexical representations for longer words may be somewhat less accurate. For 4 and 5 year-old children there was no significant interaction between speech production task and stimulus length. Performance on naming and word repetition was better than performance on NWR. However, for the 6 and 7 year-old age groups there were significant interactions between task and length. The source o f the interaction occurs as the relationship between naming and repetition performance changes with increasing stimulus length. For 2 syllable stimuli, naming and word repetition are significantly better than NWR, and there appears to be an advantage for lexical items. For 3-4 syllable stimuli, naming is not more accurate than NWR. W ord repetition is significantly better than both naming and NWR. Lexical representations for longer items, as accessed in naming, appear to be less accurate, and NWR may be affected by the amount o f novel phonological material to be processed. Word repetition is an easier task having the advantage o f existing lexical representation, and an opportunity to update an inaccurate motor program using the spoken form heard. Measures o f speech output processing in children aged 6 years and over, may be more sensitive to individual variation if multisyllabic stimuli are used.
Although there are some significant differences between task performances, the pattern o f difference varies with age. However, performance on each o f the speech production tasks was significantly correlated with performance on each o f the other tasks, so that more accurate production skills seem to affect accuracy on all speech production measures. Naming was more highly correlated with word repetition than with NWR. This may reflect a lexicality effect for the children across the
whole age range, except those aged 3 years, or the fact that the same stimuli were used for the naming and word repetition tasks.
3.6.iv Correlations between speech production and auditory