RECOMENDACIONES

This task was carried out the week after the lexical decision task. The children were reminded of the previous task and told that all the real words had been removed and that they were left with all the made-up words. They were told that their task this time was to spot all the words that could sound like a real word but had been spelled wrongly. They had to say ”Yes” if the non-word "sounded like a real word" and"7Vb" if it did not sound like a real word. They had a practice session with 8 items where any problems were discussed and then they responded to the 40 items, 20 pseudo-homophones and 20 ordinary non-words. The response times were again automatically recorded and stored in the programme and the accuracy of each response was keyed in by the author.

Stainthorp/ Chapter 5

5.6 RESULTS AND DISCUSSION

Table 5.8 shows the percentage of each of the two stimulus type that were categorized as sounding like a real word. These data are illustrated in Figure 5.2. The data were analysed using a three way factor analysis of variance (1 between and 2 within) with planned comparisons. There was a strong effect of non-word type, with pseudo-homophones being significantly more likely to be classified as sounding like a real word than ordinary non-words (F (l,12)= 41.6, p < 0 .0 0 1 ). There was no main effect of time, but an interaction between stimulus type and time (F (l,1 2 )= 8 .0 3 , p < 0 .0 5 ). This showed that subjects differentiated between the two types of stimuli more successfully on the second testing. Planned comparisons showed that the Poor Readers significantly increased the accuracy of their responses to pseudo-homophones. Table 5.8 shows that they were barely above chance in categorizing the pseudo-homophone when tested in 1990.

Stimulus Pseudo-homophone

Type

Non-word 1990 - 9 years

Good Readers Mean 72.86 32.86

sd 16.29 20.99

Poor Readers Mean 56.43 37.14

sd 13.14 20.99

1992 - 11 years

Good Readers Mean 76.43 22.14

sd 8.52 16.55

Poor Readers Mean 74.29 37.86

sd 12.39 15.24

Table to show the mean percentage of stimuli categorized as sounding like real words by each group at each time of testing

Stamthorp/ Chapter 5 M e a n p e r c e n ta g e o f p s e u d o - h o m o p h o n e s a n d n o n - w o r d s c la ssifie d as " S o u n d in g lik e a r e a l w o r d " 100 S 50 Good Good Readers age II * Pseudo-homophones O Non-words Figure 5.2

A chi square test was carried out on the "Yes" and "No" responses. This gave a value of chF (1 df)=0.91 which was not significant. Thus, when tested at age 9 years, these children were not able to discriminate between the two types of stimuli. This is indicative of their poor decoding skills at this time. A chi square test carried out on their responses on second testing gave a chF (1 df)=4.95, p < 0.025. By the time they were 11 years old they had shown a considerable improvement and were able to discriminate and make a phonological decision about the stimuli with significantly greater accuracy.

The Good Readers were able to discriminate and make accurate phonological judgements both times they were tested (chF (1 df)=4.95, p < 0.025 in 1990 and c h F (l df) = 10.03, p < 0.002 in 1992)

The median response times were calculated for the correct phonological decisions and the means of these are presented in Table 5.9.

Stainthoip/ Chapter 5 STIMULUS Pseudo-homophone TYPE Non-word 1990 - 9 years

Good Readers Mean 1752.86 1835.29

sd 543.64 670.55

Poor Readers Mean 2416.86 2229.00

sd 1422.97 1666.31

1992 - 11 years

Good Readers Mean 1311.29 1764.57

sd 332.87 753.37

Poor Readers Mean 1926.00 2710.43

sd 1181.71 1635.99

Table to show the mean phonological decision response times made by each group at each time of testing

Table 5.9

As with the lexical decision data, the range of the standard deviations was very wide which necessitated performing a square root transformation in order to analyse the data from the pseudo-homophone stimuli and from the non-word stimuli by means of a two way factor analyses of variance (1 between and 1 within). There was no difference between the two groups in their response times to decide that a pseudo­ homophone sounded like a real word (F(l,12) = 1.53, p = 0 .2 ); there was a main developmental effect, with both groups making faster decisions when tested at age 11 years (F (l,1 2 )= 6 .6 0 , p < 0.025); there was no interaction (F (l,12)= 0.02, p= 0.89). There were no significant effects when the response times to decide that a non-word did not sound like a real word were analysed (Group factor: F (l,1 2 )= 0 .9 0 , p= 0.36; Developmental factor: F (l,1 2 )= 0 .6 1 , p= 0.45; Interaction: F (l,12) = 1.40, p= 0.25)

The results from this task suggest that, on first testing, the Poor Reader group were unable to differentiate between the two types of non-words. This suggests that they were unable to use an assembled, alphabetic strategy as a route to reading. However, by the time they were tested two years later, they had improved their decoding skills and were therefore able to use them to differentiate successfully

Stainthorp/ C h u ter 5

between the two types of non-word stimuli on the basis of their phonology. Their behaviour on the ordinary non-word items is interesting. Inspection of their behaviour in discriminating between the two types of non-word shows that by 11 years they were able to differentiate. This means that they could use an assembled, alphabetic strategy, but the significantly longer response times to reject correctly these ordinary non-words suggests that this strategy was still unrefined so they adopted a conservative strategy of checking before rejecting. Their alphabetic skills could not be characterised as automatic even at this stage.

The Good Readers, on the other hand, were able to discriminate successfully between the two types of non-word stimuli on first testing. They showed a trend towards a more accurate performance, as would be predicted if they had developed a skilled decoding strategy which they could use selectively in response to differential task demands.

Table 5.10 presents a comparison of the phonological decision data from this study with that from Johnston and Thompson (1989).

STIMULUS Pseudo-homophone TYPE Ordinary Non-word Johnston & Thom pson (1989) Britain age 8 74.1 31.5 NZ age 8 61.7 46.9 Present study

Poor Readers age 9 56.4 37.1

Good Readers age 9 72.9 32.9

Poor Readers age 11 74.3 37.9

Good Readers age 11 76.4 22.1

Table showing a comparison of the data from this study with that of Johnston and Thompson (1989)

Stainthorp/Ch£çter 5

It can be seen that in 1990, the Poor Readers were showing a pattern of performance closest to that of the New Zealand children and the Good Readers were performing like the Scottish (British) children. By 1992, both cohort groups were showing a pattern of performance like Johnston and Thompson’s Scottish children.

The question raised by Johnston and Thompson regarding the effects of teaching methods on children’s reading strategies needs to be addressed. They argued that the phonics teaching received by the Scottish children facilitated their decoding skills at the early stages. The whole word teaching received by the New Zealand children, they argue, lead to them having less facility with accessing the phonology by an assembled route, but greater facility with an addressed route. The data from this study suggests that the effects of teaching are more complex than this. The Poor Readers and the Good Readers in this cohort had all received the same overall type of reading instruction, but they performed differently. The Good Readers, who had developed efficient decoding skills, were flexible enough to switch to orthographic encoding when the task demands required it. This must be seen as the aim of developing reading strategies - to have sufficient skill to access the internal lexicon by whichever route is fastest and most accurate. The results presented here suggest that the Good Readers were beginning to develop this level of flexibility.

The Poor Readers, on the other hand, showed a pattern of performance much like that of the New Zealand children when they were tested at 9 years. However, even they were beginning to show the development of a flexible approach to a task by 11 years, although they were of course still less skilled overall than the Good Readers.

Taking the results from Experiments 4 and 5 together the performance of the Good Readers at age 9 years showed that they could use a visual strategy for the lexical decision task and an alphabetic strategy for the phonological decision task. After two years their performance became faster and more accurate but the pattern did not change. It can be argued that they were beginning to develop an orthographic

Staiathoip/Qiapter 5

Strategy in 1990, and becoming increasingly sophisticated about switching strategies to suit the task demands.

On first testing the Poor Reader group showed no pseudo-homophone ^ e c t and chance performance on the phonological decision task. This pattern is that which was predicted for logographic readers in Table 5.1. The combined results indicate that this group were using a logographic strategy only. Though the visual strategy is more effective for the lexical decision task, their performance on the phonological decision task indicated that they were unable to change strategy to suit different task demands. On second testing this group were still showing no pseudo-homophone effect, but they had improved their accuracy and speed of lexical decision. Their performance on the phonological decision task showed that they were able to use an alphabetic strategy. Taken together these results suggest that the children were showing an ability to switch strategies depending on the task demands.