p<.005); (delayed post test: Kruskal-Wallis I-Way Anova: X2= 8.49, df=2, p<.05)]; the
Analogy condition [(immediate: Kruskal-Wallis I-Way Anova: X2= 12.4, df=2, p<.005);
(Delayed: ICruskal-Wallis I-Way Anova: X2= 10.5, df=2, p<.005)J; and Contrast condition
[(immediate: Kruskal-Wallis I-Way Anova: X2= 13.7, df=2, p<.005); (delayed: Kruskal-
Wallis I-Way Anova: X2= 7.7, df=2, p<.05)].
To what extent does children’s performance on the sentence generation task differ by
linguistic condition ?
If all children from each linguistic condition were correct on the two target words, the
maximum total score would be 96 for each linguistic condition and each post test. Figure
5.20 below demonstrates children’s performance on the sentence generation task by linguistic
condition for both post tests.
Figure 5.20 Total number of correct responses in the sentence generation task by
linguistic condition for both post tests
∣ Immediateposttest
∏ Delayedpoettest
As the figure above shows children in the Inference condition performed better than the
children in the other conditions. Children’s responses were analysed using two ANOVAs
with scores in the sentence generation task as the dependent factor and linguistic condition
as the independent factor. There was a trend for significance for the differences during the
immediate post test, while no significant differences were found during the delayed post test.
Post hoc analysis during the immediate post test demonstrated that children in the Inference
condition performed significantly better than the children in the Analogy condition (Mann-
Whitney: Z=2.6, p<.05). Furthermore, during the delayed post test there was a trend for the