These three composite scores were then entered as covariates into separate MANCOVAs for verbal and visual declarative memory ( Table

3, Covariate: Working Memory). These EPZ015666 research buy analyses revealed, first of all, a statistically significant multivariate group effect for the declarative memory subtests of verbal information (p = .009), though with a smaller effect size than the analogous model with no covariates. The MANCOVA on the declarative memory subtests of visual information revealed no group differences (p = .278). The univariate tests examining group differences while controlling for working memory ( Table 5, under “Covariates: Working Memory”) yielded mostly small or medium effect sizes for the verbal information subtests, with only two of the subtests showing significant group differences (Short and Delayed recall of the Stories subtest). None of the visual information subtests yielded significant univariate group differences, and all showed small effect sizes. As with the working memory subtests that involve language, any observed SLI deficits on the verbal declarative memory subtests could be due

to language problems rather than impairments with declarative memory itself. Therefore we analysed the verbal declarative memory subtests while covarying the language factor described above. The MANCOVA yielded a significant multivariate group effect (p = .042), though with a further reduction (to medium) of the effect size ( Table 3, Covariates: Language Factor). Controlling for language abilities, none of the univariate learn more analyses of the individual measures of verbal declarative memory not were significant, and all showed small to medium effect sizes ( Table 5, under “Covariate: Language Factor”). Finally, to remove confounds of both working memory and language in the declarative memory subtests of verbal information, we included the three working memory composite scores as well as the language factor as covariates in the analyses. The

multivariate group effect was not significant (p = .328, Table 3). Moreover, none of the univariate group differences ( Table 4, under “Covariates: Working Memory & Language Factor”) were significant, and all showed small effect sizes. We investigated procedural memory by examining sequence learning with the SRT task. We first probed accuracy. The average proportion of correct responses for both groups approached ceiling (SLI: M = .89, SD = .08, Min = .69, Max = .99; TD: M = .92, SD = .06, Min = .62, Max = .99). An independent samples t-test on arcsine transformed proportions, to correct for non-normality, revealed no significant group difference in accuracy [t(100) = 1.681, p = .096, partial η2 = .027]. These results suggest that the two groups were responding with comparable levels of accuracy.