Grammatical judgment
The ANOVA for the diagnostic groups examining A’ scores revealed a main effect of position, F(1,81) = 7.10, p = .01, ŋ
p
2
= 0.08. Children were better at detecting late errors (M = 0.78, SE = 0.02) than early errors (M = 0.73, SE = 0.02). There was also a main effect of group, F(2,81) = 5.63, p < .001, ŋ
p
2
= 0.12. Pairwise comparisons revealed that the TD group (M = 0.82, SE = 0.02) displayed significantly (p < .05) better sensitivity to grammatical errors than the SLI group (M = 0.69, SE = 0.03). The A’ scores for the ASD group (M = 0.76, SE = 0.03) did not differ significantly from the scores of the other two groups. The position × group interaction was not statistically significant, F(2,81) = 2.38, p = .10, ŋ
p
2
= 0.06. Analysis of B'' scores indicated that there was no significant difference for the diagnostic groups with respect to response bias, F(2, 81) = 0.54, p = .59, ŋ
p
2
= 0.01.
Analysis of RT scores for the diagnostic groups revealed a main effect of position, F(1,80) = 557.57, p < .001, ŋ
p
2
= 0.88, such that RTs for detecting early errors were longer (M = 2936.74, SE = 54.10) than RTs for detecting late errors (M = 1825.79, SE = 35.10). The main effect of group was not significant for RT, F(2, 80) = 1.23, p = .30, ŋ
p
2
= 0.03, nor was the position × group interaction effect, F(2,80) = 2.12, p = .13, ŋ
p
2
= 0.05.
For the language status groups, the analysis of A’ scores yielded a significant main effect of position, F(1,82) = 8.70, p < .001, ŋ
p
2
= 0.10. Late errors were detected more easily (M = 0.77, SE = .0.01) than early errors (M = 0.72, SE = 0.02). There was also a significant main effect of group, F(1,82) = 22.65, p < .001, ŋ
p
2
= 0.22, with the NL group performing better (M = 0.82, SE = 0.02) than the LI group (M = 0.67, SE = 0.02). The position × group interaction was significant, F(1,82) = 6.97, p = .01, ŋ
p
2
= 0.08, indicating that the LI group performed better on the late errors (M = 0.72, SE = 0.02) than on early errors (M = 0.63, SE = 0.03) whereas the NL group performed similarly on late errors (M = 0.82, SE = 0.02) and early errors (M = 0.82, SE = .02). Analysis of B'' scores revealed that the LI and NL groups did not differ significantly with respect to response bias, F(1,82) = 0.36, p = .55, ŋ
p
2
= 0.00.
Because the language status groups differed significantly on nonverbal cognition, an analysis of covariance (ANCOVA) was conducted using WISC-IV scores as the covariate in addition to the original analysis of A’ scores reported above. ANCOVA results revealed no significant main effect of position, F(1,81) = 0.58, p = .45, ŋ
p
2
= 0.01. However, there was a significant main effect of group, F(1,81) = 14.50, p < .001, ŋ
p
2
= 0.15 such that the NL group performed better (M = 0.81, SE = 0.02) than the LI group (M = 0.69, SE = 0.02). The position × group interaction was also significant, F(1,81) = 5.50, p = .02, ŋ
p
2
= 0.06. The LI group was better at detecting late errors M = 0.73, SE = 0.02) than early errors (M = 0.65, SE = 0.03) but the NL group performed equally well on late errors (M = 0.81, SE
= 0.02) and early errors (M = 0.81, SE = 0.02). Analysis of B'' scores from the ANCOVA again revealed no significant difference in response bias for the LI and NL groups, F(1,82) = 0.55, p = .55, ŋ
p
2
= 0.00.
In terms of RTs for the language status groups, ANOVA results revealed a significant effect of position, F(1,81) = 524.28, p < .001, ŋ
p
2
= 0.87, with longer RTs observed for early errors (M = 2944.84, SE = 56.13) than late errors (M = 1825.24, SE = 35.62). The main effect for language status group with respect to RTs was not significant, F(1,81) = 0.08, p = .77, ŋ
p
2
= 0.00 and the position × group interaction was also not significant, F(1,81) = 0.01, p = .94, ŋ
p
2
= 0.00. Statistically equivalent results for RTs were obtained from the ANCOVA analysis that accounted for group differences in nonverbal cognition. There was a significant position effect, F(1,80) = 9.73, p < .001, ŋ
p
2
= 0.11, such that children were slower to respond to early errors (M = 2942.23, SE = 56.72) than late errors (M = 1826.26, SE = 36.02). There was no main effect for language status group for RTs, F(1,80) = 0.04, p = .84, ŋ
p
2
= 0.00, or position × group interaction, F(1,80) = 0.03, p = .84, ŋ
p
2
= 0.00.
Nonverbal WM
The ANOVAs examining performance of the diagnostic groups indicated that there was no significant group difference on the N-back task, F(2,81) = 0.57, p = .56, ŋ
p
2
= 0.01. On the N-back task the following values were obtained: TD M = 0.78, SE = 0.04, ASD M = 0.75, SE = 0.04, and SLI M = 0.82, SE = 0.05.
Similarly, analysis of the language status groups yielded no significant differences on the N-back task, F(1,82) = 0.21, p = .65, ŋ
p
2
= 0.00. Group means were NL M = 0.79, SE = 0.03, and LI M = 0.77,
SE = 0.04.
Relation between nonverbal WM and grammatical judgment
In order to assess the association between nonverbal WM and morphosyntactic processing, we employed separate linear regression models for each group within the diagnostic and language status groupings. The predictor variable was 1-back target accuracy on the N-back task (this specific index from the N-Back task was used in a latent variable analysis of EF [
100]). The criterion variables were early
A’, late
A’, early RT, and late RT. Holm-Bonferroni sequential correction was used to account for the multiple regression analyses and adjusted
p values (
p’) are reported.
For the diagnostic groups, the N-back task did not significantly predict
A’ scores for early errors on the grammatical judgment task for the TD group,
F(1,34) = 0.74,
p’ = .79,
R
2
= 0.02 (see summary of regression findings in Table
2). For late errors, however, there was a significant relationship between WM and morphosyntactic processing for the TD group,
F(1,34) = 8.58,
p’ = .02,
R
2
= 0.20, such that N-back scores predicted performance. For the ASD group, nonverbal WM was associated with
A’ scores for early errors on the grammatical judgment task,
F(1,25) = 10.62,
p’ = .02,
R
2
= 0.30, such that the N-back task significantly predicted performance. The same pattern was observed for
A' scores for late errors in the ASD group,
F(1,25) =11.10,
p′ = .02,
R
2
= 0.31. For the SLI group, the association between visual WM and
A’ scores for early errors was not significant,
F(1,19) = 3.09,
p’ = ..29,
R
2
= 0.25, and similarly there was no significant relationship for late errors,
F(1,19) = 0.12,
p’ = .79,
R
2
= 0.01.
Table 2
Summary of nonverbal working memory and morphosyntactic processing relationships for the diagnostic groups and language status groups
TD | NS |
R
2 = 0.20* | NS |
R
2 = 0.22* |
ASD |
R
2 = 0.30* |
R
2 = 0.31* | NS | NS |
SLI | NS | NS | NS | NS |
Language status Groups | | | | |
NL | NS |
R
2 = 0.20* | NS | NS |
LI |
R
2 = 0.21* | NS | NS | NS |
Next, we examined the role of nonverbal WM in predicting RT scores for early vs. late errors for the diagnostic groups. For the TD group, there was not a significant relationship between WM and RT scores for early errors, F(1,34) = 2.95, p’ = .38, R
2
= 0.08. However, the association between nonverbal WM and RTs for late errors for the TD group was significant, F(1,34) =9.75, p’ = .02, R
2
= 0.22. For the ASD group, nonverbal WM did not predict RT on errors occurring early in the sentence, F(1,25) =5.31, p’ = .15, R
2
= 0.18 or late in the sentence, F(1,25) = 0.08, p’ = .99, R
2
= 0.00. Similarly, for the SLI group there was no significant relationship between WM and RTs for either early errors, F(1,18) = 0.02, p’ = .99, R
2
= 0.00, or late errors, F(1,19) =0.01, p’ = .99, R
2
= 0.00.
The regression analysis for the language status grouping revealed no significant relationship between nonverbal WM and A' scores for the NL group on early errors, F(1,52) = 3.69, p′ = .12, R
2
= 0.07. However, there was a significant association between WM and A’ scores for the NL group for late errors, F(1,52) = 12.65, p’ = .00, R
2
= 0.20, with the N-back task performance predicting morphosyntactic processing. The opposite profile was observed for the LI group, such that there was a significant relationship between WM and A’ scores on the grammatical judgment task for early errors, F(1,28) = 7.62, p’ = .03, R
2
= 0.21, but no significant association for late errors, F(1,28) = 0.72, p′ = .41, R
2
= 0.03.
With respect to RT scores for the language status groups, there was not a significant relationship between WM and performance for the NL group on early error items, F(1,52) = 4.98, p’ = .10, R
2
= 0.09, or on late errors, F(1,52) =5.42, p’ = .10, R
2
= 0.09. Likewise, for the LI group, there was no significant relationship between WM and RT on early errors, F(1,27) = 0.91, p’ = .70, R
2
= 0.03, or late errors, F(1,28) = 0.01, p’ = .92, R
2
= 0.00.