Descriptive data by diagnostic group are presented in . The diagnostic groups did not differ on chronological age, WISC VCI, WISC PRI, or gender distribution. As expected, children in the HFA group had significantly higher scores on the SCQ, t(59)=11.65, p < .001, d=2.99, and the ASSQ, t(59)=12.44, p< .001, d=3.20. Children in the HFA group tended to provide fewer correct responses to the Strange Stories vignettes, t(48)= −1.96, p = .055, d=0.55; however, they did not differ from children in the comparison sample on the Children’s Eyes Task, t(53)= −1.31, ns.
Sample Characteristics for HFA and Comparison Groups: Mean (SD)
Memory Task Performance
Data were analyzed using a 2 (Diagnostic Group) × 3 (Processing Condition: Self, Other, Letter) mixed ANCOVA with VCI and age as covariates. After controlling for age and VCI, there was a significant interaction between Diagnostic Group and Processing Condition, F(2, 57)=4.00, p = .02, ηp2= .12 (see ). In addition, there was an interaction between Age and Processing Condition, F(2, 57)=4.66, p = .013, ηp2 =.14.
Memory Sensitivity (d′) by Processing Condition and Diagnostic Group
In order to examine the interaction between Diagnostic Group and Processing Condition, a series of between- and within-group post hoc comparisons were conducted. Between-group comparisons revealed that children with HFA did not differ from children in the comparison group on mean level performance in any condition. Within-group analyses revealed that children in the comparison group displayed the typical SRM effect in that they had increased memory for self- compared to other-referenced words, t(30)=4.66, p < .001, d=0.65, self- compared to letter-referenced words, t(30)=9.75, p < .001, d=1.69, and other- compared to letter-referenced words, t(30)=5.61, p < .001, d=1.04. Alternatively, this pattern of differences was not apparent for children in the HFA group whose performance did not differ in the self- versus other-referenced conditions, t(30)=0.64, ns. However, children with HFA performed better in the self versus letter condition t(30)=7.13, p < .001, d=.97 and in the other versus letter condition t(30)=7.11, p < .001, d=1.02 (see ).
In order to interpret the Age by Processing Condition interaction, a series of hierarchical regression analyses were conducted to examine the effects of (1) age (as a continuous variable), (2) diagnostic group, and (3) the interaction of age by diagnostic group on memory performance. Separate regression analyses were conducted for each processing condition and for the three difference scores reflecting the magnitude of the advantage for processing in the self- or other-referenced conditions. In this way, developmental changes in memory performance as well as possible differences in developmental trajectories by diagnostic group (i.e., the interaction of age and diagnostic group) were evaluated. Memory performance in the self condition was significantly predicted by age, F(1, 60)=5.55, p =.02, ΔR2 =.09, f2 =.09, with increasing age being associated with better performance. After controlling for age, diagnostic group and the interaction of age and diagnostic did not account for significant unique variance in performance in the self-referenced condition. Memory performance in the other and letter counting conditions was not associated with age, diagnostic group, or their interaction.
The magnitude of the self-other bias (d′ self - d′ other) was significantly predicted by age, F(1, 60)=4.36, p=.04, ΔR2 = .07, f2 =.07. Again, as children got older, there was a greater difference between performance in the self and other conditions. After controlling for age, diagnostic group accounted for additional unique variance in the magnitude of the self-other bias, F (1, 59) = 5.78, p = .02, ΔR2 =.08, f2=.09, Total R2 = .15, f2 = .18. That is, children with HFA, regardless of age, showed a smaller self-other bias compared to children in the comparison group (see ). A similar pattern of results emerged for the magnitude of the self-letter bias (d′ self - d′ letter), such that age predicted a larger self-letter bias, F(1, 60)=10.75, p=.002, ΔR2 = .15, f2 =.18 and diagnostic status predicted additional unique variance, F(1, 59) = 7.49, p=.008, ΔR2 = .10, f2 =.11, Total R2 = .25, f2 = .33. The interaction of age and diagnostic group was not significant for either of these analyses, suggesting that all children, regardless of diagnostic group, showed age-related increases in the preferential processing of self-referenced information. The magnitude of the other-letter bias was not associated with age, diagnostic status, or the interaction of age and diagnostic status.
Age as a Predictor of Self-Other Bias by Diagnostic Group
Individual Differences: Self-referenced Memory and Autism Symptom Expression
The degree to which self-referenced memory performance was related to social behavior and autism symptoms on the ASSQ and SCQ was examined in a series of hierarchical regression analyses. In the first two analyses, ASSQ scores and SCQ total scores were regressed on Age, Verbal IQ, Diagnostic Group, memory performance in the self-referent condition (d′ self), and the interaction of diagnostic group and memory performance. In the second two analyses, ASSQ scores and SCQ total scores were regressed on Age, Verbal IQ, Diagnostic Group, self-other bias scores (d′ self- d′ other), and the interaction of diagnostic group and self bias.
Of the four regression models, the only one in which memory performance predicted unique variance over and above the effects of diagnostic group was the one in which ASSQ scores were regressed on the self-other bias score. Specifically, after accounting for diagnostic group differences on the ASSQ, the magnitude of the self-referent bias was significantly associated with ASSQ scores, ΔR2 = .03, F(1, 56)=6.20, p=.016, f2=.03; Total R2 = .74, f2 = 2.85. The lack of an interaction between self-referent bias and diagnostic status suggests that for all children, better memory for words in the self versus other processing condition was related to parent reports of fewer social difficulties on the ASSQ (see ).
Self-Other Bias as a Predictor of ASSQ by Diagnostic Group
Individual Differences: Self-Referenced Memory, Social Cognitive Mentalizing, and Autism Symptoms
A third set of analyses examined relations between SRM performance (d′ self, d′ other), self-other bias (d′ self- d′ other) and social cognitive mentalizing measures (Eyes Task, Strange Stories). Six regression analyses were run in which social cognitive mentalizing scores (Eyes Task or Strange Stories) were regressed on Age, Verbal IQ, diagnostic group, memory performance (self, other, or self-other), and the interaction of diagnostic group and memory performance. Eyes Task performance was positively associated with Age, F(1, 48)=8.02, p=.007, ΔR2=.14, f 2= .16, and Verbal IQ, F(1, 47)=7.30, p=.01, ΔR2=.12, f 2= .14. Diagnostic group did not account for unique variance in Eyes Task performance, however; additional unique variance was accounted for by memory performance in the self condition, F(1, 45) = 4.82, p =.03, ΔR2 = .07, f 2 = .07 and a trend for the interaction of diagnostic group and performance in the self condition, F(1, 44) = 3.38, p = .07, ΔR2 = .05, f 2 =.06; Total R2 = .39, f 2 = .64. In order to examine the trend, the associations between memory performance in the self condition and Eyes Task performance were examined separately for children in the HFA and comparison samples. For children in the HFA sample, there was a strong positive correlation between memory performance in the self condition and Eyes Task performance, r(25)=.57, p =.003, d=1.39; however, the correlation was not significant for children in the comparison sample, r(30)=.19, ns, d=0.39 (see ). After controlling for Age, Verbal IQ, and Diagnostic Group, Eyes Task performance was not associated with memory performance in the other-referenced condition or the magnitude of the self-other bias.
Self-referenced memory as a Predictor of Eyes Task by Diagnostic Group
Performance on the Strange Stories task was positively associated with Age, F(1, 48)=4.96, p=.03, ΔR2=.09, f 2= .10, and Verbal IQ, F(1, 47)=12.99, p=.001, ΔR2=.20, f 2= .25. Diagnostic Group accounted for additional unique variance, F(1, 46)=3.89, p=.055, ΔR2=.06, f 2= .06. However, none of the indices of SRM performance predicted unique variance above and beyond these effects.
A final goal of the study was to assess whether the associations between self-referenced memory and symptom scores were mediated by social-cognitive task performance. Given that the self-other difference score was associated with ASSQ scores across the entire sample, partial correlations were run controlling for Eyes Task and Strange Stories Performance. Controlling for Eyes Task performance, the association between the self-other difference scores and ASSQ scores remained significant across the entire sample, partial r(51)= −.44, p = .001, d=.98, and within the HFA sample alone r(22)= −.43, p = .04, d=.95. Controlling for Strange Stories performance, the association between the self-other difference score and ASSQ scores remained significant across the entire sample, partial r(46)= −.42, p =.003, d=.93, and was slightly reduced within the HFA sample alone r(20) = −.38, p = .08, d=.82. Thus, it did not appear that social cognitive mentalizing as assessed using the Eyes Task or the Strange Stories task mediated or accounted for the observed associations between SRM performance and symptom expression.