Hierarchical Regressions
We conducted hierarchical multiple regression analyses in order to investigate the association between ToM performance, executive functioning, and subclinical ASD traits whilst controlling for age and general cognitive ability. Beta estimates for the models are presented in Table
4. This analysis was intended to further assess the hypothesis that, in addition to a unique contribution by ASD traits, the cognitive, but not the affective, domain of empathy would be associated with executive control abilities (H3).
For the model predicting naturalistic ToM performance, total MASC scores were first regressed onto Age and IQ scores. The regression model was significant (
R = .45,
R
2 adj = .19,
F
(2,104) = 13.25,
p < .001), with the two predictors collectively explaining 20 % of the variance in MASC scores. Age (
t
(104) = 2.79,
p = .006) and IQ (
t
(104) = 2.20,
p = .030) were both positive predictors of MASC performance. At the second step ASD traits were entered. The regression model remained significant (
R = .54,
R
2 adj = .27,
F
(3,103) = 13.74,
p < .001), with the three predictors jointly explaining 29 % of the variance in naturalistic ToM performance. ASD traits (
t
(103) = 3.46,
p = .001) were uniquely and negatively associated with MASC performance, whilst Age (
t
(103) = 2.09,
p = .039) and IQ (
t
(103) = 2.53,
p = .013) revealed unique positive associations. The
R
2
change was significant (
F
(1,103) change = 11.94,
p = .001), indicating that including ASD traits explained significant additional variance in the model.
Table 4
Hierarchical regressions of naturalistic ToM and Static ToM on IQ and age (Step 1), autism traits (Step 2), and executive control (Step 3)
Step 1 |
IQ | .226 | 2.197 | .030* | .148 | 1.315 | .191 |
Age | .288 | 2.794 | .006** | .093 | .829 | .409 |
Step 2 |
IQ | .249 | 2.532 | .013* | .163 | 1.458 | .148 |
Age | .211 | 2.094 | .039* | .044 | .384 | .701 |
AQ | −.296 | 3.456 | .001** | −.190 | 1.951 | .054 |
Step 3 |
IQ | .187 | 1.866 | .065 | .058 | .503 | .616 |
Age | .174 | 1.732 | .086 | .024 | .213 | .831 |
AQ | −.243 | 2.862 | .005** | −.180 | 1.852 | .067 |
WCST | .241 | 2.778 | .007** | .272 | 2.749 | .007** |
Go/No-Go | −.127 | 1.505 | .135 | .068 | .708 | .481 |
ToL-F | .042 | .447 | .656 | .070 | .660 | .511 |
At the third stage, measures of executive function (i.e., WCST Efficiency, GNG, and ToL-F) were entered. The regression model was significant, (R = .60, R
2 adj = .32, F
(6,100) = 9.30, p < .001), and together, the six predictors explained 36 % of the variance in MASC scores. ASD traits (t
(100) = 2.86, p = .005) and WCST performance (β = .24, t
(100) = 2.78, p = .007), emerged as significant predictors of naturalistic ToM performance. The signs of the coefficients suggested that higher ASD traits and lower levels of cognitive flexibility were related to difficulties in mental state attribution in a naturalistic context. The R
2
change was significant (F
(3,100) change = 3.76, p = .013), suggesting that executive functioning explained significantly more variance in the model. None of the other predictors in the model reached statistical significance, Age (t
(100) = 1.73, p = .086), GNG (t
(100) = 1.51, p = .135), and ToL-F (t
(100) = .447, p = .656), although IQ (t
(100) = 1.87, p = .065) indicated a trend towards significance.
The same regression sequence was applied to static ToM performance. At the first step, scores on the EYES test were regressed onto Age and IQ. The regression model was non-significant, (R = .21, R
2 adj = .03, F
(2,104) = 2.47, p = .089), and neither Age (t
(104) = .829, p = .409), nor IQ (t
(104) = 1.32, p = .191) reached individual statistical significance in the model. ASD traits were entered at the second step. The regression model was significant, (R = .28, R
2 adj = .05 F
(3,103) = 2.96, p = .036). Collectively, the three predictors explained 8 % of the variance in EYES scores. The unique negative association between ASD traits (t
(103) = 1.95, p = .054) and EYES scores approached significance. The R
2 change was significant at trend level (F
(1, 103) change = 3.81, p = .054), suggesting that ASD traits explain incremental variance in the model. Once again, Age (t
(103) = .384, p = .701) and IQ (t
(103) = 1.46, p = .148) did not reach statistical significance.
Measures of executive function (i.e., WCST Efficiency, GNG, and ToL-F) were entered at the third and final step. The regression model was significant, (R = .40, R
2
adj = .11 F
(6,100) = 3.17, p = .007), and together, the six predictors explained 16 % of the variance in EYES scores. WCST performance (t
(100) = 2.75, p = .007), emerged as a significant predictor of static ToM. The R
2 change was significant (F
(3,100) change = 3.18, p = .027), indicating that including executive functioning explained significant additional variance in the model. None of the remaining predictors reached statistical significance: ASD traits (t
(100) = 1.85, p = .067), GNG (t
(100) = .708, p = .481), ToL-F (t
(100) = .660 p = .511), IQ (t
(100) = .503, p = .616), and Age (t
(100) = .213, p = .831).
Last, bivariate correlations were computed to assess the hypothesis that individuals with elevated levels of ASD traits (H4) and alexithymia (H6) would demonstrate poorer performance on tasks indexing executive control. Bivariate correlation coefficients for all variables are presented in Table
5. There was a significant positive correlation between ASD traits and commission errors on the GNG task (
r = .250,
p = .009). Analysis also revealed a significant negative association between ASD symptomatology and WCST Shift scores (
r = −.224,
p = .021). However, the negative correlations between ASD traits and WCST Efficiency scores and ToL-F performance were not statistically significant (
p > .05). A similar pattern emerged with alexithymia. Whilst there was a significant positive association with GNG scores (
r = .219,
p = .023), the negative relationship with WCST (Efficiency:
r = − .135,
p = .165; Shift:
r = −.134,
p = .170) and ToL-F did not reach statistical significance (
r = −.153,
p = .116). Once again, gender was included in all analyses and subsequently eliminated after returning non-significant results.
Table 5
Descriptive statistics and correlations between measures of autism spectrum disorder, alexithymic traits, and executive control
1. AQ | – | | | | |
2. TAS | .476** | – | | | |
3. WCST Shift | −.224* | −.134 | – | | |
4. WCST Efficiency | −.133 | −.135 | .829** | – | |
5. GNG | .250** | .219* | −.152 | −.057 | – |
6. ToL | −.013 | −.153 | .231* | .295** | .112 |
Taken together, findings from Study 2 indicated a substantial overlap between empathic processing, executive function, and ASD traits. Analysis revealed that higher scores on the naturalistic ToM task was associated with better performance across all components of executive processing, whilst static ToM was associated with planning and cognitive flexibility. By contrast, there were no statistically significant associations between the affective domain of empathy and executive function.
Our findings also demonstrated age-related improvements in naturalistic ToM as well as in the set-shifting and planning domains of executive control. However, the association between age, affective empathy, and commission errors on the response inhibition task did not reach statistical significance.
The hierarchical regressions suggested that accurately decoding mental states from video-based stimuli is associated with lower levels of autism symptomatology and flexible cognition. Of the executive function measures used in this study, accurate performance on the static EYES test was exclusively associated with set-shifting ability. In terms of the autism symptomatology and executive function relationship, findings showed that individuals with higher levels of ASD traits exhibit a profile of executive function impairments that is partially comparable to those reported in clinical ASD. Lastly, greater levels of alexithymia were also found to be associated with impaired response inhibition. However, the negative correlation between alexithymia and the executive domains of planning and set-shifting ability did not reach statistical significance. Overall, our data yielded strong support for H3, partial support for H4 and H6, and also indicate the existence of age-related advancements in mentalizing ability and executive control.