Table 1
Autism Scores per Group.
| MEAN (SD) | CRONBACH’S ALPHA (95%-CI) | ||
|---|---|---|---|
| AUTISM | COMPARISON | ||
| Principal component | 2.04 (1.04) | –0.44 (1.01) | |
| CATI | 164.98 (23.60) | 112.57 (25.02) | 0.95 [0.95,0.96] |
| AQ | 149.39 (15.09) | 114.39 (13.83) | 0.91 [0.89,0.92] |
[i] Note. As the principal component negatively correlated with the questionnaire scores, we here multiplied the means and correlation coefficients by –1 for a more intuitive mapping (higher: more symptoms). CATI and AQ were significantly correlated (r(395) = 0.86, p < .001) and Cronbach’s alpha of each questionnaire suggests excellent consistency in our sample. AQ: Autism Spectrum Quotient. CATI: Comprehensive Autistic Trait Inventory.
Figure 1
Task Procedure.
Note. Participants underwent four experimental blocks in total, of which the first half (n = 40 trials) consisted of cued tasks, as indicated by a vowel or consonant cue, and the second half of free choice trials, as indicated by the hashtag. Participants either had to decide whether a word was animate or inanimate, or smaller or larger than a basketball. If correct, participants received points (low/high) based on their reward condition (switch/repeat reinforced) in the cued phase but not in the free choice phase. Reward contingencies switched after the first experiment half. Letters and words were presented in white on a black background in the original experiment.
Figure 2
Voluntary Switch Rate in The First and Second Experiment Half.
Note. The model estimated main effect of voluntary task-switching across groups (autism or comparison group) per first (a) and second (b) experiment half. Our results suggest that, across groups, participants rewarded more on switches in the first experiment half alternate more between tasks in the first experiment half, but they do not seem to unlearn this mapping after the contingency reversal in the second half. The green, pink and black lines represent the marginal means extracted from emmeans (Lenth, 2022) averaged across gender and the means of age and ICAR.
Figure 3
Switch Costs and Interference Effect in Cued and Free Choice Trials.
Note. The model-estimated reaction time and accuracy interference effect and switch costs per (a) cued and (b) free choice experiment phase. Participants were slower on switch trials compared to repeat trials and less accurate on non-interfering compared to interfering trials. Next to and below the graphs is depicted what constitutes a repeat versus switch trial and a non-interfering versus interfering keyboard response.
Figure 4
Interference Effects and Switch Costs in Cued Choice Trials in the Group and Trait Analyses.
Note. While only reaching significance in the interference effect of the group analysis and switch costs in the trait analysis (non-transparent), our results suggest an antagonistic relationship between interference effects (stability) and switch costs (flexibility), mostly in the trait analyses, where people scoring higher on autism traits showed smaller interference effects at the cost of larger switch costs. The blue and purple lines represent the difference scores between the marginal means of repeat versus switch and interfering versus non-interfering trials (as depicted in Figure 3), extracted from emmeans (Lenth, 2022). For visualisation purposes, we show autism traits at the mean, as well as at one standard deviation below and above the mean.
Table 2
Correlations between the autism trait and the four principal components.
| COMPONENT | PEARSON’S r | p-value |
|---|---|---|
| Depressive symptoms/ Rumination | 0.449 | <.001 |
| Social withdrawal | 0.730 | <.001 |
| Alexithymia | 0.521 | <.001 |
| Compulsivity and intrusive thought | 0.572 | <.001 |
[i] Note. Since the autism component loaded negatively on each of the autism questionnaires, and the principal components positively on their respective items, we multiplied the autism component by –1 before computing correlations.