Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2024 Dec 5;18(2):281–294. doi: 10.1002/aur.3281

Cognitive flexibility in autism: How task predictability and sex influence performances

Adeline Lacroix 1,2,✉, Emma Torija 1, Alexander Logemann 3, Monica Baciu 1, Renata Cserjesi 3, Frédéric Dutheil 4, Marie Gomot 5, Martial Mermillod 1
PMCID: PMC11826006  PMID: 39635938

Abstract

While cognitive flexibility challenges are frequently reported in autistic individuals, inconsistencies in the findings prompt further investigation into the factors influencing this flexibility. We suggest that unique aspects of the predictive brain in autistic individuals might contribute to these challenges, potentially varying by sex. Our study aimed to test these hypotheses by examining cognitive flexibility under different predictability conditions in a sample including a similar number of males and females. We conducted an online study with 263 adults (127 with an autism diagnosis), where participants completed a flexibility task under varying levels of predictability (unpredictable, moderately predictable, and predictable). Our results indicate that as task predictability increases, performance improves; however, the response time gap between autistic and non‐autistic individuals also widens. Moreover, we observe significant differences between autistic males and females, which differ from non‐autistic individuals, highlighting the need to consider sex differences in research related to the cognition of autistic individuals. Overall, our findings contribute to a better understanding of cognitive flexibility and sex differences in autism in light of predictive brain theories and suggest avenues for further research.

Keywords: autism, cognitive flexibility, gender differences, predictive brain, sex differences

Lay Summary

We used an online task‐switching task to explore how task predictability and sex affect cognitive flexibility in autistic individuals. Our findings show that increased predictability improves performance for both autistic and non‐autistic individuals. However, the response time gap between these groups becomes larger as predictability increases. Additionally, we found unique differences in cognitive flexibility between autistic males and females, which do not match the patterns seen in non‐autistic individuals. These results enhance our understanding of predictive coding in autism and emphasize the importance of considering sex differences in autism research.

BACKGROUND

Cognitive flexibility is not a unified concept (Ionescu, 2012). It partly refers to the ability to switch from one task to another depending on the situation or the environment (Miyake et al., 2000), and to adapt one's cognitive processing strategies to cope with changes and unpredictability (Han et al., 2012). A key characteristic of autism includes resistance to change, a strong need for immuability (repetitive behaviors, insistence on sameness), and adapting difficulties (American Psychiatric Association, 2013). This has been linked to cognitive inflexibility (Mostert‐Kerckhoffs et al., 2015), making cognitive flexibility an important process to understand in autism, as it can affect various aspects of daily life, including adaptive behaviors (Bertollo et al., 2020), social interactions (Bertollo et al., 2020), and emotional functioning (Cai et al., 2018; Scarpa et al., 2021). It can also play a role in mental health challenges in autism, such as anxiety (Harrop et al., 2024). Thus, understanding cognitive (in) flexibility in autism can help shape targeted support to improve psychosocial functioning (Harrop et al., 2024; Scarpa et al., 2021).

Despite being extensively studied, the manifestation of cognitive inflexibility in autism presents intriguing complexities. Notably, there is a discrepancy between laboratory findings measured with tests, revealing mixed results, and real‐world experiences typically assessed through questionnaires like the Behavioral Rating Inventory of Executive Function scale, showing more consistent flexibility difficulties in autism (Geurts et al., 2009; Leung & Zakzanis, 2014; Teunisse et al., 2012). It has been suggested that this disparity may stem from the lack of ecological validity in traditional research methods (Geurts et al., 2009; Leung & Zakzanis, 2014). Daily life situations involve specific characteristics that significantly impact flexibility difficulties. In particular, the predictability of a situation could be a critical factor that could influence cognitive flexibility in autism (Van Eylen et al., 2011). This hypothesis is particularly compelling given the predictive coding theories of autism (Brock, 2012; Gomot & Wicker, 2012; Pellicano & Burr, 2012; Van de Cruys et al., 2014), which have gained interest during the past decade (for a review see Cannon et al., 2021). The latest encompass several hypotheses, proposing that autistic individuals may assign low weight to predictions (Pellicano & Burr, 2012) and/or would alternatively rely more than non‐autistic (NA) individuals on sensory input (Brock, 2012). Another hypothesis is that autistic individuals present High and Inflexible Precision to Prediction Error (HIPPEA, Van de Cruys et al., 2014), leading them to overfit environmental variability, and making unpredictable environments particularly challenging. This may partly explain adaptive difficulties and resistance to change in autism. Thus, while autistic individuals may demonstrate intact flexibility in predictable scenarios (Van Eylen et al., 2011), their performance may decline in unpredictable contexts (Lacroix et al., 2022).

In line with this hypothesis, Lacroix et al. (2022) found lower performances for autistic individuals compared with NA on a flexibility task involving unpredictable shifts, while their switch cost (i.e., the difference between the shift and non‐shift conditions) was similar to those of NA individuals in a task involving predictable shifts. However, this study utilized different types of stimuli for the two tasks. The unpredictable task (Emotional Shifting Task/EST–Biro et al., 2021; Biró et al., 2022) involved socio‐emotional stimuli, which are inherently challenging for autistic individuals to process, while the predictable task (Task Switching Task/TST–Rogers & Monsell, 1995) utilized character stimuli. This limitation prevents drawing conclusions on which specific aspect(s) of the task particularly impaired autistic individuals—whether it was the social nature of the stimuli (Lacroix, Bennetot‐Deveria et al., 2024), their emotional content, or the unpredictability of the task. Therefore, further research is needed to differentiate between these factors more specifically.

In addition, Lacroix et al. (2022) demonstrated sex differences autism, that mirror sex differences found in NA, in the task involving unpredictable shifts of socio‐emotional stimuli, with response times (RT) being shorter for females. However, there was no interaction with the shift condition, indicating that the difference is likely not related to the unpredictable nature of the stimuli but may be more associated with socio‐emotional processing, where sex differences have been reported in autism (Harrop et al., 2018, 2020; Harrop et al., 2019; Lacroix, Harquel et al., 2024). In the predictable task using character stimuli (TST), NA males were faster than NA females, while autistic males were slower than autistic females. This result could suggest faster processing speed in autistic females compared with autistic males (Lehnhardt et al., 2016) or greater flexibility (Bölte et al., 2011; Lehnhardt et al., 2016) or ability to adjust social predictions in autistic females. To fully understand these potential explanations, additional research is required to clarify and distinguish among these possibilities.

Thus, the present study had two main goals. First, it sought to address the role of unpredictability in flexibility difficulties in autism. Second, it investigated sex differences in these processes. To achieve this, we designed a modified TST task, named the Prediction Modulated Task Switching Task (P‐TST). In this task, participants were presented with one letter and one number appearing side by side on the screen, surrounded by either a yellow or blue frame. The yellow frame indicated that participants should perform the number task, pressing “V” for even numbers and “B” for odd numbers (two keys side by side). The blue frame indicated the letter task, with participants pressing “V” for consonants and “B” for vowels. There were three blocks in the task: a predictable block, in which participants repeatedly performed the letter task three times and the number task three times; a moderately predictable block, in which the task was predictable 70% of the trials; and an unpredictable block, in which the tasks appeared randomly. We preregistered five hypotheses. Our main hypothesis was that the performance gap between autistic and non‐autistic (NA) individuals would be larger in moderately predictable and unpredictable conditions than in the predictable condition, supporting the hypothesis that unpredictable situations would be more challenging for autistic individuals compared with NA individuals. We also hypothesized that we would replicate the results of Lacroix et al. (2022), with autistic females responding faster than autistic males but more slowly than NA individuals, and NA females responding more slowly than NA males. This might vary according to the level of predictability, but this was not preregistered and was tested for exploratory purposes. Additionally, we hypothesized that shifting from one task to another would decrease performance (fewer correct responses and longer response times), which is a prerequisite for confirming the task's ability to induce a switch cost, usually used to capture cognitive flexibility (Hohl & Dolcos, 2024). Finally, we hypothesized that increased predictability would enhance performance in a linear trend, and autistic individuals would be slower than NA individuals, consistent with processing speed impairments in autistic adults (Haigh et al., 2018).

METHOD

Participants

We gathered data from 263 French‐speaking participants aged 18 to 45, to capture the period when executive function maturity is optimal. As preregistered, we excluded NA participants with Autism Spectrum Quotient (AQ) scores greater than 32 (Baron‐Cohen et al., 2001) (N = 4). Additionally, transgender participants (N = 5 autistic female‐to‐male) were excluded due to insufficient gender diversity for analysis. The final sample included 127 autistic adults (65 females) and 127 NA adults (69 females), with the groups matched for age (mean age = 34 ± 7) (see Figure 1a). Autistic participants with a pre‐existing diagnosis were recruited through specialized centers, healthcare professionals, and autism‐focused organizations to ensure they had received an official diagnosis. Some participants were also recruited via social media; in these cases, individuals were required to contact us to verify their formal diagnosis. We asked specific questions about the diagnosing center, the assessment used, and the study link was only sent if they confirmed a diagnosis from a qualified healthcare professional. Additionally, the questionnaire included questions about the type of professional/center who provided the diagnosis. NA participants were recruited via social media, with the stipulation that they should not suspect themselves of being autistic nor have immediate family members (child, parent, sibling) on the autism spectrum. All participants were required to be free from medical treatments that could impact cognitive functions.

FIGURE 1.

FIGURE 1

Open in a new tab

Stimuli and procedure. (a) Number of participants in each group. (b) Procedure including the three conditions. Predictable condition = task shifting every three trials. Moderately predictable condition = task shifting every three trials 70% of the trials, and after two trials 30% of the trials. Unpredictable condition = task shifting randomly. (c) Example of a screen when a wrong answer is given (top) and enlargement of the message (bottom).

All procedures were conducted in accordance with the World Medical Association's Declaration of Helsinki. The study was also approved by the local ethics committee (CER‐Grenoble Alpes, COMUE University Grenoble Alpes, IRB00010290).

Material and procedure

The study was conducted online using the PsyToolkit platform (Stoet, 2010, 2016). Participants began the study by reviewing and agreeing to an informed consent form. They were instructed to perform the tasks in a quiet, distraction‐free environment. Demographic information, including sex, age, and diagnoses (autism and other psychiatric, neurological, or neurodevelopmental conditions), was collected at the outset. Participants completed two tasks: the P‐TST (described below) and another task related to a different research question (Lacroix, Bennetot‐Deveria et al., 2024), presented in a randomized order. Following these tasks, participants completed the Autism Spectrum Quotient (Baron‐Cohen et al., 2001). To verify the integrity of the experimental conditions, participants were asked about any potential disturbances during the tasks and were given the opportunity to provide comments.

P‐TST

This task was adapted from the Task Switching Task by Rogers and Monsell (1995) to evaluate participants' shifting abilities under different levels of predictability. During the task, participants were presented with a letter and a number side by side on the screen. These characters were enclosed in either a blue or a yellow frame. The blue frame indicated that the letter task should be performed, where participants had to press ‘V’ if they saw a consonant (G, K, M, R) and ‘B’ if they saw a vowel (A, E, I, U). The yellow frame indicated that the number task should be performed, where participants had to press ‘V’ if they saw an odd number (3, 5, 7, 9) and ‘B’ if they saw an even number (2, 4, 6, 8).

To familiarize themselves with the tasks, participants first completed 15 trials of the letter task only, followed by 15 trials of the number task. After this training, participants performed the test task, which consisted of three blocks of 36 trials each, corresponding to three conditions presented in a random order. In the predictable condition, participants performed the letter task three times, followed by the number task three times, requiring a task switch every three trials. In the moderately predictable condition, the task switch occurred every three trials for 70% of the trials, but after two trials for 30% of the trials. In the unpredictable condition, the task switch occurred randomly every 1, 2, or 3 trials (see Figure 1b). The type of condition was not indicated to the participants, making the level of predictability implicit. In case of an error, participants received a reminder of the instructions (see Figure 1c). The instructions were also displayed between each block.

For each trial, participants were given 5,000 ms to respond. If no response was given within this time or in case of a wrong answer, a reminder of the instructions was shown for 3 s. If a correct response was given, the task continued immediately without waiting for the 5,000 ms to elapse and without instruction's reminder.

Analyses

All statistical analyses were performed using R Core Team (2020) and RStudio Team (2019), version 2023.12.1 + 402. To compare the four subgroups in our sample (autistic males, autistic females, non‐autistic (NA) males, and NA females), linear models were used for continuous variables and Tukey's Honnestly Significant Difference for post‐hoc comparisons. For categorical variables, Pearson's Chi‐square test with Yates' correction was applied with Chi‐square test of association for post‐hoc comparisons.

Accuracy (correct responses [CR]) and RT for correct responses were analyzed for the P‐TST. Mixed‐effects logistic regression was performed for CR analysis, and generalized mixed‐effects models with a gamma distribution were used for RT analysis, utilizing the lme4 package (Bates et al., 2015). Fixed effects in the models included group (autism, NA), sex (males, females), shift condition (non‐shift, shift), and predictability (predictable, moderately predictable, unpredictable). We included psychiatric/neurological and neurodevelopmental diagnoses other than autism (e.g., Attention Deficit Hyperactivity Disorder), as well as age, as covariates to account for their potential impact on executive functions. Age was centered. The maximal random effects structure was kept, and reduced only in cases of model non‐convergence (Barr et al., 2013). Since there is no standardized method for calculating effect sizes in mixed models, we present unstandardized effect sizes (Rights & Sterba, 2019), represented by the β coefficient values. The dichotomous variables (group, sex, and shift condition) were contrast‐coded with values of −0.5 and 0.5 for easier interpretation. Custom contrasts for predictability included a linear trend test (−1, 0, 1) and a quadratic trend test (1, −2, 1). Detailed results are available in Data S1.

Significant interactions were illustrated and explained using estimated marginal means, which estimate the effect of each variable while controlling for other factors in the model, along with 95% confidence intervals (Cumming, 2009; Garofalo et al., 2022), employing the emmeans package (Lenth, 2021). Pairwise comparisons of estimated marginal means were also performed, with Tukey adjustment applied to estimate the differences (β values).

As pre‐registered, correlations with the AQ were conducted as complementary analyses and were presented in Data S1.

Despite the moderately uneven sample sizes, no participants were excluded, as this was not preregistered and mixed‐effects models are generally robust to unbalanced designs (Bolker, 2015; Schielzeth et al., 2020). To ensure this, we ran the models with a randomly selected equal number of participants from each group (N = 58 in each group), which yielded mostly similar outcomes. Nevertheless, we observed an additional interaction effect between group and predictability on CR in this second analysis, similar to that found in RT, which reinforces our conclusions. However, we did not report it in the results as we cannot determine whether it is a spurious effect, and further studies would be required to clarify this.

RESULTS

Participants

The demographics of the final sample are detailed in Table 1. No significant age differences were found between groups. However, differences in educational attainment were observed, with autistic males having lower levels compared with NA males (p < 0.001), NA females (p = 0.02), and autistic females (p = 0.03). As expected, AQ scores were significantly higher in autistic individuals compared with NA participants (all p < 0.001). Additionally, NA males had higher AQ scores than NA females (p = 0.02). Autistic males were diagnosed at a younger age than autistic females (p < 0.001; 2 missing data). Finally, autistic participants had more comorbid diagnoses than NA participants (all p < 0.001).

TABLE 1.

Demographics. Mean value, standard deviation (SD), and range for age, education, AQ scores, age at the diagnosis of autism, as well as the percentage of participants with a diagnosis other than autism for each group, and group comparison.

Autistic females (N = 65) Autistic males (N = 62) Non autistic females (N = 69) Non autistic males (N = 58) p value
Age 0.061
Mean (SD) 35.8 (6.7) 32.4 (7.7) 34.3 (7.6) 34.9 (7.1)
Range 19.0–45.0 18.0–45.0 19.0–45.0 20.0–45.0
Education <0.001
Mean (SD) 14.5 (1.8) 13.7 (2.0) 14.5 (1.1) 14.9 (1.6)
Range 9.0–17.0 9.0–17.0 12.0–17.0 9.0–17.0
AQ <0.001
Mean (SD) 38.4 (5.3) 37.1 (5.8) 14.3 (7.7) 17.6 (7.3)
Range 24.0–49.0 25.0–50.0 1.0–32.0 4.0–32.0
Diagnostic Age <0.001
Mean (SD) 31.3 (6.6) 25.7 (9.6) NA NA
Range 15.0–42.0 3.0–43.0 NA NA
Psychiatric/neurological diagnostic <0.001
No 49 (75.4%) 47 (75.8%) 69 (100.0%) 58 (100.0%)
Yes 16 (24.6%) 15 (24.2%) 0 (0.0%) 0 (0.0%)
Neurodevelopmental diagnostic (other than autism) <0.001
No 49 (75.4%) 44 (71.0%) 65 (94.2%) 55 (94.8%)
Yes 16 (24.6%) 18 (29.0%) 4 (5.8%) 3 (5.2%)
Open in a new tab

Abbreviations: AQ, autism quotient; N‐miss, number of missing data.

P‐TST

Main effects

Analyses revealed a significant effect of the shift condition, with the shift condition leading to lower odds of correct responses (Figure 2a) and increased RT (Figure 2c) compared with the non‐shift condition (OR: exp.(β) = 0.58, 95% CI [0.51, 0.65], p < 0.001; RT: β = 0.55, 95% CI [0.53, 0.56], p < 0.001). The analyses also revealed a significant effect of predictability, with evidence for both linear and quadratic trends. Specifically, as predictability increased, we observed a linear increase in the odds of correct responses, represented in Figure 2b (OR: exp.(β) = 1.17, 95% CI [1.07, 1.29], p = 0.001) and a decrease in RT, represented in Figure 2d (RT: β = −0.06, 95% CI [−0.07, −0.05], p < 0.001). In other words, more predictable stimuli were associated with more accurate and faster responses. Additionally, there was a significant quadratic trend for RT, also observable in Figure 2d (RT: β = 0.01, 95% CI [0, 0.02], p = 0.001), suggesting a curvilinear relationship between predictability and RT. This finding indicates that while RT generally decreased with increasing predictability, the rate of decrease varied across different levels of predictability. Post‐hoc comparisons revealed that the difference in RT between moderately predictable and highly predictable stimuli (RT: β = 0.03, 95% CI [0, 0.05], p = 0.014) was less pronounced compared with the difference between moderately predictable and unpredictable stimuli (RT: β = 0.09, 95% CI [0.06, 0.11], p < 0.001). Importantly, there was no main effect of the group on RT (RT: β = −0.04, 95% CI [−0.14, 0.06], p = 0.412).

FIGURE 2.

FIGURE 2

Open in a new tab

Significant main effects on correct response (CR) and response time (RT). The box plots on the left represent the observed median percentage of CR for figures (a) and (b) and the median observed RT in second for figures (c) and (d), with the interquartile range and individual data points for each condition. The red crosses represent the observed means. The line plots on the right represent the estimated marginal means with their 95% CI. (a) Main effect of shift on CR, (b) Main effect of predictability on CR. (c) Main effect of shift on RT, (d) Main effect of predictability on RT. pred, predictable; mod pred, moderately predictable; unpred, unpredictable; S, shift; NS, non‐shift.

Interaction effects

These main effects were qualified by several significant interaction effects on RT. The interaction between shift and predictability (linear trend), represented on Figure 3a, was significant (β = 0.02, 95% CI [0, 0.04], p = 0.024), showing that predictability had a different impact on RT depending on the shift condition. More specifically, whereas the linear trend was observed in the non‐shift condition indicating that increasing level of predictability decreased RT (unpredictable vs. moderately predictable: β = 0.09, 95% CI [0.06, 0.11], p < 0.001; moderately predictable vs. predictable: β = 0.05, 95% CI [0.03, 0.08], p < 0.001), it was not observed in the shift condition (unpredictable vs. moderately predictable: β = 0.09, 95% CI [0.05, 0.13], p < 0.001; moderately predictable vs. predictable: β = 0, 95% CI [−0.04, 0.05], p = 0.975). In addition, this interaction indicated a larger switch cost in the predictable condition compared with the unpredictable and the moderately predictable conditions (non‐shift vs. shift in the unpredictable condition: β = −0.53, 95% CI [−0.56, −0.51], p < 0.001; moderately predictable condition: β = −0.53, 95% CI [−0.56, −0.5], p < 0.001; predictable condition: β = −0.58, 95% CI [−0.61, −0.55], p < 0.001).

FIGURE 3.

FIGURE 3

Open in a new tab

Significant interactions on response time (on correct responses). The box plots on the left represent the observed median response time (RT) in second with the interquartile range and individual data points for each condition, and the red crosses represent the observed means. The line plots on the right represent the estimated marginal means with their 95% CI. (a) Interaction between shift condition and predictability. (b) Interaction between group and predictability. (c) Interaction between shift condition and sex. (d) Interaction between shift condition, predictability, group and sex. pred, predictable; mod pred, moderately predictable; unpred, unpredictable.

The results also showed an interaction between group and predictability (linear trend) represented in Figure 3b (β = −0.02, 95% CI [−0.04, 0], p = 0.018), indicating that the linear trend on RT differed between the autism and NA groups. Post‐hoc comparisons revealed an increased difference between autistic and NA individuals with the increase in predictability (autism vs. NA in the unpredictable condition: β = 0.02, 95% CI [−0.08, 0.12], p = 0.746; moderately predictable condition: β = 0.04, 95% CI [−0.06, 0.14], p = 0.427; predictable condition: β = 0.07, 95% CI [−0.04, 0.17], p = 0.203). Post‐hoc comparisons also revealed a significant difference between the unpredictable and the moderately predictable condition in NA (β = 0.1, 95% CI [0.07, 0.13], p < 0.001) and in autism (β = 0.08, 95% CI [0.04, 0.11], p < 0.001), whereas the difference between moderately predictable and predictable was significant in NA (β = 0.04, 95% CI [0.01, 0.08], p = 0.012) but not in autism (β = 0.02, 95% CI [−0.02, 0.05], p = 0.5).

In addition, the results revealed an interaction between sex and shift, represented in Figure 3c (β = −0.06, 95% CI [−0.09, −0.03], p = 0.001), indicating a larger difference in RT (with females responding slower than males) in the shift (β = 0.1, 95% CI [−0.02, 0.22], p = 0.111) than in the non‐shift condition (β = 0.05, 95% CI [−0.07, 0.16], p = 0.738). This interaction was also qualified by a four‐way interaction between group, sex, shift, and predictability (quadratic trend) represented in Figure 3d (β = 0.05, 95% CI [0.01, 0.1], p = 0.026). While this interaction seems complex, the figure and post‐hoc test help us to understand it. It showed that there is a quadratic relationship between predictability and RT, appearing only in the shift condition, for all the subgroups including autistic females (unpredictable vs. moderately predictable: β = −0.08, 95% CI [−0.15, −0.01], p = 0.02; moderately predictable vs. predictable: β = 0.02, 95% CI [−0.05, 0.1], p = 0.551), autistic males (unpredictable vs. moderately predictable: β = −0.08, 95% CI [−0.15, −0.02], p = 0.014; moderately predictable vs. predictable: β = 0, 95% CI [−0.07, 0.08], p = 0.945), NA males (unpredictable vs. moderately predictable: β = −0.15, 95% CI [−0.21, −0.08], p < 0.001; moderately predictable vs. predictable: β = 0.02, 95% CI [−0.06, 0.09], p = 0.647), but not NA females (unpredictable vs. moderately predictable: β = −0.05, 95% CI [−0.11, 0.01], p = 0.125; moderately predictable vs. predictable: β = −0.06, 95% CI [−0.13, 0.01], p = 0.111). In addition, this interaction revealed a larger difference between females and males (females being slower) appearing only in NA, in the shift and moderately predictable condition (NA females vs. NA males in the moderately predictable condition: β = 0.19, 95% CI [0.05, 0.33], p = 0.008).

Finally, there was a significant interaction between group and sex on the odds of correct responses represented in Figure 4 (exp(β) = 0.59, 95% CI [0.05, 1.13], p = 0.034) indicating that the odds of correct responses was higher in females than males in the autistic group (exp(β) = 1.39, 95% CI [0.84, 2.31], p = 0.329), and higher in males than females in the NA group (exp(β) = 0.77, 95% CI [0.47, 1.28], p = 0.55).

FIGURE 4.

FIGURE 4

Open in a new tab

Significant interaction on correct responses. The only significant interaction on correct responses (CR) was between group and sex. The box plots on the left represent the observed median percentage of CR with the interquartile range and individual data points for each condition, and the red crosses represent the observed means. The line plots on the right represent the estimated marginal means with their 95% CI.

Effects of covariates

Analyses revealed that having a neurodevelopmental diagnosis other than autism decreased accuracy at the P‐TST (exp(β) = 0.62, 95% CI [−0.86, −0.11], p = 0.011). In addition, having an older age increased RT (β = 0.01, 95% CI [0, 0.02], p = 0.002).

DISCUSSION

The present study aimed to investigate the impact of unpredictability, and its interaction with sex, in flexibility difficulties in autistic individuals. To do so, we designed the P‐TST, a modified version of the TST (Rogers & Monsell, 1995), in which we included 3 levels of predictability. The findings provided evidence that the task captures, at least partly, cognitive flexibility, as it induced the expected switch cost. The findings also showed increased difficulty with increasing unpredictability, particularly in the non‐shift condition. While the findings support the hypothesis of predictive specificities in autism, they were opposite to our hypothesis as they an suggest increased difference with NA with increasing level of predictability. Finally, the findings also support the idea of a specific profile in autistic females.

The large main effect of shift indicated increased accuracy and decreased RT for the non‐shift compared with the shift condition. These slower and more error‐prone responses after task switches show that the task induced the expected switch cost, in line with the results of the original TST (Rogers & Monsell, 1995). Although cognitive flexibility is not a unified concept and is related to other executive functions such as inhibition and working memory, switching performance is often considered as one possible behavioral measure of cognitive flexibility (Hohl & Dolcos, 2024).

The main effect of predictability (linear trend) indicated an increase in accuracy and a decrease in reaction times as predictability increased (for both groups). Thus, the more predictable the task, the better the performance. Since the predictable nature of the task was not explicitly mentioned to the participants, this suggests that they may have detected regularities in the environment to form predictions based on those patterns, aligning with predictive brain theories (Friston, 2009; Rao & Ballard, 1999). This is supported by informal discussions with some participants who reported that they were not aware that certain conditions were more predictable than others, indicating that these predictions may have been formed implicitly or automatically (Tivadar et al., 2021). This aligns with research suggesting that autistic individuals can form predictions based on implicit learning (Foti et al., 2015; Mayo & Eigsti, 2012; Obeid et al., 2016; Pesthy et al., 2023). However, further investigation is needed, as we do not possess proper metacognitive measures. Overall, these main effects are consistent with our hypotheses and tend to indicate that the task measures what it is intended to measure, namely cognitive flexibility and predictions, although further validation would be needed to confirm this.

Additionally, we observed a significant quadratic trend of predictability on RT, showing that the linear trend is modulated by a curvature. Thus, the relationship between RT and predictability is curvilinear: the decrease in RT with increasing predictability is not uniform but slows down as predictability increases. To better understand the precise relationship between predictability and RT, further studies with more levels of predictability are needed.

Notably, there was also an interaction between shift and predictability (linear trend) on RT. This interaction revealed that the linear relationship between predictability and RT differed between conditions and was more pronounced in the non‐shift condition (repeated task) than in the shift condition. It also revealed that the switch cost increased with the task's predictability, consistent with previous research (Bonnin et al., 2011; Duthoo et al., 2012). As observed in Figure 3a, this effect may be explained by the significant improvement in performance in the non‐shift condition (when the task is repeated), whereas the improvement is not as marked in the shift condition probably because, even if expected, the switch still incurs a cost.

Regarding our main hypothesis, we found an interaction effect between the group and the linear trend of predictability, but the effect was opposite to our expectations. We anticipated that autistic individuals would show less divergence from NA individuals in the predictable condition compared with the moderately predictable and unpredictable conditions. Instead, the results revealed an increasing difference between autistic individuals and NA individuals as predictability increased. This finding does not align with our original hypothesis, which was based on our previous study. In that study, we demonstrated lower performance in autism compared with NA individuals during an unpredictable task but not in a predictable one, specifically the TST (Lacroix et al., 2022). However, that study had significant limitations due to several varying factors between the two tasks, including not only predictability but also the nature of the stimuli (social vs. non‐social). Thus, while one of our interpretation of the results was that autistic individuals might struggle with unpredictable situations—consistent with the predictive coding framework that they may attribute excessive weight to prediction errors, leading them to overestimate changes in their environment (Van de Cruys et al., 2014)—the social nature of the unpredictable task might have played a greater role in the findings (Lacroix, Bennetot‐Deveria et al., 2024).

Interestingly, alternative predictive coding frameworks could nevertheless explain the present data. Indeed, the findings are consistent with the hypo‐prior hypothesis (Pellicano & Burr, 2012), which suggests that broader priors in autism lead to less reliance on predictions. According to this theory, autistic individuals may show less difference from NA individuals in unpredictable contexts, where predictions cannot be made and sensory information serves as the sole guide (Brock, 2012; Pellicano & Burr, 2012). This could be because their cognitive style may rely more on bottom‐up processing compared with that of non‐autistic individuals, who typically rely more on top‐down predictive processing. The results may also align with empirical evidence suggesting that perception in autism is influenced by longer‐term statistics, resulting in a slower updating (and forgetting) of priors (Lieder et al., 2019; Sapey‐Triomphe et al., 2023). Although the present task is not purely perceptual, the results may reflect similar mechanisms in cognitive tasks, where autistic individuals may require more time to encode implicit rules than NA individuals, needing a higher level of certainty (Latinus et al., 2019) before they can form and rely on predictions. Overall, while the present findings could align with specificities in predictive coding related to autism, the difference between the results and our hypothesis also emphasizes the need to better delineate the predictive mechanisms that differ between autistic and NA individuals, which may contribute to difficulties in cognitive flexibility. It will also be necessary to better understand the circumstances under which each of the predictive brain theories of autism applies, and why this may vary.

Regarding sex differences, we hypothesized an interaction effect where autistic females would respond faster than autistic males, though slower than NA individuals with NA females responding slower than NA males, replicating Lacroix et al. (2022)'s TST results. However, this interaction was observed not in RT but in accuracy: autistic females had higher odds of correct responses compared with autistic males, while NA females had lower odds of correct responses compared with NA males. In RT, a more complex interaction between group, sex, shift, and predictability was observed, revealing a quadratic relationship between predictability and RT (i.e., the decrease in RT slows down as predictability increases) in all subgroups (autistic males and females, NA males) except NA females. These findings might suggest different predictive processing mechanisms between males and females that are not observed in autism, supporting the idea of a unique profile for autistic females, possibly aligning with the extreme male brain theory of autism (Baron‐Cohen, 2002). They also align with studies showing better cognitive flexibility in autistic females (Demetriou et al., 2021; Lehnhardt et al., 2016) and may contribute to improved daily functioning in this subgroup (Mandic‐Maravic et al., 2015).

Nevertheless, these results should be interpreted cautiously. Notably, the discrepancy between the present study and the previous one (Lacroix et al., 2022) warrants further investigation. The differences may be attributed to the introduction of the predictability variable in the P‐TST; however, this is unlikely since, even in the predictable condition (which is closest to the TST), autistic females were not faster than autistic males. Alternatively, variations between the studies may stem from task differences (e.g., the task switch indicated by colored frames that need to be memorized) or individual differences in cognitive profiles either due to individual variations (e.g., the autistic female sample may prioritize accuracy over speed in this experiment), or because of the lack of clinical and cognitive characterization of the sample recruited in these online studies. Thus, further research is required to explore these sex‐based differences in relation to clinical and cognitive characteristics, and their implications for understanding cognitive processing in autism.

Finally, despite autistic participants responding slower than NA participants, this effect was not significant. Therefore, our hypothesis regarding decreased RT due to processing speed in autism compared with NA (Zapparrata et al., 2023) was not confirmed. This might be explained by the nature of the task. Indeed, the slowing of RT in autism compared with NA depends on the task (Ferraro, 2016) and might be more pronounced for tasks involving simple reaction times than for those involving interference control (Zapparrata et al., 2023). However, this does not explain the difference from Lacroix et al. (2022)'s findings on a similar task, which might be due to high variability in RT in the present task, possibly masking potential differences.

LIMITATIONS

Online studies have strengths, such as enabling the recruitment of a larger sample of autistic participants and allowing them to be tested in their preferred environments, reducing confounding effects like anxiety from social interactions or new places. However, they also present weaknesses. Despite a rigorous recruiting process to ensure that the diagnosis has been delivered by a professional, we lacked the clinical data necessary for more precise analyses and for relating clinical characteristics to the results. While the Autism Quotient (Baron‐Cohen et al., 2001) is sufficient for excluding NA participants with high autistic traits, it is inadequate for clinically characterizing autistic individuals due to its good sensitivity but low negative predictive value and weak correlation with gold standard diagnostic tools (Ashwood et al., 2016). In addition, the questionnaire included questions such as, ‘Have you been diagnosed with another neurodevelopmental disorder, like ADHD, dyslexia, or dyspraxia?’ which resulted in a dichotomous response. Unfortunately, we could not collect more precise data on co‐occurring diagnoses due to compliance with the ethical committee's requirements and the General Data Protection Regulation. This strategy was implemented as a compromise to mitigate the risk of collecting sensitive information online while still gathering essential data on co‐occurring conditions that could influence the results. Indeed, conditions such as depression (Nuño et al., 2021), anxiety (Hartanto & Yang, 2022), and ADHD (Boonstra et al., 2005) may all affect task switching and processing speed. By including these measures as covariates in our analyses, we aimed to account for the variance associated with other diagnoses and ensure that the observed effects were not attributable to these conditions. However, this method of reporting co‐occurring conditions limited our ability to gain a more detailed understanding of which specific comorbidities might impact the findings.

Finally, while the study excluded autistic adults with intellectual disabilities, it did not precisely characterize the cognitive profiles of the participants, and autistic males exhibited a slightly (but significantly) lower level of educational attainment than the other groups. Nonetheless, the average educational attainment among autistic males was the second year of a bachelor's degree, indicating good cognitive abilities.

Overall, the limitations primarily concern the characterization of the population due to the online nature of the study, which suggests that the present results are preliminary. Hence, future studies should include clinical and cognitive measures to better control for and examine the influence of these factors on the results, and to be able to draw more robust conclusions. Additionally, adapted versions of this experiment should be developed for other populations, such as children, teenagers, and autistic individuals with intellectual disabilities. This would enhance our understanding of the developmental trajectory of the interplay between prediction and cognitive flexibility across the entire autism spectrum.

CONCLUSION

While there is a growing body of empirical research investigating prediction in autism (Cannon et al., 2021), none have directly explored its connection to cognitive flexibility, and few have examined sex differences in this domain. The novel design of this study highlights that autistic adults may have unique prediction‐building processes that could impair their cognitive flexibility and contribute to specificities in adaptive behaviors. However, the divergence between our primary hypothesis and the present results suggests that predictive brain theories aimed at explaining the characteristics of autism, while promising, still require more precise empirical research to be better understood in terms of the insights they can provide.

Furthermore, our results reveal a distinct profile of sex differences between autism and NA individuals that could be in line with the notion of better daily life skills in autistic females without intellectual disabilities. Nevertheless, the variations observed across different studies warrant caution in drawing conclusions, while still underscoring the critical need to investigate sex differences in autism to better understand the unique characteristics of females. Although the limitations and inconsistencies should be acknowledged, they also present several opportunities for future research in both NA and autistic populations.

FUNDING INFORMATION

This work was also supported by MIAI@Grenoble Alpes (ANR‐19‐P3IA‐0003) and by the CerCoG@UGA IDEX UGA (ANR‐15‐IDEX‐02).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

ETHICS STATEMENT

All procedures were conducted in accordance with the World Medical Association's Declaration of Helsinki. The study was also approved by the local ethics committee (CER‐Grenoble Alpes, COMUE University Grenoble Alpes, IRB00010290). Participants received pre‐study information, provided online informed consent. OSF preregistration: https://osf.io/dct6w.

Supporting information

Data S1. Supplementary Information.

AUR-18-281-s001.docx (143.5KB, docx)

ACKNOWLEDGMENTS

We thank all participants for their implication in the study.

Lacroix, A. , Torija, E. , Logemann, A. , Baciu, M. , Cserjesi, R. , Dutheil, F. , Gomot, M. , & Mermillod, M. (2025). Cognitive flexibility in autism: How task predictability and sex influence performances. Autism Research, 18(2), 281–294. 10.1002/aur.3281

DATA AVAILABILITY STATEMENT

The data collected for this study and the code used for the analyses are available on the following osf repository: https://osf.io/etkfd/?view_only=808b22fa42a9479aa21a339cc6da0f1b.

REFERENCES

  1. American Psychiatric Association . (2013). Diagnostic and statistical manual of mental disorders (DSM‐5). American Psychiatric Pub. [Google Scholar]
  2. Ashwood, K. , Gillan, N. , Horder, J. , Hayward, H. , Woodhouse, E. , McEwen, F. , Findon, J. , Eklund, H. , Spain, D. , Wilson, C. E. , Cadman, T. , Young, S. , Stoencheva, V. , Murphy, C. , Robertson, D. , Charman, T. , Bolton, P. , Glaser, K. , Asherson, P. , & Murphy, D. (2016). Predicting the diagnosis of autism in adults using the autism‐spectrum quotient (AQ) questionnaire. Psychological Medicine, 46, 1–10. 10.1017/S0033291716001082 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Baron‐Cohen, S. (2002). The extreme male brain theory of autism. Trends in Cognitive Sciences, 6(6), 248–254. 10.1016/S1364-6613(02)01904-6 [DOI] [PubMed] [Google Scholar]
  4. Baron‐Cohen, S. , Wheelwright, S. , Skinner, R. , Martin, J. , & Clubley, E. (2001). The autism‐spectrum quotient (AQ): Evidence from Asperger syndrome/high‐functioning autism, males and females, scientists and mathematicians. Journal of Autism and Developmental Disorders, 31(1), 5–17. 10.1023/A:1005653411471 [DOI] [PubMed] [Google Scholar]
  5. Barr, D. J. , Levy, R. , Scheepers, C. , & Tily, H. J. (2013). Random effects structure for confirmatory hypothesis testing: Keep it maximal. Journal of Memory and Language, 68(3), 255–278. 10.1016/j.jml.2012.11.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bates, D. , Mächler, M. , Bolker, B. , & Walker, S. (2015). Fitting linear mixed‐effects models using Lme4. Journal of Statistical Software, 67(1), 1–48. 10.18637/jss.v067.i01 [DOI] [Google Scholar]
  7. Bertollo, J. R. , Strang, J. F. , Anthony, L. G. , Kenworthy, L. , Wallace, G. L. , & Yerys, B. E. (2020). Adaptive behavior in youth with autism spectrum disorder: The role of flexibility. Journal of Autism and Developmental Disorders, 50(1), 42–50. 10.1007/s10803-019-04220-9 [DOI] [PubMed] [Google Scholar]
  8. Biró, B. , Cserjesi, R. , Kocsel, N. , Galambos, A. , Gecse, K. , Kovács, L. N. , Baksa, D. , Juhasz, G. , & Kokonyei, G. (2022). The neural correlates of context driven changes in the emotional response: An fMRI study. PLoS One, 17, e0279823. 10.1371/journal.pone.0279823 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Biro, B. , Kôkônyei, G. , De Oliveira Negrao, R. , Dancsik, A. , Karsai, S. , Logemann, A. , & Cserkesi, R. (2021). Interaction between emotional context‐guided shifting and cognitive shifting: Introduction of a novel task. Neuropsychopharmacologia Hungarica, 23(3), 319–330. [PubMed] [Google Scholar]
  10. Bolker, B. M. (2015). Linear and generalized linear mixed models. In Fox, G. A. , Negrete‐Yankelevich, S. & Sosa, V. J. (eds), Ecological Statistics: Contemporary theory and application. Oxford University Press. 10.1093/acprof:oso/9780199672547.003.0014 [DOI] [Google Scholar]
  11. Bölte, S. , Duketis, E. , Poustka, F. , & Holtmann, M. (2011). Sex differences in cognitive domains and their clinical correlates in higher‐functioning autism spectrum disorders. Autism, 15(4), 497–511. 10.1177/1362361310391116 [DOI] [PubMed] [Google Scholar]
  12. Bonnin, C. A. , Gaonac'h, D. , & Bouquet, C. A. (2011). Adjustments of task‐set control processes: Effect of task switch frequency on task‐mixing and task‐switching costs. Journal of Cognitive Psychology, 23(8), 985–997. 10.1080/20445911.2011.594435 [DOI] [Google Scholar]
  13. Boonstra, A. M. , Oosterlaan, J. , Sergeant, J. A. , & Buitelaar, J. K. (2005). Executive functioning in adult ADHD: A meta‐analytic review. Psychological Medicine, 35(8), 1097–1108. 10.1017/S003329170500499X [DOI] [PubMed] [Google Scholar]
  14. Brock, J. (2012). Alternative Bayesian accounts of autistic perception: Comment on Pellicano and Burr. Trends in Cognitive Sciences, 16(12), 573–574. 10.1016/j.tics.2012.10.005 [DOI] [PubMed] [Google Scholar]
  15. Cai, R. Y. , Richdale, A. L. , Uljarević, M. , Dissanayake, C. , & Samson, A. C. (2018). Emotion regulation in autism spectrum disorder: Where we are and where we need to go. Autism Research, 11(7), 962–978. 10.1002/aur.1968 [DOI] [PubMed] [Google Scholar]
  16. Cannon, J. , O'Brien, A. M. , Bungert, L. , & Sinha, P. (2021). Prediction in autism spectrum disorder: A systematic review of empirical evidence. Autism Research, 14(4), 604–630. 10.1002/aur.2482 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Cumming, G. (2009). Inference by eye: Reading the overlap of independent confidence intervals. Statistics in Medicine, 28(2), 205–220. 10.1002/sim.3471 [DOI] [PubMed] [Google Scholar]
  18. Demetriou, E. A. , Pepper, K. L. , Park, S. H. , Pellicano, L. , Song, Y. J. C. , Naismith, S. L. , Hickie, I. B. , Thomas, E. E. , & Guastella, A. J. (2021). Autism spectrum disorder: An examination of sex differences in neuropsychological and self‐report measures of executive and non‐executive cognitive function. Autism, 25, 2223–2237. 10.1177/13623613211014991 [DOI] [PubMed] [Google Scholar]
  19. Duthoo, W. , De Baene, W. , Wühr, P. , & Notebaert, W. (2012). When predictions take control: The effect of task predictions on task switching performance. Frontiers in Psychology, 3, 282. 10.3389/fpsyg.2012.00282 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Ferraro, F. R. (2016). No evidence of reaction time slowing in autism spectrum disorder. Autism, 20(1), 116–122. 10.1177/1362361314559986 [DOI] [PubMed] [Google Scholar]
  21. Foti, F. , Crescenzo, F. D. , Vivanti, G. , Menghini, D. , & Vicari, S. (2015). Implicit learning in individuals with autism spectrum disorders: A meta‐analysis. Psychological Medicine, 45(5), 897–910. 10.1017/S0033291714001950 [DOI] [PubMed] [Google Scholar]
  22. Friston, K. (2009). The free‐energy principle: A rough guide to the brain? Trends in Cognitive Sciences, 13(7), 293–301. 10.1016/j.tics.2009.04.005 [DOI] [PubMed] [Google Scholar]
  23. Garofalo, S. , Giovagnoli, S. , Orsoni, M. , Starita, F. , & Benassi, M. (2022). Interaction effect: Are you doing the right thing? PLoS One, 17(7), e0271668. 10.1371/journal.pone.0271668 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Geurts, H. M. , Corbett, B. , & Solomon, M. (2009). The paradox of cognitive flexibility in autism. Trends in Cognitive Sciences, 13(2), 74–82. 10.1016/j.tics.2008.11.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Gomot, M. , & Wicker, B. (2012). A challenging, unpredictable world for people with autism spectrum disorder. International Journal of Psychophysiology, 83(2), 240–247. 10.1016/j.ijpsycho.2011.09.017 [DOI] [PubMed] [Google Scholar]
  26. Haigh, S. M. , Walsh, J. A. , Mazefsky, C. A. , Minshew, N. J. , & Eack, S. M. (2018). Processing speed is impaired in adults with autism spectrum disorder, and relates to social communication abilities. Journal of Autism and Developmental Disorders, 48(8), 2653–2662. 10.1007/s10803-018-3515-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Han, K. , Young Kim, I. , & Kim, J.‐J. (2012). Assessment of cognitive flexibility in real life using virtual reality: A comparison of healthy individuals and schizophrenia patients. Computers in Biology and Medicine, 42(8), 841–847. 10.1016/j.compbiomed.2012.06.007 [DOI] [PubMed] [Google Scholar]
  28. Harrop, C. , Bodfish, J. , Lecavalier, L. , Dallman, A. R. , Jones, D. , Pritchett, J. , Whitten, A. , & Boyd, B. A. (2024). Refining our understanding of anxiety in autistic youth: Examining the role of behavioral inflexibility. Journal of Autism and Developmental Disorders, 54(10), 3620–3629. 10.1007/s10803-023-06079-3 [DOI] [PubMed] [Google Scholar]
  29. Harrop, C. , Jones, D. , Zheng, S. , Nowell, S. , Schultz, R. , & Parish‐Morris, J. (2019). Visual attention to faces in children with autism spectrum disorder: Are there sex differences? Molecular Autism, 10(1), 28. 10.1186/s13229-019-0276-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Harrop, C. , Jones, D. , Zheng, S. , Nowell, S. W. , Boyd, B. A. , & Sasson, N. (2018). Sex differences in social attention in autism spectrum disorder. Autism Research, 11(9), 1264–1275. 10.1002/aur.1997 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Harrop, C. , Jones, D. R. , Sasson, N. J. , Zheng, S. , Nowell, S. W. , & Parish‐Morris, J. (2020). Social and object attention is influenced by biological sex and toy gender‐congruence in children with and without autism. Autism Research, 13(5), 763–776. 10.1002/aur.2245 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Hartanto, A. , & Yang, H. (2022). Testing theoretical assumptions underlying the relation between anxiety, mind wandering, and task‐switching: A diffusion model analysis. Emotion, 22(3), 493–510. 10.1037/emo0000935 [DOI] [PubMed] [Google Scholar]
  33. Hohl, K. , & Dolcos, S. (2024). Measuring cognitive flexibility: A brief review of neuropsychological, self‐report, and neuroscientific approaches. Frontiers in Human Neuroscience, 18. 10.3389/fnhum.2024.1331960 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Ionescu, T. (2012). Exploring the nature of cognitive flexibility. New Ideas in Psychology, 30(2), 190–200. 10.1016/j.newideapsych.2011.11.001 [DOI] [Google Scholar]
  35. Lacroix, A. , Bennetot‐Deveria, Y. , Baciu, M. , Dutheil, F. , Magnon, V. , Gomot, M. , & Mermillod, M. (2024). Understanding cognitive flexibility in emotional evaluation in autistic males and females: The social context matters. Molecular Autism, 15(1), 49. 10.1186/s13229-024-00622-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Lacroix, A. , Dutheil, F. , Logemann, A. , Cserjesi, R. , Peyrin, C. , Biro, B. , Gomot, M. , & Mermillod, M. (2022). Flexibility in autism during unpredictable shifts of socio‐emotional stimuli: Investigation of group and sex differences. Autism, 26(7), 1681–1697. 10.1177/13623613211062776 [DOI] [PubMed] [Google Scholar]
  37. Lacroix, A. , Harquel, S. , Mermillod, M. , Garrido, M. , Barbosa, L. , Vercueil, L. , Aleysson, D. , Dutheil, F. , Kovarski, K. , & Gomot, M. (2024). Sex modulation of faces prediction error in the autistic brain. Communications Biology, 7(1), 1–12. 10.1038/s42003-024-05807-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Latinus, M. , Cléry, H. , Andersson, F. , Bonnet‐Brilhault, F. , Fonlupt, P. , & Gomot, M. (2019). Inflexibility in autism spectrum disorder: Need for certainty and atypical emotion processing share the blame. Brain and Cognition, 136, 103599. 10.1016/j.bandc.2019.103599 [DOI] [PubMed] [Google Scholar]
  39. Lehnhardt, F.‐G. , Falter, C. M. , Gawronski, A. , Pfeiffer, K. , Tepest, R. , Franklin, J. , & Vogeley, K. (2016). Sex‐related cognitive profile in autism spectrum disorders diagnosed late in life: Implications for the female autistic phenotype. Journal of Autism and Developmental Disorders, 46(1), 139–154. 10.1007/s10803-015-2558-7 [DOI] [PubMed] [Google Scholar]
  40. Lenth, R. V. (2021). Emmeans: Estimated Marginal Means, aka Least‐Squares Means.
  41. Leung, R. C. , & Zakzanis, K. K. (2014). Brief report: Cognitive flexibility in autism spectrum disorders: A quantitative review. Journal of Autism and Developmental Disorders, 44(10), 2628–2645. 10.1007/s10803-014-2136-4 [DOI] [PubMed] [Google Scholar]
  42. Lieder, I. , Adam, V. , Frenkel, O. , Jaffe‐Dax, S. , Sahani, M. , & Ahissar, M. (2019). Perceptual bias reveals slow‐updating in autism and fast‐forgetting in dyslexia. Nature Neuroscience, 22(2), 256–264. 10.1038/s41593-018-0308-9 [DOI] [PubMed] [Google Scholar]
  43. Mandic‐Maravic, V. , Pejovic‐Milovancevic, M. , Mitkovic‐Voncina, M. , Kostic, M. , Aleksic‐Hil, O. , Radosavljev‐Kircanski, J. , Mincic, T. , & Lecic‐Tosevski, D. (2015). Sex differences in autism Spectrum disorders: Does sex moderate the pathway from clinical symptoms to adaptive behavior? Scientific Reports, 5(1), 10418. 10.1038/srep10418 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Mayo, J. , & Eigsti, I.‐M. (2012). Brief report: A comparison of statistical learning in school‐aged children with high functioning autism and typically developing peers. Journal of Autism and Developmental Disorders, 42(11), 2476–2485. 10.1007/s10803-012-1493-0 [DOI] [PubMed] [Google Scholar]
  45. Miyake, A. , Emerson, M. J. , & Friedman, N. P. (2000). Assessment of executive functions in clinical settings: Problems and recommendations. Seminars in Speech and Language, 21(2), 169–183. 10.1055/s-2000-7563 [DOI] [PubMed] [Google Scholar]
  46. Mostert‐Kerckhoffs, M. A. L. , Staal, W. G. , Houben, R. H. , & de Jonge, M. V. (2015). Stop and change: Inhibition and flexibility skills are related to repetitive behavior in children and young adults with autism spectrum disorders. Journal of Autism and Developmental Disorders, 45(10), 3148–3158. 10.1007/s10803-015-2473-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Nuño, L. , Gómez‐Benito, J. , Carmona, V. R. , & Pino, O. (2021). A systematic review of executive function and information processing speed in major depression disorder. Brain Sciences, 11(2), 147. 10.3390/brainsci11020147 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Obeid, R. , Brooks, P. J. , Powers, K. L. , Gillespie‐Lynch, K. , & Lum, J. A. G. (2016). Statistical learning in specific language impairment and autism spectrum disorder: A meta‐analysis. Frontiers in Psychology, 7. 10.3389/fpsyg.2016.01245 [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Pellicano, E. , & Burr, D. (2012). When the world becomes “too real”: A Bayesian explanation of autistic perception. Trends in Cognitive Sciences, 16(10), 504–510. 10.1016/j.tics.2012.08.009 [DOI] [PubMed] [Google Scholar]
  50. Pesthy, O. , Farkas, K. , Sapey‐Triomphe, L.‐A. , Guttengéber, A. , Komoróczy, E. , Janacsek, K. , Réthelyi, J. M. , & Németh, D. (2023). Intact predictive processing in autistic adults: Evidence from statistical learning. Scientific Reports, 13(1), 11873. 10.1038/s41598-023-38708-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. R Core Team . (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing. [Google Scholar]
  52. Rao, R. P. N. , & Ballard, D. H. (1999). Predictive coding in the visual cortex: A functional interpretation of some extra‐classical receptive‐field effects. Nature Neuroscience, 2(1), 79–87. 10.1038/4580 [DOI] [PubMed] [Google Scholar]
  53. Rights, J. D. , & Sterba, S. K. (2019). Quantifying explained variance in multilevel models: An integrative framework for defining R‐squared measures. Psychological Methods, 24(3), 309–338. 10.1037/met0000184 [DOI] [PubMed] [Google Scholar]
  54. Rogers, R. D. , & Monsell, S. (1995). Costs of a predictible switch between simple cognitive tasks. Journal of Experimental Psychology: General, 124(2), 207–231. 10.1037/0096-3445.124.2.207 [DOI] [Google Scholar]
  55. RStudio Team . (2019). RStudio: Integrated development environment for R. RStudio, Inc. [Google Scholar]
  56. Sapey‐Triomphe, L.‐A. , Sanchez, G. , Hénaff, M.‐A. , Sonié, S. , Schmitz, C. , & Mattout, J. (2023). Disentangling sensory precision and prior expectation of change in autism during tactile discrimination. Npj Science of Learning, 8(1), 1–16. 10.1038/s41539-023-00207-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Scarpa, A. , Swain, D. M. , Factor, R. S. , Dahiya, A. V. , & Bertollo, J. R. (2021). Enhancing flexibility: A biosocial model for resilience to adversity in youth with autism. SAGE Open, 11(3), 21582440211037997. 10.1177/21582440211037997 [DOI] [Google Scholar]
  58. Schielzeth, H. , Dingemanse, N. J. , Nakagawa, S. , Westneat, D. F. , Allegue, H. , Teplitsky, C. , Réale, D. , Dochtermann, N. A. , Garamszegi, L. Z. , & Araya‐Ajoy, Y. G. (2020). Robustness of linear mixed‐effects models to violations of distributional assumptions. Methods in Ecology and Evolution, 11(9), 1141–1152. 10.1111/2041-210X.13434 [DOI] [Google Scholar]
  59. Stoet, G. (2010). PsyToolkit: A software package for programming psychological experiments using Linux. Behavior Research Methods, 42(4), 1096–1104. 10.3758/BRM.42.4.1096 [DOI] [PubMed] [Google Scholar]
  60. Stoet, G. (2016). PsyToolkit: A novel web‐based method for running online questionnaires and reaction‐time experiments. Teaching of Psychology, 44, 24–31. 10.1177/0098628316677643 [DOI] [Google Scholar]
  61. Teunisse, J.‐P. , Roelofs, R. L. , Verhoeven, E. W. M. , Cuppen, L. , Mol, J. , & Berger, H. J. C. (2012). Flexibility in children with autism spectrum disorders (ASD): Inconsistency between neuropsychological tests and parent‐based rating scales. Journal of Clinical and Experimental Neuropsychology, 34(7), 714–723. 10.1080/13803395.2012.670209 [DOI] [PubMed] [Google Scholar]
  62. Tivadar, R. I. , Knight, R. T. , & Tzovara, A. (2021). Automatic sensory predictions: A review of predictive mechanisms in the brain and their link to conscious processing. Frontiers in Human Neuroscience, 15, 702520. 10.3389/fnhum.2021.702520 [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Van de Cruys, S. , Evers, K. , Van der Hallen, R. , Van Eylen, L. , Boets, B. , de Wit, L. , & Wagemans, J. (2014). Precise minds in uncertain worlds: Predictive coding in autism. Psychological Review, 121(4), 649–675. 10.1037/a0037665 [DOI] [PubMed] [Google Scholar]
  64. Van Eylen, L. , Boets, B. , Steyaert, J. , Evers, K. , Wagemans, J. , & Noens, I. (2011). Cognitive flexibility in autism spectrum disorder: Explaining the inconsistencies? Research in Autism Spectrum Disorders, 5(4), 1390–1401. 10.1016/j.rasd.2011.01.025 [DOI] [Google Scholar]
  65. Zapparrata, N. M. , Brooks, P. J. , & Ober, T. M. (2023). Slower processing speed in autism spectrum disorder: A meta‐analytic investigation of time‐based tasks. Journal of Autism and Developmental Disorders, 53(12), 4618–4640. 10.1007/s10803-022-05736-3 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data S1. Supplementary Information.

AUR-18-281-s001.docx (143.5KB, docx)

Data Availability Statement

The data collected for this study and the code used for the analyses are available on the following osf repository: https://osf.io/etkfd/?view_only=808b22fa42a9479aa21a339cc6da0f1b.

Articles from Autism Research are provided here courtesy of Wiley

RESOURCES