Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2026 Apr 17;117(4):1449–1468. doi: 10.1111/bjop.70074

Discrimination in autism as measured by minimal group and sheer difference experiments

Zachary Tan 1, Xizhuo Zou 1, Daniel P Skorich 2, Eliane Deschrijver 1,✉
PMCID: PMC13655983  PMID: 41996209

Abstract

Autistic individuals often show fewer social biases than neurotypical people. Whether they show fewer discriminatory tendencies is however unclear. The present study examined discriminatory tendencies in autistic versus neurotypical individuals in the minimal group paradigm and the novel ‘sheer difference’ paradigm. Seventy‐five autistic and neurotypical participants were recruited for each group, totalling 150 participants. In the ‘sheer difference’ paradigm, participants received a coin toss outcome in each trial, after which they were tasked to assign points to a single other participant who had received the same versus a different coin flip outcome. In the minimal group paradigm, participants were assigned to a group based on coin flips, and then, they assigned points to members of their own group versus the other group. The ‘sheer difference’ paradigm contributes to the study's aims by testing in autism whether discrimination can also follow from individual rather than group‐based difference versus sameness. We found that discriminatory tendencies can come about on both individual and group levels. We did not find clear differences between autistic and neurotypical populations, with implications for the way in which we conceptualize discrimination and understand autism.

Keywords: autism, discrimination, minimal group paradigm, sheer difference

BACKGROUND

Autistic individuals1 have often been described as less susceptible than neurotypical people to social biases, such as racial or gender stereotypes. For example, they might be less likely to associate playing with dolls to girls, despite an understanding of these stereotypes (Birmingham et al., 2015; Hirschfeld et al., 2007). Consistent with this, research seems to indicate that autism is associated with lesser, albeit still existing, social biases. For instance, two studies examining implicit social biases and the use of stereotypes in autistic individuals found that significant effects for autistic individuals were present in the Implicit Association Test (IAT) (Birmingham et al., 2015; Kirchner et al., 2012). Both studies, however, also found that relative to neurotypical controls, the measured effects and biases were significantly reduced. Relatedly, a propensity to conform to a group consensus has been shown to be lower in autistic compared to non‐autistic children, while autistic traits were negatively associated with conformity in a group of neurotypical children (Yafai et al., 2014). A recent systematic review highlighted studies which theorized that autistic individuals were more resistant to conforming to social norms, presumably because of being less susceptible to societal pressure (Wattel et al., 2024). While there are clear theoretical and empirical links between autism and reduced social biases, whether similar patterns of difference for autistic and neurotypical individuals arise for discriminatory tendencies remains to be seen.

Measuring psychological drivers of discrimination: The ‘minimal group’ paradigm

The term ‘discrimination’ has typically been used, both in and outside of the scientific literature, to describe unequal treatment of people on the basis of a particular group or category—such as those based on age, gender, race, ethnicity and so on. When using the term ‘discrimination’ in the context of social psychology, it often refers to favouritism of one's own group over another group, without necessarily the societal negative connotations of poor treatment towards a lower status group (Al Ramiah et al., 2010). A series of experiments exploring this phenomenon, known as ‘minimal group’ studies, implemented group division on the basis of trivial and ad hoc differences—such as painting preferences, dot estimation tasks, or a coin toss (Tajfel & Turner, 1979, 2004). They measured discrimination via the Tajfel matrices, in which participants would allocate points to ingroup and outgroup members, which would later be translated into monetary rewards (Tajfel, 1970; Tajfel & Turner, 1979). The matrices allowed for the computation of ‘pull scores’, which measured the relative strength of the allocation strategies participants used when assigning points (Bourhis et al., 1994). It was reported that the mere categorization of individuals into separate groups was sufficient to financially favour the ingroup (‘us’) versus the outgroup (‘them’). This occurred even in the absence of actual or imagined conflicts of interest, pre‐existing attitudes, or past history of hostility, between individuals in each group (Tajfel, 1970).

Early research suggested that discriminatory tendencies did not emerge without explicit group division (Billig & Tajfel, 1973), leading to the view that the main drivers of discriminatory tendencies are intergroup processes involving social identity and categorization (Tajfel, 1970, 1974; Tajfel & Turner, 1979, 2004). Tajfel and Turner indeed attempted to explain these findings by positing two closely related theories: social identity theory (SIT) and self‐categorization theory (SCT) (Tajfel & Turner, 2004; Trepte & Loy, 2017; Turner et al., 1989). Essentially, these two theories argue that self‐categorization as a group member—that is, the process of coming to conceive of oneself as psychologically interchangeable with other ingroup members—is the underlying mechanism that gives rise to group behaviour in general, including intergroup discrimination in favour of the ingroup. More specifically, this discrimination in the minimal group paradigm can take the form of maximum differentiation, in which the difference in money allocated to the ingroup and outgroup was maximized rather than allocating the most money to the ingroup. In this way, perceivers appear to be motivated to clearly differentiate between the groups in a way that results in a positive social identity.

The integrated self‐categorization model of autism (ISCA)

Recently, Skorich and Haslam (2022) have put forward a theoretical model of autism, the Integrated Self‐Categorization Model of Autism (ISCA). This model argues that the enhanced perceptual functioning known to characterize autism manifests as a reduced tendency for group‐level self‐categorization, which in turn results in reduced shared attention, and theory of mind differences. The ISCA model also then makes the novel prediction that all those phenomena emerging from self‐categorization should also manifest differently in autism, including, as relevant to the current study, ingroup favouritism in the minimal group paradigm. Putting the idea that autism could be related to discriminatory tendencies to the test for the first time, Bertschy et al. (2020) measured the degree of autistic traits of a general population sample and related these to their social identification with a minimal group based on painting preference and their pattern of points allocation to ingroup versus outgroup members. Ingroup favouritism within the paradigm was found to be positively correlated to the extent to which people identified with their ingroup (i.e., in Likert‐scale evaluations of statements like ‘I feel a bond with the Klee Group’; Leach et al., 2008, as cited by Bertschy et al., 2020, p. 3301). In turn, individuals who reported higher levels of autistic traits showed lower levels of identification with their ingroup due to self‐categorization (Bertschy et al., 2020). Furthermore, a relationship between the autistic traits in this non‐clinical population and ingroup favouritism was found, and this was also found to be mediated by social identification. In a similar vein, a study by Wu et al. (2024) invited autistic and neurotypical participants to partake in a variation of the classical minimal group paradigm. The procedure for this study involved a typical dots estimation task from the minimal group paradigm; but rather than assigning money after group allocation, participants rated the genuineness of ingroup or outgroup members' smiles as a proxy measure for ingroup favouritism. It was found that neurotypical and autistic individuals equally rated ingroup smiles as ‘more genuine’, suggesting in this case that autism was not associated with ingroup favouritism. Overall, however, neither of these studies can necessarily be taken as a direct empirical test for whether individuals with a diagnosis of autism show different or similar levels of discriminatory resource division strategies relative to neurotypical populations: Wu et al. (2024) did not include resource division measures, and Bertschy et al. (2020) relied on a measure of autistic traits in a neurotypical population, rather than diagnosis.

The ‘sheer difference’ paradigm

A recent series of studies (Deschrijver & Ramsey, 2025; Zhuang et al., 2025) has suggested that discriminatory tendencies might even occur on the basis of individual difference versus sameness—if people are not allocated to groups. Participants in these studies assigned money to a single other participant (this participant was the same across multiple trials) who showed sameness versus difference on the same outcomes seen in prior minimal group studies (painting preferences, dot estimations, coin tosses). The results in these so‐called ‘sheer difference’ experiments suggested that unequal resource division occurred on the basis of ad hoc difference versus sameness when the main manipulation of explicit group division and social identity assignment was removed from the paradigm ‐ even when based on a random outcome such as a coin toss (Deschrijver & Ramsey, 2025; Zhuang et al., 2025). The authors indeed observed that participants would use discriminatory strategies such as maximum differentiation towards a single other individual for demonstrating difference versus sameness in painting preferences, dots estimations, or coin flips. This paradigm may provide a new avenue of investigation into how discrimination may be affected in autistic individuals, especially in a dyadic context, given that discriminatory behaviour was observed outside of explicit group division in neurotypicals.

Deschrijver and Ramsey (2025) further speculated that a conflict signal in the brain of neurotypical individuals could lie at the basis of a differential resource division response towards an individual that shows such a difference versus sameness. Moreover, it has been argued that individuals on the autism spectrum would show a lesser sensitivity to such a conflict signal (Deschrijver, 2021; Deschrijver & Palmer, 2020). If autistic individuals were less sensitive to monitoring mental conflict, they could process the social differences in both the standard and individual versions of the minimal group paradigm to a lesser extent—thereby reducing the extent to which they discriminate. The next logical step would be to investigate if discrimination is elicited in autistic individuals specifically from individual‐level difference versus sameness, without explicit group allocation. Whether they would discriminate less compared to neurotypicals in this manner remains similarly unexplored.

The current study

The present study aims to investigate if clinically diagnosed autistic individuals show less discriminatory behaviour in the minimal group paradigm compared to neurotypicals (Bertschy et al., 2020; Skorich & Haslam, 2022; Tajfel & Turner, 2004). This aim also extends to re‐examining if discrimination can follow from individual difference versus sameness in neurotypicals—as demonstrated by Deschrijver and Ramsey (2025) and Zhuang et al. (2025)—and how this discrimination via ‘sheer difference’ may present itself in autistic individuals as well. The theoretical and empirical work discussed above among discrimination, other social biases and autism provides sufficient direction to make the prediction that autistic individuals may discriminate, but to a lesser extent than neurotypicals (Bertschy et al., 2020; Birmingham et al., 2015; Deschrijver & Ramsey, 2025; Wattel et al., 2024). Based on these studies, we predict the same outcome in each version of the paradigm, but with some differences across diagnosis groups.

Preregistered hypotheses

We predict that neurotypical individuals will show discrimination in both the minimal group and individual paradigms.

Autistic individuals are predicted to show discrimination in both the minimal group and individual paradigm.

Autistic individuals are expected to show less discrimination compared to neurotypical individuals in the minimal group paradigm.

Autistic individuals will show less discrimination in the individual paradigm, compared to neurotypical individuals.

Preregistered exploratory analyses

Spearman and Pearson correlational analyses will be conducted to examine the relation between participants' Autism Questionnaire (AQ) scores and their computed pull scores. This study includes several questions concerning group identification at the end of the experiment; participants' scores on these questions will be used in correlational analyses with AQ scores and pull scores separately.

MATERIALS AND METHODS

The present study adapted methods from Deschrijver and Ramsey (2025); the experiments were programmed in JsPsych and ran with the program JATOS on the server Mindprobe. Before the start of participant recruitment and data collection, the hypotheses, research questions, planned analyses, sample size and exclusion criteria were all preregistered on the Open Science Framework (https://osf.io/enrcs/overview). The questionnaires used, raw data, stimuli and analysis code for the experiment were preregistered and are available at the same link. The study received approval by the ethical committee of the School of Psychology at the University of Sydney (2023/HE000684).

Participants

Participants were recruited via the platform Prolific (www.prolific.co). Eligibility criteria for participant recruitment were as follows. Country of residence as the United Kingdom; Prolific acceptance rate of at least 95%; and having indicated ‘yes’ in the prescreening to being comfortable with experiments involving deception. We sampled equally across genders for both the autism and the neurotypical group. Participants in the neurotypical control group must also have reported in the Prolific prescreening survey to not having been diagnosed with autism, not be in the process of receiving a diagnosis of ASD, not identify as being on the autism spectrum and to have no other mental health issues. Participants in the autism group must have reported to have received a diagnosis of autism as a child or adult. Participants who did not use Google Chrome as the web browser to participate were precluded from the experiment. In addition, participants who failed the two attention checks were automatically excluded and replaced.

An a‐priori power analysis was conducted using the software G*Power to determine the appropriate sample size. Assuming an effect size of 0.15 and 95% power for a within‐between ANOVA with 3 measures and 2 groups, we required a minimum of 116 participants. Considering available resources, 150 participants were recruited—with an equal number of autistic and neurotypical individuals. The effect size was chosen based on the considerations outlined by Lakens (2022), including the smallest effect size of interest and using estimates from previous studies.

We preregistered that participants in the neurotypical group who indicated in our study to have received a diagnosis of ASD would be excluded from the group analyses and replaced. Vice versa, participants in the autism group who indicated in our study to have not received a diagnosis would also be excluded and replaced. Furthermore, participants in the neurotypical group who scored above the 32 point cut‐off on the AQ were excluded from the data analysis; this is in line with prior autism research as indicative of someone who displays a clinical level of autistic traits and thus might confound our analysis if their scores were to be included in the control group (Atherton et al., 2022; Baron‐Cohen et al., 2001; Berent et al., 2022). However, individuals in the autism group who scored below this cut‐off were not excluded from the analyses if they indicated to have received a formal diagnosis. This rationale is in line with other studies including autistic participants which used the AQ in a similar manner to exclude participants (Atherton et al., 2022; Galvin & Richards, 2023). Seven participants in the neurotypical group scored above the AQ cut‐off. Six participants from the autism group and one from the neurotypical group were replaced due to mismatching diagnoses. Three autistic and one neurotypical participant encountered technical difficulties which prevented them from completing the study. In total, 17 participants were excluded and replaced (the above count is 18, but one participant was counted twice; they both failed to meet diagnostic criteria and encountered technical difficulties), arriving at the final planned sample of 150 responses.

Measures and design

Autism spectrum quotient (AQ)

The Autism Spectrum Quotient (AQ) is a 50‐item self‐report questionnaire commonly used to assess autistic traits in adolescents or adults over the age of 16 (Baron‐Cohen et al., 2001). The AQ was used in the current study to verify that autistic traits were indeed higher in the autistic population, compared to the non‐autistic population, in line with our exclusion criteria listed above. These scores were also used in preregistered exploratory analyses.

Procedure

The following procedure has been carried out previously by Deschrijver and Ramsey (2025). Participants were briefed on the responsible researchers, tasks to be completed in the study and their payment for participating in the experiment; participants were made aware of their right to withdraw from the study at any time, and what it would mean to consent to taking part in the study. Following the briefing, participants were presented with an Informed Consent form. Demographic data were then taken; this involved a question asking whether participants had received a diagnosis of autism and included two attention checks. After providing participants with the relevant information and collecting their demographic data, the experiment began.

The study was split into two blocks, participants underwent an ‘individual’ block followed by a ‘group’ block—this was done to avoid priming them with the notion of groups from the start, which previous research has suggested could have an impact (Billig & Tajfel, 1973). In the ‘individual’ block, participants were informed that they would be engaging in a visual judgement task where they would repeatedly receive the outcome of a heads‐or‐tails coin toss, visualized by a picture of an Australian 50‐cent coin—they were told that these outcomes were random and equally likely. Participants were told that precisely one other participant had already completed the trial and received coin tosses as well—this was, however, a deception. Participants were then tasked with using the matrices to assign points to the other ‘participant’, which (they were told) would be translated to monetary reward after data collection. They were also deceived into thinking that this other participant had ostensibly done the same for them, to replicate as close as possible original experiments (Billig & Tajfel, 1973). After the outcome of each coin toss, a matrix was presented to capture discriminatory tendencies towards the individual that demonstrated difference versus sameness (see further below). Before the beginning of the actual test trials, participants engaged in several practice trials in order to familiarize themselves with answering on the matrices. A visual representation of this procedure can be seen in Figure 1.

FIGURE 1.

FIGURE 1

Open in a new tab

‘Individual’ discrimination block procedure; coin tosses followed by point allocation. This figure demonstrates the procedure of flipping a coin followed by individual point allocation.

Once participants finished answering on the ‘individual’ block trials, they moved on to the ‘group’ block; the procedure and design were the exact same as the ‘individual’ block with a few key differences. Participants were once again told that they would repeatedly receive the outcome of a coin toss, visualized this time by a New Zealand 50‐cent coin. In this ‘group’ block, participants were instead told that many other participants had already completed the trial. Rather than allocating points after each coin toss, participants were assigned to a New Zealand head or New Zealand tail group, depending on which one they received most after 13 tosses—after which they would assign points to other ‘members’ of each group. Then, matrices were presented to capture discriminatory tendencies towards the outgroup versus the ingroup (see further). A visual representation of this procedure can be seen in Figure 2. The rationale behind the use of Australian and New Zealand coins is that they are neutral stimuli which UK participants likely have not seen before. These coins might bear some visual resemblance to UK coins, but as the autism and neurotypical groups are both sampled from the UK, we expect any potential resemblance to play out similarly across groups.

FIGURE 2.

FIGURE 2

Open in a new tab

‘Group’ discrimination block procedure. This figure demonstrates the procedure of flipping 13 coins followed by group assignment and point allocation.

After the test trials for both blocks concluded, participants answered three questions about group identification before being tasked with filling out the AQ. The three group identification questions presented were: ‘How firmly do you stand behind your group?’, ‘How strongly do you identify with your group?’ and ‘How convinced are you about your group membership?’. Participants rated these statements on a 7‐point scale, with higher ratings indicating stronger group identification. Once these questionnaires were completed, participants were asked open‐ended questions about technical issues, their strategies assigning points, if they noticed anything in particular, and if they were suspicious about anything. Participants were then debriefed on the aim of the experiment, and the deception was revealed.

Tajfel's matrices and pull scores

Discriminatory tendencies were measured in the current study using Tajfel's matrices (Bourhis et al., 1994). The original 6 types of matrices as described by Bourhis et al. (1994) were presented once in the original format, and once in a flipped version, resulting in 12 total trials. The matrices in the ‘group block’ asked: ‘How many points will you offer in this trial to…’, followed by ‘…member X of the New Zealand head group?’ and ‘…member Y of the New Zealand tail group?’ (X/Y indicating a random number). The following was presented below each matrix: ‘You are member 63 of the New Zealand head/tail group. Think carefully.’ The matrices depicted 13 columns with 2 numbers in each. Participants indicated their choice of a matrix column using the letter buttons A to M. These matrices allowed us to quantify the relative strength by which participants allocated points based on group‐level differences, in the form of pull scores (Bourhis et al., 1994). The two main discriminatory strategies measured by the matrices are Maximum Ingroup Profit (MIP)—awarding the highest possible points to an ingroup member regardless of outgroup rewards; and Maximum Differentiation (MD)—maximizing the difference in point allocation in favour of the ingroup even if MIP suffers (Bourhis et al., 1994; Tajfel & Turner, 1979). Ingroup favouritism (FAV) is a term used to denote choices which use both MIP and MD strategies; therefore, discriminatory behaviour can be understood as either MIP + MD, or as ingroup favouritism (Bourhis et al., 1994). Prosocial strategies include Maximum joint profit (MJP) and Parity (P). The former signifies the maximizing the total amount of points assigned across in‐ and outgroups. The latter is a prosocial strategy involving equal distribution of points to in‐ and outgroup members.

Deschrijver and Ramsey (2025) adapted the group‐based matrices to assess discriminatory tendencies towards an individual. Here, the question was asked: ‘How many points will you offer in this trial if the person had an…’, followed by ‘…Australian tail as a coin toss?’ on one row of the matrix and ‘…Australian head as a coin toss?’ on the other. The following sentence was presented below the matrix: ‘The coin toss for you was: Australian head/tail. Think carefully.’ These matrices allow us to quantify the relative strength by which participants allocated points based on individual‐level difference versus sameness, in the form of pull scores (Bourhis et al., 1994). While we use the same original abbreviations for discriminatory strategies, their meaning is different within the ‘individual block’: MIP signifies awarding the highest possible points to the other when they demonstrate sameness regardless of rewards for the other when they demonstrate difference. MD suggests maximizing the difference in point allocation in favour of the other when they demonstrate sameness even if MIP suffers (Bourhis et al., 1994; Tajfel & Turner, 1979). The prosocial strategy MJP here signifies maximizing the total amount of points assigned to the other for showing sameness or difference. P involves an equal distribution of points to the other for showing sameness versus difference.

The computation of pull scores was carried out as described by Bourhis et al. (1994). The three main discriminatory pull scores to be measured were:

  1. Favouritism (FAV; when the pulls of MD and MIP are simultaneously assessed) versus MJP,

  2. MD versus MIP + MJP and

  3. FAV (MIP + MD) versus P.

Ingroup Favouritism (in the ‘group block’) or favouritism of the other that demonstrates sameness (in the ‘individual block’) is represented by positive discriminatory pull scores. Outgroup Favouritism (in the ‘group block’) or favouritism of the other that demonstrates difference (in the ‘individual block’), in contrast, would be represented by negative values (Bourhis et al., 1994). In line with our a‐priori power analysis and proposed hypotheses, we conducted a 3 (Matrix Type: FAV vs. MJP, MD vs. MIP + MJP, FAV vs. P) x 2 (Diagnosis: autism, neurotypical) Mixed ANOVA with Matrix Type as the within‐subjects factor and Diagnosis as the between‐subjects factor.

RESULTS

The data were cleaned and collated in Microsoft Excel (v16.57) via a Python script; all statistical analyses were conducted using IBM SPSS Statistics software (Version 29.0.2.0). Figures 3, 4, 5 were plotted and generated using RStudio (Version 2024.04.2 + 764).

FIGURE 3.

FIGURE 3

Open in a new tab

Violin plots of the 3 pull scores in the ‘individual’ block for each diagnosis group. Orange = Autism Diagnosis group, Grey = Neurotypical Diagnosis group. Higher scores represent more discrimination; a score of 0 indicates no discriminatory behaviour. The mean pull scores are all above 0, indicating discriminatory behaviour was present in both groups in the ‘individual’ block. Error bars denote the standard error of the mean.

FIGURE 4.

FIGURE 4

Open in a new tab

Violin plots of the 3 pull scores in the ‘group’ block for each diagnosis group. Orange = Autism Diagnosis group, Grey = Neurotypical Diagnosis group. Higher scores represent more discrimination; a score of 0 indicates no discriminatory behaviour. The mean pull scores are all above 0, indicating discriminatory behaviour was present in both groups in the ‘group’ block. Error bars denote the standard error of the mean.

FIGURE 5.

FIGURE 5

Open in a new tab

Scatterplot of AQ scores and MD versus MIP + MJP pull scores in the ‘group’ discrimination block. A dashed regression line is noted in red. Higher Autism Quotient Scores indicate greater levels of autistic traits. Larger pull scores indicate greater levels of discriminatory tendencies. The plot illustrates the correlational finding that people in the autism group with higher levels of autistic traits demonstrate lesser discriminatory tendencies in the Group Block for the MD vs. MIP+MJP matrix.

Demographics

Participant demographics for each diagnosis group are shown in Table 1 below. Independent sample proportion tests and independent sample t‐tests were conducted to examine if participants displayed any significant difference between groups across demographic variables. The results of these tests can be seen in the ‘Test Statistic’ and ‘p‐value’ columns in Table 1. The difference in mean AQ scores between autistic (M = 34.67, SD = 8.65) and neurotypical (M = 18.52, SD = 6.84) groups was significant, t(148) = 12.68, p < .001. A significant difference in age was also found, with autistic people on average being younger (M = 35.09, SD = 11.78) than the neurotypical (M = 42.53, SD = 14.15) group.

TABLE 1.

Participant demographic data.

Sample characteristic Autism Neurotypical Test statistic p‐value
Age −3.5 <.001
Mean 35.09 42.53
Standard deviation 11.78 14.15
Minimum 18 21
Maximum 72 73
Gender 1.00 .316
Female 34 45
Male 36 30
My gender does not fit these categories 4 0
Prefer not to say 1 0
Handedness −1.34 .181
Right 60 66
Left 11 6
Ambidextrous 4 3
Education −1.37 .172
Bachelor's degree 28 32
Doctorate 4 3
Elementary school 1 1
High school 12 12
Master's degree 9 10
Post‐secondary diploma 7 5
College or further education (no degree) 13 8
Trade or vocational school 1 4
AQ 12.68 <.001
Mean 34.67 18.52
Standard DEVIATION 8.65 6.84
Open in a new tab

Pull score analysis

As highlighted in the Methods section above, the matrices measure the relative strength, or pull, of three main discriminatory strategies (FAV vs. MJP, MD vs. MIP + MJP, FAV vs. P) through the allocation of points (Bourhis et al., 1994). In order to measure discrimination in this study and analyse if these pull scores are significantly different from zero, Wilcoxon signed‐rank tests were conducted for each score individually (Bourhis et al., 1994, p. 231). This results in 12 Wilcoxon tests in total, three scores from each of the two test blocks (‘individual’ and ‘group’), for each diagnosis group (autism and neurotypical). These tests follow from our preregistered analyses; the p‐values of these tests are presented in Table 2. All Wilcoxon signed‐rank tests conducted were significant (p < .05), except for the test analysing the significance of the pull of MD over MIP + MJP in autistic participants in the ‘individual’ block, which was marginally significant. A visual representation of these data can be seen in Figures 3 and 4. No statistical correction for multiple comparisons was preregistered; in line with reviewers' comments, FDR‐controlling procedures (specifically the Benjamini & Hochberg, 1995 procedure) was used to control for this post‐hoc, among the Wilcoxon signed‐rank tests. Upon correcting for multiple comparisons, all previously significant p‐values remained statistically significant at the adjusted FDR level. Only the pull of MD over MIP + MJP among autistic participants in the ‘individual block’, was found to be not significant before and after controlling the FDR. In sum, the results of the Wilcoxon tests remain unchanged.

TABLE 2.

Wilcoxon signed‐rank test significance (p) values.

Autism Neurotypical
Individual
FAV vs. MJP <.001 <.001
MD vs. MIP + MJP .065 .032
FAV vs. P <.001 <.001
Group
FAV vs. MJP <.001 <.001
MD vs. MIP + MJP .003 .003
FAV vs. P <.001 <.001
Open in a new tab

ANOVA analyses

A 3 (Matrix Type: FAV vs. MJP, MD vs. MIP + MJP, FAV vs. P) × 2 (Diagnosis: autism, neurotypical) Mixed ANOVA was conducted for the ‘individual’ and ‘group’ blocks separately—in line with our planned analyses. Because of the significant difference in age between the autistic and the neurotypical diagnosis groups (see Demographics), the ANOVAs were first conducted with age as a covariate, even while these were not preregistered analyses. These analyses revealed that age did not show a main effect and did not interact with any factors. The main effect of Matrix Type was no longer significant when age was included, likely due either to reduced statistical power or because age‐related variance was partialled out. However, this effect only reveals whether discrimination differs between different Matrix types, not whether discrimination itself has occurred (which was confirmed via the Wilcoxon tests). We therefore mostly discuss the analyses without covariates, as the interpretation of Matrix Type effects is not the focus of our hypotheses. For transparency, the statistical outputs of these analyses are provided in Table S1.

In the ‘individual’ block, Mauchly's Test indicated that the data violated the assumption of sphericity, χ2 (2) = 55.61, p < .001. Mauchly's Test in the ‘group’ block revealed a similar violation of sphericity, χ2 (2) = 90.15, p < .001. Therefore, degrees of freedom were corrected using the Huynh‐Feldt estimates of sphericity in the ‘individual’ block—as the Greenhouse–Geisser epsilon was greater than .75 (ε = .76)—and the Greenhouse–Geisser method in the ‘group’ block (ε = .69). The following ANOVA test statistics reported were estimated using these methods.

A significant effect of matrix type was observed in the ‘individual’ block, F (1.54, 228.49) = 30.22, p < .001, ηp2 = .17. Averaged across both diagnosis groups, participants showed higher levels of discrimination on FAV vs. MJP (M = 3.27, SD = 4.95) and FAV vs. P (M = 3.03, SD = 5.26) pulls, compared to the pull of MD vs. MIP + MJP (M = 0.60, SD = 2.31). Although these within‐subjects effects were significant, they are not as relevant to our proposed hypotheses as the between‐subjects or interaction effects. Interpretation of these results simply inform us that the strength of the types of discriminatory strategies used is not equal across all Matrix Types. There was no significant effect of Diagnosis on pull scores observed in the ‘individual’ block, F (1, 148) = 0.97, p = .327, ηp2 = .007. No significant interaction effect was observed; indicating the differences in pull scores across matrix types did not differ significantly between autistic and neurotypical groups, F (1.54, 228.49) = 0.24, p = .73, ηp2 = .002. This indicates that we did not find evidence for autistic and neurotypical participants to discriminate in significantly different manners.

Analysis of within‐subject effect of Matrix Type in the ‘group block’ was significant, F (1.37, 202.96) = 49.43, p < .001, ηp2 = .25. Participants again showed higher levels of discrimination on FAV vs. MJP (M = 4.10, SD = 5.59) and FAV vs. P (M = 4.46, SD = 5.44) pulls, compared to the pull of MD vs. MIP + MJP (M = 1.03, SD = 2.81), across both diagnosis groups. No significant effect for the between‐subjects variable of Diagnosis was observed in the ‘group’ block, F (1, 148) = .05, p = .819, ηp2 < .001. The interaction effect was also not significant in the ‘group’ block, F (1.37, 202.96) = .149, p = .779, ηp2 = 001. Again, we did not find evidence for autistic and neurotypical participants to discriminate in significantly different manners.

Because we sampled equally across genders for both the autistic and the neurotypical group, it was possible to assess whether any different results would emerge when accounting for gender. We therefore conducted another two ANOVAs (one for the ‘individual block’ and one for the ‘group block’) with gender as an additional factor (see Table S2). None of the findings varied with gender, even while the inclusion of gender diminished the effect of Matrix Type in the individual block (in the same way as did age). We therefore further discuss the results without accounting for gender.

Correlational analyses

Reliability analyses were run on the AQ and Group Identification questionnaires. The AQ consisted of 50 items, and Cronbach's Alpha showed high internal consistency overall (α = .93). Group Identification consisted of three items, and the overall measure showed high internal consistency as well (Cronbach's Alpha α = .88). Averaged across the three questions, the mean Group Identification score for the neurotypical group was 4.81 (SD = 1.52), and 4.25 for the autistic group (SD = 1.83). Although we did not preregister the following analysis, we did test whether the Group Identification score was significantly smaller in the autism as compared to the neurotypical group. This was indeed the case (t(148) = −2.04, p < .05, Cohen's d = .33).

Following our planned exploratory analyses, correlations were computed to assess the relationships between autistic participants' AQ scores and their measured pull scores. Tests of normality were run on the AQ and Group identification scores to determine which type of correlation to report; Kolmogorov–Smirnov tests were significant (p < .05) for both variables, indicating that the data for both these variables were not normally distributed. Therefore, Spearman correlations are reported in the tables below. Correlational analyses were also conducted between autistic participants' scores on the Group Identification questions and their pull scores, and between their AQ scores and Group Identification scores. The results of these analyses are presented in Tables 3 and 4.

TABLE 3.

Spearman correlations between the pull scores from ‘individual’ and ‘group’ blocks and the scores of relevant questionnaires for the autism diagnosis group.

AQ Group ID
Individual
FAV vs. MJP .097 .057
MD vs. MIP + MJP .084 .097
FAV vs. P −.081 .127
Group
FAV vs. MJP .152 .468**
MD vs. MIP + MJP −.233* .239*
FAV vs. P .076 .466**
Open in a new tab

Note: Abbreviations for questionnaire from left to right: Autism Quotient, Group Identification. The values used for the Group Identification correlations are averaged scores across the three questions.

*

p < .05.

**

p < .01.

TABLE 4.

Spearman correlations between AQ scores and Group Identification questions, in the autism diagnosis group.

Group ID Q1 Group ID Q2 Group ID Q3 Group ID average
AQ −.053 −.140 −.060 −.093
Open in a new tab

There was a significant positive correlation between overall Group Identification scores and the pull scores observed in the ‘group’ discrimination block; indicating that the more participants identified with their group, the more discrimination in terms of point allocation was observed. Scores on the AQ also showed a significant inverse correlation with pull scores as measured by the MD versus MIP + MJP matrix in the ‘group’ discrimination block (see Figure 5). No other significant correlations were found between pull scores and AQ or Group Identification questionnaires. No significant correlations were found between autistic participants' AQ scores and their scores on the Group Identification questions.

Unlike the Wilcoxon tests, no statistical correction for multiple comparisons was conducted: As each test was interpreted independently, the Type I error rate would be defined for each Wilcoxon test individually, rather than across a set of tests (Rubin, 2021). Nonetheless, we exercise caution when describing these correlations in the discussion.

DISCUSSION

The present study intended to investigate discriminatory biases of autistic versus neurotypical participants, as measured in the minimal group and sheer difference paradigms (Deschrijver & Ramsey, 2025; Tajfel, 1970). Based on previous literature, we predicted that the autistic group would demonstrate discriminatory tendencies that would be present, albeit reduced, relative to neurotypical participants. Discriminatory tendencies in individuals with a diagnosis of autism versus neurotypical participants had, to our knowledge, not been assessed prior to this study.

Main findings

The significance of the Wilcoxon signed‐rank tests evidences that discriminatory tendencies were present in both autistic and neurotypical individuals across both the ‘individual’ and the ‘group blocks’. All pull scores were significantly different from zero, except for the pull of MD versus MIP + MJP within the ‘individual block’ in the autistic sample—yet, this pull score was still marginally significant with a p‐value of .065. The pull score of MD versus MIP + MJP being of a lower magnitude than the pull scores of the other matrices is a common finding (see Figures 3 and 4; see also Deschrijver & Ramsey, 2025; Zhuang et al., 2025). While the other two pull scores in the ‘individual block’ for the autistic group were highly significant (p‐value <.001), we conclude that we did find evidence for intact discriminatory tendencies in the autistic group overall. All pull scores for the neurotypical group were significant across ‘individual’ and ‘group blocks’, presenting evidence for discriminatory tendencies in the group of neurotypical individuals as expected.

Hypothesis H1, which predicted discriminatory tendencies in neurotypical individuals, was intended to be a confirmatory hypothesis based on research in minimal group studies, as well as on recent work on discriminatory tendencies after individual difference versus sameness (Deschrijver & Ramsey, 2025; Zhuang et al., 2025). The finding that neurotypical individuals show discrimination at the group level is unsurprising given prior literature (Bertschy et al., 2020; Tajfel, 1970; Tajfel & Turner, 1979). Evidence for unequal resource division by neurotypical individuals at the individual level likewise provides support for Deschrijver and Ramsey's (2025) ‘sheer difference’ account—which suggests that discrimination can be driven by an interindividual comparison that treats sameness more positively than difference.

As we observed significant pull scores also in the autistic group across both versions of the task, we provide support for our second hypothesis H2: discriminatory tendencies are present in autistic individuals. This adds to findings that social biases can be observed in autistic individuals (Birmingham et al., 2015; Kirchner et al., 2012; Yafai et al., 2014). Overall, our findings seem to provide support in favour of the idea that both neurotypical and autistic individuals display discriminatory tendencies in minimal group and sheer difference contexts.

Speculations from other studies in the literature which point to reduced discriminatory biases for autistic individuals in group contexts (H3) and in dyadic contexts (H4) are not strongly supported by our current findings (Birmingham et al., 2015; Yafai et al., 2014). The results of the two ANOVAs showed no significant difference between autistic and neurotypical participants in terms of discriminatory tendencies, as measured by the pull scores, at both individual and group levels. As such, the lack of effects observed does not support our third and fourth hypotheses. Some other research findings support the account that social biases in autism are intact but reduced; however, we did not detect a difference between autistic and neurotypical individuals in both of our paradigms. Our results are therefore effectively null. There are no statistical techniques that can prove a null hypothesis (Harms & Lakens, 2018). Therefore, we interpret the lack of evidence for reduced discrimination in autism with caution.

It should be noted that the effect of Matrix Type was weakened when controlling separately for age and gender (as can be seen in Tables S1 and S2). However, as mentioned in the Results section above, a significant effect of Matrix Type does not provide evidence for or against discrimination per se—rather, it suggests only that participants' responses differed significantly as a function of the type of matrix. Evidence of discrimination as such can instead be found in the Wilcoxon test results. In contrast to the effect of Matrix Type, the inclusion of gender and age did not change the (non‐significant) main effect of Diagnosis group. In light of this, and given that these age and gender‐related findings are not directly relevant to our main hypotheses, we do not discuss them in further detail here.

Implications of the lack of evidence for reduced discrimination in autism

The lack of support for our third hypothesis H3, that there would be lower ingroup favouritism in the autistic than the neurotypical group in the minimal group paradigm, is also by extension a lack of support for the same prediction within the ISCA model (Bertschy et al., 2020; Skorich & Haslam, 2022). Specifically, Skorich and Haslam (2022) proposed that one difference in autistic individuals versus neurotypicals is an overall tendency to self‐categorize more at a local level—that is, at the individual level in a typical minimal group paradigm—which could manifest as a decrease in ingroup favouritism, because the nominal ingroup would be less salient for autistic individuals on average, all else being equal. The lack of difference observed in this study cannot support this prediction. Nevertheless, non‐preregistered exploratory analyses of the Group Identification score did show that the autistic individuals identified less strongly with their assigned group than neurotypical individuals. The present study also observed a significant inverse correlation between autistic participants' AQ scores and their pull scores as measured by the MD versus MIP + MJP matrix in the ‘group block’, even while no effect of Diagnosis was observed for pull scores in this block overall. A similar correlation between AQ scores and pull scores was found by Bertschy et al. (2020), using a matrix with discriminatory strategies grouped together instead (MD + MIP versus MJP). Despite the lack of support for the ISCA from our main effects, the similar correlational findings between the two studies can be argued to be broadly supportive of some elements of the model—as both correlations suggest a relation between autistic traits and a measure of group salience. Nevertheless, their findings cannot be taken as strong evidence for the relation between autism and discriminatory behaviour, in part because the correlations were conducted based on autistic traits in neurotypical participants rather than those with formal autism diagnoses (Bertschy et al., 2020). The findings also dovetail with an earlier study that used a minimal group manipulation in autistic individuals: Wu et al. (2024) found that autistic individuals were equally susceptible to ingroup bias while rating others' smiles, even while some more subtle rating differences with the neurotypical group were also observed.

We also found a lack of strong support for our hypothesis H4, which predicted lower discriminatory tendencies in the ‘individual block’ for the autistic than for the neurotypical group. Our results provide a measure of support for Deschrijver and Ramsey's (2025) ‘sheer difference’ framework of discrimination, in the sense that it replicates in neurotypical groups and extends to a clinical participant group the initial findings that the use of discriminatory tendencies can rise even if people are not allocated to groups. Deschrijver and Ramsey (2025) speculated that difference versus sameness based on social information may come with a conflict signal in the brain, and Deschrijver and Palmer (2020) had previously argued that autistic individuals would be less sensitive to this signal. This motivated, in part, the rationale for the current study. The proposed explanatory mechanism within the neurotypical brain in terms of a neural conflict signal after detecting social difference is however still to be substantiated (Deschrijver & Ramsey, 2025). Since the current task did not directly test the involvement of such a neural conflict signal, the results cannot speak directly to any neural differences, or lack of neural differences, between neurotypes. Overall, the findings indicate the need for further research in the proposed neural mechanisms underlying performance in the task, both in neurotypical and autistic populations.

Discrimination based on ‘sheer difference’

The finding that neurotypical individuals display discriminatory tendencies in our study is not surprising, given established research on the minimal group paradigm (Tajfel, 1970; Tajfel & Turner, 1979, 2004). However, the fact that discriminatory tendencies were demonstrated outside explicit group allocation—by both autistic and neurotypical individuals—is novel. It could be noted that SIT did emphasize group‐based social categorization as a crucial condition to observe effects such as maximal differentiation (Billig & Tajfel, 1973), as explained by Deschrijver and Ramsey (2025). Taken together, the present findings can add to a new dimension to the field of discrimination research—suggesting that individual‐level difference versus sameness outside of explicit group categorization may facilitate discriminatory behaviour (Deschrijver & Ramsey, 2025; Zhuang et al., 2025). As differences on an individual level can likely contribute to effects observed in the minimal group paradigm as well (Deschrijver & Ramsey, 2025), the theoretical framework of ‘sheer difference’ can help partially explain results from both group and individual settings. If one facilitatory mechanism of discrimination is potentially the treatment of sameness as more positively than difference (in social information), then the results of the ‘group’ block could in part be explained in a similar fashion. In addition, the ‘sheer difference’ paradigm poses an intriguing and novel method of conceptualizing discrimination, which had not yet been examined in autism prior to this study. It suggests in particular that current interventions against discrimination, which are often based on group‐processes (e.g., Pettigrew & Tropp, 2006), may benefit from considering the fact that discriminatory tendencies can also be driven by individual‐level differences. The finding that autistic individuals may also discriminate based on individual‐level differences adds new knowledge to our current conceptions of discrimination—improving our understanding of how it may or may not differ in autism.

Limitations and constraints on generality (Simons et al., 2017)

Critiques of the minimal group paradigm, on which our findings are based, can be taken as limitations to our study. Some criticisms argue that the Tajfel matrices do not measure allocation strategies independently, while others question if the arbitrary allocations actually call upon fundamental intergroup processes (Otten, 2016). Although studies have replicated the minimal group findings of intergroup bias on the basis of arbitrary group allocation, a portion of research argues that the measurement of various related outcomes and how they impact effects on the minimal group paradigm is lacking (Otten, 2016; Schmidt & Drake, 2023). From a theoretical and methodological standpoint, Schiffmann and Wicklund (1992) argue that because of its design, findings in the minimal group paradigm can only be framed and understood within the paradigm itself. They debate that arbitrary group identification cannot be equated to social identity and that it is not robust as an account of real‐world systemic social phenomena (Schiffmann & Wicklund, 1992). Considering these criticisms, it appears that the current findings can only be understood when based in the minimal group paradigm. However, these arguments align with our current suggestion, pointing towards an understanding of discrimination outside of ideas of group categorization as situated on the minimal group paradigm. Future research should investigate these findings further, testing for discriminatory tendencies in autism using measures aside from those used in minimal group studies.

Our online data collection assured a broad participant reach and collection from a wider community sample. However, the online nature of the study also warrants some consideration of potential threats to data integrity, namely the challenges in confirming autism diagnoses, and the reduced experimental control compared to in‐person studies. Participants attempting to pose as autistic individuals in order to receive monetary compensation is an issue that could affect the rigour of our study (Pellicano et al., 2024). In order to mitigate this, additional steps were taken to confirm participants' diagnoses; these steps are highlighted in detail in the ‘Participants’ and ‘Exclusion Criteria’ sections above. Furthermore, a study by Peer et al. (2022) has shown that Prolific provides the highest quality data overall; participants are in general more attentive, and behave more honestly. While this does not fully mitigate the issues of experimental control, it does lend some credibility to our study through the use of Prolific. Nonetheless, it is still possible that some participants could have avoided our attempts to exclude mismatching or false diagnoses, and it is important to consider this when discussing the implications of our study. Lastly, the generalizability of our results is minorly constrained; all participants in our sample were recruited from the United Kingdom, therefore our conclusions can only be generalized to this population. Evidence of discrimination via sheer difference has yet to be examined outside of WEIRD populations as well (Deschrijver & Ramsey, 2025; Zhuang et al., 2025).

CONCLUSION

This study aimed to investigate if autistic individuals would discriminate on the minimal group paradigm, and at an individual level outside of group allocation. It subsequently examined if this discrimination by autistic individuals was significantly reduced as compared to neurotypical individuals—following trends in other social biases examined in autism research. This study found direct empirical evidence that autistic individuals display discriminatory behaviour in both minimal group and ‘sheer difference’ paradigms; demonstrating that this level of discrimination is not significantly different from that of neurotypical individuals. Overall, the study provides further evidence for the fact that discriminatory tendencies can also arise from individual‐level differences versus sameness, and both in a neurotypical and a clinical autism group. This contributes to recent ideas of the ‘sheer difference’ framework in that discriminatory tendencies may be also grounded also in individual comparisons. Considering the criticisms of the minimal group paradigm as well, further research on whether autistic individuals may discriminate outside of this paradigm may prove fruitful. This would lead to a deeper understanding of how discrimination may function in autism, holding implications for real‐world settings and society more broadly.

AUTHOR CONTRIBUTIONS

Zachary Tan: Conceptualization; methodology; investigation; formal analysis; data curation; writing – original draft; writing – review and editing; visualization. Xizhuo Zou: Formal analysis; data curation; writing – review and editing; visualization. Daniel P. Skorich: Writing – review and editing; supervision. Eliane Deschrijver: Conceptualization; methodology; investigation; data curation; writing – review and editing; funding acquisition; supervision.

Supporting information

Table S1. Table of the Mixed ANCOVA for the Individual and Group Blocks (Age as Covariate).

Table S2. Table of the Mixed ANOVA for the Individual and Group Blocks (Gender as Between‐Subjects Factor).

BJOP-117-1449-s001.docx (18.6KB, docx)

ACKNOWLEDGEMENTS

This research was supported by a Discovery Early Career Researcher Award (DECRA: DE220100087) by the Australian Research Council, awarded to ED. Open access publishing facilitated by The University of Sydney, as part of the Wiley ‐ The University of Sydney agreement via the Council of Australasian University Librarians.

Footnotes

1

In contemporary research, there is some debate as to the appropriate way to refer to individuals with autism. The present study will respect the preference of many autistic individuals by using identity‐first language. For more information on this, see Kenny et al. (2016), Bury et al. (2023), Keating et al. (2023), and Taboas et al. (2023).

DATA AVAILABILITY STATEMENT

Our preregistered hypotheses and data are openly available on OSF: https://osf.io/enrcs. The authors declare that they have no conflict of interest.

REFERENCES

  1. Al Ramiah, A. , Hewstone, M. , Dovidio, J. F. , & Penner, L. (2010). The social psychology of discrimination: Theory, measurement and consequences. In Making equality count: Irish and international research measuring equality and discrimination (pp. 84–112). Dublin: The Liffey Press. [Google Scholar]
  2. Atherton, G. , Edisbury, E. , Piovesan, A. , & Cross, L. (2022). Autism through the ages: A mixed methods approach to understanding how age and age of diagnosis affect quality of life. Journal of Autism and Developmental Disorders, 52(8), 3639–3654. 10.1007/s10803-021-05235-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. 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]
  4. Benjamini, Y. , & Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series B: Statistical Methodology, 57(1), 289–300. 10.1111/j.2517-6161.1995.tb02031.x [DOI] [Google Scholar]
  5. Berent, I. , Theodore, R. M. , & Valencia, E. (2022). Autism attenuates the perception of the mind‐body divide. Proceedings of the National Academy of Sciences, 119(49), e2211628119. 10.1073/pnas.2211628119 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bertschy, K. , Skorich, D. P. , & Haslam, S. A. (2020). Self‐categorization and autism: Exploring the relationship between autistic traits and Ingroup Favouritism in the minimal group paradigm. Journal of Autism and Developmental Disorders, 50(9), 3296–3311. 10.1007/s10803-019-04149-z [DOI] [PubMed] [Google Scholar]
  7. Billig, M. , & Tajfel, H. (1973). Social categorization and similarity in intergroup behaviour. European Journal of Social Psychology, 3(1), 27–52. 10.1002/ejsp.2420030103 [DOI] [Google Scholar]
  8. Birmingham, E. , Stanley, D. , Nair, R. , & Adolphs, R. (2015). Implicit social biases in people with autism. Psychological Science, 26(11), 1693–1705. 10.1177/0956797615595607 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Bourhis, R. , Sachdev, I. , & Gagnon, A. (1994). Conducting intergroup research with the Tajfel matrices: Some methodological notes. In Zanna M. & Olson J. (Eds.), The psychology of prejudice: The Ontario symposium (Vol. 7, pp. 209–232). Erlbaum Publishers. [Google Scholar]
  10. Bury, S. M. , Jellett, R. , Haschek, A. , Wenzel, M. , Hedley, D. , & Spoor, J. R. (2023). Understanding language preference: Autism knowledge, experience of stigma and autism identity. Autism, 27(6), 1588–1600. 10.1177/13623613221142383 [DOI] [PubMed] [Google Scholar]
  11. Deschrijver, E. (2021). Relational mentalizing after any representation. Behavioral and Brain Sciences, 44, e148. 10.1017/S0140525X20001417 [DOI] [PubMed] [Google Scholar]
  12. Deschrijver, E. , & Palmer, C. (2020). Reframing social cognition: Relational versus representational mentalizing. Psychological Bulletin, 146(11), 941–969. 10.1037/bul0000302 [DOI] [PubMed] [Google Scholar]
  13. Deschrijver, E. , & Ramsey, R. (2025). Unequal resource division occurs in the absence of group division and identity. Proceedings of the National Academy of Sciences of the United States of America, 122(7), e2413797122. 10.1073/pnas.2413797122 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Galvin, J. , & Richards, G. (2023). Health anxiety in autistic adults. Research in Autism Spectrum Disorders, 104, 102146. 10.1016/j.rasd.2023.102146 [DOI] [Google Scholar]
  15. Harms, C. , & Lakens, D. (2018). Making “null effects” informative: Statistical techniques and inferential frameworks. Journal of Clinical and Translational Research, 3(Suppl 2), 382–393. [PMC free article] [PubMed] [Google Scholar]
  16. Hirschfeld, L. , Bartmess, E. , White, S. , & Frith, U. (2007). Can autistic children predict behavior by social stereotypes? Current Biology, 17(12), R451–R452. 10.1016/j.cub.2007.04.051 [DOI] [PubMed] [Google Scholar]
  17. Keating, C. T. , Hickman, L. , Leung, J. , Monk, R. , Montgomery, A. , Heath, H. , & Sowden, S. (2023). Autism‐related language preferences of English‐speaking individuals across the globe: A mixed methods investigation. Autism Research, 16(2), 406–428. 10.1002/aur.2864 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Kenny, L. , Hattersley, C. , Molins, B. , Buckley, C. , Povey, C. , & Pellicano, E. (2016). Which terms should be used to describe autism? Perspectives from the UK autism community. Autism, 20(4), 442–462. 10.1177/1362361315588200 [DOI] [PubMed] [Google Scholar]
  19. Kirchner, J. C. , Schmitz, F. , & Dziobek, I. (2012). Brief report: Stereotypes in autism revisited. Journal of Autism and Developmental Disorders, 42(10), 2246–2251. 10.1007/s10803-012-1460-9 [DOI] [PubMed] [Google Scholar]
  20. Lakens, D. (2022). Sample size justification. Collabra: Psychology, 8(1), 33267. 10.1525/collabra.33267 [DOI] [Google Scholar]
  21. Otten, S. (2016). The minimal group paradigm and its maximal impact in research on social categorization. Current Opinion in Psychology, 11, 85–89. 10.1016/j.copsyc.2016.06.010 [DOI] [Google Scholar]
  22. Peer, E. , Rothschild, D. , Gordon, A. , Evernden, Z. , & Damer, E. (2022). Data quality of platforms and panels for online behavioral research. Behavior Research Methods, 54(4), 1643–1662. 10.3758/s13428-021-01694-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Pellicano, E. , Adams, D. , Crane, L. , Hollingue, C. , Allen, C. , Almendinger, K. , Botha, M. , Haar, T. , Kapp, S. K. , & Wheeley, E. (2024). Letter to the editor: A possible threat to data integrity for online qualitative autism research. Autism: The International Journal of Research and Practice, 28(3), 786–792. 10.1177/13623613231174543 [DOI] [PubMed] [Google Scholar]
  24. Pettigrew, T. F. , & Tropp, L. R. (2006). A meta‐analytic test of intergroup contact theory. Journal of Personality and Social Psychology, 90(5), 751–783. 10.1037/0022-3514.90.5.751 [DOI] [PubMed] [Google Scholar]
  25. Rubin, M. (2021). When to adjust alpha during multiple testing: A consideration of disjunction, conjunction, and individual testing. Synthese, 199(3–4), 10969–11000. 10.1007/s11229-021-03276-4 [DOI] [Google Scholar]
  26. Schiffmann, R. , & Wicklund, R. A. (1992). The minimal group paradigm and its minimal psychology: On equating social identity with arbitrary group membership. Theory & Psychology, 2(1), 29–50. 10.1177/0959354392021002 [DOI] [Google Scholar]
  27. Schmidt, K. , & Drake, R. G. (2023). Minimal group procedures and outcomes. Collabra: Psychology, 9(1), 90187. 10.1525/collabra.90187 [DOI] [Google Scholar]
  28. Simons, D. J. , Shoda, Y. , & Lindsay, D. S. (2017). Constraints on Generality (COG): A Proposed Addition to All Empirical Papers. Perspectives on Psychological Science, 12(6), 1123–1128. 10.1177/1745691617708630 [DOI] [PubMed] [Google Scholar]
  29. Skorich, D. P. , & Haslam, S. A. (2022). The integrated self‐categorization model of autism. Psychological Review, 129(6), 1373–1393. 10.1037/rev0000385 [DOI] [PubMed] [Google Scholar]
  30. Taboas, A. , Doepke, K. , & Zimmerman, C. (2023). Preferences for identity‐first versus person‐first language in a US sample of autism stakeholders. Autism, 27(2), 565–570. 10.1177/13623613221130845 [DOI] [PubMed] [Google Scholar]
  31. Tajfel, H. (1970). Experiments in intergroup discrimination. Scientific American, 223(5), 96–103. [PubMed] [Google Scholar]
  32. Tajfel, H. (1974). Social identity and intergroup behaviour. Social Science Information/Sur les Sciences Sociales, 13(2), 65–93. 10.1177/053901847401300204 [DOI] [Google Scholar]
  33. Tajfel, H. , & Turner, J. (1979). An Integrative Theory of Intergroup Conflict. 94–109. https://www.semanticscholar.org/paper/An‐integrative‐theory‐of‐intergroup‐conflict.‐Tajfel‐Turner/3f6e573a6c1128a0b0495ff6ebb9ca6d82b4929d
  34. Tajfel, H. , & Turner, J. C. (2004). The social identity theory of intergroup behavior. In Jost J. T. & Sidanius J. (Eds.), Political psychology (pp. 276–293). Psychology Press. 10.4324/9780203505984-16 [DOI] [Google Scholar]
  35. Trepte, S. , & Loy, L. S. (2017). Social identity theory and self‐categorization theory. In Rössler P., Hoffner C. A., & Zoonen L. (Eds.), The international encyclopedia of media effects (1st ed., pp. 1–13). Wiley. 10.1002/9781118783764.wbieme0088 [DOI] [Google Scholar]
  36. Turner, J. C. , Hogg, M. A. , Oakes, P. J. , Reicher, S. D. , & Wetherell, M. (1989). Rediscovering the social group: A self‐categorization theory. Contemporary Sociology, 18, 645. [Google Scholar]
  37. Wattel, L. L. , Walsh, R. J. , & Krabbendam, L. (2024). Theories on the link between autism Spectrum conditions and trans gender modality: A systematic review. Review Journal of Autism and Developmental Disorders, 11(2), 275–295. 10.1007/s40489-022-00338-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Wu, R. , Hamilton, A. F. D. C. , & White, S. J. (2024). Can group membership modulate the social abilities of autistic people? An intergroup bias in smile perception. Cortex, 173, 150–160. 10.1016/j.cortex.2023.12.018 [DOI] [PubMed] [Google Scholar]
  39. Yafai, A.‐F. , Verrier, D. , & Reidy, L. (2014). Social conformity and autism spectrum disorder: A child‐friendly take on a classic study. Autism: The International Journal of Research and Practice, 18(8), 1007–1013. 10.1177/1362361313508023 [DOI] [PubMed] [Google Scholar]
  40. Zhuang, M. , Deschrijver, E. , Ramsey, R. , & Turel, O. (2025). Comparing discriminatory behavior against AI and humans. Scientific Reports, 15, 10894. 10.1038/s41598-025-94631-9 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

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

Supplementary Materials

Table S1. Table of the Mixed ANCOVA for the Individual and Group Blocks (Age as Covariate).

Table S2. Table of the Mixed ANOVA for the Individual and Group Blocks (Gender as Between‐Subjects Factor).

BJOP-117-1449-s001.docx (18.6KB, docx)

Data Availability Statement

Our preregistered hypotheses and data are openly available on OSF: https://osf.io/enrcs. The authors declare that they have no conflict of interest.

Articles from British Journal of Psychology are provided here courtesy of Wiley

RESOURCES