Whether a suspect appears to be telling the truth or lying during an investigative interview has far-reaching implications; from the perceived reliability of their statement by police to how they are viewed by jurors in court (Denault, 2020; Haworth, 2020). However, accuracy for detecting deception is broadly at chance level, often based upon faint or unreliable nonverbal and paraverbal cues such as eye contact and vocal pitch (Bond and DePaulo, 2006; Sporer and Schwandt, 2006). Verbal deception cues have proved a more reliable and promising direction for investigative interviewing research. Richness of detail is one such verbal indicator of veracity (Nahari, 2016) as liars use self-regulation strategies such as keeping a story simple and avoiding verifiable details that could reveal deceit (Hartwig et al., 2014; Nahari, 2018). Further, in the criminal justice system (CJS), evidence typically links a suspect to an alleged offence, meaning a suspect’s account can be compared against available evidence (Police and Criminal Evidence Act: PACE. Home Office, 1984/2008). Thus, lying suspects may display verbal deception cues such as contradicting this evidence (‘statement-evidence inconsistencies’) (Hartwig et al., 2005; Vredeveldt et al., 2014), viewed as indicative of deception by investigative officers in the field (Deeb et al., 2018).

However, the impact of such verbal deception cues within investigative interviews is, to date, based upon neurotypical population samples. Thus, whether such deception cues are applicable to neurodivergent adults is yet to be examined. Concerningly, individuals on the autism spectrum (henceforth autism) appear to be overrepresented in the CJS (Justice Inspectorates, 2021). Autism is a lifelong neurodevelopmental condition characterized by persistent difficulties with social communication and interaction, as well as restricted and repetitive behaviors, interests and activities (American Psychiatric Association, 2013; World Health Organization, 2018). Relatedly, social communication and memory differences in autism present substantial challenges for providing best evidence’ during investigative interviews (Maras, 2021). Autistic mock-witnesses often provide less detailed free-recall accounts (Maras and Bowler, 2011, 2012; Henry et al., 2020) due in part to autism-common difficulties with episodic memory retrieval, exacerbated by insufficiently specific, structured questioning (Maras, 2021). More broadly, autistic individuals may produce less coherent and causally connected narrative versions of events with fewer key contextualizing details (Barnes and Baron-Cohen, 2012; Baixauli et al., 2016; but see Henry et al., 2020). Indeed, autistic adults more often fail to recognize and report extricating details that would help demonstrate their innocence of mock-criminal offences (Young and Brewer, 2020). Thus, even when being truthful, autistic suspects may display verbal cues associated with deception in neurotypical populations, such as statement-evidence inconsistencies (Vredeveldt et al., 2014), a lack of verifiable extricating information (Nahari, 2018) and sparsely detailed accounts (Vrij et al., 2014). Verbal responses containing insufficient information may be interpreted as evasive or deceptive, leading to repeated questioning and challenges by investigators (Gudjonsson et al., 2007) which, in turn may lead to further breakdown in communication (O’Mahony et al., 2012) and even false confessions (Gudjonsson, 2021). These issues are likely to be further exacerbated by the stress of a suspect interview experience, as autistic people may experience investigative interviews as highly socially and cognitively demanding (Herrington and Roberts, 2012; Maras et al., 2020).

Understanding the verbal behavior of autistic suspects who are actively deceptive is also crucial for effective investigative interviewing practice. While there is a substantial body of research showing that autistic children have difficulty with lying, relatively little is known about deception in autistic adulthood (Bagnall et al., 2022). Over the past few years, there have, however, been high-profile criminal cases in the United Kingdom (UK) in which autistic defendants have deceived others (e.g., Murray, 2020; De Simone, 2021). Like non-autistic individuals, some autistic adults without co-occurring intellectual disability tell verbal lies for self-protective purposes (Davidson and Henderson, 2010; Jaarsma et al., 2012) and can successfully deceive in computerized paradigms (van Tiel et al., 2021). Autistic adults also report an inclination to lie in everyday situations comparable with non-autistic adults, though such deception may require greater cognitive effort than neurotypical peers (Bagnall et al., under review). Deception, during even mock-suspect interviews, can be highly cognitively demanding (Caso et al., 2005). Common (though not universal) autism difficulties in taking others’ perspectives—or Theory of Mind (ToM: Baron-Cohen, 1997; but see Milton, 2012)—and social decision-making (Woodcock et al., 2020; Brosnan and Ashwin, 2022) may suggest that autistic adults’ verbal deception cues are more pronounced than those of non-autistic adults. Indeed, while many autistic children can and do tell spontaneous verbal lies, they tend to have greater difficulty than non-autistic children maintaining these lies during subsequent follow-up statements (Li et al., 2011). Identifying how verbal deception cues are displayed by (both truthful and lying) autistic mock-suspects is crucial for the development of best practice investigative interviewing.

In summary, socio-cognitive and sensory processing differences in autism raise numerous concerns relating to the investigative interviewing of autistic suspects. Ensuring that police suspect interviews are conducted fairly and ethically requires understanding if (and how) autistic peoples’ accounts are affected, and if this depends on whether they are being truthful or deceptive. Recognizing how those on the autism spectrum experience investigative interviews is also crucial for identifying relevant areas of support (e.g., supportive interviewing practices or adjustments to custody – see Holloway et al., 2020; Maras et al., 2020). We address these issues in the present study. We predicted that, during a mock-suspect interview, autistic adults’ deceptive accounts would present more pronounced verbal cues to deception (i.e., greater inconsistencies, sparser accounts) than those of non-autistic adults. We also expected that autistic adults’ truthful accounts would more frequently display verbal deception cues (i.e., greater inconsistencies, sparser accounts and fewer verifiable extricating details) than neurotypical adults’ truthful accounts. We also anticipated that experiencing the mock-suspect interview process would be more challenging for autistic than non-autistic adults (i.e., difficulty understanding questions, level of anxiety, how supported they feel).

In the present study, we examined verbal deception cues displayed by autistic adults during a mock-police suspect interview. We found that autistic mock-suspects displayed similar verbal deception cues (in terms of statement-evidence consistency and investigation-relevant information) to the non-autistic group regardless of whether they were telling the truth or being deceptive. To our knowledge this was the first study in which autistic adults (without co-occurring intellectual disability) actively demonstrated the ability to tell verbal lies (see Bagnall et al., 2022). Specifically, the degree to which mock-suspects lied through omitting incriminating actions during the VE task (therefore contradicting the incriminating evidence) was similar regardless of diagnostic group. Further, a similar proportion of investigation-relevant information (IRI) was reported by autistic and non-autistic mock-suspects who lied, with both diagnostic groups providing sparser verbal accounts than mock-suspects who told the truth. This is consistent with reality monitoring theory (Johnson and Raye, 1981), in which information derived from internal sources (e.g., imagination) is less detail-rich than information from external sources (e.g., experienced events). These findings therefore suggest an at least surface-level consistency between autistic and non-autistic verbal deception.

However, we also found that innocent autistic-mock-suspects displayed verbal cues associated with deception. Consistent with findings from Young and Brewer (2020), autistic-mock-suspects who told the truth reported fewer items of verifiable, extricating information to support their innocence. This is concerning given that providing fewer verifiable details during interview is not only consistent with liars’ strategies to avoid disprovable claims (Nahari, 2018), it also makes veracity judgments more difficult (Porter and Salvanelli, 2020) and may narrow the options for further investigation and elimination from enquiries (College of Policing, 2021). As such, our findings emphasize that investigative interviewers should be cautious when interpreting gaps or missing elements in autistic suspects’ accounts. Indeed, previous studies report that expressive language capacity predicts autistic adults’ verbal specificity during interviews (Norris and Maras, 2022) and ToM (ability to take others’ perspectives) relates to their likelihood of providing extricating innocence-supporting detail (Young and Brewer, 2020). Consequently, more supportive interview techniques appear necessary to help autistic suspects report all information relevant to an investigation (more on this point later in the “Discussion”).

Contrary to our expectations, autistic mock-suspects’ statements were not significantly less consistent with the seven pieces of incriminating evidence nor contained significantly less IRI than those of non-autistic mock-suspects. It is of note that innocent autistic-mock-suspects’ statement-evidence consistency was indeed lower than the non-autistic group (with a small effect size of partial η² = 0.017), as was proportion of IRI (a medium effect size of partial η² = 0.066). The present study was powered to detect medium-large effects of interview performance, meaning that a larger sample may have been necessary to detect these smaller effects. Overall, however, our findings indicate that autistic adults may display certain verbal deception cues when telling the truth during police suspect interviews.

Further, autistic participants (in both innocent and guilty conditions) found interview questions harder to answer, felt more anxious and perceived the interview as less supportive to their needs. While investigative interviews elicit anxiety in neurotypical populations (Vanderhallen et al., 2011) this may be particularly problematic for autistic suspects. Elevated anxiety is associated with poorer executive functioning in autistic adolescents (Hollocks et al., 2014), and broader socio-cognitive processing difficulties (Velikonja et al., 2019) may impact autistic suspects’ ability to provide best evidence. It should be noted that, although autistic participants reported significantly poorer interview experiences than non-autistic participants, average scores still tended to fall in a ‘neutral’ rating. However, the mock-interviewer was specifically trained and instructed to adopt an encouraging and non-confrontational questioning style. Interviews with vulnerable suspects carried out during genuine investigations are often less accommodating. Inappropriate (e.g., forced choice) questions and minimisation tactics have been found to be more commonly used with suspects who have mental health conditions than suspects without such conditions (Farrugia and Gabbert, 2020b). Autistic people have also described feeling overwhelmed by the frequency and length of real-life suspect interviews, as well as difficulty concentrating on questions and experiencing pressure from investigators (Allen et al., 2008). Our present findings therefore likely underplay the difficulty of a real-life suspect interview for an autistic person and the degree to which subsequent verbal behaviors (associated with deception) may be exacerbated. Future research should further investigate the experience of police suspect interviews for autistic people, and the factors which contribute to atypical behavior as well as the elicitation of accurate and reliable information.

It should also be noted that increased anxiety may potentially contribute to autistic people displaying stress-adaptive (though atypical) behaviors during a suspect interview. For example, autistic people (automatically and voluntarily) use gaze aversion and repetitive movement to self-regulate hyperarousal (Collis et al., 2022; Stuart et al., 2022). Concerningly, these nonverbal behaviors are also stereotypically associated with deception (Hartwig and Bond, 2011; Vrij, 2019). Lim et al. (2022) examined whether the presence of specific autism-typical verbal, paraverbal and nonverbal behaviors (gaze aversion, repetitive movement, literal interpretation of figurative language, poor reciprocity and flat affect) predicted truthful autistic mock-suspects as being (incorrectly) perceived as deceptive. Autistic mock-suspects were rated by observers as more deceptive and less credible than non-autistic controls, though none of the hypothesized autism-typical behaviors predicted deception judgments (nor did behaviors significantly differ in prevalence between diagnostic groups). Understanding which verbal, paraverbal and nonverbal characteristics displayed by autistic people are most influential for inaccurate deception judgments, and how these may be exacerbated by police suspect interviews, is a key direction for future research.

The present research is not without limitations. Our participant sample was of above average intelligence (based upon IQ scores), and as such the autistic group is not reflective of the full heterogeneity of the autism spectrum. Indeed, it is estimated that between 13 to 20% of autistic people have co-occurring intellectual disability (Ghirardi et al., 2018; Rydzewska et al., 2019). Given that people with intellectual disabilities may be overrepresented in the CJS (Hellenbach et al., 2017; Chester, 2018), better understanding of specific vulnerabilities for autistic people with co-occurring developmental conditions during investigative interviews is needed. However, given the greater social and cognitive difficulties associated with intellectual disability (Smith et al., 2020), the issues raised in the present study may only be more pronounced for autistic people with co-occurring intellectual disability. It is also important to acknowledge that, unlike in a ‘real’ investigative interview in the UK, autistic mock-suspects were not entitled to support via an Appropriate Adult nor legal advisor (Home Office, 1984/2008) either of which may have helped improve interview performance and reduce anxiety. Though as appropriate support is often not provided to autistic adults in custody (Slavny-Cross et al., 2022), our findings emphasise the vulnerabilities of autistic suspects when support is absent.

The potential effects of employing a virtual environment (VE) paradigm should also be considered. It is possible that undertaking a mock-crime or a non-criminal task in a VE (rather than via an ‘in-person’ task) contributed to participants perceiving the task and subsequent mock-suspect interview as more simulative. In which case, guilty participants may have perceived their mock-crime to be less transgressive and innocent participants felt it less important they convey their innocence. However, previous studies have shown computer game-based (mock-crime and non-criminal) tasks to be effective when participants are required to generate truthful and deceptive accounts during a subsequent mock-investigative interview (e.g., Dando and Bull, 2011). Further, virtual recreations of tasks such as public speaking can induce comparable levels of stress response to real-life equivalent tasks (Kothgassner et al., 2016) and such immersive technologies often provide valid alternatives to in vivo (real-life) exposure (Wechsler et al., 2019; Liberatore and Wagner, 2021). It should be noted that we used a desktop-based VE paradigm in the present study, and VEs presented on head-mounted displays (e.g. 3D virtual reality) can produce greater spatial presence and immersion (Shu et al., 2019). The potential effects of using a desktop-based VE (rather than 3D virtual reality) in the present study is unclear. A more realistic criminal and non-criminal task (i.e., in 3D virtual reality) may elicit a greater sense of ecological validity and influence mock-suspects’ perception of a subsequent mock-suspect interview (e.g., feel a greater desire to convey their innocence). However, given that autistic participants in the present study experienced the mock-interviews as more demanding than non-autistic participants, such group differences may only be more pronounced if a more realistic simulation were used. Future research may therefore benefit from presenting VEs using head-mounted displays (i.e., in 3D virtual reality) to further examine potential vulnerabilities for autistic people during suspect interview settings.

These limitations notwithstanding, the present research has several implications for practice. Our findings highlight the additional complexity for investigators when interviewing autistic suspects, as verbal deception cues may be displayed whether the interviewee is truthful or lying. Existing witness and suspect interview models provide future directions for more supportive practice to benefit both interviewer and interviewee. The Witness-Aimed First Account (WAFA) approach reduces social and cognitive demand through autistic mock-witnesses generating segmented event memories prior to ‘free recall’, resulting in more detailed and accurate accounts while also being making autistic mock-witnesses feel more socially comfortable (Maras et al., 2020). In turn, this approach may aid recall of relevant, verifying information while reducing stress-induced paraverbal and nonverbal behaviors (i.e., deception cues). Further, the ‘Model Statement’—an example of a detailed statement on an unrelated topic presented to a suspect pre-interview (Leal et al., 2015)—may help account for autistic peoples’ difficulty in gauging relevance and quantity of required information. An interviewer being more explicit and specific about what is expected of autistic suspects during interview may assist the suspect’s understanding of questioning. Further research is needed to ascertain the validity of such models for detecting truth and lies in autistic suspects.

However, adaptations to interview practice with autistic suspects are dependent upon pre-interview identification of ‘vulnerability’. Although custody staff and interviewing officers are guided to consider signs of potential vulnerability (College of Policing, 2022), an autistic suspect may not be correctly identified due to a lack of specific questions about autism during the ‘booking in’ phase in custody (Sims, 2017) or because a detainee chooses not to disclose being autistic out of concern of stigma (Crane et al., 2016). A lack of awareness of a suspect’s autism may lead to harsher interpretations of their behavior during interview (Maras et al., 2019; Logos et al., 2021; Lim et al., 2022). Custody staff should therefore make additional efforts to identify potential autistic detainees (Holloway et al., 2022) and interviewers should be conscious of avoiding a guilt-presumptive questioning style (Lidén et al., 2018). Despite PACE guidelines (section 11C) in the UK stating that vulnerable suspects’ accounts may be inadvertently “unreliable” or “misleading,” and that “corroboration of any facts admitted” should be obtained and an appropriate adult provided, there is little further specific guidance for interviewers relating to this issue. Extending this guidance with evidence-based examples highlighting the heterogeneity of autistic verbal, paraverbal and nonverbal behavior and embedding it in policy and training is an important future direction for CJS practice.

In the present study, we found that investigative interviews are more socially and cognitively demanding for autistic than neurotypical mock-suspects. In addition, verbal cues associated with deception can be displayed by autistic mock-suspects even when truthful. The development of autism-focused suspect interview techniques is therefore crucial to resolve the (interrelated) issues of interviewee welfare and provision of best evidence. Discriminating between difficulty and deception in autistic suspects’ interview accounts is a challenging though necessary task for researchers and practitioners alike.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Psychological Research Ethics Committee at the University of Bath. The participants provided their written informed consent to participate in this study.

RB, AR, MB, MO, and KM contributed to the conceptualization and design of the study. RB and AC collected and coded the data. RB performed the statistical analysis and wrote the first draft of the manuscript. All authors contributed to the article and approved the submitted version.

This work was conducted as part of PhD research undertaken by RB, funded by the Economic and Social Research Council (2096910).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.