Reasoning on the Autism Spectrum: A Dual Process Theory Account

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder of unknown underlying etiology characterised by persistent deficits in social communication and social interaction combined with restricted, repetitive patterns of behaviour, interests, or activities (APA 2013) with prevalence estimates up to one in 68 children (CDC 2015). A continuum of autism traits extend throughout the general population until they become clinically significant under ASD diagnostic criteria to form part of an ‘autism spectrum’ of autistic presentation (Constantino and Todd 2003; Plomin et al. 2009; Posserud et al. 2006; Wing 1988; see Ruzich et al. 2015, for systematic review). Ruzich et al. report that, within the general population, males have significantly higher levels of autism traits than females, and those with ASD have significantly higher autism traits than males from the general population (with no sex differences in autism traits within the ASD population).

Reasoning and decision making are core human capabilities that enable effective participation within society, yet have received relatively little attention within the autism spectrum literature. Reasoning in a manner which is normatively logical and subsequent rational decision making are specialised higher cognitive functions attributed to common processing mechanisms (see Evans and Stanovich 2013). Luke et al. (2012) identified three core features of reasoning and decision making that were particularly problematic for people with ASD using a self-report methodology. Decisions were difficult to make for people with ASD if they involved talking to others; involved a change in routine; or if the decision has to be made quickly. Whilst difficulties talking with others and changes in routine reflect the core diagnostic features of ASD (respectively), the difficulty with rapid decision making for those with ASD does not, and may provide added insight into reasoning on the autism spectrum.

A more deliberative approach to reasoning has been proposed to characterise people with ASD compared to the general population. For example, De Martino et al. (2008) report that people with ASD reason in a more logically consistent manner than matched controls. People with ASD also request more information prior to making a decision upon a probabilistic reasoning task compared to controls, a style of reasoning that has been termed a ‘circumspect reasoning bias’ (Brosnan et al. 2014a). Similarly, those from the general population who self-reported being higher in autism traits also require more information prior to making decisions when compared to those lower in autism traits (Brosnan et al. 2013). This behavioural data is consistent with self-reports from people with ASD about their difficulties with rapid decision-making and reasoning and a preference for more deliberation (Luke et al. 2012). Contrasting rapid decision-making/reasoning processes with deliberative decision-making/reasoning processes forms the basis of Dual Process Theory. When considering human reasoning, Dual Process Theory has been a dominant model within cognitive psychology for almost 50 years (Evans and Frankish 2009). The dual processes are referred to as Type 1 and Type 2 and will be referred to as intuition and deliberation (respectively) for convenience. Intuition involves rapid, effortless, parallel, non-conscious processing that is independent of working memory and cognitive ability. Deliberation, on the other hand, involves slower, effortful, sequential, conscious processing and is heavily dependent on working memory and related to individual differences in cognitive ability (see Evans 2011; Evans and Stanovich 2013; Kahneman 2011; Stanovich and West 2000, 2008; for reviews; see Keren and Schul 2009 for critique; see Kruglanski and Gigerenzer 2011 for an alternative view).

Within Dual Process Theory, rapid autonomous processes (‘intuitive reasoning’) are assumed to yield default responses unless intervened upon by distinctive higher order reasoning processes (‘deliberative reasoning’). Intuitive reasoning preceding deliberative reasoning is known as the default-interventionist position (see Evans and Stanovich 2013; Kahneman 2011). One of the most widely used behavioural assessments of intuition and deliberation is the Cognitive Reflections Test (CRT: Frederick 2005). The CRT comprises of three reasoning questions that have both an intuitive (incorrect) and deliberative (correct) response. A majority of intuitive responses are typically provided for the CRT questions (Frederick 2005), which is theorised to reflect the output from initial intuitive reasoning which has not been over-ridden by deliberative reasoning. The over-riding of initial intuitive reasoning by subsequent deliberative reasoning is demonstrated by achieving the correct answer. In support of this, experimental manipulations designed to encourage participants to engage in deliberative reasoning reduces intuitive responses (Evans and Curtis-Holmes 2005).

Intuitive reasoning is also argued to be evidenced by ‘the framing effect’ (Tversky and Kahneman 1974) in which logical decision making is influenced by the context of the reasoning task. De Martino et al. (2008: 10746) report a decreased susceptibility to the framing effect in people with ASD, who demonstrate an ‘unusual enhancement in logical consistency’. Within the context of Dual Process Theory, De Martino et al. (2008) hypothesise that individuals with ASD have an increased tendency towards deliberation, attributable to impairment within intuitive reasoning systems. This is consistent with Klin and Volkmar (1997: 102) observations of people with Asperger’s Syndrome as having ‘a deficient intuition and lack of spontaneous adaptation’ (see Allman et al. 2005). Klin et al. (2003) propose an embodied cognition ‘Enactive Mind’ approach for understanding social adaptation, which is argued to have important temporal constraints. Social adaptation is reasoned to require the processing of salient stimuli based upon split-second environmental demands with moment-by-moment disregard of stimuli perceived as irrelevant. Under this approach, people with ASD do not reflect the typical processing bias towards socially relevant stimuli.

In addition to the CRT behavioural measure, the propensity to engage in intuitive and deliberative reasoning can be assessed through self-report. The Rational-Experiential Inventory (REI) is a widely used measure of intuition and deliberation (Epstein et al. 1996). The Rational component is based upon a ‘need for cognition’ (Cacioppo and Petty 1982) which measures engagement in, and enjoyment of, cognitive activities. The Experiential component was developed to measure engagement and confidence in one’s intuitive abilities and is termed ‘faith in intuition’ (Epstein et al. 1996; Pacini and Epstein 1999). Epstein et al. argue that these two information processing styles are independent of one another, such that one can be high or low in either or both dimension.

Deliberative responses on the CRT have been found to positively correlate with REI self-reported deliberation and negatively with REI self-reported intuition. Additionally, intuitive responses on the CRT have been found to positively correlate with REI self-reported intuition and negatively with REI self-reported deliberation (Pennycook et al. 2015). However, other studies have only reported the positive relationship between deliberation on the CRT and REI (Liberali et al. 2012; Thoma et al. 2015). Thus, the variability between self-reported preference for intuition and behavioural intuition needs to be borne in mind. Freeman et al. (2012) found that combinations of high and low intuition with high and low deliberation, as measured by the REI, best predicted clinically relevant traits (schizotypy) in a general non-clinical population. The literature above would suggest that autism traits in a general population would best be predicted by a combination of high deliberative and low intuitive reasoning styles.

The aim of the present research was to investigate intuitive and deliberative reasoning across two studies that focus on the autism spectrum; one involving a non-clinical sample, and the other involving a clinical sample. As a continuum of autism traits is proposed to extend throughout the (general and ASD) population, the relationship with reasoning was examined in relation to relatively higher and lower levels of autism traits. In the non-clinical sample it was predicted that higher levels of autism traits would relate to a profile characterised by greater deliberative and reduced intuitive reasoning. Study 2 compared self-report and behavioural measures of intuitive and deliberative reasoning between people with and without ASD, and it was predicted that the ASD group would show a more deliberative and less intuitive profile than the comparison group.

Two studies explored the relationship between Dual Process Theory of human cognition and reasoning on the autism spectrum. People with high autism traits and those diagnosed with ASD showed a pattern of having a combination of lower intuitive and greater deliberative reasoning styles. Both those with high autism traits and those with a diagnosis of ASD consistently responded less intuitively and more deliberatively, when compared to those with low autism traits and without a diagnosis of ASD, on behavioural and self-report assessments of reasoning. Taken together, the results suggest that Dual Process Theory provides a useful framework for considering the strengths and weaknesses in reasoning on the autism spectrum. De Martino et al. hypothesised that those with ASD may have compromised intuitive reasoning, based upon those with ASD demonstrating enhanced logical consistency. Consistent with this, this is the first study to demonstrate enhanced deliberative responding explicitly in those with ASD and those with high autism traits, which is consistent with enhanced logical consistency or a circumspect reasoning bias (Brosnan et al. 2014a; De Martino et al. 2008).

Within the default-interventionist perspective of dual process theory (see Evans and Stanovich 2013; Kahneman 2011), intuitive processes are assumed to yield default responses unless intervened upon by distinctive higher order reasoning processes (deliberative). There are at least three potential explanations for the present findings: (1) those with ASD have impaired intuitive mechanisms and consequently deliberative reasoning is dominant; (2) intuitive mechanisms are intact but dominated by deliberative reasoning; or (3) Intuitive mechanisms are intact in different contexts but this context triggers deliberative reasoning in those with higher levels of autistic traits. Under this Dual Process model, a slower, effortful, sequential, deliberative reasoning style would be dominant in ASD as a consequence of impairment in rapid, effortless, parallel intuitive mechanism (De Martino et al. 2008), though the possibility remains that this may be due to dominant deliberative reasoning (possibly within the context of the task). Either way, higher autism traits would be associated with a propensity to engage in deliberative reasoning rather than being associated with deliberative reasoning abilities per se (e.g. a greater preference for deliberative reasoning within an individual does not necessarily entail greater deliberative skills within that individual). This is consistent with Luke et al. (2012) who found that those with ASD did not self-report a greater reliance upon a rational/deliberative reasoning style. However, people with ASD self-reported that they have difficulties making decisions quickly, which is consistent with the idea of an impairment in the rapid reasoning that characterises intuition.

It should be noted that the intuitive response is actually a wrong choice, despite being the dominant response in highly educated American college students (Frederick 2005). A propensity to engage in deliberative reasoning within a context that typically triggers erroneous intuitive reasoning can be seen as an advantage associated with ASD (and potentially higher autism traits). Toplak et al. (2011, 2014) provide an extensive analysis of the correlates of the CRT with other reasoning tasks, however there are no tasks which assess intuition independently of deliberation. Within the CRT, intuitive reasoning results in erroneous responses and could therefore be regarded as a deficit within this context that people with ASD do not demonstrate to the same degree as the general population.¹ Toplak et al. (2011, 2014) argue that the CRT is a unique predictor of susceptibility to biases. As biases occur through the application of heuristics rather than engaging in further analytic processing, Toplak et al. describe the CRT as a particularly potent measure of ‘miserly’ cognitive processing (or ‘lazy thinking’, Kahneman 2011: 48). Defaulting to the reduced processing demands of being a cognitive miser has been argued to be typical in many contexts (cf. Fiske and Taylor 1991). From this perspective, ASD is associated with not being cognitively miserly (or lazy thinking), and the proposed reasoning bias is best characterised as being unbiased (within the context of the CRT at least).

Thus, in the present study, the ASD group obtained more correct answers than the TD group which is consistent with the literature identifying a diminished framing effect in ASD (De Martino et al. 2008). The deliberative score on the CRT relates to a wide range of rational reasoning tasks (Lesage et al. 2013; Sirota et al. 2014; Toplak et al. 2011, 2014). However, whilst self-reported intuition can negatively correlate with rational reasoning (Shiloh et al. 2002), apart from the intuitive subscale of the REI, assessments of intuition are rare (see Pennycook et al. 2015). The proposed advantages in a greater propensity towards deliberative reasoning in ASD may have an associated cost on intuitive reasoning. However, this was not independently evidenced in this study as we did not have a behavioural measure of intuition independent of deliberation. Recall also, the variation in findings regarding the relationship between behavioural and self-reported intuition in the general population identified in the literature, which may indicate that the intuitive subscale of the REI is not a reliable index of intuitive behaviour.

Despite issues with assessment, potential biases away from intuitive reasoning in other contexts, such as social contexts, may be pertinent to ASD as they may be associated with deficits in empathy (e.g. Baron-Cohen and Wheelwright 2004). The social world often requires rapid processing of social cues occurring in an uncertain context rather than an overt ‘rule-based’ system. Within their Enactive Mind approach, Klin et al. (2003) characterise the social world as an ‘open domain task’ requiring an understanding of the relative significance of a multitude of elements, the importance of which are dependent upon the context of the situation. Future research can develop the extent to which a propensity towards deliberative processing within naturalistic social settings is too slow and effortful to allow for effective participation. As Darius (2002: 25; cited in Davidson 2008) recounts: ‘There is no such thing as adequate delayed social reactions. One is either quick enough to keep up, or one is weird and socially disabled’. Rapidly and automatically extracting emotional information from social environments is argued to be an intuitive process that feeds ‘downstream’ empathy processes and related social–emotional functioning (Clark et al. 2008; Kahneman 2011; see also Rump et al. 2009; Tracy et al. 2011). Thus, the unbiased use of deliberation identified in the present research may relate to the social-emotional weaknesses which form part of the diagnostic criteria for ASD. Consistent with this, the measures on intuition used in the present study have been found to correlate with measures of empathy in a general population (Brosnan et al. 2014b). Dual Process Theory may therefore extend beyond reasoning to provide a fuller account of the social cognition that characterises ASD (see Evans 2008; Sherman et al. 2014).

Thus a bias towards deliberative reasoning within Dual Process Theory may provide an account of the strengths associated with ASD. Dual Process Theory therefore might usefully bring to bear additional cognitive research from non-clinical groups pertaining to how combinations of intuition and deliberation may relate to ASD. For example, the default-interventionist position could characterise the application of deliberative reasoning in ASD to typically intuitive tasks such as emotion recognition (‘corners of mouth turned down, lowered eyebrows = sad’: Rutherford and McIntosh 2007; Walsh et al. 2014; see also Brosnan et al. 2015b; Golan and Baron-Cohen 2006; Golan et al. 2010). Within Dual Process Theory, intuitive tasks such as rapid emotion recognition, would be expected to be independent of working memory and cognitive abilities in a general population, but not in an ASD group who were utilising deliberative reasoning (see Harms et al. 2010, for review of the evidence for this). Physically slowing stimuli would also be predicted to enhance the performance of those utilising deliberative reasoning strategies (see Gepner et al. 2001; Tardif et al. 2007; Gepner and Féron 2009). Typically, intuitive and deliberative reasoning can be applied as appropriate to the perceived demands of the reasoning context. The present study is consistent with the idea that those with ASD do not have the balance of reasoning styles but have a bias towards deliberative reasoning and away from intuitive reasoning across contexts. There may be contexts where this is beneficial (e.g. mathematics) and contexts where this is detrimental (e.g. social).

Interestingly, Freeman et al. (2012) also propose an imbalance in reasoning styles is related to schizotypy traits in a non-clinical population sample. Freeman et al. found that higher intuitive reasoning combined with lower deliberative reasoning was related to the degrees of schizotypy traits, which is the opposite pattern to that associated with autism traits in the present study. Those with high levels of schizotypy have also been found to bias towards making decisions rapidly (‘jumping to conclusions’; Freeman 2007; Freeman et al. 2008; Garety et al. 2005, 2007) again reflecting the opposing pattern to higher autism traits and ASD identified in the present study. Schizotypy was not assessed in the present study, but it is interesting to speculate that variation in the relative biases towards intuition and deliberation within Dual Process Theory may represent a framework within which similarities and differences between these clinical conditions and associated traits across the general population can be further explored (see Crespi and Badcock 2008; see also Brosnan et al. 2010; Chisholm et al. 2015 for review). Within the non-clinical population, the extent to which individuals engage in intuitive or deliberative reasoning has been found to be susceptible to manipulation. For example, being told to ‘think carefully’ or to write down details of how you came to a decision have been found to elicit more deliberative responses, where as being instructed to go with a ‘gut-feeling’ has been found to elicit more intuitive responses (e.g. Dijkstra et al. 2012; Usher et al. 2011). However, if reasoning in ASD is characterised by diminished intuitive mechanisms, whether such manipulations would affect those with ASD is an open question future research. Further, there is also the intriguing possibility that even if people with ASD were shown not to employ rapid, effortless, parallel, non-conscious processing (‘intuition’) in social contexts they may default to them in other contexts (e.g. there are alternative ‘open domain tasks’; Klin et al. 2003).

There were limitations to the present research. All of the participants were at university or intending to go to university and were therefore not reflective of either the ASD or non-clinical populations as a whole. This may relate to the lack of sex differences in Study One. The sample was also not specified in terms of further demographics, such as social economic status, which limits generalisability. The participants in Study 2 were all considering attending university, and the degree of ASD symptomology was likely mild and the findings may not extend to the whole autism spectrum. Study 2 only compared male participants which is another limitation of the study, especially given the relatively little understanding of female populations with ASD (Halladay et al. 2015). In addition, only 11 participants with ASD completed the REI in Study 2, and the potential for a Type 1 error needs to be borne in mind. Whilst the CRT is a widely used measure of intuitive and deliberative responding, it should be noted that the intuitive response is wrong. Other wrong responses are not considered intuitive and are not typically analysed (hence the task is not purely ipsative). Analysing errors may provide useful insights into whether they emerge from intuitive or deliberative reasoning (Brosnan et al. 2015a). In addition, the terms intuition and deliberation have been used for stylistic convenience, but it should be noted that they refer to two clusters of concepts (Type 1 and Type 2 respectively, see Evans 2008 for a review; see Keren and Schul 2009 for critique; Kruglanski and Gigerenzer 2011 for an alternative view) that share similarities, but also have differences, within each cluster. Finally, an independent assessment of IQ was not undertaken, which is a major limitation of the study. Whist intuitive reasoning is argued to be independent of cognitive abilities, deliberative reasoning is not. The participants were studying at the same educational level (A-levels, examinations typically taken at 18 years of age for University entry) though may not attain the same grades. Future research can address the assessment of cognitive ability. Although college students do not necessarily have higher levels of autism traits than random control groups, those studying sciences (including mathematics) have been shown to have higher levels of autism traits than those studying humanities and social sciences (Baron-Cohen et al. 2001a, b), which can also be explored in future research.