The Impact of Exam Stress on the Relationship Between Autistic Traits and Disordered-Eating Attitudes in a Non-clinical Population

Previous research has demonstrated that individuals with anorexia nervosa display elevated autistic traits in comparison to individuals who do not have the disorder. However, it remains unknown as to whether this relationship is a stable trait of individuals or if it is caused by a psychological or physical state. This study investigated the state vs trait nature of the relationship between disordered eating and autistic traits in a non-clinical sample, while additionally exploring the ability of cognitive in�exibility to predict this relationship. Thirty-four undergraduate students from the University of St Andrews completed questionnaires regarding their affect, disordered-eating attitudes and autistic traits alongside a measure of set-shifting ability in the Brixton Spatial Anticipation test during a period of examination stress and a period of no stress. It was revealed that on both occasions, individuals who scored higher on a test of disordered-eating attitudes also scored higher on a test measuring autistic traits. This relationship was stronger when participants were stressed. Autistic traits were also found to signi�cantly predict a change in eating attitudes between a stress-induced state and pre-existing trait conditions. In�exibility, as re�ected by Brixton scores, did not appear to relate to either measure. These results indicate that slightly elevated autistic traits in those exhibiting disordered-eating attitudes may be state-dependent, as opposed to re�ecting the engrained traits of the individual. Findings are discussed in relation to future research replicating the study with a larger sample and implications for the diagnosis and treatment of anorexia.


Introduction An overview of the relationship between anorexia nervosa and autism spectrum disorders
Anorexia nervosa (AN) is a mental illness characterised by a persistent fear of weight gain, disturbance in body image and continuous restriction of energy intake.Autism spectrum disorder (ASD) is a neurodevelopmental disorder, diagnosed by dif culties regarding social communication, restricted interests and repetitive behaviour (American Psychiatric Association, 2013).In recent years, an in uential line of research exploring the cognitive similarities and clinical relationship between AN and ASD has emerged.The rst case study in this area was published in 1988, depicting a female with infantile autism who developed AN upon entering puberty (Rothery and Garden, 1988).Since then, Huke, Turk, Saeidi, Kent and Morgan (2013) have estimated the prevalence of ASD in disordered-eating populations at 23 per cent, with Gillberg and Råstam (1992) also stating that 23 per cent of girls with AN exhibit an outward manifestation of autistic behaviour, such as 'social problems and communication peculiarities'.Such evidence has led to the exploration of shared endophenotypes between the two disorders (Zucker et al., 2007), the identi cation of atypical eating behaviours in ASD (Karlsson, Råstam and Wentz, 2013) and even the suggestion that AN may Reinvention: an International Journal of Undergraduate Research 16:2 (2023) be a female presentation of ASD (Treasure, 2013).However, the notion that AN and ASD may result from the same underlying endophenotype is constrained due to the matter that ASD is a pervasive, lifelong disorder, whereas AN is a mental illness that can be recovered from.Currently, deciphering whether the pathophysiological similarities between AN and ASD are state-dependent or trait-dependent has been overlooked in literature, but may have implications regarding diagnostic process, prevention and treatment.

Cognitive and behavioural similarities
The proposed relationship between AN and ASD was formed on the basis of clinical observations, particularly food refusal and selective eating in those with ASD (Gillberg, 1983;Råstam, 2008).Such observations sparked further investigation into the cognitive de cits of AN populations, as even fully recovered, weight-restored individuals tended towards test pro les similar to those with ASD (Gillberg, Gillberg, Råstam and Johansson, 1996).'Weak central coherence theory' -initially proposed by Frith (1989) and referring to failure to integrate local details into a global entity -is a highly credited explanation for the global dif culties (especially restricted interests) observed in ASD.It has also been empirically noted as associated with AN (Coombs et al., 2011) through use of tasks such as the Rey-Osterrieth Complex Figure Test (Osterrieth, 1944), whereby participants are required to locate hidden shapes.Those with AN and ASD both demonstrate superior detail focus of this task, indicated by faster reaction times relative to healthy controls (Lang et al., 2015).

Set-shifting dif culties in anorexia nervosa and autistic spectrum disorders
As such, empirical research has focused largely on the cognitive processing styles that are common across the two disorders, such as those described above.However, one area of difference that has been overlooked thus far is set-shifting -the ability to shift from one course of action to another with little dif culty (Huke, Turk, Saeidi, Kent and Morgan, 2013).Switching 'back and forth between multiple tasks, operations, or mental sets' (Miyake et al., 2000) is a major component of executive functioning and is essential for the adaptation of behaviour in line with changing demands of the environment.Individuals with AN are often described as persistent, with rigid or obsessional personalities, representing an outward manifestation of a set-shifting dif culty (American Psychiatric Association, 2013).This is re ected further by low scores from samples of those with AN on the Wisconsin Card Sorting Task (WCST) , whereby individuals must match cards based on an unknown rule (Roberts, Tchanturia, Stahl, Southgate and Treasure, 2007).
Dysfunctional set-shifting is also a hallmark of ASD (Yerys et al., 2009).A more general inability to shift attention during daily activities exists in autistic samples (Gioia, Isquith, Kenworthy and Barton, 2002), yet individuals also respond with less accuracy and at a slower rate than healthy controls on the WCST (Dichter et al., 2009).While set-shifting has never been directly compared in AN and ASD samples, Oldershaw, Treasure, Hambrook, Tchanturia and Schmidt (2011) statistically compared the neuropsychological pro les of currently ill patients with anorexia with published scores of individuals with ASD on the WCST and ascertained the two groups to be statistically similar.
Despite Theory of Mind (ToM) dif culties becoming less atypical when AN groups are recovered and weight restored (Oldershaw, Hambrook, Tchanturia, Treasure and Schmidt, 2010), dif culties in set-shifting have been found to persist following recovery (Tchanturia et al., 2004), suggesting they are a durable trait of those with the disorder.For example, Tchanturia et al. ( 2004) not only found no improvement on a battery of setshifting tasks following weight recovery, but also hypothesised that poor performance on tests of set-shifting Reinvention: an International Journal of Undergraduate Research 16:2 ( 2023) is associated with childhood rigidity -implying that this dif culty exists pre-morbidly to the onset of AN.
Similarly, Holliday, Tchanturia, Landau, Collier and Treasure (2005) suggest that poor performance on setshifting tasks has been observed in the unaffected sisters of individuals with AN, indicating a genetic linkage.Findings from genetic and longitudinal studies such as these imply that cognitive in exibility persists following weight restoration, and thus may constitute a familial trait that is associated with an increased risk of developing AN.

The state vs trait conundrum
Data regarding whether general autistic traits persist in individuals with anorexia following recovery is contradictory and inconclusive.The question of whether the relationship between AN and ASD is caused by the permanent traits of the individual or a eeting state (Chaplin, John and Goldberg, 1988) has been highly debated, with many suggesting that individuals with AN may come to resemble those with ASD due to the chronic starvation or psychological stress associated with the disorder (Hiller and Pellicano, 2013;Treasure, 2013).

The psychological effects of stress on anorexia nervosa and autistic spectrum disorders
This study will aim to accurately isolate whether the relationship between AN and ASD is dependent on the state the individual is in or if it is a durable trait of the individual by utilising a naturally occurring state of stress.Despite the pervasive nature of ASD, individuals with the disorder often experience a worsening or lessening of symptoms under conditions of stress (Baron et al., 2006).Groden et al. (2001) hypothesised that persons with autism may be even more vulnerable to the psychological effects of stress than the general population as they lack the repertoire of appropriate coping mechanisms that others possess.For example, dif culties in communication and social skills, alongside heightened perceptual sensitivity, have been found to lead to an increased vulnerability to stress and elicit an even greater stress response than average (Groden, Prince, Cautela and Berryman, 1994).
Studies have also revealed that stress and eating-disorder symptomatology are positively correlated (Shatford and Evans, 1986;Wolf and Crowther, 1983), even going as far as to suggest that disordered eating in itself may be a manifestation of maladaptive coping styles in response to stress in student populations (Mayhew and Edelmann, 1989).For example, Macht, Haupt and Ellring (2005) found that emotional regulation through eating is experienced in student populations during stressful periods, causing an increased tendency to over-eat or under-eat.These studies suggest that utilising a naturally occurring condition to heighten stress will elicit a pseudo-anxious state that will allow us to measure how autistic traits and disordered-eating attitudes change in response to stress.This study will elicit enduring features of stress in a student sample by testing during the examination period.

The relationship between autistic traits and disordered eating in non-clinical populations
This line of research has the potential to make an important contribution towards the clinical eld and could potentially in uence how AN is diagnosed and treated (Baron-Cohen et al., 2013).However, it also holds extensive implications regarding how the association between disordered eating and autistic traits occurs in non-clinical populations.The notion of the autistic spectrum, in lieu of a single de ned disorder, has been extended from clinical samples to the general population, with it now being recognised that there are individual differences in 'autistic traits' (Constantino and Todd, 2003).Similarly, dieting and weight-related Reinvention: an International Journal of Undergraduate Research 16:2 (2023) concerns are extremely prevalent in the general population without an individual meeting the diagnostic criteria for AN, with such behaviours tentatively considered to be risk factors for the subsequent development of eating disorders (Jacobi, Hayward, de Zwann, Kramer and Agras, 2004).

Aims and hypotheses
The primary aim of the current study was to identify whether the relationship between disordered-eating attitudes and ASD symptomatology is state-or trait-dependent in a non-clinical sample.In keeping with previous literature that states that those who scored high on tests of disordered-eating attitudes also scored higher on tests of autistic traits (Baron-Cohen et al., 2013), we hypothesised that there would be a relationship between scores on measures of the two constructs in our sample of students.As research also states that symptoms from both spectrums of disorders are heightened during a state of stress, we predicted that scores on both measures would be higher during the examination period when participants were stressed.
A secondary aim of the study was to explore how cognitive in exibility -an area in which research is mixedwas associated with disordered-eating attitudes and autistic traits.This could reveal the role of set-shifting as a causal and maintenance factor in both disorders.It was predicted that cognitive in exibility would also be elevated by stress, predicting more errors on a visuospatial test of cognitive exibility during examinations.

Methodology Participants
Thirty-four rst and second year undergraduate students from the University of St. Andrews volunteered to participate in this experiment using a PsychSONA online registration system.Of this cohort, 9 males and 25 females participated, with ages ranging between 18 and 21 (M=19.88;SD=2.24).Participants were similar in terms of age and education level; however, they were from a range of ethnicities and cultures.All participants were required to have been taking December examinations at an undergraduate level in order to participate.The same sample participated at both Time 1 (in December during the exam period) and Time 2 of the study (in January when exams had ceased).

Eating Disorder Inventory (EDI-2)
The 23-item EDI-2 (Garner, 1991; α= 0.95) self-report questionnaire is a widely used measure of attitudes associated with body image and eating-related issues (see Appendix A).The items are divided into three individual sub-scales, including 'Drive for Thinness' (α= 0.63) depicting behaviour and feelings related to weight loss, 'Bulimia' (α= 0.87) depicting behaviour and attitudes consistent with Bulimia Nervosa (seven items) and 'Body Dissatisfaction' (α= 0.88) depicting thoughts and feelings regarding one's own body (nine items).Participants are asked to rate each item on a 6-point Likert scale in accordance with how applicable the statement is to their general eating attitudes, ranging from 'always' (scoring 3 points) to 'never' (scoring 0 points).Higher scores indicate greater psychopathology and increased negativity in relation to eating attitudes and behaviours.This test was chosen due to its reliability for measuring eating psychopathology in non-clinical samples (Schoemaker, van Strien and van der Staak, 2010;Tachikawa et al., 2004).
Reinvention: an International Journal of Undergraduate Research 16:2 ( 2023) The Autism Spectrum Quotient The Autism Spectrum Quotient (AQ) was originally developed as a screening instrument for non-clinical populations to assess how individuals of normal intelligence vary on autistic traits (Stewart and Austin, 2009; see Appendix D).The scale has high test-retest reliability and sound internal reliability with Cronbach's α for all sub-scales rating between .80 and .84.It is a self-report questionnaire containing 50 items relating to social skills, attention switching, attention to detail and communication.The questionnaire uses a forced-choice format , with each item having four possible responses ('de nitely agree', 'slightly agree', 'slightly disagree', 'de nitely disagree').Respondents are instructed to select one response per item.A point is assigned for each response characteristic of ASD, giving a resulting overall score between 0 to 50.
Higher scores indicate more autistic traits.The AQ has scored high for inter-reliability and test-retest reliability in a student sample (Baron-Cohen et al., 2001).

Depression, Anxiety and Stress Scales (DASS-21)
The Depression Anxiety Stress Scales 21 is a shortened form of Lovibond and Lovibond's (1995) 42-item instrument.It is a 21-item (α= 0.91) self-report test measuring three related negative-affect conditions over the previous week (see Appendix B).It is divided into three sub-scales, each containing seven items designed to assess the psychological features of depression (D-sub-scale; α= 0.94), anxiety (A-sub-scale; α= 0.87) and stress (S-Sub-scale; α= 0.91).
Each statement is rated on a 4-point Likert scale regarding applicability to their mood, from 'not at all' (0 points) to 'always' (3 points).The higher an individual scores, the greater their negative affect has been over the previous week.It is a reliable measure of state stress, with all scales being recorded as correlating moderately to highly with pre-existing measures of DASS (Antony et al., 1998).

Brixton Spatial Anticipation Test
The Brixton is a visuospatial sequencing task that assesses rule-shifting abilities (Burgess and Shallice, 1997; See Appendix C).Participants are presented with a 5 × 2 grid of circles, with one circle lled in blue.They are asked to predict where the blue dot will appear in the next grid.In order to successfully complete the task, the participants must abstract the currently active rule (Craik and Levine, 2012).The pattern by which the blue circle moves changes nine times.Responses for all 56 trials a (56 patterns whereby the rule changes 9 times) are recorded.Total number of errors across the trials is converted into a scaled score, which is then used as an indication of set-shifting ability.Therefore, the lower the number of errors, the greater the setshifting ability (Lounes, Khan and Tchanturia, 2011).scoresheet.The Brixton was always completed last so as to act as a distraction task and reduce any of the potentially harmful effects that may have been caused by being confronted with potentially dif cult issues in the three screening self-report questionnaires.All participants were then debriefed on the purpose of the study.Scores on all measures were calculated and statistical analyses were run.This experiment was run twice: Time 1 (T1) was in November, then it was repeated at Time 2 (T2) in January using an identical procedure.

Descriptive statistics and Wilcoxon signed-rank tests
Firstly, as a Shapiro-Wilk test revealed that assumptions of normality were not met, a series of Wilcoxon signed-rank tests were conducted to analyse if the mean scores observed in Table 1 on all measures were signi cantly different between T1 when participants were stressed and T2 when they were not.It was found that scores on all measures were signi cantly higher at T1 than T2 at an alpha level of .05,indicating that there was a decrease in scores when participants were not stressed.

Correlations
Based on the original hypotheses of this study, we used a Spearman's rho to identify the relationship between scores on the AQ and EDI at T1 and T2.As shown in Table 2, scores on the AQ and EDI were positively correlated at both T1, (r s = .594)and T2 (r s = .412).While the correlation was considerably stronger at T1 than T2, using the Fisher r-to-z transformation, it was found not to be statistically signi cant (Z= .95,p= .342).Secondly, no correlation was found between the AQ and the Brixton at either T1 (r s =. 115) or T2 (r s = .023), nor the EDI and the Brixton at T1 (r s =-.019) or T2 (r s = .063).
Reinvention: an International Journal of Undergraduate Research 16:2 (2023) A further aim was to determine how the sub-scales of the EDI were correlated with the AQ.On both occasions, all sub-scales of the EDI were positively correlated with the AQ, as can be observed in Table 2.At T1, the strongest correlation was between BD and AQ scores and at T2 between Desire for Thinness (DFT) and AQ scores.All of these correlations were stronger at T1 than T2; however, a Fishers r-to-z transformation revealed that these differences in magnitude were not signi cant: DFT (Z=0.37,p= .711),B (Z=0.94, p= .347)and B (Z=1 .42,p= .156).

Hierarchal regression
Firstly, a series of hierarchical multiple regressions were conducted to determine whether negative affect (as measured by the DASS) and autistic traits (as measured by the AQ) were signi cant predictors of eatingdisordered attitudes (as measured by EDI).A Shapiro-Wilk revealed that assumptions of normality were met for all except the stated test.At T1, DASS (model 1) was entered at the rst step and explained 32 per cent of the variance in scores on the EDI and was a signi cant predictor.
Reinvention The inclusion of AQ scores (model 2) accounted for an additional 21 per cent of variance in the dependent variable, leading to a signi cant increase in r 2 .In this nal model, both predictor variables were statistically signi cant (see Figure 1 for a visual representation).
Reinvention: an International Journal of Undergraduate Research 16:2 (2023) In the nal model, AQ was not found to be a signi cant predictor whereas DASS was.These results imply that AQ scores may be more predictive of EDI scores when participants were in a state of stress as opposed to when they were not.However, the Shapiro-Wilk test for this second regression violated assumptions of normality; therefore, these results should be treated with caution.
The nal hierarchical regression aimed to explore the extent to which negative affect and AQ scores at T1 would be able to predict EDI scores at T2, after controlling for initial EDI scores at T1. EDI scores at T1 (model 1) were entered at the rst step and were found to signi cantly predict scores on the EDI at T2, as was expected.While the inclusion of DASS scores and AQ scores (model 2) did not lead to a signi cant increase in r 2 change -only contributing an additional 4 per cent variance in the dependent variable -AQ scores at T1 entered as a signi cant predictor of EDI scores at T2, whereas DASS scores at T1 did not.This preliminarily indicates that it was the elevated autistic traits, as caused by the stress state at T1, that contributed towards the change in EDI scores between T1 and T2.
Reinvention Hierarchical regressions were also performed to investigate the ability of AQ to predict scores on each subscale of the EDI after controlling for negative affect, and to determine if this relationship would differ between T1 and T2.
As shown in Table 3, Desire for Thinness was seen to have the strongest correlation with the AQ at T2.At T1, DASS scores (model 1) were entered at the rst step and explained 24 per cent of variance, with the inclusion of AQ scores (model 2) accounting for an additional 15 per cent.In the nal model, DASS was not a signi cant predictor, whereas AQ was.Similar ndings were revealed at T2.DASS scores (model 1) were entered at the rst step and accounted for 20 per cent of variance, with the inclusion of AQ scores (model 2) accounting for an additional 14 per cent of variance.In the nal model, DASS was not a signi cant predictor, whereas AQ was.
As shown in Table 3, Body Dissatisfaction was seen to have the strongest correlation with AQ scores at T1.
DASS scores (model 1) entered at the rst step accounted for 30 per cent of the variance in EDI scores, with the inclusion of AQ scores (model 2) accounting for an additional 20 per cent of the variance and leading to a signi cant increase in r 2 .AQ and DASS scores entered as signi cant predictors.At T2, DASS scores (model 1) were entered at the rst step and accounted for 41 per cent of variance.However, the inclusion of AQ scores (model 2) accounted for 3 per cent of additional variance, leading to no signi cant increase in r 2 .AQ scores did not enter as a signi cant predictor.
For Bulimia at T1, DASS scores (model 1) were entered at the rst step and explained 24 per cent of variance.
The inclusion of AQ scores (model 2) accounted for an additional 15 per cent of variance, meaning in the nal model, DASS was not a signi cant predictor, whereas AQ was.At T2, DASS scores (model 1) were entered at the rst step and accounted for 5 per cent of variance in the dependent variable.While the inclusion of AQ scores (model 2) accounted for an additional 10 per cent of variance, neither DASS nor AQ scores were signi cant predictors in the nal model.

Summary of ndings
Previous research has demonstrated that clinical AN and ASD share cognitive and behavioural similarities (Zucker et al., 2007;Treasure, 2013), with 50 per cent of those with AN exhibiting a broader autism Reinvention: an International Journal of Undergraduate Research 16:2 (2023) phenotype on the AQ (Baron-Cohen et al., 2013).The primary purpose of the present study was to expand upon these ndings by investigating whether an association exists between disordered-eating attitudes and autistic traits in a non-clinical population, and if so, whether this association would strengthen or weaken in response to a naturally occurring state of examination stress.Our ndings support the existence of this relationship, as levels of autistic traits during stress entered as a signi cant predictor of the change in disordered-eating attitudes, and were found to be positively correlated with eating attitudes on two distinct occasions.As this relationship was greater in magnitude during a state condition whereby the stress of participants was elevated, this supports the hypothesis that, in a non-clinical population, the relationship between disordered-eating attitudes and autistic traits is dependent upon the state of the individual as opposed to a lasting trait.
Firstly, based upon previous literature that indicates clinical AN and ASD may share some underlying endophenotype (Westwood and Tchanturia, 2017;Baron-Cohen et al., 2013;Zucker et al., 2007), we predicted that the relationship between the two disorders could be extended to a non-clinical sample of students.In line with Coombs et al.'s (2011) research linking ED symptomatology and autistic traits in a student population, we found that if participants scored higher on a measure of disordered-eating attitudes, they also scored higher on a test of autistic traits.
However, it currently remains unknown as to whether the existence of heightened autistic traits in individuals with anorexia ameliorate following recovery, or whether they are dependent upon their current ill state.In clinical samples, the relationship between ASD and AN is pervasive, yet can be exasperated through extreme starvation (Hiller and Pellicano, 2013), with studies revealing negative correlations between body mass and autistic traits (Zakzanis et al., 2010;Bolton et al., 2014).While this implies that the cognitive dif culties experienced by the individual should normalise upon weight restoration, contrasting studies have found that ToM dif culties and weak central coherence persist in individuals with anorexia post recovery (Oldershaw et al., 2010;Nilsson et al., 1999).Our study aimed to contribute towards knowledge in this area by exploring how the relationship between disordered-eating attitudes and autistic traits would differ during a state of elevated stress and when individuals were exhibiting their natural traits alone.Our analyses revealed that the correlation between these two constructs was considerably stronger when participants were experiencing exam stress, suggesting that this association, in a non-clinical sample, is dependent upon a psychological state.Additionally, scores on the AQ were found to signi cantly predict scores on the EDI, independent of negative affect, during the state condition alone, thus supporting the notion that the aforementioned association may be heightened by stress as opposed to being a permanent characteristic of the individual.
Expanding upon this, a regression analysis revealed that the level of autistic traits exhibited by participants when in a state of elevated stress was a signi cant predictor of lower disordered-eating attitudes when participants were not in this state.This means that the higher an individual scored on the AQ at T1, the lower they scored on the EDI at T2.This predictive ability also held even when controlling for original levels of disordered-eating attitudes and negative affect.These ndings support the hypothesis that the relationship between disordered-eating attitudes and autistic traits is dependent upon the state of the individual, as stress-induced autistic traits were found to be a driving force in the changes in levels of disordered-eating attitudes.However, with no previous research exploring this notion, these ndings are tentative and must be regarded with caution.
Reinvention: an International Journal of Undergraduate Research 16:2 (2023) While the main investigation described above supports ndings from previous literature, our exploration into the effects of cognitive in exibility on disordered-eating attitudes and autistic traits appeared to be antithetical.Research has demonstrated that autistic individuals consistently perform poorly on empirical tests of set-shifting, including the Wisconsin Card Sorting Test (Ozonoff and Jensen, 1999;Prior and Hoffmann, 1990) and the Intradimensional-Extradimensional Shift Task (Hughes, Russell and Robbins, 1994).Similarly, the Brixton test itself has been used to con rm set-shifting dif culties in individuals with anorexia (Tchanturia et al., 2011;Tchanturia et al., 2004).These studies strongly indicate that there is a shared de cit regarding cognitive in exibility between both spectrums of disorders.However, no correlation was found on any occasion between disordered-eating attitudes or autistic traits and set-shifting ability.
There are two possible accounts for this unexpected nding.Firstly, a previous study conducted by Murphy (2011) also failed to nd a signi cant correlation between the AQ and the Brixton task.Although the Wisconsin and Intra-Extradimensional tasks have been shown to adequately measure set-shifting dif culties in autistic samples, it is possible to hypothesise that, due to its inability to replicate such results, the Brixton may not be an adequate measure of cognitive in exibility in autistic-trait individuals.Secondly, Coombs et al.'s (2011) study exploring the relationship between eating-disorder symptomatology and autistic traits in a non-clinical sample found that the sub-scales of the AQ that represented set-shifting were not wellcorrelated with scores on the EAT-26.This led the authors to hypothesise that the association between impaired set-shifting and disordered-eating behaviour did not emerge due to the use of a non-clinical sample.It could therefore be proposed that the relationship between eating attitudes, autistic traits and cognitive in exibility may only emerge in severe cases of AN and ASD.
The nal purpose of the present study was to examine how speci c aspects of eating-disorder symptomatology, as measured by sub-scales of the EDI, were related to autistic traits and change in response to stress.This is of particular interest as, although the relationship between AN and ASD has been established, limited research has been conducted into the relationship between unique aspects of disordered-eating attitudes and ASD.Firstly, bulimic individuals have been shown to score higher than controls on the AQ (Gesi et al., 2017), with Lopez et al. ( 2009) also demonstrating that they exhibit weak central coherence and dif culties regarding set-shifting on a battery of tests.However, our ndings revealed that the bulimia sub-scale had the weakest correlation with autistic traits on both occasions and held little predictive value.In reference to body dissatisfaction, there is support for the notion that individuals with ASD may be at an increased risk to negatively experience and evaluate their bodies (Krumm, Ferraro and Ingvalson, 2017), implying that scores on this construct should positively correlate with AQ scores.This is supported by our results, as body dissatisfaction was found to show the strongest correlation to autistic traits when participants were stressed.However, with limited research existing, these ndings can only be applied to our speci c sample of university students, who have been found to exhibit higher levels of body dissatisfaction in general (Ansari et al., 2014).

Limitations
The limitations associated with the present study may account for the fact that, although some statistical tests were approaching signi cance, they failed to reach it at a .05alpha level.Without signi cance, the observed relationships cannot be considered highly credible.However, as our analyses did reveal sizeable differences between testing periods, it is possible to argue that the use of such a small sample greatly increased the margin of error, whereas a larger sample size would narrow the distribution of the test statistic Reinvention: an International Journal of Undergraduate Research 16:2 ( 2023) and increase its ability to detect an effect.Further, the differences in scores could also be accounted for by practice effects.
Similarly, using a small non-clinical sample limits the spread of results on the EDI and AQ.As few individuals presented high scores on these measures, the results bear little application to those with a diagnosis of AN, and cannot be generalised to a clinical eld.While it is important that university students are studied as they are at an increased risk for the development of eating disorders (Hoek and van Hoeken, 2003), the use of a student sample does not render great application.
Finally, this was a self-report study, relying on participants to accurately detail their emotions retrospectively.While the AQ, EDI and DASS are valid and reliable measures of their respective constructs (see Methodology for alpha levels), self-report scales can be easily in uenced by social desirability and retrospective reporting bias.Similarly, asking individuals with heightened cognitive dif culties to report on their own attitudes can also become concerning as this may impair their self-awareness.Further, Likert scales, an ordinal psychometric measure of attitudes and behaviour, do not re ect the multi-dimensional nature of emotional states, and thus the distance between ratings may not accurately represent reality.

Implications and future research
Firstly, replicating this study using a larger number of participants may increase the effect size and enable the current ndings of this study to reach the point of statistical signi cance.This should therefore be the primary focus of future research.Secondly, replicating this study using a clinical sample could have great implications regarding the diagnostic process and treatment of both AN and ASD.Should it be found that elevated autistic traits are a lasting psychological feature in individuals with AN, then treatments could be developed to target these characteristics so as to reduce overall symptoms.In contrast, should this relationship be revealed as state-dependent, preventative measures could be put in place at times when individuals are most vulnerable to the psychological effects of stress.Regardless, the implications of the present study indicate that treatment of AN should also prioritise the cognitive pro le (autistic traits) of patients, not just their behaviour.
Finally, it would be interesting to explore, through longitudinal research, how the state vs trait relationship between AN and ASD differs between males and females.An altered phenotypic pro le of autism for females is supported by 78.8 per cent of studies (Kirkovski, Enticott and Fitzgerald, 2013), and researchers have gone as far as to hypothesise that AN may be the female phenotype of ASD (Treasure, 2013).Recording genderbased differences in research such as this may inform knowledge regarding the differing presentation of ASD in males and females and how the symptomatology presented by both genders may differ in response to stress.

Conclusions
This study has demonstrated that there is a relationship between disordered-eating attitudes and autistic traits in a non-clinical, university student cohort.Further, it was found that this relationship was stronger during a state of elevated stress as opposed to when individuals were presenting their natural traits, and that stress-induced (state) autistic traits were signi cantly predictive of changes in eating disordered attitudes.
However, this relationship was not found to be affected by cognitive in exibility, as measured by the Brixton, which was not related to either disordered-eating behaviour or autistic traits.
Reinvention: an International Journal of Undergraduate Research 16:2 (2023) These ndings suggest that elevated autistic traits in individuals who are exhibiting disordered-eating attitudes may be a re ection of their current state, and may therefore ameliorate when eating-disorder symptomatology does so.Further, they demonstrate that the association found in clinical patients exists in non-clinical populations also.This suggests that exploring these populations can provide critical insight into mechanisms underpinning disordered-eating psychopathology before the clinical onset of anorexia.This project has further highlighted the need for research exploring the exact mechanisms shared between AN and ASD so as to provide insight into more effective diagnosis and treatment interventions.
Participants volunteered their own appointment time between 9 a.m. and 5 p.m. via PsychSONA, with each participant undergoing testing individually in an interview at the University of St. Andrews School of Psychology.After being made aware of the nature of the experiment, subjects signed an informed consent form.They were rst requested to complete the three self-report questionnaires independently -namely, the DASS-21, EDI-2 and the AQ.This took approximately 15 minutes, and the questionnaire order was randomised to minimise potential order effects.The participants then completed the Brixton task.This test took approximately 10 minutes, and the experimenter recorded the responses of the participants by pen and Reinvention: an International Journal of Undergraduate Research 16:2 (2023)

Figure 1 :
Figure 1: Partial regression plot showing the relationship between residual EDI scores and AQ scores after controlling for negative affect at T1.
feel satis ed with the shape of my body A U O S R N 11) I exaggerate or magnify the importance of weight A U O S R N 12) I have gone on eating binges where I have felt that I could not stop A U O S R N 13) I like the shape of my buttocks A U O S R N 14) I am preoccupied with the desire to be thinner A eat moderately in front of others and stuff myself when they're gone A U O S R N 18) If I gain a pound, I worry that I will keep gaining A U O S R N 19) I have the thought of trying to vomit to lose weight A U O S R N 20) I think that my thighs are just the right size A