The self-reference effect on memory is not diminished in autism: Three studies of incidental and explicit self-referential recognition memory in autistic and neurotypical adults and adolescents.

Procedure and materials

Participants completed implicit and explicit encoding conditions of the trait memory task, always completing the implicit condition first. The stimuli comprised 144 trait words (half of which were psychological traits, such as “intelligent”, and half of which were physical traits, such as “tall”). All items were rated on a 5-point scale for valence by 10 independent adults. The mean ratings across raters were used to split the items into six equal lists of 12 traits (50% psychological, 50% physical), which were balanced for valence. A one-way analysis of variance (ANOVA) show no effect of list or trait type (or interaction between them) on valence, F s ≤ 0.02, p s ≥ .90. These individual lists were then organized into 12 different combinations to produce 12 different versions of the experiment, ensuring that each list of psychological and physical words was present within each of the six possible referent/encoding conditions (self-implicit, other-implicit, lure-implicit, self-explicit, other-explicit, lure-explicit). These 12 versions were used so that stimuli varied in relation to condition across participants.

Before beginning the tasks, participants were shown a picture of a famous person of the same gender and asked to identify them. Correctly identifying their name selected that name as the “other name” to be used during both the trait memory tasks. If the participant incorrectly identified the first person, they were shown a second famous person to identify. Females were first shown a picture of Queen Elizabeth, followed by Theresa May. Males were shown a picture of Prince Charles, followed by David Cameron. For each participant, the experimenter checked that the first name on the consent form was their preferred name. If it was not, the verbally stated preferred name was used. Participants were informed that throughout the experiment they would be seeing either their own name or the name of the correctly identified famous person, along with some other words and would have some simple responses to make in relation to the stimuli.

Each trial began with a name (participant’s or celebrity’s) presented centrally on a computer monitor for 500ms, followed by the presentation of a trait word either to the left or right of the name for 1500ms, with the name remaining on screen throughout. The words then disappeared, and the participant was asked to press the “Z” or “M” key to indicate whether the trait word had appeared to the left or right, respectively, of the central name. Participants were instructed that they did not need to pay attention to either the name or the meaning of the trait word on each trial. Rather, their task was merely to concentrate on the spatial location of the trait word on each trial. Each participant performed 48 trials (fixed pseudorandomized order), 24 with their own name and 24 with the name of the celebrity.

Recognition test phase

Following completion of the encoding/study phase, participants completed the surprise recognition test phase. On each trial, participants were presented with a trait word along with the question, “Did you see this word during the study phase?” and a “yes” box and a “no” box. Participants had unlimited time to click in either response box to indicate whether or not they recognized the word from the study phase. Each participant performed 72 trials (fixed pseudorandomized order), 48 of which included a previously studied/old trait presented during the study phase and 24 of which included a previously unseen lure/new item. The fixed pseudorandomized order ensured that no more than 3 items from any category (self, other, lure) appeared in a row.

Recognition test phase

The procedure for the recognition test phase in the explicit condition was identical to the procedure used in the implicit condition.

Variables and scoring

d′ (d-prime) scores were calculated as a standard measure of recognition memory accuracy, using the formula, d ′ = z (HR) – z (FAR). Here, HR refers to hit rate (proportion of old item correctly identified as old); FA refers to false alarm rate (proportion of new items incorrectly identified as old); and z refers to z transformation. Higher d ′ scores indicate greater discrimination between “old” and “new” items, and hence greater recognition memory accuracy. As a measure of the size of the self-reference effect, self-bias scores (self-other difference scores) were calculated by subtracting average other/celebrity trial d ′ scores from average self trial d ′ scores.

Data analysis

An increasingly used supplement to null hypothesis significance testing in general is to calculate a Bayes factor for each key analysis. Bayesian analyses provide an estimation of the relative strength of a finding for one hypothesis over another (i.e., the alternative hypothesis over the null, or vice versa), which allows a more graded interpretation of the data than is possible using p values or effect sizes alone (e.g., Dienes, 2014 ; Rouder, Speckman, Sun, Morey, & Iverson, 2009 ). According to Jeffreys’ (1961) criteria, Bayes factors (BF 10 ) > 3 provide firm evidence for the alternative hypothesis (with values >10, >30, and >100 providing strong, very strong, and decisive evidence, respectively) and values under 1 provide evidence for the null (with values <0.33 providing firm evidence). BF 10 values can be considered to reflect the likelihood that the alternative hypothesis is more likely to be true than the null hypothesis. Hence, a BF 10 of 3 suggests the alternative hypothesis is three times more likely to be true than the null hypothesis. Bayesian analyses were conducted using JASP 0.8.1 ( JASP Team, 2016 ).

Association analyses

Kendall’s tau correlations were carried out to explore the relationships among AQ scores, implicit self-bias scores (self d ′ minus other d ′) and explicit self-bias scores. The correlation matrix is reported in Table 1 and accompanying scatterplots can be found in Supplementary Figure S1 in the online supplemental materials. All correlations were small and not statistically significant, with Bayes factors indicating firm support for the null hypothesis.

Discussion

As expected, NT individuals showed a typical self-reference effect (self-bias) in memory, recognizing trait words processed in relation to self significantly more reliably than trait words processed in relation to others in both the implicit and explicit conditions of the trait memory task. In keeping with our a priori prediction, AQ score was not significantly associated with the size of the self-bias in the implicit condition. Contrary to our prediction, however, AQ score was not significantly associated with the size of the self-bias in the explicit condition either. There was sufficient statistical power to detect even small associations if they existed, and Bayesian analyses consistently suggested that the data provided firm support for the null hypotheses in association analyses. Of course, the majority of the sample was female and the participants tested had a relatively narrow age range. A sample more representative of the general population would have been advantageous and increased confidence in the generalizability of results. However, post hoc analyses (see Footnote 3 ) showed that sex/gender did not influence the size of the self-bias or size of the correlation with AQ in Experiment 1 and so it is not clear that a sample with a higher proportion of males would have changed the results.

The current results are out of keeping with those of Henderson et al. (2009) who observed a significant association between number of ASD traits (measured using the Autism Spectrum Screening Questionnaire; ASSQ; Ehlers, Gillberg, & Wing, 1999 ) and the size of the self-bias on an explicit trait memory paradigm that was very similar to the explicit condition used in the current study. Of course, Henderson et al.’s study used a case-control design and observed a significant association between ASSQ score and size of self-bias (after controlling for group differences in the size of the self-bias) among their 59 participants (note that groups were collapsed on the basis that a Group × Self-Bias interaction effect emerged in their regression analysis, although the statistics associated with the interaction effect were not reported).

Given that ASD features are likely to be distributed continuously throughout the general population (e.g., Frazier et al., 2014 ), studying individual differences in ASD traits and their relation to psychological abilities in the NT population can make an important contribution to our understanding of ASD itself. However, there can still be qualitative differences in the mechanisms/processes that underpin those traits in each population (e.g., Mandy et al., 2012 ; Peterson, Wellman, & Liu, 2005 ). As such, a full understanding requires the study of diagnosed cases, as well as traits in the NT population. Thus, even though we found no evidence of an association between ASD traits and size of the self-bias in individuals from the general population in Experiment 1, it does not rule out the possibility that between-groups differences in the size of the self-bias (and/or significant AQ × size of self-bias associations) would emerge in a case-control study. Therefore, in Experiment 2, we gave the same implicit and explicit conditions of the trait memory task used in Experiment 1 to a group of adults with a full diagnosis of ASD, as well as a closely matched group of NT comparison adults. At the outset of the study, we predicted that the size of the self-bias would be significantly smaller among ASD participants than comparison participants in the explicit condition only.

Procedure and materials

Participants from each group completed each condition (implicit/explicit) of the trait memory paradigm used in Experiment 1.