The Self-Reference Effect on Perception: Undiminished in Adults with Autism and No Relation to Autism Traits.

Background: It is widely agreed that the self plays an important role in human cognition and perception, exerting influence across a range of domains and situations. The clearest empirical demonstration of this influence is the “self-reference effect” (SRE; Rogers et al., 1977), whereby information encoded in relation to the self has a mnemonic advantage over information encoded in other ways. This effect is apparent in a number of different paradigms and across different domains of processing. This SRE in memory appears diminished in ASD and related to the number of ASD traits manifested by neurotypical individuals (the fewer ASD traits, the larger the SRE). Here, we report the first experiments exploring the relation between ASD and the SRE in perception.

Objectives: To investigate the role of self-reference on perception in ASD.

Methods: Using a “Shapes” Task (Sui et al., 2012), participants learned to associate three different shapes (triangle, circle, square) with three different labels representing self, a familiar other, or an unfamiliar other (e.g., “you”, “mother”, “stranger”). Participants then completed trials during which they were presented with one shape and one label for 100ms, and made judgements about whether each shape-label pairing matched the learned contingency. Participants also completed measures of mentalizing (Reading the Mind in the Eyes and Animations tasks) and ASD traits/severity (Autism-spectrum Quotient). Bayesian analyses were employed to complement null hypothesis significance testing (BF₁₀ < 1 = evidence for the null; BF₁₀ > 3 = evidence for the alternative hypothesis).

Results: In Experiment 1, neurotypical participants (n=124) showed the expected SRE, detecting self-related matches more reliably and quickly than matches involving the familiar or unfamiliar other (all ps<.001, ds>0.97, BF₁₀>100). Most important, number of ASD traits was unrelated to the size of the SRE for either accuracy or RT (all rs <.10, ps >.27, BF₁₀ <0.21). Hence, Bayesian association analyses strongly supported the null hypothesis. In Experiment 2, there were no differences between 22 adults with ASD and 21 age-, sex, and IQ-matched comparison adults in performance on the Shapes Task (see Figure 2). Despite showing large and significant mentalizing impairments (all ps <.01, ds >0.80, BF₁₀ >23.00), participants with ASD showed the typical SRE for accuracy and RT, and there were no between-group differences in this respect (ps >.66, ds <0.13, BF₁₀ <0.33). Moreover, there were no associations with ASD traits (measured using AQ or ADOS) in either group. Bayesian analyses favoured the null hypothesis.

Conclusions: These findings suggest that self-representation influences perception in a normal way in ASD. Considered alongside previous findings, the current results are relevant for theories about self-representation in ASD. Our interpretation is that first-order representations of self are typical in ASD and available to bias perception/cognition in a typical manner, contrary to the “experiencing-I” theory espoused by Millward et al. (2000). In contrast, we suggest that second-order (meta-) representations of self are impoverished among people with ASD (in keeping with the “absent self” theory; Frith, 2003), which leads to a diminished SRE on tasks that require reflections on/evaluations of self.