Emotional Inferencing in the Reading Comprehension of Persons with Autism

Background: Persons with Autism Spectrum Disorders (ASD) often have problems in reading comprehension associated with difficulties in inferencing. However, previously published studies provide conflicting data: studies using off-line tasks tend to find differences between the autism and control groups, whereas others using on-line measures do not. In addition, most tasks used in the literature assess local coherence more often than global text comprehension. 

Objectives: With this series of experiments we aimed at exploring if reading difficulties persist when tasks require global coherence, specifically in the construction of a situation model based on the emotional status of the main character in a story. We also wanted to compare the performance of readers with autism on both types of measures (off- and online) to poor comprehenders (PC) and typically developing readers (TD).

Methods: Three equal sized groups of adolescents and young adults with ASD, PC, and TD (N = 66), were matched on chronological age, reading speed, working memory, and gender. They all had a normal nonverbal IQ. There were significant differences in reading comprehension between the control group and the two other groups.

Three experiments were carried out in which it was necessary to construct a situation model that included determining the emotional status of the protagonist. In Experiments 1 and 2 (on-line tasks), participants were presented with stories that included a target phrase that explicitly mentioned an emotion. This sentence could be consistent or inconsistent with the previous text. In Experiment 1, the target appeared immediately after the text necessary to infer the emotion, while in Experiment 2 we included filler sentences between the section of the text describing the implicit emotion and the target. In Experiment 3 (off-line), the emotion was not made explicit in any target sentence.  Participants were asked to rate four emotions according to how well they described the feelings of the main character in the story.

Results: In Experiments 1 and 2, mean reading times (RT) for targets were significantly longer when the explicit emotion was inconsistent with the previous text. However, in the first experiment no interaction of consistency and group was found. In the second, there was an interaction effect, resulting from no differences between consistent and inconsistent target RTs in the PC group. In Experiment 3, there was a significant main effect of group in accuracy scores. Participants with ASD had the lowest accuracy rate of all three groups.  

Conclusions: Although poor comprehenders show difficulties in the production of automatic inferences when these involve increased working memory load, this is not the case for readers with autism. However, ASD readers do show impaired performance in tasks that require the use of information available in the text in order to respond to explicit questions. The results point to difficulties in the comprehension of ASD readers that could be specific and different from poor comprehenders.