Transitive Inference in Children with Autism Spectrum Disorder and Limited Verbal Ability

Background:  

Memory functioning in Autism Spectrum Disorder (ASD) follows a characteristic pattern, including good rote memory (Kanner, 1943) and cued recall (Bowler, Matthews & Gardiner (1997).  Impairments have been demonstrated in the free recall of semantically related items (Tager-Flusberg, 1991), and the recognition of combinations of features (Bowler, Gardiner & Gaigg, 2008).  This suggests a difficulty with relational binding - the ability to encode items and events, and the relationships between them, to allow for adaptive use of the information (Opitz, 2010).

Objectives:  

Studies of relational binding have mainly been carried out with verbally able adults with ASD, using verbalisable stimuli (e.g. lists of words.  The current study aims to extend these findings to lower-functioning individuals with ASD.  A non-verbal, transitive inference paradigm is adapted (Maclean, 2008); (if A > B, and B > C, then A > C).  Typical adults tend to use a relational strategy during these tasks (Moses, Villate, and Ryan, 2006), whereas non-human animals tend to use an associative, pair-by-pair strategy (von Fersen et al, 1991), suggesting that language is important in the construction of relational strategies.

Methods:  

20 children with ASD and limited verbal ability, and 20 typically developing children matched on non-verbal IQ.  Replicating Maclean et al’s paradigm, pairs of stimuli are presented in hierarchical order, with participants required to find the “higher-ranked” of the two by touching it onscreen; feedback is given regarding their choice.  A test phase is then presented, comprising two trial types: adjacent pairs of stimuli from the training phase, and re-pairings of the stimuli into novel combinations of non-adjacent stimuli (e.g. B-D).  Participants must use information from the adjacent training pairs to infer which is the higher-ranked stimulus of each non-adjacent pair, which requires the flexible processing of the relations between pairs.  The current study predicts no group differences in the training phase or in the test phase where adjacent stimuli are presented, whereas participants with ASD will be impaired in the non-adjacent pairs, due to the flexible relational binding required to solve those trials.

Results:  

Preliminary data show no group differences throughout the training blocks, although a main effect of training block was found, F(2,8) = 5.03, p = .04, η_p² = .56, indicating a decrease in performance as the blocks become more complex, although performance was > 70% for all participants throughout training.  During the test phase no difference in performance was found between the two trial types (adjacent and non-adjacent); there was a significant group difference, F(1,4) = 10.0, p = .03, η_p² = .71, indicating that the typically-developing participants performed significantly better overall, regardless of trial type.

Conclusions:  

Results indicate that, during the test phase, children with ASD and limited verbal ability perform significantly worse, despite having performed above chance during training.  This in part confirms the idea that language ability is important in this type of task; the fact that performance in the ASD group was worse overall during the test phase will also be discussed in light of theoretical perspectives on transitive inference.