Transitive Inference Learning in Children and Adolescents with ASD

Background : Individuals with ASD exhibit impairments in generalizing (transferring) learning to new situations. Transitive inference – learning a series of ordered stimulus pairs (AB, BC, CD, EF where A>B>C>D>E), and then transferring this learning about order to novel pairs (AC, AD, AE, BD, BE, CE) -- is a form of generalization. Previously, we examined TI in young adults with ASD and TYP (Solomon, Frank, Smith, Ly, & Carter, 2011), and found that ASD used a strategy involving rote memorization, whereas TYP employed a more flexible strategy using information about the reinforcement value of the hierarchy end-items. Due to the use of a rote memory strategy, the ASD group performed comparably to TYP on the harder BD pair, but worse on the easier AE pair—essentially “missing the forest for the trees” in grasping that A is always right while E never is.

 Objectives : We extend the investigation of this pattern of findings to children and adolescents by examining their strategy use; its cognitive correlates; and its relationship to school achievement and age.  Given that problems with generalizing learning negatively impact the academic performance of those with ASD, this investigation holds the potential to inform the development of important and widely-implemented remediation strategies.

Methods : Participants included 23 high functioning children and adolescents ages 8-16 with ASD, qualified using community diagnosis plus ASSQ, SCQ, and DSM-IV criteria, and 23 age, IQ, and gender matched children with TYP. They completed a TI task previously used in TYP children (Townsend, Richmond, Vogel-Farley, & Thomas, 2010), which included training on four stimulus pairs of colored ovals, with subsequent testing on novel pairs. WASI and WIAT-II, were used to assess cognitive abilities and academic performance.

 Results : Replicating prior results, during training, a 2x2 ANOVA demonstrated the ASD group used more of a rote memory strategy as evidenced by a significant group by pair type interaction, with the ASD group performing worse on outer pairs (F(1, 43) = 4.85, p = .043). At test, the ASD group showed intact performance on the BD inference pair, but significantly worse performance on the BE (t(48) = 2.29, p = .03), CE (t (43) = 2.82, p = .03), and AE pair at a trend (t(43) = 1.9, p = .06). For both groups, using a reinforcement learning strategy was positively associated with Verbal IQ and age, while using a rote memory strategy, was associated with only Performance IQ.  In both groups, training trial accuracy was positively related to Math Problem Solving (r = .529, p =.001).

 Conclusions: Most of the high functioning children with ASD in this study used a TI problem solving strategy that was sub-optimal and relied more on visuo-spatial versus verbal information. Furthermore, performance relying on this strategy did improve with age. TI strategy may be an important individual difference variable to consider when teaching math and other subjects to children and adolescents with ASD, most of whom employ a visuo-spatial, versus a verbal problem solving approach, relative to TYP.