![[ Visit Client Website ]](images/banner.JPG)
Friday, May 8, 2009
Northwest Hall (Chicago Hilton)
2:30 PM
Background: Many symptoms that are pathognomic of autism spectrum disorders (ASDs) can be considered to reflect deficits in learning. Although the majority of empirically supported autism interventions are premised on learning theory, there has been little experimental research about this clinically significant topic. One major learning problem for individuals with ASDs is the failure to “generalize” or to transfer past learning to new situations.
Objectives: To examine performance of adults with ASDs on two reinforcement learning tasks that assess processes involved in generalization. In the first, a probabilistic selection (PS) task, three stimulus pairs, AB, CD, and EF were presented in random order. Participants learned to choose one of the two stimuli based on probabilistic feedback following each trial that was accurate 80%, 70%, and 60% of the time. In the second, a transitive inference (TI) task, participants were trained on a partially-overlapping stimulus hierarchy with four pairs: A+B-, B+C-, C+D-, and D+E-. During a subsequent test phase, the novel combinations BD and AE were tested.
Methods: Twenty five participants each were enrolled in the ASD and typically developing (TYP) groups. Sixteen adults with ASDs (Mean age = 22.7) and 17 TYP (Mean age = 23.8) could complete the PS, and 18 (Mean age = 22.4) ASDs and 23 TYP (Mean age = 22.7) could complete the TI. All participants had Full Scale IQs of > 70 on the Wechsler Abbreviated Scales of Intelligence. Participants with ASDs had prior diagnoses and met criteria for ASD on ADOS-G (Lord et al., 2000), and a DSM-IV-R checklist.
Results: Early in probabilistic learning, individuals with ASDs were unimpaired at acquiring the simplest AB stimulus pair, however, they exhibited slower learning for the more difficult CD pair, and deficits using positive feedback. Interestingly, ASDs performed significantly better than TYP on the DE stimulus pair for which accurate feedback was rarely provided. After training on a TI task, individuals with ASDs learned simple stimulus response associations comparably to TYP and exhibited a U-shaped learning curve (better performance on outer pairs compared to inner pairs), which signals efficient use of both associative and elemental learning strategies. At test, there were no significant differences in performance on the novel BD pair, although performance on the AE pair, was significantly worse in the ASD group.
Conclusions: Although, ASDs performed comparably to TYP after training, early in learning they exhibited a pattern of relative impairment and relative strength on the PS task. On TI, they learned the training pairs comparably to TYP and were able to complete the BD pair. Frank et al. (2004, 2005, 2006) have proposed a systems-level computational model of reinforcement learning that focuses on interactions between the basal ganglia and the prefrontal cortex. Taken in the context of this computational model, our findings suggest intact functioning of sub-cortical structures, with deficits in frontal functioning and/or fronto-striatal connectivity. Treatment implications also will be discussed.
Objectives: To examine performance of adults with ASDs on two reinforcement learning tasks that assess processes involved in generalization. In the first, a probabilistic selection (PS) task, three stimulus pairs, AB, CD, and EF were presented in random order. Participants learned to choose one of the two stimuli based on probabilistic feedback following each trial that was accurate 80%, 70%, and 60% of the time. In the second, a transitive inference (TI) task, participants were trained on a partially-overlapping stimulus hierarchy with four pairs: A+B-, B+C-, C+D-, and D+E-. During a subsequent test phase, the novel combinations BD and AE were tested.
Methods: Twenty five participants each were enrolled in the ASD and typically developing (TYP) groups. Sixteen adults with ASDs (Mean age = 22.7) and 17 TYP (Mean age = 23.8) could complete the PS, and 18 (Mean age = 22.4) ASDs and 23 TYP (Mean age = 22.7) could complete the TI. All participants had Full Scale IQs of > 70 on the Wechsler Abbreviated Scales of Intelligence. Participants with ASDs had prior diagnoses and met criteria for ASD on ADOS-G (Lord et al., 2000), and a DSM-IV-R checklist.
Results: Early in probabilistic learning, individuals with ASDs were unimpaired at acquiring the simplest AB stimulus pair, however, they exhibited slower learning for the more difficult CD pair, and deficits using positive feedback. Interestingly, ASDs performed significantly better than TYP on the DE stimulus pair for which accurate feedback was rarely provided. After training on a TI task, individuals with ASDs learned simple stimulus response associations comparably to TYP and exhibited a U-shaped learning curve (better performance on outer pairs compared to inner pairs), which signals efficient use of both associative and elemental learning strategies. At test, there were no significant differences in performance on the novel BD pair, although performance on the AE pair, was significantly worse in the ASD group.
Conclusions: Although, ASDs performed comparably to TYP after training, early in learning they exhibited a pattern of relative impairment and relative strength on the PS task. On TI, they learned the training pairs comparably to TYP and were able to complete the BD pair. Frank et al. (2004, 2005, 2006) have proposed a systems-level computational model of reinforcement learning that focuses on interactions between the basal ganglia and the prefrontal cortex. Taken in the context of this computational model, our findings suggest intact functioning of sub-cortical structures, with deficits in frontal functioning and/or fronto-striatal connectivity. Treatment implications also will be discussed.