Behavioral Flexibility and the Effect of Various Feedback Types: A Developmental Study

Friday, May 12, 2017: 5:00 PM-6:30 PM
Golden Gate Ballroom (Marriott Marquis Hotel)
E. Oberwelland Weiss1,2, J. A. Kruppa3,4, G. R. Fink2,5, B. Herpertz-Dahlmann1, K. Konrad1,2 and M. Schulte-Ruther2,3, (1)Child and Adolescent Psychiatry, University Hospital Aachen, Aachen, Germany, (2)Institute of Neuroscience and Medicine (INM-3), Research Center Jülich, Jülich, Germany, (3)Department of Child and Adolescent Psychiatry, Psychosomatics and Psychotherapy, University Hospital RWTH Aachen, Aachen, Germany, (4)Cognitive Neuroscience, Institute of Neuroscience and Medicine (INM-3), Research Center Jülich, Jülich, Germany, (5)Neurology, University Hospital Cologne, Cologne, Germany

In an ever-changing environment it is essential to shift strategies or adapt response patterns based on obtained feedback. Such behavioral flexibility is linked to executive functioning and cognitive control, but it has also been shown to have implications for social interactions. Accordingly it has also been linked to two core deficits of autism spectrum disorder (ASD): (1) deficits in social interactions and communication and (2) restricted, repetitive behaviors. Recent studies have used probabilistic reversal learning tasks to examine behavioral flexibility in adolescents and adults. By means of computational modeling (e.g., Hierarchical Bayesian learning models), the underlying processes could be examined more systematically and in depth1,2. However, developmental data is still scarce.


In our present study we aim to investigate behavioral flexibility in children and adolescents with and without ASD and systemically examine the effect of various feedback types (i.e. social, individualized and control feedback).


The study is still ongoing. Until now 19 typically developing (TD) children (8 and 12 years of age), 14 TD adolescents (13 – 18 years of age) and 12 individuals with ASD (8 and 18 years of age) completed three runs of a probabilistic reversal learning task with either social, individualized or control feedback.


First, TD children needed more trials to reach the learning criteria (i.e., three consecutive correct trials) compared to TD adolescents. Second, TD adolescents profited more from control feedback than TD children. Third, TD adolescents made more preservative errors (i.e., an incorrect trial after reversal whereby participants have still chosen the previously reinforced response before they at least once chose the new/correct target) than TD children, whereas TD children made more regressive errors (i.e. an incorrect trial after reversal whereby participants choose the previously reinforced response after having already chosen the new/correct target at least once) than TD adolescents. Both type of errors are also highly negatively correlated. This might indicate that children are more susceptible to “false feedback” (which inevitably occurs due to the probabilistic task), and consequently may adapt their response choices more immediately and thus be less systematic than adolescents. The learning behavior of individuals with ASD was comparable to age- and gender-matched TD individuals, but children with ASD rated social feedback as less rewarding than their TD peers. In a next step, we aim to implement computational modeling of the behavioral data to pinpoint individual learning strategies. Additionally, the model parameters will be compared to various background variables (e.g., IQ, autistic traits, attention spam). This again would allow for a more systematic comparison of changes in behavioral flexibility during the course of development as well as its effects in ASD.

Conclusions:  N/A