|
| Contemporary Behavioral and Neuroscience Perspectives on Transitive Inference, Relational Reasoning and Instructional Control |
| Tuesday, May 26, 2009 |
| 10:30 AM–11:50 AM |
| North 228 |
| Area: EAB; Domain: Experimental Analysis |
| Chair: Simon Dymond (Swansea University) |
| CE Instructor: Mandy J. Rispoli, Ph.D. |
| Abstract: This symposium brings together leading researchers from behavior analysis, neuroscience and cognitive science to present their work on relational reasoning, transitive inference and neurocomputational investigations of instructional control. The four papers each address a specific topic from these dynamic, multi-disciplinary research areas. The first presentation provides a critical review of nonhuman research on transitive inference, which, it is argued, is best explained in terms of reinforcement history. The second presentation describes the findings of a brain imaging study conducted with a novel paradigm drawn from research on derived relational responding and relational frame theory that was designed to examine human transitive inference-like performance. The third presentation describes a series of studies aimed at establishing patterns of relational responding in accordance with derived comparative and opposition relations through multiple exemplar training. The final presentation describes the findings of a neurocomputational study on the effects of instructional control on human probabilistic reinforcement learning. |
| |
| Transitive Inference Without the “Inference” in Non-Human Animals |
| MARCO VASCONCELOS (University of Oxford) |
| Abstract: Research with non-human animals shows that learning a relatively small number of partially overlapping simultaneous discriminations can immediately lead to other novel and apparently transitive discriminations. These findings have prompted a flourishing empirical and theoretical search for the mechanism(s) mediating this ability. I will critically analyze the most prominent models proposed to explain transitive-like behavior in non-human animals. Some models are cognitive, proposing for instance that animals use the rules of formal logic or form mental representations of the premises to solve the task; others models appeal to reinforcement mechanisms to explain such behavior. I will argue that transitive inference in non-human animals is best considered as a property of reinforcement history rather than of inferential processes. |
| |
| fMRI Study of Relational Reasoning with Derived Comparative Relations |
| SIMON DYMOND (Swansea University), Elanor Hinton (Cardiff University), Ulrich von Hecker (Cardiff University), Anita Munnelly (Swansea University) |
| Abstract: Considered a hallmark of human reasoning, “transitive inference” is typically studied either with procedures that train overlapping simultaneous discriminations or that present premise pairs based on pre-existing stimulus relations. Contextually controlled derived comparative relations (more than/less than) may provide a model of the behavioral processes involved in this complex behavior. The present study describes the findings of two experiments designed to test this relational reasoning model by synthesising procedures from research on relational frame theory with behavioral neuroscience research on “transitive inference”. First, a behavioral study compared the effects of two training schedules on subsequent novel performance. Next, the neural correlates of this behavior were examined with fMRI. Results demonstrated no differences between training schedules on subsequent novel probe performance, but an overall improvement in accuracy and decrease in response latencies from trained to tested relations in both groups. Imaging findings broadly supported those of previous studies. Hippocampal activation was correlated with accuracy on some test trial-types, and activity in PFC and parietal cortex showed the same trend as the behavioral data (i.e., ‘distance effect’). Implications of the relational reasoning model for behavior-analytic accounts of complex human behavior are discussed. |
| |
| Acquisition and fluency of arbitrarily applicable derived relational responding in accordance with opposition and comparison contexts |
| ROSA MARÁA VIZCAÁNO (University of Almeria), Carmen Luciano Soriano (University Almer&íacute;a, Spain), Vanessa SÁnchez (University of Almeria), Francisco José Ruiz Jiménez (University of Almer&íacute;a) |
| Abstract: The aims of the present study were twofold. On the one hand, to implement a brief multiple-exemplar-training (MET) to establish derived responding according to opposition and comparison. On the other hand, to show the process to establish fluency and flexibility across several relational responding. The study was conducted with a single four-year-old child whose language and cognitive abilities were evaluated before and after the implementation of MET. The process began with the evaluation of derived relational responding according to coordination. Secondly, brief MET involving different dimensions was implemented to establish derived relational responding according to the opposition contextual cue which was followed by a similar multi-dimension but brief MET to establish derived relational responding according to comparison. In addition across these phases, fluency and flexibility was promoted using new examples for arbitrary application contextual responding. Results showed the emergence of complex relational specific patterns that involved the transformation of functions across many examples and different contextual cues. Results are discussed in terms of the novelty and effective of this procedure to establish not only the contextual relational responding of opposition and comparison but, most importantly, to promote fluency and flexibility. |
| |
| Instructional control over reinforcement learning: Behavioral and neurocomputational investigations |
| BRADLEY DOLL (University of Arizona), Michael J Frank (University of Arizona) |
| Abstract: Humans learn how to behave directly through environmental experience and indirectly through rules and instructions. Research has shown that instructions can control behavior, even when such behavior leads to sub-optimal outcomes (Hayes, 1989). We examine the control of behavior through instructions in a reinforcement-learning task known to depend on striatal dopaminergic function. Participants selected between probabilistically reinforced stimuli, and were (incorrectly) told that a specific stimulus had the highest reinforcement probability. Despite experience to the contrary, instructions drove choice behavior. We present neural network simulations that capture the interactions between instruction-driven and reinforcement-driven behavior via two potential neural circuits: one in which the striatum is inaccurately trained by instruction representations coming from prefrontal cortex/ hippocampus (PFC/HC), and another in which the striatum learns the environmentally based reinforcement contingencies, but is "overridden" at decision output. We attempt to distinguish between the proposed computational mechanisms governing instructed behavior by fitting a series of abstract "Q-learning" and Bayesian models to subject data. The best-fitting models support the network model in which the PFC/HC system trains the striatal reinforcement system. |
|
| |