Reinforcement learning of associative-rules in the absence of awareness

BARTTFELD P, JOURDAN I, YORIO A AND ZANUTTO S

Front. Hum. Neurosci. 2:1. doi:10.3389/neuro.09/001

A. Bunge, University of California Berkeley, USA

Lugar: Berkeley, USA; Año: 2008

Reinforcement learning of associative-rules in the absence of awareness Pablo Barttfeld1,2,*, Iván Jourdan1,2, Alberto A. Yorio2 and B. Silvano Zanutto1,21,2,*, Iván Jourdan1,2, Alberto A. Yorio2 and B. Silvano Zanutto1,2 1. Instituto de Ingeniería Biomédica, Facultad de Ingeniería, Universidad de Buenos Aires, Buenos Aires, Argentina 2. Instituto de Biología y Medicina Experimental, Consejo Nacional de Investigaciones Cientifi cas y Técnicas, Buenos Aires, Argentina Edited by: Silvia A. Bunge, University of California Berkeley, USA Reviewed by: Tobias Egner, Northwestern University, USA Silvia A. Bunge, University of California at Berkeley, USA The mechanisms and bounds of implicit learning are issues of great interest in behavioral and brain sciences. Here we provide the fi rst evidence not only that human subjects can learn arbitrary relationships with subliminally presented stimuli, but additionally that this learning occurs at a level of processing that incorporates semantic information and shows rule-like properties (such as transitivity and symmetry). We assess this by use of an operantly learned match to sample task in which subjects had to explore the relation between simple shape and letter stimuli. Subjects were provided positive feedback for correct responses. However, unbeknownst to them, there was no rule linking the shapes and letters but, instead, rules linked the shapes with subliminally primed words. We found that subjects quickly developed patterns of response times in which correct responses (as determined by the rule linking the subliminal prime and the shapes) were signifi cantly faster than those of incorrect ones. Variations of this protocol suggests that subjects can learn without awareness tasks of semantic access, transitive inference and symmetry; tasks often used to characterize higher order cognition. These results extend implicit learning mechanisms to an area thought to be the exclusive domain of explicit cognition and raise the question of whether all rule learning is performed fi rst implicitly and our explicit knowledge of these rules comes after. Keywords: associative-rules, awareness, masked priming, match to sample