THE INFLUENCE OF INSTRUCTION AND TASK COM PLEXITY ON TRANSITIVE DECISIONS

Transitive inference is the ability to draw a conclusion B>D from the overlapping premises A >B, B >C, C >D and D >E of a series ABCDE. For a nonverbal presentation premise pairs are converted into simultaneous discriminations (A+B-, B+C-, C+D- and D+E- where a '+' stands for rew ard and a '-' stands for penalization). Transitive responding is demon strated if in test pair ED stimulus B is preferred over D. In two expe riments transitive responding in adults was investigated using a 5-ter m series of geometric figures presented in a nonverbal task in order t o gain information about the individual strategies. In the first exper iment uninformed and informed subjects learned the pairs A+B, B+C-, CD and D+E-. After a binary transitivity test (ED) they were confronted with a triadic test, where they had to transfer their knowledge to gr oups of three items instead of two (e. g. BCD, CDE). Eleven out of 20 uninformed subjects solved the transitivity task above chance level an d showed a transfer to the subsequent triadic test. Only 4 of these 11 subjects could describe explicitly how they had managed to solve the binary test. The remaining 7 solvers could not decribe their strategy and had obviously solved the task implicitly. Informed subjects requir ed less trials to criterion, but their performance was comparable with that of the uninformed explicit solvers. In a second experiment 22 su bjects simultaneously learned the 4 training pairs A+B-, B+C-, C+D-, D +E- (red) but additionally the same pairs with reversed reinforcement allocations (A-B+, B-C+, C-D+, D-E+; presented in grey). Which end of the series was the positive one was indicated by the stimulus color (r ed vs. grey). This corresponds to a nonverbal translation of a verbal bidirectional questioning form. On average the percentage of correct c hoices was comparable to the first experiment. Most of the subjects fo llowed a stable strategy with both of the serial ''response directions '': they either chose transitively or antitransitively. The results ar e simulated with a simple learning model.