Effects of Schedule and Length of Training on Reward-Induced Stereotypy and Negative Transfer in Humans

Abstract

Prior research by Schwartz (1982) indicates two things: (1) that human subjects are as lilkely to show reward-induced sequence stereotypy with 100% continuous reward (CRF) as with 50% partial reward (PR50) rates, and (2) that experienced subjects with a prior history of stereotypy take more trials than inexperienced hold subjects to identify rules governing solution of a novel problem on the stereotypy matrix (i.e., negative transfer). In conjunction with the latter results, Schwartz reports that stereotyped subjects test fewer hypotheses during the solution period and that the hypotheses themselves are inefficient. The results of the present study are only partially consistent with Schwartz’s. In Experiment 1, 11 of 11 CRF subjects showed stereotypy in comparison to only 4 of 13 PR50 subjects. In Experiment 2, subjects with 3 days of prior training showed negative transfer relative to hold subjects, whereas subjects with only 1 or 2 days of prior training performed at least as well as hold subjects. Additionally, presence vs. absence of negative transfer was not strongly related to prior schedule (i.e., CRF vs. PR50) or prior performance (i.e., presence vs. absence of stereotypy). Consistent with Schwartz’s findings, however, subjects showing negative transfer tended to use fewer hypotheses and inefficient ones.