Luck and learning: feedback contingencies and initial success in verbal discrimination learning

Ninety undergraduate volunteers learned a 12-pair, low frequency, verbal discrimination list. The independent variables were feedback (positive only, negative only, or both) and initial success (17%, 50%, or 83% correct on trial one). With the total errors on nine trials as the dependent variable, the main effect of feedback was found to be not significant. The main effect of initial success was significant. Subjects who started at 50% correct had the most difficulty learning the list and subjects in the 83% correct condition learned quickest. The predicted interaction between feedback and initial success was not significant, even though the means for the three types of feedback at the 17% correct starting level reflect the expected differences. The results are interpreted as supporting the rule-use component of frequency theory. The relationship between initial success and subsequent performance appears to be curvilinear, rather than linear, as many theorists have assumed.