Cue Competition in Human Incidental Learning

McLaren, I. P., Jones, F. W., McLaren, R. and Yeates, F. (2013) Cue competition in human incidental learning. In: Reinforcement Learning and Decision Making Meeting 2013, 25th-27th October, Princeton, New Jersey, USA .There is a question as to whether cue competition effects can be observed in incidental learning paradigms in humans. Some authors have reported that cue competition is not observed, and that previous demonstrations of cue competition have relied on explicit awareness of the task in hand. This would imply that these effects are more likely to be the product of cognitive inference than associative learning. We addressed this question by using two paradigms previously shown to produce associative learning under incidental conditions. One was a standard SRT task in which the preceding two trials of a run of three predicted the third 2/3 of the time, and the other was based on another predictive cue, a colored square, which could also stochastically predict the next response required.We have demonstrated in other studies that both cues would support learning under incidental conditions in the absence of any verbalisable knowledge of the rules involved. The question was to what extent would these two cues compete if run concurrently, as assayed by their ability to make the next response faster and more accurate than controls? We assessed this by comparing a dual cue group to a color only control and a sequence only control. Our results showed that all three groups learned, but that during a test phase where each cue could be assessed independently, the dual group showed a marked decline in performance relative to the color control, and very similar performance to the sequence control. We interpret this as evidence for overshadowing occurring between the two predictive cues in the dual group, such that when combined their performance would be equivalent or superior to either control, but when assessed independently, the color cue actually has a weaker association to the outcome than the equivalent cue in the control group. We conclude that the sequence cues overshadowed the color cues in this task, and discuss possible theoretical accounts of this phenomenon