Interaction between positional but not between non-positional cues in human predictive learning.

Four experiments with human subjects examined the cue-interaction effects using a computer-controlled predictive learning task. In Phase 1, subjects learned that cue P was consistently associated with the occurrence of an outcome (P+), whereas cue N was never followed by the outcome (N-). In Phase 2, two neutral cues, R and I, were compounded with P and N, respectively. Each compound was followed by the outcome (PR+ and NI+). Thus, cue R was compounded with the already predictive cue P, whereas cue I was compounded with the non-predictive cue N. In each phase, subjects rated the contingency between the different cues and the outcome. In experiments 1 and 2, the spatial position of the cues was fixed, whereas it was variable in experiments 3, 4a and 4b. Verbal cues were used in experiments 1-3, whereas the cues consisted of geometrical figures in experiments 4a and 4b. Evidence for cue interaction, as indicated by giving cue I a higher contingency rating than cue R after or during Phase 2, was only found under the conditions of experiments 1 and 2. The results indicate that the use of positional cues facilitates the occurrence of cue-interaction effects. Possible reasons for this finding are discussed.