Abstract
Previous studies have demonstrated that in a visual search task, observers are faster to locate targets when these are presented in repeated rather than random contexts, an effect termed contextual-cuing. Using a random mapping condition and a consistent mapping condition, in which targets and distractor-sets were paired throughout the experiment, we replicated this classic effect (Experiment 1), and conducted two experiments aimed at clarifying how contextual cuing interacts with general task improvement (i.e., learning). We used either repeated contexts only (Experiment 2) or random contexts only (Experiment 3), and found that learning curves were similar in both experiments and, surprisingly, much faster than in Experiment 1. Our results suggest that contextual-cuing does not facilitate general learning performance. Instead, an advantage for a repeated context condition seems to be observed only when it is intermixed within a random context condition. To better understand whether our results are specific to contextual cuing, or can be generally applied to other kinds of implicit learning, we conducted two more experiments with a discrimination task. Participants were instructed to answer as fast as possible which image (out of four possibilities) is presented. In Experiment 4, images appeared at random locations, while in experiment 5 images appeared according to a specific 9-elements spatial order. Results were similar to the contextual cuing paradigm experiments. In general, our results question the current view of statistical learning as a mechanism that supports general learning improvements. We argue that the contextual-cuing effect is not a result of facilitation from repeated contexts, but rather reflects overcoming interference caused by mixing regularities and randomness.
Meeting abstract presented at VSS 2016