Context-dependent control over attentional capture

Abstract

A number of studies have demonstrated that the likelihood of a salient item capturing attention is dependent on the "attentional set" an individual employs in a given situation. The instantiation of an attentional set is often viewed as a strategic, voluntary process, relying on working memory systems that represent immediate task priorities. However, influential theories of attention and automaticity propose that goal-directed control can operate more or less automatically on the basis of longer term task representations, a notion supported by a number of recent studies. Here, we provide evidence that longer term contextual learning can rapidly and automatically influence the instantiation of a given attentional set. Observers learned associations between specific attentional sets and specific task-irrelevant background scenes during a training session, and in the ensuing test session, simply reinstating particular scenes on a trial-by-trial basis biased observers to employ the associated attentional set. This directly influenced the magnitude of attentional capture, suggesting that memory for the context in which a task is performed can play an important role in the ability to instantiate a particular attentional set and overcome distraction by salient, task-irrelevant information.

Figure 1

An example of the stimuli and design employed in Experiment 1. During training, observers searched for either the circle (feature search condition) or the “different shaped” item (singleton search condition) in separate blocks of trials. Each set was associated with a specific class of task-irrelevant scene (forest vs. city street scenes), counterbalanced across observers. During testing observers searched for a constant target (a circle) among homogeneous non-targets, and thus either attentional set could be employed. Of interest was whether the presentation of scenes associated with specific attentional sets during training would cause observers to employ the associated set when the scene was encountered during the testing session.

Figure 2

Reaction time data and error rates (bottom of each column) for each condition in the testing session of Experiment 1. During the testing session, observers always searched for a circle among homogeneous diamond non-targets. The search arrays were embedded either within scenes that had been associated with a feature set or singleton set during the training session. Error bars represent 95% within-subjects confidence intervals (Loftus & Masson, 1994; Morey, 2008).

Figure 3

Reaction time data and error rates (bottom of each column) for each condition in the testing session of Experiment 2. During the testing session, observers always searched for a circle among homogeneous diamond non-targets. The search arrays were embedded either within scenes that had been associated with a heterogeneous or homogeneous search display during the training session. Error bars represent 95% within-subjects confidence intervals (Loftus & Masson, 1994; Morey, 2008).