Journal of Experimental Psychology: Human Perception and Performance, 2014. Paper abstract bibtex
Four experiments studied the interfering effects of a to-be-ignored " stimulus suffix " on cued recall of feature bindings for a series of objects. When each object was given equal weight (Experiment 1) or rewards favored recent items (Experiments 2 and 4), a recency effect emerged that was selectively reduced by a suffix. The reduction was greater for a " plausible " suffix with features drawn from the same set as the memory items, in which case a feature of the suffix was frequently recalled as an intrusion error. Changing payoffs to reward recall of early items led to a primacy effect alongside recency (Experiments 3 and 4). Primacy, like recency, was reduced by a suffix and the reduction was greater for a suffix with plausible features, such features often being recalled as intrusion errors. Experiment 4 revealed a tradeoff such that increased primacy came at the cost of a reduction in recency. These observations show that priority instructions and recency combine to determine a limited number of items that are the most accessible for immediate recall and yet at the same time the most vulnerable to interference. We interpret this outcome in terms of a labile, limited capacity " privileged state " controlled by both central executive processes and perceptual attention. We suggest further that this privileged state can be usefully interpreted as the focus of attention in the episodic buffer. The role of attention in visual working memory is controversial, and this is well illustrated by the problem of memory for feature bindings. Luck and Vogel (1997) showed that storage capacity is limited to a small number of multifeature objects, with no corre-sponding limit on the number of individual features. Wheeler and Treisman (2002) followed this with evidence that attention is required for encoding and maintaining feature bindings but not individual features. These observations suggest a close relationship between storage capacity and attention. However, our own exper-iments led us to question this, because we found equal involvement of attention in memory for features and their bindings (Allen, Baddeley, & Hitch, 2006, 2014; Baddeley, Allen, & Hitch, 2011). These contrasting conclusions about the role of attention in maintaining feature bindings in visual working memory are based on different kinds of evidence and, we suggest, different compo-nents of attention. Wheeler and Treisman (2002) used change-detection methodology and compared a whole-display test, repre-senting all the memorized items, with a single-probe test, presenting only a single item. They found that binding memory was selectively impaired with a whole-display test, whereas fea-ture memory was unaffected. This led them to conclude that a multiple-item test display required a reallocation of general atten-tional resources needed to maintain binding information in mem-ory, whereas this was not the case for a single-probe test. In contrast, our own experiments studied attention using dual-task methodology and showed that a demanding concurrent task did not disrupt memory for feature bindings any more than memory for individual features (Allen et al., 2006, 2014; Allen, Mate, Hitch, & Baddeley, 2012; Baddeley et al., 2011). In the present article, we develop the idea that resolution of this apparent contradiction and the key to further progress lies in a broad distinction between perceptual selective attention and executive attentional control (see, e.g., Chun, Golomb, & Turk-Browne, 2011; Lavie, 2010; Petersen & Posner, 2012). We present evidence that executive control and perceptual selection combine to determine the contents of working memory, and we use this to suggest that dual-task interference reflects competition for executive processes, whereas effects of test display reflect perceptual attention. Our previous dual-task studies stemmed from the hypothesis that encoding and maintaining any type of binding information in working memory is a function of a multimodal " episodic buffer "