Attentional blink magnitude is predicted by the ability to keep irrelevant material out of working memory

In order to avoid ceiling effects on behavioural accuracy set sizes of three and six were used here instead of the set sizes two and four used by Vogel et al. (2005). Ceiling effects were not a concern in Vogel et al. where the dependent variable was CDA amplitude.

Nineteen of the 60 participants had filtering efficiency scores larger than 1.0 with the vast majority of these 19 having scores just slightly above 1, suggestive of slight measurement error. To be conservative, these participants were excluded from the analyses, leaving N = 41. In each case, a filtering score of greater than 1.0 was the result of the participant having slightly higher accuracy in the 3 red + 3 blue condition than in the 3 red condition. This pattern suggests that that these participants were fully able to filter out the irrelevant blue distractors, but that their scores in the 3 red condition were slightly underestimated and/or the scores in the 3 red + 3 blue condition were slightly overestimated. Following the assumption that scores in the 3 red condition should be just as high as scores in the 3 each condition, when we corrected the 3 red scores so that they were equal to the 3 each scores, the pattern of results observed when using all 60 participants was the same as the patterns reported here with N = 41. Namely, a significant correlation was observed between filtering efficiency and AB magnitude (r = −0.29, p = 0.025), and filtering efficiency predicted AB magnitude over and above WM capacity and T2 sensitivity (semi-partial r = − 0.32, p = 0.012).

It is possible that the slightly higher estimate of K obtained here resulted from the fact that the displays in the present study allowed for triplets of the same colour in a single display, whereas previous studies have often allowed only singles and doubles (see Awh, Barton, & Vogel, 2007). The use of triplets may have encouraged participants to group the items on some trials, and therefore increased their estimates of K. Note below that K shows no relationship with any AB variables, so if grouping was present, this strategy does not appear to be related to AB magnitude.

It is possible that poorer filtering may not only predict better long lag T2 performance (as we found here), but also higher T1 accuracy (which we did not find here).Greater overinvestment may not always lead to greater target accuracy, as the greater overinvestment to targets would also be accompanied by a greater overinvestment to adjacent distractors, and therefore it is difficult to know when one should expect greater overinvestment to lead to greater target performance. Our speculation is that this may depend on the nature of the target task, the target, and the distractor.