Abstract
Extant attentional blink (AB) theories implicitly assume serial processing in which first target (T1) identification impedes second target (T2) identification when T1–T2 lag is short (200–500 ms). Most AB studies report T2 identification conditionalized on T1 identification (T2|T1). Tan and Dark (2006) also examined T1 identification conditionalized on T2 (T1|T2). Prior to each trial, they cued T1 identity with 50% validity (valid vs. invalid cue); subjects were categorized according to whether they used or ignored the cue (confirm vs. ignore strategy). They found lag effects on both T2|T1 (the AB) and T1|T2, suggesting a more dynamic view of target processing. The AB effect was attenuated and the T1|T2 lag effect was eliminated for “confirm” subjects on valid trials, i.e., when T1 consolidation should be facilitated. We further examined this effect by varying T1 identity cue validity. Cue validity was low (25%) for Experiment 1 and high for Experiment 2 (75%). Both the AB and the T1|T2 lag effect were attenuated when cue validity was high but not when it was low, regardless of the subjects' strategy. However, because subjects knew T1 identity in the condition showing attenuation, the attenuation could be a task artifact. Therefore, we cued T2 identity with 50% validity in Experiment 3. If attenuation of the T1|T2 lag effect were due to a task artifact associated with knowing T1 identity, one predicts that cueing T2 identity would not affect T1|T2. We found that both the AB and T1|T2 performance were worse for “confirm” subjects on invalid trials. If more attentional resources are devoted to identifying T2 when the cue is being used but is invalid, the results suggest that devoting resources to T2 identification is detrimental to T1 identification. We propose a model in which T1 consolidation and T2 selection compete dynamically for a higher-order attentional resources.