Abstract
The Attentional Blink (AB) phenomenon has been at the epicentrum of attention research for over two decades. However, the neural mechanisms underlying the phenomenon have still eluded understanding. How is it possible that subjects can easily report two targets (T1 and T2) embedded in a visual stream of distractors? Yet, when they are separated in time by 200–500 ms, T2 is often unavailable for conscious report. Most prevalent theories posit that T1 occupies attentional resources necessary for T2’s conscious access. Here, we provide evidence that the AB can be understood in terms of target-target activity-pattern similarity. To define similarity, 18 subjects completed two fMRI sessions performing a working memory (WM) task, and four sessions of EEG performing an AB task. In fMRI, similarity was defined as the voxel-pattern similarity, while in EEG, similarity was based on the T1 scalp-patterns. We show that in higher-tier visual areas, target-target similarity is detrimental for T2 performance. ROI-based analyses revealed robust negative correlations between similarity and T2 performance, most prominently in the Lateral Occipital Cortex (LOC) and Ventral Stream (VS). In contrast, similarity in V1 increased the probability of conscious access to T2. In EEG, similarity in the time window of 140–300ms negatively correlated with performance, consistent with our LOC and VS results. The negative correlations were present both when T2 was presented 200ms (Lag-2) and 700ms (Lag-7) after T1, demonstrating that the WM maintenance of T1 interfered with T2 encoding. Meanwhile, similarity in V1 only improved performance during Lag-2, implying a short-lasting priming effect. Our findings indicate that conscious access in AB depends on two contending target-interaction processes that are temporally and spatially separated. These insights into how, when and where T1 affects the probability of conscious access to T2 highlight important neural correlates for visual conscious processing.