The human brain's capacity for conscious processing is limited, leading to phenomena like the attentional blink (AB). The AB refers to the reduced ability to detect a second target (T2) when it appears within 200-500ms of a first target (T1) in a rapid stream of visual stimuli. While behavioral studies have extensively documented the AB, its neural underpinnings remain unclear due to limitations in temporal resolution of conventional neuroimaging techniques. This study aimed to characterize the precise effects of the AB on behavior and neural activity by utilizing multivariate encoding analysis to extract feature-selective information from EEG data recorded during an RSVP task. Understanding the neural basis of the AB is crucial for gaining insights into the mechanisms of selective attention, visual processing, and the neural correlates of consciousness. Previous research has proposed various theories to explain the AB, including resource depletion models, which posit that processing the first target exhausts limited attentional resources, leaving insufficient capacity for processing the second. Other accounts involve temporary loss of control over attentional processes or an active inhibition of the second target representation to prevent interference with the first. This study aimed to disentangle these competing hypotheses by directly measuring the neural representations of both targets and distractors, allowing for a precise analysis of their dynamic interactions during the attentional blink.

Existing literature on the attentional blink (AB) suggests a complex interplay of attentional and perceptual processes. Early studies using rapid serial visual presentation (RSVP) paradigms established the core phenomenon: a significant drop in the correct identification of the second target (T2) when it follows the first target (T1) within a specific temporal window (typically 200-500ms). Several hypotheses have attempted to explain the AB, including: resource depletion models, suggesting that processing T1 consumes attentional resources needed for T2; inhibition models proposing that an active suppression mechanism prevents T2 from gaining conscious access to avoid interference with T1 processing; and models focusing on the role of distractors in disrupting T2 processing. Neuroimaging studies, while providing some insights, often lack the temporal resolution to fully capture the rapid dynamics involved in the AB. Existing studies that used EEG or MEG have reported ERP components related to the AB but didn't precisely pinpoint the stage of processing affected. This study leveraged a more advanced multivariate encoding approach which offers higher temporal precision and the capability to analyze individual item representations within the RSVP stream.

Two experiments were conducted. Experiment 1 (N=22) focused on behavioral aspects of the AB using an RSVP task. Participants viewed a stream of oriented gratings and reported the orientation of two target gratings. The lag between targets was manipulated. Gaussian functions were fit to quantify the behavioral response accuracy. Regression analysis was used to assess the influence of distractor items on target reports. Experiment 2 (N=23) replicated the behavioral task while simultaneously recording EEG. A multivariate forward encoding model was developed to decode orientation information from EEG patterns for each item in the RSVP stream. This method allowed for the separation of neural responses to targets and distractors. The model was used to assess how the neural representation of T2 was affected by the AB (correct vs. incorrect T2 reports). Univariate analysis was also performed to examine the scalp topographies associated with target and distractor processing. Data were pre-processed using EEGLAB and SASICA toolbox to remove artifacts and noise. For statistical analysis, Gaussian functions were fitted to both behavioral and neural data; ANOVAs and t-tests were employed to compare conditions and Bonferroni corrections were used to adjust for multiple comparisons. Nonparametric sign permutation tests were employed for time-course analyses to account for non-normality. Cluster-based permutation testing was used for correcting for multiple comparisons across electrodes and time.

Experiment 1 replicated the classic AB effect: T2 accuracy was significantly lower than T1 accuracy at short lags (2 and 3), but not at longer lags (5 and 7). The reduction in T2 accuracy was primarily due to a decrease in response gain, not precision (i.e. width), and an increase in baseline guessing rates, supporting global workspace theory of consciousness. Crucially, distractors had minimal influence on target orientation judgements, except for the item immediately following a target which significantly biased the perceived target orientation. Interestingly, a significant interaction between the orientations of T1 and T2 was found, indicating long-range temporal integration of target representations. Experiment 2 replicated the behavioral findings and revealed neural correlates of the AB. Forward encoding of EEG data showed robust orientation selectivity for each item in the RSVP stream. For T2 at Lag 3, there was significantly reduced feature-selective information in trials with incorrect T2 reports compared to trials with correct reports. The reduction was mainly in the gain of the neural response, not the width. Importantly, distractors' neural representations were unaffected by the AB, contrasting with distractor-based accounts. The suppression of T2 representation happened rapidly (100-150ms) after T2 appeared, suggesting early sensory inhibition. Univariate analysis showed higher orientation selectivity for targets compared to distractors over occipital and parietal areas.

The findings provide strong support for T1-based theories of the AB, demonstrating that the second-target deficit is primarily driven by processing competition between target items rather than distractor interference. The observed long-range integration of target representations suggests that the visual system integrates features from successive targets even when separated by multiple items, challenging existing theoretical accounts. The reduction in gain of neural representations, without changes in the width, implies that the AB affects the amplitude, not the precision of target processing. This is consistent with the notion that temporal attention, similar to spatial attention, affects the gain of neural responses. The rapid suppression of T2 neural representations at early stages of processing suggests an active inhibitory mechanism to prevent interference with T1 processing. The results suggest a model where the detection of T1 triggers a local integration window incorporating the immediate subsequent item, followed by a global integration process possibly involving working memory.

This study offers novel insights into the neural and behavioral mechanisms of the attentional blink using a combination of innovative behavioral paradigms and advanced multivariate EEG analysis techniques. The findings strongly support T1-based theories, highlighting the importance of processing competition between targets rather than distractor interference. The observed long-range target integration challenges existing theoretical frameworks, calling for revisions of current models. Future research could investigate the specific brain regions and neurotransmitters involved in the observed neural dynamics. Furthermore, applying the methodology to other related phenomena, such as repetition blindness, could yield valuable insights into the dynamics of visual attention and conscious processing.

The study utilized a specific RSVP paradigm with oriented gratings, potentially limiting the generalizability of findings to other stimuli or task conditions. Although the study used a robust method to reduce artifacts, residual noise in EEG data could have affected the results. The specific mechanism(s) responsible for the observed long-range target integration remain to be elucidated. The study focused on a specific temporal window, and the findings may not extend to other temporal lags.