Electrophysiological Correlates of Visual Crowding
Introduction: Flankers can strongly deteriorate performance on a visual target (crowding). For example, vernier offset discrimination is strongly affected by neighboring flankers. Interestingly, performance for longer and shorter flankers is better than performance for equal length flankers. We previously found that crowding is strongest when the vernier and the flankers group (equal length flankers) and weaker when they ungroup (shorter or longer flankers, Malania, Herzog, & Westheimer, 2007; Manassi, Sayim, & Herzog, 2012; Figure 1A). Here, we used high density EEG to investigate the mechanisms underlying crowding. Methods: The vernier target consisted of two vertical lines displaced either in the left or right direction (Figure 1A, a right offset vernier is shown). Ten observers discriminated the offset direction by pushing the corresponding button. Flankers could be either short, equal length, or long lines surrounding the vernier. We recorded high-density event-related potentials (ERPs). We used the source-localization technique LAURA to localize cortical sources of crowding-related activity. To dissociate target processing from flanker processing, we conducted a second experiment using the EEG frequency-tagging technique. Eight new observers discriminated the offset direction of a vernier that was slowly increasing in size either to the left or right. The vernier and the flankers were either green or red and flickered at two different frequencies. We recorded steady-state visually evoked potentials (ssVEPs) to the vernier and flanker flicker. Results: In the ERP study (Figure 1), the P1 amplitude correlated with the length of the flankers (stimulus energy, F(1.6, 14.1) = 11.2, p = 0.002). The N1 amplitude correlated with performance, with the highest amplitudes occurring for long flankers, intermediate ones for short flankers, and low amplitudes for medium flankers (F(2, 18) =10.1, p = 0.001, Figure 1C). Source localization showed that the reduced N1 amplitudes were caused by reduced activation of the lateral occipital cortex and posterior temporal and parietal cortices without a significant contribution of the V1 (Figure 1D). In the ssVEP study, flankers of the same color as the vernier (green-green or red-red) crowded more strongly than flankers of a different color (green-red or red-green) because the former, as we propose, grouped with the vernier. EEG responses to the vernier were suppressed during crowding (same color flankers) compared to uncrowding (different color flankers). EEG responses to the flankers were slightly larger when the flankers grouped with the target compared to when they ungrouped from the target. Conclusions: The N1 ERP component is a good predictor for perceptual grouping and visual crowding. Crowding-related processes seem to occur after the P1 wave, i.e., after basic feature extraction, in higher-level visual and non-visual areas of the brain. Frequency tagging dissociated target and flanker processing in crowding and showed that the target is suppressed when it groups with the flankers while flanker-related activity increases or stays constant.