Neural Determinants of Task Performance during Feature-Based Attention in Human Cortex

Behavioral results in the scanner. Error bars indicate ± within-subject SEM following the method of Cousineau (2005).

Univariate results. A, Group r ² map of the attention task shown on an inflated Caret atlas surface. The approximate locations of retinotopically-defined (V1-7, MT+, IPS1, and IPS2) and task-defined (FEF and IFJ) areas are indicated by lines. B, Mean BOLD response in the attention task from two ROIs (V1 and IPS1). The error bar on the first time point is the average ± within-subject SEM across all time points.

Schematic of the correlation analysis. Each matrix represents the spatial pattern of response amplitudes from voxels within a ROI (amplitude is color coded according to the scale bar at the bottom). The middle column of matrices (shaded area labeled “baseline”) illustrates the baseline neural response pattern to each direction (CW and CCW) during the baseline task. The other two columns illustrate the neural response pattern to an attended motion direction during correct and incorrect trials during the attention task. The black double-sided arrows between matrices represent the correlations that were calculated (Pearson’s r) and the bounded lines represent the averaging of correlation coefficients across directions to obtain an overall index of attentional priority quality for correct and incorrect trials.

Average neural patterns for prioritized features in occipital and frontoparietal areas vary with task performance. Group-average Fisher-transformed correlation coefficients (averaged across motion directions) are shown, which reflect the similarity of neural patterns of activity between the attention and baseline tasks for correct and incorrect trials. The ExS label represents extrastriate visual areas. Error bars are ± within-subject SEM following the method of Cousineau (2005). Asterisks indicate the significance level in paired t tests (**p < 0.01, *p < 0.05).

Feature coding in PPC tracks trial-by-trial fluctuations in task performance. Group-average classification accuracies are shown and plotting conventions are the same as Figure 5.