Article Figures & Data
Figures
- Figure 1.
Exemplars of intact, locally scrambled, and globally scrambled images from the different object categories.
- Figure 2.
Category-specific patterns of EEG response to intact and scrambled images. A, For each time point, normalized patterns of response to odd and even trials of each category were compared across 64 electrodes. The correlation coefficients were then represented in a similarity matrix for that time point. Distinct category-specific patterns of response were defined by higher within-category (e.g., face–face) compared to between-category (e.g., face–bottle) correlations. B–D, Correlation time-courses are shown for the intact (B), locally scrambled (C), and globally scrambled (D) image types. The shaded region represents 95% CIs obtained by bootstrapping across participants. Group mean correlation matrices at 100 ms intervals are shown above the plot. Gray box at the base of the plot represents the time points at which the stimulus was present. Blue bar at the base of the plot represents time points at which the lower bound of the CI is >0, indicating significantly higher within-category correlations than between-category correlations.
- Figure 3.
A–C, Classifier accuracy for between-category discrimination (blue line) with intact (A), locally scrambled (B), and globally scrambled (C) images (chance = 33%, gray line). The blue-shaded regions represent 95% CIs obtained through bootstrapping across participants. The blue bar at the top of the plot represents time points at which the lower bound of the CI is above chance. The gray box on the axes of the plot represents the stimulus duration.
- Figure 4.
Similarity between patterns of EEG response to intact images and locally scrambled (A) or globally scrambled (B) images from the same object category. Blue-shaded regions represent 95% CIs across participants. The blue bar at the top of the plot indicates time points at which the correlation is significantly >0. The orange bar indicates the time points at which the correlation is not significantly different from the noise ceiling. The gray box represents the stimulus duration.
- Figure 5.
Classifier performance across different image types. A, Accuracy in classifying responses to either intact or locally scrambled images when trained on locally scrambled or intact images, respectively. B, Accuracy in classifying responses to either intact or globally scrambled images when trained on globally scrambled or intact images, respectively. The blue line indicates classifier accuracy across time, with shaded regions representing 95% CIs obtained through bootstrapping across participants. The blue bar at the top of the plot represents the time points at which the lower bound of the CI is above chance. The gray box shows stimulus duration.
- Figure 6.
Temporal cross-correlation matrices for each image type. Responses to trials of the same condition were correlated over each combination of time points. A–C, Correlations were collapsed across categories to give one matrix per image type [intact (A), locally scrambled (B), globally scrambled (C)]. The colorbar represents Pearson’s correlation coefficients. Matrices were thresholded by obtaining 95% CIs at each coordinate by bootstrapping across participants. Coordinates at which these intervals overlapped with 0 are shown in white. The gray box represents the stimulus duration.
In this issue
