Distinct cortical codes and temporal dynamics for conscious and unconscious percepts
- Institut National de la Santé et de la Recherche Médicale, France
- Institut d'Imagerie Biomédicale, France
- Ben-Gurion University of the Negev, Israel
- Aalto University School of Science, Finland
- University of Oxford, United Kingdom
- Collège de France, France
- University of Paris-Sud, France
Figures
Figure 1
Experimental design and behavioral results.
(A). Sequence of events presented on a trial. Subjects attempted to localize a brief target, which could appear at one of eight locations. Mask contrast was adjusted to ensure ∼50% of unseen trials. …
Figure 2 with 4 supplements
Time course of location information.
(A). Average posterior probability of a correct classification of target location, as a function of time. Chance = 12.5% (1/8). Decoding confusion matrices are shown at the two decoding peaks. (B) …
Figure 2—figure supplement 1
Event-Related Fields and potentials for ‘Seen–Correct’ vs ‘Unseen–Correct’.
ERF and ERP components time windows were determined according to the global field power (upper panel). Two time windows were chosen; on each time window, cluster analysis was performed. In the …
Figure 2—figure supplement 2
Event-Related Fields and potentials for ‘UnSeen–Correct’ vs ‘Unseen–InCorrect’.
ERF and ERP components time windows were determined according to the global field power (upper panel). Two time windows were chosen; on each time window, cluster analysis was performed. No …
Figure 2—figure supplement 3
Classifying visibility.
The black line portrays the mean classification probability of classifier trained to classify ‘Seen–Correct’ and ‘Unseen–Correct’. The lighter line portrays classifier trained on the dataset with …
Figure 2—figure supplement 4
Controls for eye movements and motor-based decoding.
(A) Average posterior probability of a correct classification of target location, across time, separately for ‘Seen–Correct’ and ‘Unseen–Correct’ trials, for classifiers trained on EOG channels …
Figure 3
Asymmetrical cross-condition generalization.
A classifier trained in one condition and then tested on new data either from the same condition or the other condition (e.g. trained on ‘Seen–Correct’ trials and tested on new ‘Seen–Correct’ trials …
Figure 4 with 1 supplement
Source-based decoding and cross-condition generalization.
Classifiers were trained as in Figure 3, but using a restricted subset of cortical sources: pericalcarine (A), superior parietal (B), rostro-medial frontal (C), or superior frontal (D). Note, how …
Figure 4—figure supplement 1
Time course of location information for the different cortical sources.
Average posterior probability of a correct classification of target location, as a function of time for in 68 regions of interest. Chance = 12.5% (1/8).
Figure 5 with 3 supplements
Generalization of location decoding over time.
8-location classifiers trained at a specific time were then tested on data from all other time points. (A) Average classification probability as a function of testing time for each training time …
Figure 5—figure supplement 1
Analysis of variance (ANOVA) on Classification Endurance (CE).
The table gives the statistics and significance values for the main effects of Visibility (Seen–Correct vs Unseen–Correct), Direction of generalization (forward, backward), and Timeframe (5 levels), …
Figure 5—figure supplement 2
ANOVAs on CE with factors of Visibility (Seen–Correct vs Unseen–Correct) and Timeframe (5 levels), separately for forward and backward generalization.
(A) Forward generalization (B) Backward generalization.
Figure 5—figure supplement 3
