The movies below show match/nonmatch information is coded by changing patterns of neural activity that are time locked to the onset of the stimulus/trial (essentially these are the right panels of Figs. 3a, and S4a in movie format). The black non-changing curve in the movies show the decoding accuracy when the training and testing were done at the same time period (i.e., this is the same curve as shown in red in Figs. 2b and S3b). The changing blue curve shows the decoding accuracy when training with data at one particular time period and testing at all other times (the training time is indicated by the blue moving line segment at the bottom of the movie). Match/nonmatch information from the feature task shown in the left movie and from the spatial task in the right movie. As can be seen, the best decoding always occur when the training and testing is done at the same time period relative to stimulus/trial onset, indicating that information is coded by different patterns of activity at different time periods.
To save the movies to your computer, right-click the links below and choose "Save Link As..."
Download dynamic population coding movie for the feature task
Download dynamic population coding movie for the spatial task