Fig. 5: Median-split analysis from Experiment 2.
From: Ensemble representations reveal distinct neural coding of visual working memory
a We split a response distribution in the VO condition of Experiment 2 by its circular median. As a result, one set was closer to the mean orientation, while the other set was closer to the target orientation. The SO decoder was built and generalized to the two sets, separately. b Reconstruction from the mean-close responses showed stable coding. c Reconstruction from the target-close responses showed no evidence of either stable nor dynamic coding. Top, the x- and y-axes show the time of test and training sets after stimulus onset, respectively. Colors represent the t-statistics. Black areas show significant modulation over time, determined by a one-sided Monte Carlo randomization test, p < 0.05. Bottom, the x-axis shows the time after stimulus onset. The y-axis shows the magnitude of stable and dynamic coding. Green lines show stable coding. Red dotted lines show dynamic coding. Shaded areas represent ± 1 bootstrapped standard error. SO same orientation; VO varied orientation.
