Classification performances and time resolution of code for stimulus comparison. A-F: Classification performance for linear classifier on test data (not used for training) as a function of the number of neurons on a logarithmic scale in ensembles #1 (green), #2 (blue), and #3 (red) and combination (neurons pooled over all ensembles; dotted black) for decoding stimulus perception (A), stimulus retention (B), motor selection (C), stimulus comparison (D), correct-error comparison (E) and particular-stimulus comparison (F). The input from each neuron was the spike count in consecutive 25 ms bins. The intervals are set at -10 - -5 s (A), -5 - 0 s (B) and 0 - 1 s (C-F) after the test tone onset. G: Time resolution of code for stimulus comparison. Average classification performance over three ensembles as a function of bin size (2-200 ms, temporal resolution) to count spikes within a 0-1 s window after the test tone onset for stimulus comparison. Error bars show the SEM. Dashed lines show chance levels (= 50%).