Fig. 3

Scatter diagrams and histograms. In panel a, we recall how a scatter diagram is constructed for each trial (here black crosses for trial 1 and red crosses for trial 2) and then superposed. Various histograms can be built with those points, using different partitions. Panel b gives the partition corresponding to the classical cross-correlogram. Panel c gives the partition of the JPSTH. The diagonal squares are filled in gray: if a point of the scatter diagram lies in one of those square, its coordinates correspond to spikes that are very close to each other. Two very close spikes (one on N1, the other one on N2) and their corresponding point in the scatter diagram are added, showing that the reverse is wrong. Therefore, there are some couples that are close to each other and not counted by the JPSTH as such. Panel d gives the partition used for the computation of the vectors ${\overline{\mathbf{n}}}_{m,\ell }$ of dimension K ($K=2$ here). More precisely, ${\overline{\mathbf{n}}}_{1,2}$ corresponds to the vertical part $2\to 1$, whereas ${\overline{\mathbf{n}}}_{2,1}$ corresponds to the horizontal part $1\to 2$. The same spikes are added, and they are now counted as close to each other