Table 1.

Accuracy of pairwise and pairwise–adjacent models

ConditionCell typeΔtPairwisePairwise–adjacentEmpirical
Constant illuminationON parasol10 ms98.6 ± 0.5%98.3 ± 1.0%99.8 ± 0.1%
5 ms97.8 ± 0.5%97.5 ± 1.0%99.8 ± 0.1%
20 ms99.2 ± 0.4%98.9 ± 0.9%99.7 ± 0.1%
OFF parasol10 ms98.9 ± 0.3%98.2 ± 1.3%99.4 ± 0.4%
5 ms98.5 ± 0.4%97.4 ± 1.5%99.0 ± 0.5%
20 ms99.2 ± 0.3%98.6 ± 1.3%99.4 ± 0.3%
White-noise stimulusON parasol10 ms98.6 ± 0.4%98.4 ± 0.9%99.6 ± 0.2%
5 ms97.8 ± 0.5%97.5 ± 1.0%99.6 ± 0.2%
20 ms99.1 ± 0.3%98.9 ± 0.9%99.5 ± 0.2%
OFF parasol10 ms98.6 ± 0.5%98.3 ± 0.9%98.8 ± 0.7%
5 ms98.2 ± 0.8%98.0 ± 1.3%98.4 ± 1.0%
20 ms98.6 ± 0.7%98.3 ± 1.0%98.6 ± 0.9%
  • Each numerical entry indicates the percentage of departures from statistical independence captured by a specific model, for a given stimulus condition, cell type tested, and time bin size (Δ t). Results cited in text were obtained with Δ t = 10 ms. The range indicates the mean ± 1 SD across several hundred cell groups tested, each consisting of three to seven cells (in the case of the pairwise model) or four to seven cells (in the case of the pairwise–adjacent model). For each model, the quantities shown are as follows: pairwise model, (DindDpair)/Dind; pairwise–adjacent model, (DindDadj)/Dind; empirical model, (DindDemp)/Dind.