Skip to main content
. 2015 Dec 8;4:e10163. doi: 10.7554/eLife.10163

Figure 5. Heterogeneity preceding stimulus onset predicts behavioral reaction time.

(a, b) Example traces from one example animal for population dF/F0 (a) and heterogeneity (b) of fast (F, green), slow (S, purple), and miss (M, red) responses (mean ± standard error over trials) also showing mean stimulus offsets. (c, d) As (a, b) but showing mean ± standard error over animals. (e) Fast behavioral responses are predictable before stimulus onset using heterogeneity [FDR-corrected one-sample t-tests vs. chance level (0); S–M, p=0.799; F–S, p<0.001; F–M, p<0.01] but not using dF/F0 (FDR-corrected one-sample t-tests vs. chance level; S–M, p=0.157; F–S, p=0.811; F–M, p=0.924; FDR-corrected paired t-test for heterogeneity vs. dF/F0; S–M, p=0.477; S–F, p<0.01; F–M, p<0.05). (f) The population heterogeneity during stimulus presentation does not merely reflect a continuation of pre-stimulus neural state; detected stimuli (slow and fast responses) elicit a faster rise to the maximum heterogeneity level than undetected stimuli (miss trials) (paired t-tests, n=8 animals, p<0.05). Slow and fast responses do not differ significantly (p>0.05). All panels: error bars/shaded areas indicate standard error of the mean. Statistical significance: *p<0.05; **p<0.01; ***p<0.001.

DOI: http://dx.doi.org/10.7554/eLife.10163.014

Figure 5.

Figure 5—figure supplement 1. Behavioral response is predictable on single trials when using heterogeneity, but not when using dF/F0.

Figure 5—figure supplement 1.

(a) Predictive decoding of miss trials using mean population dF/F0 (blue) or heterogeneity (black) during 3 s preceding stimulus onset. Each point is the mean prediction for a single animal (see ‘Materials and methods’). (b) same as (a) but for slow trials. (c) same as (a) but for fast trials. (d) Quantification of predictability shows chance-level prediction using dF/F0 (one-sample t-test, p=0.643) but above-chance prediction of behavioral responses using heterogeneity (p<0.05) as well as a significant difference in prediction performance between dF/F0 and heterogeneity (paired t-test, p<0.05). Points with error bars are mean ± standard error (n=24; three points per animal: miss/slow/fast).
DOI: http://dx.doi.org/10.7554/eLife.10163.015