(A) Linear filters (black) and corresponding exponential fit (cyan) for ORNs in four glomeruli at four concentrations of the stimulus. Two odors were used (methyl acetate and benzaldehyde) that capture the response of two non-overlapping sets of glomeruli. Responses below 10% ΔF/F were removed from further analysis because of an insufficient signal-to-noise ratio (n = 4–11). (B) Example of the linear filters obtained after deconvolution of the calcium sensor, color indicates glomerulus identity and different curves correspond to different stimulus intensities. (C) Mean responses to the fluctuating stimulus averaged over time for 10 glomeruli (color coded) at four stimulus intensities (same color). Error bars indicate standard deviations. Dashed lines indicate a threshold of 10% ΔF/F. (D) Timescale τ1 of the linear filter resulting from the exponential fit. Error bars indicate 95% confidence intervals. Timescale of the linear filters is significantly anticorrelated to mean response (R = −0.6; p<0.0004), glomerulus dependent (p<10−6) and slightly dependent on stimulus intensity (p<0.02; two-way analysis of variance). Sample size per glomerulus was: nD = 6, nDC1 = 7–8, nDL1 = 7–8, nDL5 = 7–8, nDM1 = 9–11, nDM2 = 9–11, nDM4 = 9–11, nDM5 = 7–8, nDM6 = 7–8, and nVM7d=4–6. (E) Linear filters (black) and corresponding exponential fit (purple) for PNs in four glomeruli at four concentrations of the stimulus. For each glomerulus, three models were fitted (single exponential, single exponential plus a constant, and double exponential, see Materials and methods). The model with the lowest BIC was selected. The shape of the linear filter in PNs depends on stimulus intensity. (F) Linear filter obtained from calcium responses after deconvolution of the sensor kinetics, showing that PNs filters are concentration dependent. Color indicates concentration as in Figure 2D. (G) Same as in C for PNs. (H) Timescales of the linear filter resulting from the exponential fit. Stars (*) indicate data sets that were fitted by a double exponential (model 3): in this case, a second timescales τ2 is shown, which is associated to the negative lobe pf the double exponential. τ1 is slightly correlated with the mean response (R = 0.4, p<0.02), whereas τ2 is independent from it (p>0.05). Differences between glomeruli are not significant (Kruskal-Wallis test, p>0.05). The sample size per glomerulus was as follows: nD = 7, nDC1 = 3–4, nDL1 = 6–7, nDL5 = 7–8, nDM1 = 6–7, nDM2 = 6–7, nDM4 = 6–7, nDM5 = 5–7, nDM6 = 7–8, and nVM7d = 6–7.