Fig. 3.

Variability and signal-to-noise ratio (SNR) of response features to single-facet stimuli. A: representative traces for each cell type showing the trial-by-trial variability of responses to an instantaneous luminance decrease at a single facet (0-ms duration). The stimulus occurs at the time the traces begin. Trial averaged responses are shown in black and 3 individual trials in gray. Measured features are depicted in red for each trial: for photoreceptor traces, they are the response slopes from 25–75% of the response height; for LMC and LGMD traces, they are the peaks (+) and response widths (full width at half-height, FWHH). The SNR is noted for the peak and slope features for the experiments shown. It is defined as the mean divided by the SD of each measured feature. B: relative response strengths (photoreceptor slopes; peak heights for other recording types) to single-facet luminance changes of varying duration. For each box plot, the central horizontal line denotes the median and the box extent depicts the 25th (p₂₅) to 75th percentile (p₇₅) data range. Whiskers show the range of data not considered outliers, and plus signs represent outliers, i.e., points >1.5 (p₇₅ − p₂₅) times away from lower (p₂₅) or upper (p₇₅) data quartile. Responses are normalized to the mean of the population distribution (at 0-ms duration) for each recording type (photoreceptors, 5.5 mV/10 ms; LMCs, 7.2 mV; LGMD I_m, 2.0 nA; LGMD V_m, 9.3 mV). C: SNR as a function of luminance change duration for the recording types. Box plots conventions are as described in B. The dashed line shows the mean SNR of the LGMD V_m peak for pseudolooming stimuli, a stimulus type that targets 45 facets in a pattern that mimics some aspects of a looming stimulus. There is a significant effect of recording type at 0-ms luminance change duration (*P_KW = 0.0023, corrected for multiple comparisons using HSD criterion). See Data analysis and statistics for definitions of probability statistics.