Skip to main content
. 2017 Sep 5;6:e26117. doi: 10.7554/eLife.26117

Figure 6. A Drosophila’s saccadic turns and fixation periods generate bursty high-contrast time series from natural scenes, which enable R1-R6 photoreceptors (even when decoupled from visual selection) to extract information more efficiently than what they could by linear or shuffled viewing.

(A) A prototypical walking trajectory recoded by Geurten et al. (2014) (B) Angular velocity and yaw of this walk. Arrows indicate saccades (velocity ≥ |±200| o/s). (C) A 360o natural scene used for generating light intensity time series: (i). ) by translating the walking fly’s yaw (A–B) dynamics on it (blue trace), and (ii) by this walk’s median (linear: 63.3 o/s, red) and (iii) shuffled velocities. Dotted white line indicates the intensity plane used for the walk. Brief saccades and longer fixation periods ‘burstify’ light input. (D) This increases sparseness, as explained by comparing its intensity difference (first derivative) histogram (blue) to that of the linear walk (red). The saccadic and linear walk histograms for the tested images (Appendix 3; six panoramas each with 15 line-scans) differed significantly: Peaksac = 4478.66 ± 1424.55 vs Peaklin = 3379.98 ± 1753.44 counts (mean ± SD, p=1.4195 × 10−32, pair-wise t-test). Kurtosissac = 48.22 ± 99.80 vs Kurtosislin = 30.25 ± 37.85 (mean ± SD, p=0.01861, pair-wise t-test). (E) Bursty stimuli (in Figure 1, continuous) had sparse intensity difference histograms, while GWN (dotted) did not. (F) Saccadic viewing improves R1-R6s’ information transmission, suggesting that it evolved with refractory photon sampling to maximize information capture from natural scenes. Details in Appendix 3.

Figure 6.

Figure 6—figure supplement 1. Drosophila R1-R6 photoreceptors generate responses with higher information transfer rates to saccadic (bursty) naturalist light intensity time series (NS) than to corresponding linear or shuffled stimulation.

Figure 6—figure supplement 1.

(A) Mean intracellular voltage responses of a R1-R6 photoreceptor to Naturalistic light intensities that have been modulated by saccadic (blue), linear (red) and shuffled (gray) yaw signals. (B) Simulations of biophysically realistic Drosophila R1-R6 photoreceptor model (Appendix 1) to the same stimuli. (C) Mean information transfer rates of seven R1-R6 photoreceptor outputs to the same stimuli and their population means. These information rates are further compared to the corresponding model output rates. Note that every photoreceptor sampled most information from the NS with saccadic modulation.