Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Comparative Study
, 28 (11), 2845-55

Directional Selectivity in the Simple Eye of an Insect

Affiliations
Comparative Study

Directional Selectivity in the Simple Eye of an Insect

Joshua van Kleef et al. J Neurosci.

Abstract

Among other sensory modalities, flight stabilization in insects is performed with the aid of visual feedback from three simple eyes (ocelli). It is thought that each ocellus acts as a single wide-field sensor that detects changes in light intensity. We challenge this notion by providing evidence that, when light-adapted, the large retinal L-neurons in the median ocellus of the dragonfly respond in a directional way to upward moving bars and gratings. This ability is pronounced under UV illumination but weak or nonexistent in green light and is optimal at angular velocities of approximately 750 degrees s(-1). Using a reverse-correlation technique, we analyze the functional organization of the receptive fields of the L-neurons. Our results reveal that L-neurons alter the structure of their linear spatiotemporal receptive fields with changes in the illuminating wavelength, becoming more inseparable and directional in UV light than in green. For moving bars and gratings, the strength of directionality predicted from the receptive fields is consistent with the measured values. Our results strongly suggest that, during the day, the retinal circuitry of the dragonfly median ocellus performs an early linear stage of motion processing. The likely advantage of this computation is to enhance pitch control.

Figures

Figure 1.
Figure 1.
Responses of an L-neuron to colored bars moving vertically upward or downward. a, UV, positive contrast. b, UV, negative contrast. Responses to upward bars (below the upward arrows) are larger than those to downward bars (below the downward arrows). The different rows show the responses for different angular velocities as indicated by the numbers to the left side of the figure in units of degrees per second. Each trace is the average of two trials. c and d show the corresponding responses to green bars. In this case, there is little difference in the response. Calibration: vertical, 5 mV; horizontal, 250 ms. e–h indicate the directional indices obtained from the responses shown in the column above them (a–d). Only a quarter of the angular velocities used to determine the directional indices in e–h are shown in a–d. Pooled directional indices of 48 L-neurons are shown for negative contrast bars (i) and positive contrast bars (j). Triangles indicate that UV bars were used, and circles indicate green bars were the stimuli. Error bars indicate SEM. Differences between positive and negative contrasts indicate nonlinearities in the response.
Figure 2.
Figure 2.
Responses of an L-neuron to colored gratings moving vertically up and down (directions indicated by arrows). a–c show the responses to UV gratings with the same velocity of 1250° s−1 but three different spatial wavelengths (a, 72°; b, 36°; c, 24°). The directionality of each pair of responses is indicated by the DIs. d–f show the responses to the same patterns as those in a–d, but in this case the patterns were green. Calibration: vertical, 5 mV; horizontal, 200 ms. Note that a negative DI indicates there is a greater root mean squared response to the downward moving grating than the upward one. g, The pooled directional indices are shown for gratings colored UV (triangles) and green (circles) moved at 625° s−1. A total of n = 24 cells were shown both UV and green. h, Responses for the same cells as in g but for a faster grating speed of 1250° s−1. In the cases in which UV gratings were used, all the DIs were significantly different from zero (speed, 625° s−1: λ = 24°, p < 4 × 10−5, n = 7, unpaired t test; λ = 36°, p < 0.016, n = 7, Wilcoxon's rank test, and λ = 36°, p < 0.016, n = 7, Wilcoxon's rank test; speed, 1250° s−1: λ = 24°, n = 18, p < 4 × 10−9, unpaired t test; λ = 36°, p < 2 × 10−6, n = 18, unpaired t test, and λ = 72°, p < 6 × 10−5, n = 18, unpaired t test). The DI values for green gratings were only significantly different from zero for λ = 24° and speed of 625° s−1 (p < 3 × 10−4; n = 6; Wilcoxon's rank test); the other DI values were not significantly different from zero (speed, 1250 °s−1: λ = 24°, p > 0.25, n = 18, unpaired t test; λ = 36°, p > 0.2, n = 18, unpaired t test; λ = 72°, speed, 1250° s−1, p > 0.2, n = 18, Wilcoxon's rank test; and speed, 625° s−1: λ = 36°, p > 0.6, n = 6, unpaired t test, and λ = 72°, p > 0.55, n = 6, Wilcoxon's rank test).
Figure 3.
Figure 3.
The reverse-correlation method for mapping spatiotemporal receptive fields. a, A representation of six frames of the random UV stimulus. Each frame is separated by 1.6 ms. b, The intracellular response of an L-neuron to random stimulation by UV light (thick gray line). The thinner dark line indicates the response predicted by the derived UV spatiotemporal receptive field (UV STRF) shown in c. The mean squared prediction error for this cell is 10.3% and the separability index is α = 0.13. Only a small portion of the 40 s stimulus is shown. Calibration: vertical, 2 mV; horizontal, 100 ms. c, The STRF is shown from time t = 11.2 ms to t = 33.6 ms. The time of each two-dimensional profile is indicated in milliseconds, in the top right of each frame. The blue shaded areas represent depolarizations to light increments and the red shaded areas hyperpolarizations to light increments. Contours are at 10% increments of the maximum absolute value of the STRF, which is 0.24 mV/(C ms).
Figure 4.
Figure 4.
STRFs measured using UV and green pseudorandom stimulation. a, Four slices of the two-dimensional angular profile of the full UV STRF at times 12.8, 22.4, 27.2, 36.8 ms (from left to right). The minimum occurred at t = 22.4 ms. The blue-shaded areas represent depolarizations to light increments and the red-shaded areas hyperpolarizations to light increments. Contours are at 10% increments of the maximum absolute value of the STRF which is 0.25 mV/(C ms). The mean squared prediction error for this cell is 5.7% and the separability index is α = 0.08. b, The measured (gray line) and STRF predicted (black line) responses to upward and downward moving bars (indicated by arrows) traveling at 937.5° s−1. c shows the separable component of the full UV STRF, and d is the response predicted by this component to the same bars in a (notice the directionality disappears). e is the residual after subtracting c from a, and f is the response predicted by this residual. Contours in c and e have the same levels as shown in a. The green STRF is shown in g, i, and k that are analogous to a, c, and e. The STRF is shown at the same times and contours are at 10% of maximum absolute value of the STRF, which is this case is 0.26 mV/(C ms). h, j, and l show the same responses as shown in b, d, and f, respectively, but for responses to green bars and responses predicted by the green STRF. The MSPE for this cell is 9.3%, and the separability index is α = 0.04. Calibration: 2 mV, 100 ms.
Figure 5.
Figure 5.
The STRF predicted (dark lines) and measured (thick gray lines) DIs for negative (a) and positive (b) contrast bars pooled for 47 L-neurons. DIs were predicted from the full UV (triangles) or green (circles) STRF. c shows the directional index of the separable components of the STRFs as shown in Figures 4, e and k, with triangles and squares as in a and b. STRFs were better predictors for the DIs measured with UV positive contrast bars, indicating that directional nonlinear mechanisms respond mainly to negative contrast UV bars. These nonlinear mechanisms enhance directionality at high velocities and inhibit it at lower velocities.
Figure 6.
Figure 6.
Scatter plot of maximum DIs for positive contrast bars (45° crosses), negative contrast bars (90° crosses), and gratings (triangles) with UV DIs along the ordinate and green DIs along the abscissa (a). b, Scatter plot of the green versus UV separability index α for all cells studied in the paper. An α of zero indicates that the STRF can be perfectly reconstructed by its spatial and temporal components.

Similar articles

See all similar articles

Cited by 7 PubMed Central articles

See all "Cited by" articles

Publication types

LinkOut - more resources

Feedback