Directional Preference in Avian Midbrain Saliency Computing Nucleus Reflects a Well-Designed Receptive Field Structure

Abstract

Neurons responding sensitively to motions in several rather than all directions have been identified in many sensory systems. Although this directional preference has been demonstrated by previous studies to exist in the isthmi pars magnocellularis (Imc) of pigeon (Columba livia), which plays a key role in the midbrain saliency computing network, the dynamic response characteristics and the physiological basis underlying this phenomenon are unclear. Herein, dots moving in 16 directions and a biologically plausible computational model were used. We found that pigeon Imc's significant responses for objects moving in preferred directions benefit the long response duration and high instantaneous firing rate. Furthermore, the receptive field structures predicted by a computational model, which captures the actual directional tuning curves, agree with the real data collected from population Imc units. These results suggested that directional preference in Imc may be internally prebuilt by elongating the vertical axis of the receptive field, making predators attack from the dorsal-ventral direction and conspecifics flying away in the ventral-dorsal direction, more salient for avians, which is of great ecological and physiological significance for survival.

Keywords: biological plausibility; computational model; isthmi pars magnocellularis; motion direction; pigeon.

Figure 1

The special structure of the isthmi pars magnocellularis (Imc) receptive field. (a) A two-dimensional grid shows an example Imc unit response strength to stimuli in visual space, while the purple ellipse represents the receptive field (RF, range approximately 30° horizontal and 75° vertical) fitted by a two-dimensional Gaussian model. The strength was normalized by the maximum. (b) A dot moving across the RF center in 16 directions spaced by 22.5°, was used to measure the Imc directional tuning curve, and the n-t direction was set as 0°. The gray oval represents the fitted RF. (c) Bright field photomicrograph of 40 μm Nissl-stained coronal section, the red arrow indicates the electrolytic lesioned recording site of a sample unit. Scale bar-500 μm. d—dorsal; n—nasal; t—temporal; v—ventral.

Figure 2

The directional tuning curve of the isthmi pars magnocellularis (Imc). (a) Spike raster plots of an example Imc unit to motion at 30°/s in 16 directions, and each direction was repeated for five trials. (b) Post-stimulus time histograms (PSTHs) of an example Imc unit to dots moving at 30°/s in the ventral-dorsal (90°), dorsal-ventral (270°), temporal-nasal (180°), and nasal-temporal (360°/0°) directions. (c,d) The directional tuning curve for the example Imc unit measured at 30°/s and 60°/s, respectively. Each motion direction was repeated in five trials, and the response strength for each direction was averaged across trials. The black line represents the mean response strength, while the width of the shadow indicates the standard deviation. Directional tuning curves were smoothed for display purposes (by spline interpolation). (e,f) The mean directional tuning curve of the population Imc units measured at 30°/s and 60°/s, respectively. Each motion direction was repeated in five trials, and the response strength for each direction was averaged across trials. The same format as (c).

Figure 3

Dynamic response characteristics of isthmi pars magnocellularis (Imc) units to motion at 30°/s. (a–d) Population Imc units’ post-stimulus time histograms (PSTHs) to dots moving at 30°/s in the ventral-dorsal, dorsal-ventral, temporal-nasal, and nasal-temporal directions. The PSTH for each Imc unit was normalized to the respective maximum, and the mean PSTHs shown here are the mean dynamic responses for the direction v-d direction, from the stimuli onset to 200 ms after the finish. Error bars show 95% confidence intervals. (e) Scatterplot of population Imc units’ response characteristics. The response duration was plotted against the maximal firing rate, for each Imc unit to motion in the ventral-dorsal (red dot), dorsal-ventral (blue dot), nasal-temporal (purple dot), and temporal-nasal (gray dot) directions at 30°/s. (f) Plots of directional tuning (gray curve), maximal instantaneous firing rate (orange curve), and response duration (blue curve), normalized by respective maxima for population Imc units, suggesting close relationships with each other. d—dorsal; n—nasal; t—temporal; v—ventral.

Figure 4

Dynamic response characteristics of isthmi pars magnocellularis (Imc) units to motion at 60°/s. (a–d) Population Imc unit post-stimulus time histograms (PSTHs) to dots moving at 60°/s in the ventral–dorsal, dorsal–ventral, temporal–nasal, and nasal–temporal directions. PSTH for each Imc unit was also normalized using the respective maxima, and the mean PSTHs showed the mean dynamic response for the v-d direction, from the onset of stimuli to 200 ms after the finish. Error bars show 95% confidence intervals. (e) Scatterplot of population Imc units’ response characteristics. The response duration was plotted against the maximal firing rate, for each Imc unit to motions in the ventral-dorsal (red dot), dorsal-ventral (blue dot), nasal-temporal (purple dot), and temporal-nasal (gray dot) directions at 60°/s, respectively. (f) Plot of directional tuning, maximal instantaneous firing rate, and response duration, normalized to the respective maximum for each population of Imc units, which shows a close relationship with each other. d—dorsal; n—nasal; t—temporal; v—ventral.

Figure 5

A computational model introduced to predict the preference for motion directions. (a) The receptive field of isthmi pars magnocellularis (Imc) units was modeled with an ellipse, while the moving path was simulated by an oblique line (α denotes the counterclockwise angle from nasal-temporal) going through the receptive field (RF) center. We assumed the response strength for each direction was linearly related with the intercept, D. (b) Predicted (orange) directional tuning curve followed the actual directional tuning (black) curve at 30°/s (Mean square error = 0.007). The width of the shadow indicates the standard deviation of actual data. (c) The modified model was added with a gap to simulate the “null” direction around the n-t direction. Thus, the motion path was reduced around n-t direction (such as the red path), but had no effect on other directions (such as the blue path). (d) The predicted (orange) directional tuning curve followed the actual directional tuning (black) curve at 30°/s (mean square error = 0.003). (e) The original model predicts the directional tuning curve (orange), when compared with the actual directional tuning (black) curve at 60°/s (mean square error = 0.003). (f) The modified model predicted the directional tuning curve (orange), when compared with the actual directional tuning (black) curve at 60°/s (mean square error = 0.001). d—dorsal; n—nasal; t—temporal; v—ventral.

Figure 6

Physiological evidence supporting the biological plausibility of the computational model. (a) The structure predicted for each Imc unit was plotted against the actual data that was measured with an ellipse function, showing a good fitting performance (mean square error = 0.3662). (b) The receptive field structure predicted with data collected at 30°/s was very close (mean square error = 0.0723) to results predicted with data obtained at 60°/s from the same Imc unit, which implied the speed independence of directional tuning in Imc.