BRIEF COMMUNICATIONS Contextual tuning of direction- selective retinal ganglion cells Chuan-Chin Chiao 1,2 & Richard H Masland 1 A direction-selective (DS) retinal ganglion cell responds well to a small object moving within its receptive field center, but less well when there is also a moving stimulus in the surrounding area; this has been described as tuning for local motion. We show here an additional selectivity, such that the surround has less effect if there is a discontinuity—that is, a difference in spatial phase, spatial frequency or velocity—between the center stimulus and that present in the surround. Many studies have investigated the behavior of on-off DS cells in the rabbit retina, but most have focused on the possible mechanism of the directional discrimination 1–6 . In the current study, we focused instead on the behavior of these cells in response to extended stimuli, with the ultimate aim of understanding their role in a rabbit’s natural vision. We found unexpected contextual effects on the responses of the on- off DS cells. These would not be predicted from the previously described ‘silent surround’ 1,2,7,8 and suggest a novel tuning of the cells’ responses to natural stimuli. On-off DS cells are strongly stimulated by a small stimulus crossing the receptive field, but a large, high-frequency grating stimulates the cell only weakly 2,9 . This selectivity has two separate causes (Fig. 1). First, a stimu- lus traversing the receptive field in the preferred direction leaves behind it a trail of inhibition 2,10 . This prevents responses to a second stimulus that follows close behind. The net effect of this is a rapid fall-off in response at moderate spatial (or temporal) frequencies. Second, the on-off DS cells have a powerful, movement-sensitive surround (Fig. 1b). When the bars of a grating extend into the surround, they suppress the response to stim- ulation of the center, in accord with the previous finding 1,2,7 that motion of any kind in the surround suppresses the center’s response (for a differ- ent optic flow field, see Supplementary Fig. 1 online). However, we also found that DS cells are influenced by differences in phase, spatial frequency and velocity between the receptive field center and the surrounding field. The response to patterned stimuli thus depended on the local character of the pattern. An optimal grating was drifted across the receptive field center and a second grating was moved in the surround, with the same direction and velocity, but varying phase. The DS cells were sensitive to the spatial phase of the center relative to surround stimulus. When the sur- round grating was in phase with the center, the response to the cen- ter stimulus was depressed to 20% of the center-alone response. When it was 180° out of phase, the depression of the center response was significantly relieved (Fig. 2a–c). A similar result was obtained when a discontinuity between the center and surround stimuli was induced by a difference in spatial fre- quency. When the spatial frequency and phase of the center and sur- round stimuli matched (velocity held the same), the response to the optimal center stimulus was depressed by more than half. With an increasing difference in spatial frequency, the response was restored. Another form of contrast between the center and surround is cre- ated by giving the center and surround the same spatial frequency and varying the velocity (temporal frequency) of the surround grating. When the temporal frequency of center and surround were the same (the stimuli were presented in phase under this condition), the response to the center stimulus was depressed. When the surround had higher or lower temporal frequency, the suppression of the response was relieved. The on-off DS cell thus appears to be sensitive to discontinuities between a stimulus moving in the receptive field center and stimuli mov- ing in the receptive field surround: the same total amount of stimulation to the surround can create greater or lesser suppression of the center response, depending on the pattern and spatial relationship between cen- 1 Howard Hughes Medical Institute, Massachusetts General Hospital, Harvard Medical School, 50 Blossom Street, Wellman 429, Boston, Massachusetts 02114, USA. 2 Department of Life Science, National Tsing Hua University, Hsinchu, 300, Taiwan. Correspondence should be addressed to R.H.M. (masland@helix.mgh.harvard.edu). Published online 2 November 2003 ; doi:10.1038/nn1147 NATURE NEUROSCIENCE VOLUME 6 | NUMBER 12 | DECEMBER 2003 1251 Figure 1 Spatial frequency tuning curves of the on-off DS cells for restricted or extended stimuli. Extracellular recording from isolated rabbit retinas was carried out by conventional techniques 11 that were approved by the animal use committee of the Massachusetts General Hospital (Supplementary Methods online). A square wave grating moving in the preferred direction at 1.88 Hz was used for all cells. The square wave grating covered (a) the receptive field center of the DS cell or (b) both center and surround of the receptive field. We measured the mean rate of firing of the cells during a 3-s exposure to the moving stimulus. Responses in both conditions (c) were normalized to the maximum of each cell’s response to stimulation of the receptive field center alone (mean ± s.e.m.; n = 16 cells). The peak response was 35.5 spikes/s. In all figures, the orange arrow indicates the preferred direction, and the green arrow shows the direction of stimulus movement. Scale bar, 500 µm. © 2003 Nature Publishing Group http://www.nature.com/natureneuroscience