Orientation maps of subjective contours in visual cortex

Science

Volumn 274, Issue 5295, 1996, Pages 2110-2115

  Sheth, Bhavin R ^a   Sharma, Jitendra ^a   Rao, S Chenchal ^a   Sur, Mriganka ^a  

^a MASSACHUSETTS INSTITUTE OF TECHNOLOGY   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

ANIMAL CELL; ARTICLE; CAT; GENE MAPPING; LUMINANCE; NONHUMAN; ORIENTATION; PERCEPTION; PRIORITY JOURNAL; SINGLE UNIT ACTIVITY; STIMULUS RESPONSE; VISUAL CORTEX;

ANIMALS; BRAIN MAPPING; CATS; FORM PERCEPTION; IMAGE PROCESSING, COMPUTER-ASSISTED; LIGHT; NEURONS; VISUAL CORTEX;

ANIMALIA; FELIS CATUS;

EID: 0030468373     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.274.5295.2110     Document Type: Article

References (46)

1. L. Schumann, Z. Psychol. 23, 1 (1900); G. Kanizsa, Riv. Psichol. 49, 7 (1955); Organization in Vision: Essays on Gestalt Perception, G. Kanizsa, Ed. (Praeger, New York, 1979), p. 192; J. J. Koenderink, Perception 13, 321 (1984).
2. L. Schumann, Z. Psychol. 23, 1 (1900); G. Kanizsa, Riv. Psichol. 49, 7 (1955); Organization in Vision: Essays on Gestalt Perception, G. Kanizsa, Ed. (Praeger, New York, 1979), p. 192; J. J. Koenderink, Perception 13, 321 (1984).
3. L. Schumann, Z. Psychol. 23, 1 (1900); G. Kanizsa, Riv. Psichol. 49, 7 (1955); Organization in Vision: Essays on Gestalt Perception, G. Kanizsa, Ed. (Praeger, New York, 1979), p. 192; J. J. Koenderink, Perception 13, 321 (1984).
4. L. Schumann, Z. Psychol. 23, 1 (1900); G. Kanizsa, Riv. Psichol. 49, 7 (1955); Organization in Vision: Essays on Gestalt Perception, G. Kanizsa, Ed. (Praeger, New York, 1979), p. 192; J. J. Koenderink, Perception 13, 321 (1984).
5. R. von der Heydt, E. Peterhans, G. Baumgartner, Science 224, 1260 (1984); R. von der Heydt and E. Peterhans, J. Neurosci. 9, 1731 (1989).
6. R. von der Heydt, E. Peterhans, G. Baumgartner, Science 224, 1260 (1984); R. von der Heydt and E. Peterhans, J. Neurosci. 9, 1731 (1989).
7. D. H. Grosof, R. M. Shapley, M. J. Hawken, Nature 365, 550 (1993).
8. M. Bravo, R. Blake, S. Morrison, Vision Res. 28, 861 (1988).
9. P. De Weerd, P. E. Vandenbussche, B. De Bruyn, G. A. Orban, Behav. Brain Res 39. 1 (1990).
10. 2 volume at ∼4% at a partial pressure of 30 ± 3 mm Hg. The animal's electroencephalogram and heart rate were monitored continuously to ensure adequate anesthesia. Craniotomy followed by durotomy was performed to expose visual cortex. For imaging area V2, we centered our chamber at A4 and recorded from an area that extended approximately from A3 to A7 in the anteroposterior direction and between L0.5 to L3.5 in the medidateral direction. For imaging V1, the chamber was centered at P5 and a similar expanse of cortex was exposed. For imaging V1 and V2 simultaneously, the chamber was centered at AO [R. J. Tusa, L. A. Palmer, A. C. Rosenquist, J. Comp. Neurol. 177, 213 (1978); R. J. Tusa, A.C. Rosenquist, L. A Palmer, ibid. 185, 657 (1979)].
11. 2 volume at ∼4% at a partial pressure of 30 ± 3 mm Hg. The animal's electroencephalogram and heart rate were monitored continuously to ensure adequate anesthesia. Craniotomy followed by durotomy was performed to expose visual cortex. For imaging area V2, we centered our chamber at A4 and recorded from an area that extended approximately from A3 to A7 in the anteroposterior direction and between L0.5 to L3.5 in the medidateral direction. For imaging V1, the chamber was centered at P5 and a similar expanse of cortex was exposed. For imaging V1 and V2 simultaneously, the chamber was centered at AO [R. J. Tusa, L. A. Palmer, A. C. Rosenquist, J. Comp. Neurol. 177, 213 (1978); R. J. Tusa, A.C. Rosenquist, L. A Palmer, ibid. 185, 657 (1979)].
12. Techniques for intrinsic signal imaging were similar to those described by Grinvald et al. [A. Grinvald, E. Lieke R. D. Frostig, C. D. Gilbert, T. N. Wiesel, Nature 324, 351 (1986)] and used by us previously (26). A stainless-steel recording chamber was attached to the skull surrounding the craniotomy, filled with sili-cons oil, and then sealed with a quartz plate. A video camera (CCD-5024N Bishke, Japan, RS-170, >60 dB signal to noise ratio) consisting of a 655 by 480 array of pixels equipped with a tandem-lens macroscope [E. H. Ratzlaff and A. Grinvald, J. Neurosci. Methods 36, 127 (1991)] was positioned over the cortex. This arrangement gave a magnification of 75 pixels per millimeter. Data were collected by means of an imaging system (Optical Imaging). The camera signal was amplified by a video enhancement amplifier; a baseline image was subtracted from each stimulus response image in analog form and then digitized by an 8-bit analog-todigital converter (Matrox) installed on a 486-66 PC. Initially, a reference map of the blood vessel pattern at the surface of the cortex was obtained by means of light filtered at 550 ± 40 nm (Ealing). The camera was then focused 300 μm below the surface of the cortex. Light from a 100-W tungsten-halogen light source driven by a dc power supply (Kepco) was passed through a 610-nm filter and used to illuminate the cortex during data collection. Frames were summed between 0.9 to 3.6 s after stimulus onset, corresponding to the time of maximum signal as determined by our previous experiments (26). Data were analyzed with the use of in-house programs written in C++ (Borland) and IDL Research Systems.
13. Techniques for intrinsic signal imaging were similar to those described by Grinvald et al. [A. Grinvald, E. Lieke R. D. Frostig, C. D. Gilbert, T. N. Wiesel, Nature 324, 351 (1986)] and used by us previously (26). A stainless-steel recording chamber was attached to the skull surrounding the craniotomy, filled with sili-cons oil, and then sealed with a quartz plate. A video camera (CCD-5024N Bishke, Japan, RS-170, >60 dB signal to noise ratio) consisting of a 655 by 480 array of pixels equipped with a tandem-lens macroscope [E. H. Ratzlaff and A. Grinvald, J. Neurosci. Methods 36, 127 (1991)] was positioned over the cortex. This arrangement gave a magnification of 75 pixels per millimeter. Data were collected by means of an imaging system (Optical Imaging). The camera signal was amplified by a video enhancement amplifier; a baseline image was subtracted from each stimulus response image in analog form and then digitized by an 8-bit analog-todigital converter (Matrox) installed on a 486-66 PC. Initially, a reference map of the blood vessel pattern at the surface of the cortex was obtained by means of light filtered at 550 ± 40 nm (Ealing). The camera was then focused 300 μm below the surface of the cortex. Light from a 100-W tungsten-halogen light source driven by a dc power supply (Kepco) was passed through a 610-nm filter and used to illuminate the cortex during data collection. Frames were summed between 0.9 to 3.6 s after stimulus onset, corresponding to the time of maximum signal as determined by our previous experiments (26). Data were analyzed with the use of in-house programs written in C++ (Borland) and IDL Research Systems.
14. 2. All stimuli were shown binocularly. Subjective grating stimuli of four different orientations (0°, 45°, 90°, and 135°) were randomly interleaved with square-wave luminance grating stimuli and presented 80 to 100 times. Luminance-defined inducing lines, 1 to 2 pixels wide, were orthogonal to the subjective orientation for all subjective grating stimuli. Gratings were drifted normal to the subjective edge orientation and parallel to the orientation of the inducing lines in both directions separately. Eye position and area centrali of both eyes were checked at the start of imaging by use of a reverse opthalmoscope to project an image of the retinal vasculature onto the screen.
15. Orientation maps obtained with luminance gratings composed of thin lines (of the same width as the subjective grating inducing lines) were identical to orientation maps obtained with thicker bars (Fig. 2C) and relatively independent of grating spatial frequency.
16. -3 units) for these pixels.
17. We vectorially summed the responses of each pixel to all orientations of luminance and subjective gratings separately, obtained the pixel's resultant orientation preferences for both, and derived the difference between the two values.
18. Single-unit experiments were carried out in 12 cats. Single units were recorded with parylene-insulated tungsten microelectrodes. Responses were conventionally amplified, displayed, and stored. Stimulus conditions were the same as in the imaging studies.
19. Receptive field properties of a subset of V2 cells whose luminance and subjective orientation preferences differed by ±45° or less (17 cells) were studied in detail. Most of these cells (13 cells) were complex; four cells were simple. Only 2 of the cells were end-stopped; the remainder (15 cells) were non-end-stopped. Twenty-four cells whose locations were identified were encountered at a variety of depths (160 to 1800 μm); 11 cells were encountered in the superficial layers and 13 cells in the deep layers.
20. We correlated the responses of single neurons (n = 7) to luminance and subjective gratings with the optically imaged maps. Each of the regions in the orientation difference map - in which orientation preferences for luminance and subjective orientations are similar, orthogonal, or in-between - contained neurons with matching responses.
21. J. A. Movshon, I. D. Thompson, D. J. Tolhurst, J. Physiol. 283, 101 (1978); K. Albus, Exp. Brain Res. 24, 159 (1975).
22. J. A. Movshon, I. D. Thompson, D. J. Tolhurst, J. Physiol. 283, 101 (1978); K. Albus, Exp. Brain Res. 24, 159 (1975).
23. -3 units) for these pixels. Thus, the response to subjective gratings in V1 was stronger for subjective gratings than for luminance gratings for this subset of pixels, although they occupied a small portion of V1.
24. Receptive field properties of a subset of V1 cells whose luminance and subjective orientation preferences differed by ±45° or less were studied in detail. Three cells were complex; one was a simple cell. Two cells were end-stopped, four cells were non-end-stopped. Six cells whose locations were identified were encountered at depths ranging from 300 to 1200 μm from the surface.
25. Responses to subjective gratings cannot be explained by responses to line terminations. We have shown that cells that share the same orientation preference for luminance and subjective gratings may respond optimally to a grating composed of a grid of dots (line ends) with the same parameters (spatial frequency, temporal frequency) as the subjective grating but of an entirely different orientation. Some cells that are tuned to the same orientation of luminance and subjective gratings as well as the dot grid show a sharper tuning, or a higher response, or both, to a subjective grating than to a dot grid. A response to Fourier energy alone cannot explain why the response of a cell to an intermediate subjective orientation is higher than the response at 90° orientation difference (orientation of the inducing lines in the subjective grating), because the energy along these orientations is less than the energy along the inducing-line Qrientation. Moreover, most such cells respond optimally to a single intermediate orientation, despite the presence of equal Fourier energy in the stimulus along both the optimal orientation and its orthogonally oriented counterpart.
26. V2 and V1 in cats are anatomically distinct [R. Otsuka and R. Hassler, Arch. Psychiatr. Nervenkr. 203, 212 (1962); D. J. Price, Exp. Brain Res. 58, 125 (1985); G. H. Kageyama and M. Wong-Riley, J. Comp. Neurol. 243, 182 (1986); B. R. Payne, Vis. Neurosci. 6, 445 (1990); A. L. Humphrey, M. Sur, D. J. Uhlrich, S. M. Sherman, J. Comp. Neurol. 233, 190 (1985)]. To physiologically assess the boundary between V1 and V2, we used luminance gratings with different parameters: a grating of high spatial frequency (0.5 cycle per degree) and low drift velocity (4° per second), which V1 neurons prefer, and a grating of low spatial frequency (0.15 cycle per degree) and high drift velocity (12° per second), which V2 neurons prefer [(21)]. By distinguishing the portion of the optically imaged region that responded preferentially to the slower drift rate, higher spatial frequency stimulus (V1) compared to the higher drift rate, lower spatial frequency stimulus (V2), we could locate the V1/V2 boundary accurately. The physiological border coincided with the anatomical border between areas 17 and 18 as demonstrated by marker lesions and histology.
27. V2 and V1 in cats are anatomically distinct [R. Otsuka and R. Hassler, Arch. Psychiatr. Nervenkr. 203, 212 (1962); D. J. Price, Exp. Brain Res. 58, 125 (1985); G. H. Kageyama and M. Wong-Riley, J. Comp. Neurol. 243, 182 (1986); B. R. Payne, Vis. Neurosci. 6, 445 (1990); A. L. Humphrey, M. Sur, D. J. Uhlrich, S. M. Sherman, J. Comp. Neurol. 233, 190 (1985)]. To physiologically assess the boundary between V1 and V2, we used luminance gratings with different parameters: a grating of high spatial frequency (0.5 cycle per degree) and low drift velocity (4° per second), which V1 neurons prefer, and a grating of low spatial frequency (0.15 cycle per degree) and high drift velocity (12° per second), which V2 neurons prefer [(21)]. By distinguishing the portion of the optically imaged region that responded preferentially to the slower drift rate, higher spatial frequency stimulus (V1) compared to the higher drift rate, lower spatial frequency stimulus (V2), we could locate the V1/V2 boundary accurately. The physiological border coincided with the anatomical border between areas 17 and 18 as demonstrated by marker lesions and histology.
28. V2 and V1 in cats are anatomically distinct [R. Otsuka and R. Hassler, Arch. Psychiatr. Nervenkr. 203, 212 (1962); D. J. Price, Exp. Brain Res. 58, 125 (1985); G. H. Kageyama and M. Wong-Riley, J. Comp. Neurol. 243, 182 (1986); B. R. Payne, Vis. Neurosci. 6, 445 (1990); A. L. Humphrey, M. Sur, D. J. Uhlrich, S. M. Sherman, J. Comp. Neurol. 233, 190 (1985)]. To physiologically assess the boundary between V1 and V2, we used luminance gratings with different parameters: a grating of high spatial frequency (0.5 cycle per degree) and low drift velocity (4° per second), which V1 neurons prefer, and a grating of low spatial frequency (0.15 cycle per degree) and high drift velocity (12° per second), which V2 neurons prefer [(21)]. By distinguishing the portion of the optically imaged region that responded preferentially to the slower drift rate, higher spatial frequency stimulus (V1) compared to the higher drift rate, lower spatial frequency stimulus (V2), we could locate the V1/V2 boundary accurately. The physiological border coincided with the anatomical border between areas 17 and 18 as demonstrated by marker lesions and histology.
29. V2 and V1 in cats are anatomically distinct [R. Otsuka and R. Hassler, Arch. Psychiatr. Nervenkr. 203, 212 (1962); D. J. Price, Exp. Brain Res. 58, 125 (1985); G. H. Kageyama and M. Wong-Riley, J. Comp. Neurol. 243, 182 (1986); B. R. Payne, Vis. Neurosci. 6, 445 (1990); A. L. Humphrey, M. Sur, D. J. Uhlrich, S. M. Sherman, J. Comp. Neurol. 233, 190 (1985)]. To physiologically assess the boundary between V1 and V2, we used luminance gratings with different parameters: a grating of high spatial frequency (0.5 cycle per degree) and low drift velocity (4° per second), which V1 neurons prefer, and a grating of low spatial frequency (0.15 cycle per degree) and high drift velocity (12° per second), which V2 neurons prefer [(21)]. By distinguishing the portion of the optically imaged region that responded preferentially to the slower drift rate, higher spatial frequency stimulus (V1) compared to the higher drift rate, lower spatial frequency stimulus (V2), we could locate the V1/V2 boundary accurately. The physiological border coincided with the anatomical border between areas 17 and 18 as demonstrated by marker lesions and histology.
30. V2 and V1 in cats are anatomically distinct [R. Otsuka and R. Hassler, Arch. Psychiatr. Nervenkr. 203, 212 (1962); D. J. Price, Exp. Brain Res. 58, 125 (1985); G. H. Kageyama and M. Wong-Riley, J. Comp. Neurol. 243, 182 (1986); B. R. Payne, Vis. Neurosci. 6, 445 (1990); A. L. Humphrey, M. Sur, D. J. Uhlrich, S. M. Sherman, J. Comp. Neurol. 233, 190 (1985)]. To physiologically assess the boundary between V1 and V2, we used luminance gratings with different parameters: a grating of high spatial frequency (0.5 cycle per degree) and low drift velocity (4° per second), which V1 neurons prefer, and a grating of low spatial frequency (0.15 cycle per degree) and high drift velocity (12° per second), which V2 neurons prefer [(21)]. By distinguishing the portion of the optically imaged region that responded preferentially to the slower drift rate, higher spatial frequency stimulus (V1) compared to the higher drift rate, lower spatial frequency stimulus (V2), we could locate the V1/V2 boundary accurately. The physiological border coincided with the anatomical border between areas 17 and 18 as demonstrated by marker lesions and histology.
31. D. H. Hubel and T. N. Wiesel, J. Physiol. (London) 165, 559 (1963); ibid. 195, 215 (1968); T. Bonhoeffer and A. Grinvald, J. Neurosci. 13, 4157 (1993).
32. D. H. Hubel and T. N. Wiesel, J. Physiol. (London) 165, 559 (1963); ibid. 195, 215 (1968); T. Bonhoeffer and A. Grinvald, J. Neurosci. 13, 4157 (1993).
33. D. H. Hubel and T. N. Wiesel, J. Physiol. (London) 165, 559 (1963); ibid. 195, 215 (1968); T. Bonhoeffer and A. Grinvald, J. Neurosci. 13, 4157 (1993).
34. T. Bonhoeffer, D.-S. Kim, D. Malonek, D. Shoham, A. Grinvald, Eur. J. Neurosci. 7, 1973 (1995).
35. C. Redies, J. M. Crook, O. D. Creutzfeldt, Exp. Brain Res. 61, 469 (1986).
36. See also A. F, Rossi, C. D. Rittenhouse, M. A. Paradiso, Science 273, 1104 (1996); V. A. Lamme, J. Neurosci. 15, 1605 (1995); D. A.Leopold and N. K. Logothetis, Nature 379, 549 (1996).
37. See also A. F, Rossi, C. D. Rittenhouse, M. A. Paradiso, Science 273, 1104 (1996); V. A. Lamme, J. Neurosci. 15, 1605 (1995); D. A.Leopold and N. K. Logothetis, Nature 379, 549 (1996).
38. See also A. F, Rossi, C. D. Rittenhouse, M. A. Paradiso, Science 273, 1104 (1996); V. A. Lamme, J. Neurosci. 15, 1605 (1995); D. A.Leopold and N. K. Logothetis, Nature 379, 549 (1996).
39. V1 and V2 in cats differ in their thalamic inputs as well: V1 receives input from X and Y cells located in the A-laminae of the lateral geniculate nucleus, whereas V2 receives inputs from Y cells alone [reviewed in S. M. Sherman and P. D. Spear, Physiol. Rev. 62, 738 (1982)].
40. R. L. Gregory, Nature 238, 51 (1972); ibid. 199, 678 (1963); I. Rock and R. Anson, Perception 8, 665 (1979).
41. R. L. Gregory, Nature 238, 51 (1972); ibid. 199, 678 (1963); I. Rock and R. Anson, Perception 8, 665 (1979).
42. R. L. Gregory, Nature 238, 51 (1972); ibid. 199, 678 (1963); I. Rock and R. Anson, Perception 8, 665 (1979).
43. S. C. Rao, L. J. Toth, B. R. Sheth, M. Sur, Soc. Neurosci. Abstr. 20, 836 (1994); L. J. Toth, S. C. Rao, D.-S. Kim, D. Somers, M. Sur, Proc. Natl. Acad. Sci. U.S.A. 93, 9869 (1996).
44. S. C. Rao, L. J. Toth, B. R. Sheth, M. Sur, Soc. Neurosci. Abstr. 20, 836 (1994); L. J. Toth, S. C. Rao, D.-S. Kim, D. Somers, M. Sur, Proc. Natl. Acad. Sci. U.S.A. 93, 9869 (1996).
45. S. Nelson, L. J.Toth, B. Sheth, M. Sur Science 265, 774 (1994).
46. We thank D.-S. Kim and L. Toth for assistance and E Adelson, B. Anderson, and P. Sinha for critical reading of the manuscript. Supported by NIH grant EY07023. J.S. was supported by a fellowship from the Fogarty International Center of the NIH.