Mechanisms of heading perception in primate visual cortex

Science

Volumn 273, Issue 5281, 1996, Pages 1544-1547

  Bradley, David C ^a   Maxwell, Marsha ^a   Andersen, Richard A ^a   Banks, Martin S ^b   Shenoy, Krishna V ^a  

^a CALIFORNIA INSTITUTE OF TECHNOLOGY   (United States)

^b UNIVERSITY OF CALIFORNIA   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

ARTICLE; DIAGNOSTIC ACCURACY; EYE MOVEMENT; IMAGE ANALYSIS; PERCEPTIVE DISCRIMINATION; PRIMATE; PRIORITY JOURNAL; VISUAL CORTEX; VISUAL FIELD;

EID: 0029778317     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.273.5281.1544     Document Type: Article

References (30)

1. J. J. Gibson, Perception of the Visual World (Houghton Mifflin, Boston, 1950).
2. W. H. Warren and D. J. Hannon, Nature 336, 162 (1988).
3. C. S. Royden, M. S. Banks, J. A. Crowell, ibid. 360, 583 (1992).
4. The eye-movement signal could be a copy of the smooth-pursuit command sent to the heading computation site, or a proprioceptive signal from the eye muscles and the periorbital tissue.
5. H. Saito et al., J. Neurosci. 6, 145 (1986).
6. M. S. Graziano, R. A. Andersen, R. J. Snowden, ibid. 14, 54 (1994).
7. H. Sakata, H. Shibutani, K. Kawano, J. Neurophysiol. 49, 1364 (1983).
8. H. Komatsu and R. H. Wurtz, ibid. 61, 580 (1988).
9. W. T. Newsome, R. H. Wurtz, H. Komatsu, ibid., p. 604.
10. C. J. Duffy and R. H. Wurtz, J. Neurosci. 15, 5192 (1995).
11. The monkey was seated in a chair with its head fixed, facing a 100° by 100° projection screen. After maintaining fixation within a 5° by 5° window for the required time (3 s), the monkey was rewarded with a drop of water. Eye position was monitored with a scierai search coil [S. J. Judge et al., Vision Res. 20, 535 (1980)]. All procedures with animals were approved by the Caltech Institutional Animal Care and Use Committee.
12. Expanding random dot fields were generated by simulating the approach to a wall at 38 cm/s at a distance of 57 cm. This condition generated a radial speed of 15.7° per second at an eccentricity of 30° from the focus. Dot lifetimes were 300 ms (to limit acceleration cues), after which they were renewed at random locations. The stimulus was 50° by 50° square and contained 77 dots. The focus position was varied in 10° steps from -40° to 40° along an axis parallel to the neuron's preferred pursuit direction (distances are relative to the center of the stimulus, which was approximately centered in the neuron's spatial receptive field). For the ±30° and ±40° focus positions, the focus was situated outside the visible stimulus. We generated the stimulus for the simulated eye movements by moving the entire stimulus on the screen (by computer computation of the dot positions); this is equivalent to adding a constant-velocity component to each point in the flow field, which shifts the focus and causes the stimulus borders to move across the screen. The room was completely dark except for the stimulus.
13. Before studying a neuron's focus tuning, we determined its preferred pursuit direction by having the monkey follow a point moving at 0°, 45°, 90°, and so on, up to 315°. The direction eliciting the strongest response was taken as the preferred direction; the opposite direction was designated antipreferred.
14. Single neurons were monitored extracellularly with varnish-coated tungsten microelectrodes advanced dorsoventrally through the dura and into area MSTd, which was identified on the basis of receptive field size (most were >50° wide and crossed over the vertical meridian), flow preference (for example, expansion and rotation), position invariance, and modulation during smooth pursuit. Spike times were stored on a personal computer for subsequent analysis. All data are expressed as the mean firing rate during the middle 500 ms of a 1-s stimulus-presentation period. Results are based on a sample of 139 neurons from one hemisphere. These results are corroborated by recent data from a second animal (n = 18; K. S. Shenoy, D. C. Bradley, R. A. Andersen, Soc. Neurosci. Abstr., in press).
15. In practice, all stimuli were presented in pseudorandom order.
16. "Screen coordinates" is an operational term. Because the monkey's eyes and head were in a constant position (except for the small eye displacement during pursuit), we cannot tell whether MSTd cells code for heading in eye, head, body, or world coordinates. Our results imply simply that these neurons account for eye movement, the first step necessary for computing the heading.
17. If we arbitrarily define retinal cells as shifting by -20°, -10°, or 0°, intermediate cells by 10° or 20°, and heading cells by 30°, 40°, or 50°, the percentages of the different cell types are about 42%, 31%, and 27%, respectively, where the shift estimate was based on maximum cross-correlation between the fixed- and moving-eye curves.
18. H. C. Longuet-Higgins and K. Prazdny, Proc. R. Soc. London Ser. B 208, 385 (1980).
19. W. H. Warren, in Perception of Space and Motion (Academic Press, New York, 1995), chap. 3, pp. 263-325.
20. Because we did not correct for tangent error on the screen, dot speeds (for the pursuit target as well as the stimulus dots) were slightly underestimated at eccentric positions. However, this error was only ∼8% at 12.5° eccentricity, ∼20% in the corner (35° eccentricity), and 0 in the center. Therefore, the real and simulated pursuit conditions differed slightly in terms of the distribution of speeds across the retina.
21. c for a given relative shift between curves, we correlated the responses on one curve with the responses on the other curve, pairing by x-axis value. Only overlapping regions of the curves could be cross-correlated. MSD was computed between corresponding data points on the two curves, again, only for overlapping regions.
22. A. Mack and E. Herman, Vision Res. 18, 55 (1978).
23. T. C. A. Freeman, J. A. Crowell, M. S. Banks, Invest. Ophthalmol. Vis. Sci. 37, S454 (1996).
24. Heading cell receptive fields also showed gain changes. In fitting our model (Fig. 4) to the heading cell data, these gain changes were taken into account, along with the receptive field shifts. It is uncertain whether these gain changes encode information about the heading; it is possible that they simply average to zero.
25. Sine functions were used because different parts of a sine function can approximate either a gaussian or a sigmoid function. Each sine function was characterized by amplitude, frequency, and phase, as well as by two "gain" parameters that were applied in the pursuit condition (one gain for each pursuit direction). All parameters were adjusted simultaneously, fitting three receptive fields (fixed-eye and two pursuit directions) concurrently, with the use of nonlinear least-squares regression. Analysis was done on a subset of 36 neurons, the receptive fields of which shifted during eye movements.
26. The speed of rotating patterns was scaled so that dot speed was identical to the expanding patterns (15.7°/s) at 30° eccentricity. Eye movements cause the focus in a rotating pattern to shift orthogonally to the eye movement. Therefore, for rotating patterns the focus was also varied orthogonally - otherwise it would not be possible to measure relative shifts between fixed- and moving-eye receptive fields.
27. Shifts were calculated on the basis of MSD.
28. Because the retinal focus shift depends on the pursuit speed and the rate of image expansion, receptive field shifts of different sizes are required to compute heading under different conditions. Because we tested only one pursuit and one expansion rate, we do not know whether individual receptive fields shift by varying amounts or whether a population code is used to read out the heading from different neurons depending on the pursuit and expansion rates.
29. Preliminary psychophysical experiments in our lab suggest that humans compensate at least partly for eye movements while pursuing across rotating stimuli.
30. We are grateful to D. Ward and B. Gillikin for technical assistance and to W. Warren and J. Crowell for helpful comments. This work was funded by the National Eye Institute, the Sloan Foundation for Theoretical Neurobiology at Caltech, the Office of Naval Research, and the Air Force Office of Scientific Research.