articles© 2003 Nature Publish<strong>in</strong>g Group http://www.nature.com/natureneuroscienceficial respiration was ma<strong>in</strong>ta<strong>in</strong>ed with 30% O 2 and 70% N 2 O at25 breaths/m<strong>in</strong>, and <strong>the</strong> stroke volume was adjusted accord<strong>in</strong>g <strong>to</strong> <strong>the</strong>cat’s weight. All procedures complied with <strong>the</strong> National Institutes ofHealth Guide for <strong>the</strong> Care and Use of Labora<strong>to</strong>ry Animals and wereapproved by <strong>the</strong> University of California Animal Care Committee. Visualstimuli were generated by a computer with two high-resolution graphicsboards that ran cus<strong>to</strong>m software. Images were displayed on a pair ofvideo moni<strong>to</strong>rs that <strong>the</strong> cat views dichoptically by means of beam splitters(mean lum<strong>in</strong>ance, 23 cd/m 2 ). Action potentials were discrim<strong>in</strong>atedby cus<strong>to</strong>m-made software and time-stamped with 40 µs resolution.Receptive field mapp<strong>in</strong>g. The results from drift<strong>in</strong>g grat<strong>in</strong>g runs wereused <strong>to</strong> determ<strong>in</strong>e optimal stimulus parameters for receptive field mapp<strong>in</strong>gwith dichoptic, one-dimensional, b<strong>in</strong>ary m-sequence noise. Sixteenadjacent bars were presented <strong>to</strong> each eye at optimal orientation. Thewidth of <strong>the</strong> bars was approximately one-fourth <strong>the</strong> period of <strong>the</strong> optimalspatial frequency. The length of <strong>the</strong> bar was equal <strong>to</strong> sixteen times <strong>the</strong>width. This square pattern was centered over <strong>the</strong> receptive field. Each of<strong>the</strong> sixteen bars was ei<strong>the</strong>r bright or dark, and <strong>the</strong> mean lum<strong>in</strong>ance of<strong>the</strong> bars constituted <strong>the</strong> background. One stimulus pattern lasted forthree frames (39 ms).Each spike tra<strong>in</strong> was cross-correlated with <strong>the</strong> stimulus sequence ateach location of b<strong>in</strong>ocular comb<strong>in</strong>ation of stimuli (which varies <strong>in</strong> <strong>depth</strong>and position along a fron<strong>to</strong>-parallel plane) <strong>to</strong> obta<strong>in</strong> a two-dimensionalb<strong>in</strong>ocular <strong>in</strong>teraction map for a particular time delay. Excita<strong>to</strong>ry response<strong>to</strong> bars of <strong>the</strong> same contrast or of opposite contrast was represented as apositive or a negative number, respectively. The two-dimensional b<strong>in</strong>ocular<strong>in</strong>teraction maps were reduced <strong>to</strong> one-dimensional disparity tun<strong>in</strong>gdata by <strong>in</strong>tegration along l<strong>in</strong>es of equal disparity. This was repeated forall time delays of <strong>in</strong>terest (0–200 ms) <strong>in</strong> <strong>in</strong>crements of 5 ms. Time delaywas measured relative <strong>to</strong> <strong>the</strong> middle po<strong>in</strong>t <strong>in</strong> <strong>the</strong> three frame pattern. Theoptimal time delay was def<strong>in</strong>ed as that which produced <strong>the</strong> greatest rootmean squared (RMS) signal strength. The time slices that preceded andfollowed optimal by 20 ms were chosen for fur<strong>the</strong>r analysis. These timeslices were chosen because <strong>the</strong> temporal resolution of <strong>the</strong> stimulus is40 ms, and symmetrical slices on ei<strong>the</strong>r side and close <strong>to</strong> optimal yieldhigh signal-<strong>to</strong>-noise values and best represent receptive field dynamics.A detailed description of <strong>the</strong>se procedures is provided elsewhere 2–4,25 .Curve fitt<strong>in</strong>g. The disparity tun<strong>in</strong>g curves were fitted with a Gabor function:S(d) = exp(–(d – d o ) 2 /2σ s 2 )cos(2πf ds (d – d o ) + φ),where d is disparity, d o is <strong>the</strong> center position, σ s is <strong>the</strong> disparity range(size) parameter, f ds is <strong>the</strong> disparity frequency, and φ is <strong>the</strong> phase. If <strong>the</strong>rewere <strong>in</strong>sufficient spikes (<strong>the</strong> Gabor fit was not good), that neuron wasnot <strong>in</strong>cluded <strong>in</strong> <strong>the</strong> data set. The criteria for a good fit were that (i)R 2 > 0.8, where R 2 is <strong>the</strong> coefficient of determ<strong>in</strong>ation, and (ii) <strong>the</strong> disparityfrequency and disparity range parameters were well-constra<strong>in</strong>edbased on <strong>the</strong> confidence <strong>in</strong>terval provided by <strong>the</strong> Levenberg-Marquardtalgorithm 18 . These criteria are more suitable than <strong>the</strong> signal/noise ratioor <strong>the</strong> <strong>to</strong>tal quantity of spikes. The percentage changes <strong>in</strong> disparity resolution(frequency) and range (size) with correlation were calculated as[100 × (P +20ms – P –20ms )/P –20ms ]where P is ei<strong>the</strong>r a frequency or a size parameter used <strong>to</strong> measure dynamicalterations <strong>in</strong> <strong>the</strong> <strong>coarse</strong>ness of disparity tun<strong>in</strong>g. The maximum amplitudeof <strong>the</strong> signal at optimal time delay was used <strong>to</strong> normalize allamplitudes from <strong>the</strong> o<strong>the</strong>r time delays (Figs. 1 and 2). Summary data <strong>in</strong>scatter plots were fit with robust regressions, which m<strong>in</strong>imize <strong>the</strong> <strong>in</strong>fluenceof outliers 26 .Neural cross-correlation. Two or more cells were recorded from ei<strong>the</strong>r<strong>the</strong> same electrode or adjacent electrodes for which <strong>the</strong> difference <strong>in</strong> cortical<strong>depth</strong> did not exceed 500 µm. Data for this study <strong>in</strong>cludes two ormore b<strong>in</strong>ocular disparity tuned cells with similar orientation and spatialfrequency tun<strong>in</strong>g, and adjacent or overlapp<strong>in</strong>g receptive field locations.Cell pairs <strong>in</strong>cluded <strong>in</strong> <strong>the</strong> analysis had clear structure <strong>in</strong> <strong>the</strong>ir neuralcross-correlograms and were ei<strong>the</strong>r mono- or polysynaptic 27 .Disparity tun<strong>in</strong>g curves were fit with a Gabor function at optimal delayonly. Differences <strong>in</strong> <strong>the</strong> range or frequency parameters between pre- andpost-synaptic cells were normalized by <strong>the</strong> average value as follows:100 × (S post – S pre )/((S post + S pre )/2),where S is ei<strong>the</strong>r <strong>the</strong> size or frequency parameter of <strong>the</strong> pre (pre-synaptic)or post (post-synaptic) neuron. This is a measure of <strong>the</strong> magnitude of<strong>the</strong> <strong>coarse</strong>-<strong>to</strong>-f<strong>in</strong>e and f<strong>in</strong>e-<strong>to</strong>-<strong>coarse</strong> connections.All raw neural correlograms were shuffle-subtracted <strong>to</strong> elim<strong>in</strong>atestimulus-based correlations 28 . Shuffle-subtraction was obta<strong>in</strong>ed by firsttak<strong>in</strong>g <strong>the</strong> cross-correlogram between pairs of cells from all <strong>the</strong> repetitionsof a stimulus. Cross-correlation between repetitions, which can onlybe stimulus-based, were <strong>the</strong>n subtracted out. The result was a neurallybased cross-correlogram. In this study, we need <strong>to</strong> dist<strong>in</strong>guish common<strong>in</strong>put connections from those of a mono- or polysynaptic nature. Inmono- and polysynaptic connections, <strong>the</strong>re was a relatively narrow peakshifted away from zero so that one cell (e.g. spike 1) can be said <strong>to</strong> consistentlyfire before ano<strong>the</strong>r (e.g. spike 2) 27 . From this, we <strong>in</strong>fer that spike1 is presynaptic <strong>to</strong> spike 2. If <strong>the</strong> peak was broad and straddled zero, <strong>the</strong>n<strong>the</strong> pair of cells was assumed <strong>to</strong> be receiv<strong>in</strong>g a common <strong>in</strong>put 27 . Thesecategories are somewhat arbitrary, as <strong>the</strong>re is a cont<strong>in</strong>uum of <strong>in</strong>teractions.It is also possible for one cell <strong>to</strong> be pre-synaptic <strong>to</strong> ano<strong>the</strong>r while <strong>the</strong>y bothreceive common <strong>in</strong>put. Three quantitative criteria were used for elim<strong>in</strong>at<strong>in</strong>ga common <strong>in</strong>put type of cross-correlogram from <strong>the</strong> analysis. First,<strong>the</strong> correlogram asymmetry <strong>in</strong>dex [(AI = (R – L)/(R + L), where R andL are <strong>the</strong> areas of <strong>the</strong> b<strong>in</strong>s <strong>to</strong> <strong>the</strong> right and <strong>to</strong> <strong>the</strong> left of zero (by ± 5 ms,respectively)], is a measure of how much <strong>the</strong> correlogram peak was shiftedfrom zero 29 . Second, <strong>the</strong> latency of <strong>the</strong> cross-correlogram peak is <strong>the</strong> b<strong>in</strong>of maximum amplitude. Third, <strong>the</strong> width of <strong>the</strong> peak is def<strong>in</strong>ed as <strong>the</strong>width at half-maximum amplitude. Common <strong>in</strong>put connections weredef<strong>in</strong>ed as hav<strong>in</strong>g a latency of ≤5 ms and an AI of
articles© 2003 Nature Publish<strong>in</strong>g Group http://www.nature.com/natureneuroscience6. Wilson, H.R., Blake, R. & Halpern, D.L. Coarse spatial scales constra<strong>in</strong> <strong>the</strong>range of b<strong>in</strong>ocular fusion on f<strong>in</strong>e scales. J. Opt. Soc. Am. A 8, 229–236(1991).7. Fleet, D.J., Wagner, H. & Heeger, D.J. Neural encod<strong>in</strong>g of b<strong>in</strong>ocular disparity:energy models, position shifts and phase shifts. Vision Res. 36, 1839–1857(1996).8. Rohaly, A.M. & Wilson, H.R. Nature of <strong>coarse</strong>-<strong>to</strong>-f<strong>in</strong>e constra<strong>in</strong>ts onb<strong>in</strong>ocular fusion. J. Opt. Soc. Am. A 10, 2433–2441 (1993).9. Anderson, C.H. & Van Essen, D.C. Shifter circuits: a computational strategyfor dynamic aspects of <strong>visual</strong> <strong>process<strong>in</strong>g</strong>. Proc. Natl. Acad. Sci. USA 84,6297–6301 (1987).10. Nishihara, H.K. Practical real-time imag<strong>in</strong>g stereo matcher. Opt. Eng. 23,536–545 (1984).11. Nomura, M. A model for neural representation of b<strong>in</strong>ocular disparity <strong>in</strong>striate <strong>cortex</strong>: distributed representation and ve<strong>to</strong> mechanism. Biol. Cybern.69, 165–171 (1993).12. Quam, L. H. Hierarchical warp stereo. <strong>in</strong> Read<strong>in</strong>gs <strong>in</strong> Computer Vision (eds.Fischler, M.A. & Firsche<strong>in</strong>, O.) 80–86 (Kauffman, Los Al<strong>to</strong>s, California,1987).13. Smallman, H. S. F<strong>in</strong>e-<strong>to</strong>-<strong>coarse</strong> scale disambiguation <strong>in</strong> stereopsis. VisionRes. 35, 1047–1060 (1995).14. Smallman, H.S. & MacLeod, D.I. Spatial scale <strong>in</strong>teractions <strong>in</strong> stereosensitivity and <strong>the</strong> neural representation of b<strong>in</strong>ocular disparity. Perception 26,977–994 (1997).15. Sanger, T.D. Stereo disparity computation us<strong>in</strong>g gabor filters. Biol. Cybern.59, 405–418 (1988).16. Qian, N. & Zhu, Y. Physiological computation of b<strong>in</strong>ocular disparity. VisionRes. 37, 1811–1827 (1997).17. Ohzawa, I., DeAngelis, G.C. & Freeman, R.D. <strong>Stereoscopic</strong> <strong>depth</strong>discrim<strong>in</strong>ation <strong>in</strong> <strong>the</strong> <strong>visual</strong> <strong>cortex</strong>: neurons ideally suited as disparitydetec<strong>to</strong>rs. Science 249, 1037–1041 (1990).18. Press, W.H., Teukolsky, S.A., Vetterl<strong>in</strong>g, W.T. & Flannery, B.P. NumericalRecipes <strong>in</strong> C (Cambridge Univ. Press, Cambridge, UK, 1992).19. R<strong>in</strong>gach, D.L., Hawken, M.J. & Shapley, R. Dynamics of orientation tun<strong>in</strong>g <strong>in</strong>macaque primary <strong>visual</strong> <strong>cortex</strong>. Nature 387, 281–284 (1997).20. Bredfeldt, C.E. & R<strong>in</strong>gach, D.L. Dynamics of spatial frequency tun<strong>in</strong>g <strong>in</strong>macaque V1. J. Neurosci. 22, 1976–1984 (2002).21. Mazer, J.A., V<strong>in</strong>je, W.E., McDermott, J., Schiller, P.H. & Gallant, J.L. Spatialfrequency and orientation tun<strong>in</strong>g dynamics <strong>in</strong> area V1. Proc. Natl. Acad. Sci.USA 99, 1645–1650 (2002).22. Pack, C.C. & Born, R.T. Temporal dynamics of a neural solution <strong>to</strong> <strong>the</strong>aperture problem <strong>in</strong> <strong>visual</strong> area MT of macaque bra<strong>in</strong>. Nature 409,1040–1042 (2001).23. Parker, D.M., Lishman, J.R. & Hughes, J. Evidence for <strong>the</strong> view thattemporospatial <strong>in</strong>tegration <strong>in</strong> vision is temporally anisotropic. Perception 26,1169–1180 (1997).24. Rohaly, A.M. & Wilson, H.R. Disparity averag<strong>in</strong>g across spatial scales. VisionRes. 34, 1315–1325 (1994).25. DeAngelis, G.C., Ohzawa, I. & Freeman, R.D. Spatiotemporal organization ofsimple-cell receptive fields <strong>in</strong> <strong>the</strong> cat’s striate <strong>cortex</strong>. I. General characteristicsand postnatal development. J. Neurophysiol. 69, 1091–1117 (1993).26. Li, G. Robust regression. <strong>in</strong> Explor<strong>in</strong>g Data Tables, Trends and Shapes (eds.Hoagl<strong>in</strong>, D.C., Mosteller, F. & Tukey, J.W.) 281–340 (Wiley, New York, 1985).27. Moore, G.P., Segundo, J.P., Perkel, D.H. & Levitan, H. Statistical signs ofsynaptic <strong>in</strong>teraction <strong>in</strong> neurons. Biophys. J. 10, 876–900 (1970).28. Perkel, D.H., Gerste<strong>in</strong>, G.L. & Moore, G.P. Neuronal spike tra<strong>in</strong>s ands<strong>to</strong>chastic po<strong>in</strong>t processes. II. Simultaneous spike tra<strong>in</strong>s. Biophys. J. 7,419–440 (1967).29. Alonso, J.M. & Mart<strong>in</strong>ez, L.M. Functional connectivity between simple cellsand complex cells <strong>in</strong> cat striate <strong>cortex</strong>. Nat. Neurosci. 1, 395–403 (1998).30. Aertsen, A.M. & Gerste<strong>in</strong>, G.L. Evaluation of neuronal connectivity:sensitivity of cross-correlation. Bra<strong>in</strong> Res. 340, 341–354 (1985).nature neuroscience • volume 6 no 1 • january 2003 65