Organization of Ocular Dominance and Orientation Columns in the Striate Cortex of Neonatal Macaque Monkeys

Organization of ocular dominance and orientation columns in the striate cortex of neonatal macaque monkeys

GARY BLASDEL,1'2 KLAUS OBERMAYER,3-4 AND LYNNE KIORPES5 'Department of Physiology, University of Calgary, Calgary Alberta, Canada T2N-1N4 2Department of Neurobiology, Harvard Medical School, Boston 3The Salk Institute, La Jolla "The Rockefeller University, New York 'Center for Neural Science, New York University, New York (RECEIVED May 13, 1994; ACCEPTED November 30, 1994)

Abstract Previous work has shown that small, stimulus-dependent changes in light absorption can be used to monitor cortical activity, and to provide detailed maps of ocular dominance and optimal stimulus orientation in the striate cortex of adult macaque monkeys (Blasdel & Salama, 1986; Ts'o et al., 1990). We now extend this approach to infant animals, in which we find many of the organizational features described previously in adults, including patch-like linear zones, singularities, and fractures (Blasdel, 19926), in animals as young as 3| weeks of age. Indeed, the similarities between infant and adult patterns are more compelling than expected. Patterns of ocular dominance and orientation, for example, show many of the correlations described previously in adults, including a tendency for orientation specificity to decrease in the centers of ocular dominance columns, and for iso-orientation contours to cross the borders of ocular dominance columns at angles of 90 deg. In spite of these similarities, there are differences, one of which entails the strength of ocular dominance signals, which appear weaker in the younger animals and which increase steadily with age. Another, more striking, difference concerns the widths of ocular dominance columns, which increase by 20% during the first 3 months of life. Since the cortical surface area increases by a comparable amount, during the same time, this 20% expansion implies that growth occurs anisotropically, perpendicular to the ocular dominance columns, as the cortical surface expands. Since the observed patterns of orientation preference expand more slowly, at approximately half this rate, these results also imply that ocular dominance and orientation patterns change their relationship, and may even drift past one another, as young animals mature. Keywords: Development, Primate striate cortex, Ocular dominance, Orientation selectivity

Introduction appear to be organized in slab-shaped regions, 0.25-0.5 mm wide, that run vertically between pia and white matter, that lie Neurophysiological studies have shown that most neurons in in register with bands of afferents from the appropriate eye in striate cortex are binocular and selective for orientation (Hubel layer 4c, and that interdigitate with slabs dominated by the other &Wiesel, 1962, 1968, 1972, 1974a,b). Their binocularity is fur- eye (Hubel & Wiesel, 1972; LeVay et al., 1975; Hubel & Wiesel, ther characterized by a preference for one eye, referred to as 1977). Preferences for orientation also appear to be organized "ocular dominance," and their preference for orientation is fur- in slabs, but ones that are much narrower and shorter in length. ther characterized by a degree of selectivity, which is frequently The main reason these slabs are shorter is that they converge referred to as "orientation selectivity" or "orientation tuning." periodically in the centers of ocular dominance columns, at As numerous studies have shown, these response properties are points, or singularities, which take on the appearance of not strewn about randomly, but are highly organized across the "rosettes" or "pinwheels" when preferences for different ori- cortical surface. Cells preferring a particular eye, for example, entations are illustrated in color, as they are in Fig. la (Blasdel & Salama, 1986; Blasdel, 19896, 1992; Ts'o et al., 1990; Bon- hoeffer & Grinvald, 1991). Because most orientation slabs Reprint requests to: Gary Blasdel, Department of Neurobiology, Harvard Medical School, 200 Longwood Avenue, Boston, MA 02115, extend between singularities in adjacent ocular dominance col- USA. umns, they tend to cross their borders at right angles (Obermayer

589 Downloaded from https://www.cambridge.org/core. NYU Medical Center: Ehrman Medical Library, on 23 Feb 2018 at 18:59:26, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0952523800008476 590 C. Blasdel, K. Obermayer, and L. Kiorpes

et al., 1992a; Blasdel, 19926; Bartfeld & Grinvald, 1992; Ober- orientation signals, in young animals, as well as a tendency for mayer & Blasdel, 1993), as illustrated in Fig. lc. ocular dominance bands to lie closer together than one might We were interested in looking at how these organizations expect from their adult spacing corrected for expected rates of emerge in young animals. Previous work had indicated that vari- postnatal growth (Purves & LaMantia, 1993). Some of these ations in orientation preference and ocular dominance are results have been presented previously in abstract or seminar present at birth (Wiesel & Hubel, 1974). But the fact that bands form (Kiorpes & Blasdel, 1987; Blasdel, 1989; Obermayer et al., of afferents are not fully segregated in layer 4c until 8-10 weeks 1994). of age, makes it unlikely that organizations associated with them (e.g. ocular dominance and orientation) can mature before this time. We began investigating these issues by using optical imag- Materials and methods ing techniques to explore the organizations of ocular dominance and orientation at discrete ages during the first 14 weeks of life. In so doing, we found that patterns of ocular dominance and Infant monkeys orientation are both present in all animals (as young as 3| These studies were conducted on four Macaca nemestrina (BN- weeks of age), with adult characteristics and interactions (e.g. 5.5, BN-7.5, BN-9, and BN-14) and one Macaca fascicularis linear zones, fractures, singularities, saddle points, and instances (BF-3.5) infants. The Macacae nemestrina were all born in the of orientation slabs crossing ocular dominance borders at right animal vivarium at the University of Calgary, where they were angles) clearly evident at all times. We also observed a tendency reared by their mothers. BF-3.5 was born at the New England for ocular dominance signals to appear weaker, in relation to Regional Primate Center and reared by hand. At the time of

linear zone a)

singularity

saddle point fracture

linear zone b) singularity

saddle point fracture

C) L R L

Downloaded from https://www.cambridge.org/core. NYU Medical Center: Ehrman Medical Library, on 23 Feb 2018 at 18:59:26, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0952523800008476 Development of dominance and orientation preference 591

study, the mother was tranquilized with an intramuscular injec- catheter for the infusion of electrolytes and drugs. It was ven- tion of ketamine/xylazine, to facilitate removal of the infant, tilated with a 2:1 mixture of nitrous oxide and oxygen, and anes- which was then anesthetized with halothane (0.5-1.5%) in thesia was switched gradually to Pentothal (0.1-1.0 mg/kg/h, nitrous oxide/oxygen, and prepared for neurophysiological and i.v.), as the residual halothane was expired. optical recording. Each animal was placed on a heated waterbed with its head Experiments were limited to 18-h duration, after which the in a stereotaxic frame. The scalp was then reflected and a tre- animals were either allowed to recover and return to their moth- phine or dental drill used to bore a 25-mm hole in the cranium ers, or killed with an overdose of Pentobarbitol and perfused overlying the opercular cortex, just behind the lunate sulcus, through the heart with 4% paraformaldehyde in 0.1 M phos- as close to the midline as possible. In some cases, especially for phate buffer. Animals that were allowed to recover were first the very young infants whose cranial sutures were immature, weaned of drugs. After it became apparent they could breathe the cranial bones were additionally secured by an acrylic cap on their own, they were given supplemental doses of pyrido- constructed from a thin layer of Grip Cement. After this, a stigmine, to reverse any residual effects of curariform drugs, stainless-steel chamber suitable for either microelectrode or and methamphetamine, to counteract the residual effects of optical recordings was inserted and cemented in place. For the thiopental sodium administered during the experiment. After electrode recordings, this chamber was sealed by an O-ring sup- animals had recovered fully, and demonstrated their ability to porting a large glass disk that was pressed against it by a micro- vocalize, they were returned to their mothers who on some, but manipulator. The glass contained a guide tube that allowed an not all, occasions accepted them. Infants that were not accepted electrode to be inserted and positioned anywhere on the corti- were reared in an incubator and nursed on a 1:1 mixture of cal surface under visual guidance. For the optical recordings a Similac and water until the conclusion of the experiment, at threaded stainless-steel plug, equipped with an 18-mm glass win- which time they were killed with an overdose of barbiturate and dow, was inserted into the chamber and adjusted until the glass perfused as outlined above. rested a millimeter above the cortical surface. A minimum sep- aration of 0.5-1.0 mm was maintained between glass and cortex to allow the dye solution (NK2367, 0.1% in saline) to Preparation circulate. The results in this article are based on optical and single-unit After all surgical procedures had been completed the animal recordings from five infant macaque monkeys, that ranged in was paralyzed partially with Vecuronium bromide or Tubocu- age between 3j and 14 weeks. The surgical preparation of each rarine to stabilize the eyes. The eyes were protected with hard one was similar to that described previously for adults (Bias- gas-permeable contact lenses with curvatures chosen to bring del, 1992a,b). For simplicity, the most important of these pro- the eyes into focus on the screen of a 19-inch monitor placed cedures are described again, along with variations relevant to 2 m away. Anesthesia was maintained with Pentothal (0.1-1.0 the acquisition of optical recordings from infant animals. mg/kg/h) and verified to be adequate with frequent reference Instead of ketamine/xylazine, however, anesthesia was induced to the electrocardiogram (EKG), blood pressure, and end-tidal with halothane, delivered in a mixture of nitrous oxide and oxy- carbon dioxide (CO2), all of which were monitored continu- gen through a mask that was placed over the animal's face. The ously. The adequacy of anesthesia could also be verified fre- animal was then intubated and provided with an intravenous quently by the absence of reflexes (e.g. lateral canthal) since the

Fig. 1. Organizations of orientation preference in striate cortex of adult macaque monkeys: (a) In the upper figures, orientation preferences are indicated by opponent colors, with red and green representing horizontal and vertical and blue and yellow representing left and right oblique, respectively. In principle, we find that laterally displaced regions are mapped either continuously or discontinuously. In the continuously mapped regions, we find linear zones where orientation preferences rotate continuously along one axis and remain constant along the other (orthogonal) axis, which appear as small rainbow-colored patches, and saddle points, where the sequence of orientation preferences reverses direction, creating the appearance of a solid color that changes symmetrically on all sides. In the discontinuously mapped regions, we find fractures, where orientation preferences change abruptly across a fault line, that extends in one dimension, and singularities, where orientation preferences change continuously through 180 deg around a point, creating a zone of extreme transition in the center. One can think of linear zones and the regions surrounding singularities as Cartesian and Polar representations of similar trends: In the linear zones, orientation preferences remain constant along one (straight) axis and rotate linearly along the other, orthogonal, axis which, generat- ing rows of short rainbow colored stripes along the way. In regions surrounding singularities, on the other hand, orientation preferences remain constant radially, and rotate continuously along tangential axes, in circles, leading to the generation of swirls, rosettes, or pinwheels along the way. In the case of linear zones, the regions of linear transition are circumscribed by a border of discontinuous transition (ox fracture) on the outside, while in regions surrounding singularities, the outer borders are mapped continuously, and the discontinuity is found at the center (from Obermayer & Blasdel, 1993). (b) In this representation (of the same cortical region), contours of iso-orientation (i.e., lines along which orientation preferences remain constant) are indicated instead of orientation preferences. Accordingly, linear zones appear as patches of evenly spaced, parallel lines while saddle points take on the appearance of blank, unlined regions around which iso-orientation contours bow conspic- uously outward. Fractures appear as clumps of closely spaced lines, and singularities occur at points where lines converge (from Obermayer and Blasdel, 1993). (c) In the lower representation, ocular dominance borders have been added to a contour map of orientation preferences (taken from a different region of adult striate cortex). From this it is obvious: (1) that singularities cluster in the centers of ocular dominance columns (i.e., mid-way between the dark lines), and (2) that where the iso-orientation contours cross the borders of ocular dominance columns, they do so at angles of approximately 90 deg (from Obermayer & Blasdel, 1993).

level of neuromuscular blockade, which was assessed every 2 h Optical recording from muscle twitches induced by electrical stimulation of the Light from a 100-W quartz halogen lamp was focused by a median nerve, was never allowed to exceed 50% —a level that, even though it allows reflexes to be elicited, eliminates all but parabolic cold mirror through heat absorbing glass and an inter- the smallest eye movements (see Blasdel, 1992a). ference filter (720 nm, ±20 nm) onto the end of a 0.25-inch- diameter fiberoptic bundle. Monochromatic light from this Sequential receptive-field measurements, made during single- bundle was then collimated and deflected through the edge of unit recordings from the infants, verified that even in the pres- a 50-mm objective lens onto the cortical surface. Reflected light ence of 50% paralysis, residual eye movements were minimal was then captured by the same objective and relayed to a pro- as long as the animal remained adequately anesthetized. As pointed out in previous experiments in cats and monkeys (Bias- jection lens that focused it onto the face plate of a Newvicon del, 1992a), any movements that remained were generally slow TV camera (COHU model 5300). More detailed descriptions are and transient, and limited to 0.25 deg. Larger movement of a provided in previous publications (Blasdel, 1989, 1992a; Blas- degree or more were observed occasionally over extended peri- del & Salama, 1986). ods of time (> 12 h), as noted previously (Pettigrew et al., 1979); but these are unlikely to have any effect on the optical record- Visual stimulation ings since the visual stimuli used are presented on a monitor screen subtending 11.4 deg of the animal's visual field. Visual stimuli in all cases consisted of four superimposed square- wave gratings at nonharmonic frequencies, moving back and On those occasions when it was necessary for animals to forth at speeds of 1.5 deg/s, on the screen of a 20-inch Mit- recover, the paralytics were discontinued and the Pentothal was subishi monitor (C-6910 or C-3910) placed 2 m away. The aver- replaced by general inhalation agents (halothane in nitrous oxide 2 age screen luminance and contrast were set to 3 cd/m and and oxygen), with methamphetamine given to counteract the 80%, respectively. Between periods of visual stimulation, grat- effects of residual pentothal, accumulated in the fat, as well as ing patterns were replaced by a blank screen that also had a lumi- to raise blood pressure and increase renal clearance. The dural 2 nance of 3 cd/m . flap was sutured with 6-0 surgical silk, and, after it had been treated with antibiotics (chloramphenicol) and steroids (Deca- dron), the chamber was sealed with a sterile Teflon plug. Once Differential images of ocular dominance the animal demonstrated its ability to breathe on its own for at least an hour, any residual paralysis was reversed with a sin- All images were recorded at a pixel resolution of 512 x 480, after gle intramuscular dose of pyridostigmine, preceded by atropine. which they were compressed to sizes of 128 x 120 by binning. To obtain differential images of ocular dominance each eye was After the animal had been returned to its mother, or an incu- stimulated alternately with contours at one of four orienta- bator, antibiotics (chloramphenicol, 50 mg/kg/d, and ampicil- tions—0 deg, 45 deg, 90 deg, or 135 deg. Video images of stri- lin, 50 mg/kg/d) and steroids (Decadron, 0.1 mg/kg/d) were ate cortex obtained during visual stimulation of the left eye given prophylactically until the next recording session or until (4-6 presentations for 1.5-3.0 s, at intervals of 10-15 s) were 3 days had elapsed without incident. No area of cortex was subtracted from images obtained during stimulation of the right investigated more than three times and comparisons of activity eye. Difference images obtained in this manner, at each of four patterns were restricted to images obtained within a few hours, basic orientations, were then averaged to render the final images from the same locations. of ocular dominance, which have been shown previously to cor- respond closely in gray scale to ocular dominance values that would be determined electrophysiologically with single-unit Microelectrode recordings recordings (Blasdel & Salama, 1986). Single units were recorded with the aid of a Pyrex disk, 2 inches in diameter that was used to seal the chamber yet allow an unob- Differential images of orientation structed view of the cortical surface. A glass-insulated platinum- iridium electrode (Wolbarsht et al., 1960) was then directed, To map orientation preference and selectivity, the eyes were under visual guidance, to a particular location and advanced aligned and stimulated binocularly, with four to 12 different rapidly to a depth of 200 jim, where its location relative to sur- pairs of orthogonal contours presented 4-6 times each, in rounding blood vessels was registered by digitizing a single video pseudorandom order. Differential images were obtained for frame that was then stored on disk. At each location, record- each pair by subtracting the averaged image of cortex respond- ings were obtained from at least three visually responsive units, ing to contours at one orientation from the averaged image of isolated from the upper layers with separations of at least it responding to the orthogonal orientation. The differential 100 /*m. Receptive-field properties (e.g. orientation preference, images obtained in this manner were then converted to vector selectivity, direction selectivity, color selectivity, end-stopping, fields that were added to yield two-dimensional maps of orien- etc.) were determined separately for each eye, along with the tation preference and selectivity, as elaborated previously (Bias- locations of their receptive fields, which were then used to align del, 1992ft). the eyes with the center of the video monitor. Actual alignment was achieved by displacing the TV monitor and tilting the animal's head, to control azimuth and elevation for the right eye, Fourier analyses and by then positioning a device consisting of two servo- Fourier spectra of the orientation and ocular dominance maps controlled prisms before the left eye, which allowed the azimuth were obtained by applying a two-dimensional fast Fourier trans- and elevation of its visual field to be adjusted in increments of form (Press et al., 1988) to the final images. For this purpose, 0.01 deg. images were embedded in an array of size 128 x 128 with the

values of the additional pixels set to zero. Orientation vectors in neutral gray zones, as one would expect for regions prefer- were represented as complex quantities (Swindale, 1982), whose ring vertical (or horizontal). As the stimulus contours rotate amplitudes and phases denote orientation selectivity and pref- gradually through another 45 deg, in Figs. 2b2, 2b3, and 2ci, erence, respectively, while the strength of the ocular dominance moreover, the crosses shift gradually into the centers of white difference signal was represented by real values. After trans- zones, which also represent vertical since the image was obtained formation, images were compressed to arrays of size 128 x 103 by subtracting responses to vertical from responses to horizontal. in order to account for the different spacing of camera pixels For continued rotations through another 75 deg, in Figs. 2c2, along the horizontal and vertical directions. Power spectra and 2c3, 2d|, 2d2, and 2d3, the crosses move back towards the cen- typical wavelengths were then obtained from the Fourier trans- ters of the dark zones in which they were drawn initially. forms using procedures described in Obermayer and Blasdel, 1993. The right-hand column of Fig. 2 depicts differential images of ocular dominance, obtained with contours at different orientations. Since each of these images reflects the difference Parallelism and intersection angle between orientation between activities generated by the right and left eyes (see Mate- and ocular dominance bands rials and methods), regions dominated by the right and left eyes For a quantitative analysis of orientation preference and ocu- appear dark and light, while regions showing no particular pref- lar dominance, it is useful to have measures for the degree to erence for either eye appear gray or neutral. As one can see, which regions of similar orientation preference and ocular dom- all the images in this column appear strikingly similar, even inance are aligned as parallel slabs. We call these measures of though they were obtained with different orientations. In the parallelism and we abbreviate them by POP for orientation pref- image obtained with horizontal, for example, there are no obvi- erence and by POD for ocular dominance. ous neutral or pale regions, where the ocular dominance bands POP is derived by taking the gradient of orientation pref- fade out due to a preference for vertical. This is similar to find- erence (Obermayer & Blasdel, 1993) at every pixel location, nor- ings reported earlier for images obtained from adult cortex malizing its length, and multiplying its angular component by (Blasdel, 1992a), which were interpreted to indicate separate and two. All unit vectors located in a circular region of radius sg independent circuits subserving patterns of ocular dominance are then averaged producing a vector POP, whose length and orientation. The fact that such an independence is clearly becomes a measure Pop of parallelism. Values of 1.0 indicate evident in even our youngest animal suggests that circuits such regions where the iso-orientation lines are perfectly parallel; val- as this are present and operational by 3^ weeks of age. ues smaller than 1.0 indicate regions where the iso-orientation Even though ocular dominance patterns are not markedly lines are less aligned, while values of 0.0 indicate a singularity. sensitive to the orientation of stimuli used to drive activity (see The radius sg was chosen to be 150 fim in order to match the above), differential images of orientation are sensitive to bin- approximate radius of upper layer pyramidal dendrites. The ocularity. As one can see in Figs. 3e,f and 3h,i, where the cen- measure POD for ocular dominance can be calculated in a sim- ters of right and left ocular dominance columns have been ilar way (Obermayer & Blasdel, 1993). Throughout this paper, superimposed on summed differential images (see below) of ver- we will call a region in monkey striate cortex a linear zone if tical/horizontal and left/right oblique, differential images of OP P exceeds a value of 0.6 for sg = 150 jtm. orientation weaken periodically in the centers of ocular domi- The local angle d of intersection between lines of equal ori- nance columns, and appear strongest in the linear zones, entation preference and ocular dominance is given by the angle between the centers of ocular dominance columns, where cells between the vectors POP and POD, divided by two. In regions are most binocular. The reason for this could be that neurons where Pop and POD approach 1.0, contours of constant ori- in the linear zones are more selective for orientation or that they entation preference and constant ocular dominance are defined, are activated more strongly by binocular stimulation. In either making it possible to calculate their local angles of intersection. case, these differences, which also characterize adults (see Bias- In our calculations this local angle of intersection, denoted by del, 1992a) are clearly evident by 3^ weeks of age. d, is obtained by taking the angle of intersection between vec- As Blasdel and Salama(1986) showed initially, it is possible op OD tors P and P and dividing by 2 (see Obermayer & Bias- to combine differential images of orientation by converting each del, 1993, for further details). one to a field of vectors that are then summed at every location. In so doing the orientation of each vector is obtained by doubling the positive stimulus orientation, so that similar Results responses to orthogonal orientations cancel (see Blasdel, 19926 Fig. 2 depicts differential images of orientation and ocular dom- for details). Once these are added, the orientations of the result- inance from the striate cortex of our youngest animal, who was ing vector sums indicate the preferred orientation (multiplied studied at 3| weeks of age. The orientation dependency of the by 2), while their lengths reflect the specificity of response. In dark and light regions can be verified easily in these images by addition to providing specific values for orientation preference comparing values at similar locations. In Fig. 2a,, for exam- and selectivity at every location, which are not available from ple, where responses to vertical and horizontal have been com- any differential image by itself (Blasdel, 1992a,b), this calcu- pared, small crosses have been added at two locations having lation can be thought of as an example of orientation-dependent a clear preference for vertical. As one can see in adjacent images, averaging which improves the signal-to-noise ratio at every loca- the dark regions which indicate preference peaks for the pri- tion. For an initial sample of 12 differential orientation images, mary stimulus orientation, represented by a black bar in the for example, each of the two resulting vector fields have sig- lower right-hand corner of each frame, move laterally as the nal/noise ratios that are approximately \[6 (= 2.45) times orthogonal contours rotate counterclockwise in increments of greater than that of any single differential image used as input. 15 deg (see Figs. 2a2 and 2a3). In Fig. 2b|, for example, where As noted previously (Blasdel, 19926), these vector fields can be the contours have been rotated through 45 deg, the crosses lie displayed in either Cartesian or polar coordinates. Cartesian rep-

Downloaded from https://www.cambridge.org/core. NYU Medical Center: Ehrman Medical Library, on 23 Feb 2018 at 18:59:26, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0952523800008476 594 G. Blasdel, K. Obermayer, and L. Kiorpes BF-3.5

1 mm Fig. 2. Differential images of orientation and ocular dominance from a S^-week-old fascicularis macaque: [1-3] The first three columns depict differentially imaged responses to orthogonal contours (indicated by crossed black and white bars in the lower right hand corners) that were rotated, in increments of 15 deg, between each frame. Hence the differential image in a, depicts responses to 0°/90° (where responses to horizontal have been subtracted from responses to vertical); the image in a2 depicts differential responses to 15°/1O5°; the image in a3 depicts responses to 30°/120°; and so on. Note that, as the pairs of orthogonal contours (indicated by dark and light crosses in the lower right corners) rotate in successive frames, the dark spots indicating peak responses to primary stimulus orientations (indicated by the dark bar), move laterally with respect to the two stationary crosses that have been added at locations that preferred vertical. This is largely as one would expect from Hubel and Wiesel's previous observations in adult and baby monkeys (Hubel & Wiesel, 1974; Wiesel & Hubel, 1974). [4] Images in the right-hand column depict differential images of ocular dominance that were obtained by subtracting responses to stimulation of the left eye from responses to stimulation of the right by contours at a single orientation (for both eyes, as indicated in the lower right- hand corner). As one can see, stimulus orientation has little or no effect on the pattern of ocular dominance that emerges. As described previously for the adult (Blasdel, 1992a), the absence of apparent interaction suggests a relative independence of the cortical circuits involved in the expression of differential orientation and ocular dominance maps.

reservations of orientation-dependent information from BF-3.5, the right-hand column of Fig. 4, are compared with the differ- for example, appear in the top row of Fig. 4 (2nd and 3rd col- ential images of orientation in columns 2 and 3, it becomes umns) where they indicate averaged preferences for vertical/hor- apparent that ocular dominance signals become stronger in rela- izontal and left/right oblique. Except for their greater tion to orientation signals as animals mature. The trend is more signal/noise ratios, these images contain identical information obvious in Table 1, where signal strengths are calculated from to that available from single differential images of vertical/hor- the variance of individual differential images (after filtering to izontal and left/right oblique. Comparable summaries of the remove high- and low-frequency noise—see Obermayer & Blas- differential images of orientation and ocular dominance del, 1993). From this table, it becomes obvious that the ratio obtained from four older animals, studied at 5|, 7j, 9, and of ocular dominance to orientation preference (OD/OP) sig- 14 weeks of age, appear in Fig. 4, along with data from one nals rises from a low of approximately 0.92, for the youngest adult (NM-1) that appeared previously (Blasdel, 1992a). animal, to 1.23, for the oldest infant, and to a value of approx- If the averaged images of ocular dominance, appearing in imately 1.36 for a young adult.

1 mm

Fig. 3. Interactions between orientation and ocular dominance: The first three images (a-c) in this figure illustrate the way in which ocular dominance centers are calculated from regions of maximum curvature in the averaged ocular dominance image from our youngest animal (BF-3.5), as described previously (Blasdel, 1992a). The solid black and white regions indicate the centers of right and left eye dominance bands, respectively. In images (e) and (h), the centers of right eye bands appear superimposed on averaged differential images of 0°/90° and 45V135°, respectively, revealing a clear interaction between orientation and ocular dominance. Note how the most intense dark and light regions tend to avoid and lie along the centers of right eye columns, and how these interactions become more obvious in images (f) and (i), where the centers of left eye columns are indicated in white.

From the right-hand column of Fig. 4, one can see that the zones with greater selectivity, while images on the right indi- spacing of ocular dominance bands also increases steadily with cate the orientations that were preferred. As one can see from age. This can be seen qualitatively by comparing the spacing these images, the basic patterns of orientation preference are of bands in successive images. And it can also be demonstrated apparent from very early on. Even at 3| weeks of age, for quantitatively by taking the Fourier transform (see Materials example, one can detect the linear zones which are apparent as and methods) of each pattern and calculating the period of small rainbow colored patches. One can also detect singulari- repetition along each axis. The results of this quantitation are ties, which form the points where all orientation preferences (col- summarized in Table 2, where there is a clear increase, of ors) converge. These observations can also be documented by approximately 21%, between 3| and 14 weeks of age —an the quantitative measures applied below. increase that corresponds approximately to the growth in opercular surface area that is expected during the same period (Hubel et al., 1977a; Purves & LaMantia, 1993). Fourier analysis Fig. 5 shows polar representations of the orientation maps As Swindale (1982) pointed out initially, and Obermayer and for each animal. The images on the left reflect selectivity (tun- Blasdel (1993) demonstrated more recently, Fourier analysis pro- ing) for orientation, with brighter regions corresponding to vides a useful and global way of quantifying the periodicities

Vasculature 0° 45° OD

BF-3.5

BN-5.5

BN-7.5

BN-9

BN-14

NM-1

1 mm Fig. 4. Averaged orientation and ocular dominance patterns for all animals, in chronological order: For any particular row, the first image depicts the vascular pattern, which we photographed under green light (to maximize contrast) at the time of imaging. The second and third images are combined vectorial representations of orientation selectivity, displayed in Cartesian coordinates. As described previously (Blasdel, 1992ft), these representations are formally equivalent to the Polar representations that appear in Fig. 5. As one can verify for BF-3.5 (see Fig. 2), the first of these (under the 0° heading) is equivalent to a single differential image of 0°/90° (see a], Fig. 2) while the second (under the 45° heading) is equivalent to a single differential image of 45°/135° (see bi, Fig. 2). The main difference between them arises from the higher signal/noise ratios of the images that are combined, which reflect the benefits of greater averaging. The right-hand image in each row depicts a combined image of ocular dominance that was obtained by averaging 4 separate ocular dominance images, obtained with contours at four different orientations. As one can see by comparing combined images of orientation (columns 2 and 3) with combined images of ocular dominance (column 4), there is a steady, age-dependent increase in the strength of signals producing ocular dominance patterns, relative to those producing orientation patterns, as neonatal animals mature. These are shown quantitatively in Table 1. Even more pronounced is the steady increase in ocular dominance spacing that occurs with age. As one can see in Table 2, this increases by nearly 21%, and is linearly correlated with age (with a correlation coefficient of 0.93), as one might expect from the growth in surface area during this time (Purves & LaMantia, 1993). Since there is some, but much less, increase in the periodicity of orientation preferences (determined from the same cortical region), as one can see by comparing the second and third images with the fourth image in different rows, there is a possibility that patterns of orientation preference are not coupled rigidly to patterns of ocular dominance, or even to particular cortical regions, and that ocular dominance and orientation patterns may actually drift past one another as young animals mature.

of orientation and ocular dominance patterns along different along the axis of repetition. The Fourier spectra of infant ocu- axes. When Fourier analysis is applied to ocular dominance pat- lar dominance patterns, which appear on the right of Fig. 6, terns, for example, it generates spectra with two clusters of resemble those shown previously for adults (Obermayer & Bias- modes displaced perpendicular to the ocular dominance bands, del, 1993). The wavelength estimates (right-hand column of

Table 1. Development of ocular dominance Table 2. Characteristic wavelengths of orientation and signals with age" ocular dominance columns in microns"

Age Age Monkey (weeks/days) OD (rms) OP (rms) OD/OP Animal (weeks) \°p \°D \OPj_ i BF-3.5 3.5/25 26.54 28.85 0.92 BF-3.5 3.5 660 724 741 612 BN-5.5 5.5/39 22.93 24.00 0.96 BN-5.5 5.5 714 795 772 664 BN-7.5 7.5/53 15.29 23.01 0.66 BN-7.5 7.5 615 817 694 571 BN-9 9/63 23.63 23.71 1.00 BN-9 9 658 842 800 695 BN-14 14/98 31.84 25.81 1.23 BN-14 14 700 875 811 710 NM-1 >52/365 27.00 19.89 1.36 Adult (nemestrina) 682 ± 59 825 ± 52 776 + 89 642 ± 57 Adult (fascicularis) 662 815 722 612

aAs one can see in Fig. 4, the signals giving rise to ocular dominance patterns (right-hand column) steadily increase with age. This is not a aThe symbols XOP and XOD stand for the average wavelengths of ori- result of averaging or vagaries of the imaging technique since a corre- entation preferences and ocular dominance columns in the imaged sponding change is not evident for patterns associated with orientation regions of cortex, and may be taken to indicate average spacing. The (second and third columns). These correlations are quantified in the symbols \op> and \op± denote the wavelength of orientation prefer- above table where rms values of the ocular dominance patterns have ences parallel and perpendicular to the ocular dominance bands, respec- been calculated for 10,302 points (out of 19,200) in the central portion, tively. The estimated error for the wavelength measurements is ±50ftm. after convolving with a 0.25-mm diameter Gaussian to remove high- The wavelengths XOP| and \OPj- corresponding to animal BN-9 could frequency noise. As one can see in the OD/OP column, the strength not be estimated reliably from the Fourier spectrum, because a primary of signals giving rise to the ocular dominance patterns increases steadily axis of elongation could not be defined. The values for "adult Macaca with age, with a linear correlation coefficient of 0.75. This could reflect nemestrina" are averages over four animals (with error margins derived the segregation of LGN afferents, which reaches completion during the from standard deviations); the values for "adult Macaca fascicularis" first 8-10 weeks of age. Note that the OD/OP ratio appears surpris- are averages over two recording sites obtained from the same animal. ingly normal for BN9, even though this animal had a pronounced stra- Adult data are taken from Obermayer and Blasdel (1993). Note that bismus and no binocular units at the time of physiological recording. the error margins indicated for adult data are likely to overstate the variability (and understate the average values) since the combined patterns were obtained from several animals at a variety of locations and eccentricities, where the spacing of ocular dominance columns may well be different (LeVay et al., 1980). These factors are less likely to com- Table 1) for young animals, however, reflect a nearly linear, plicate the infant data, since all of these images were obtained from age-dependent increase in bandwidth between 3| and 14 weeks approximately the same cortical locations (immediately posterior to the of age. 17/18 border where the ocular dominance columns are most aligned). The Fourier spectra of orientation preferences in Fig. 6 (left- These data are therefore more reliable for inter-animal comparisons. hand column) contain elliptical bands of modes that resemble spectra from older animals (Obermayer et al., 1991, I992a,b; Obermayer & Blasdel, 1993). The elliptical form of these modes indicates a variation in the average frequency of repetition along zones, saddle points, fractures, and singularities (see Fig. 1) can different axes. Accordingly, axes corresponding to the highest all be seen by 3^ weeks of age. Fig. 7 shows enlarged sections rates of repetition are indicated by solid black lines in each exam- of the maps from BN-5.5 and BF-3.5, both of which contain ple. As one can see by comparing the inclinations of these lines linear zones, saddle points, singularities, and fractures. The for the ocular dominance and orientation spectra from each ani- sketches in the right-hand column of Fig. 7 indicate sample loca- mal, these patterns repeat at shorter intervals along orthogo- tions of these features in each map. To gain insight into their nal axes. In other words, orientation preferences repeat more development, it is necessary to evaluate their densities and dis- frequently along axes parallel to the ocular dominance bands, tributions at different ages, which can be done most easily by which is what one would expect if the short iso-orientation slabs calculating the degree of alignment, or parallelism, between iso- that are observed intersect ocular dominance borders at right orientation contours in discrete regions, 300 ^m across. As one angles. Such an arrangement was reported previously for adults can see in columns 4,5, and 6 of Table 3, the densities of posi- (Obermayer & Blasdel, 1993). The intervals over which orien- tive and negative singularities, as well as the relative areas of tation preferences repeat along axes parallel and perpendicular linear zones (column 2), do not change markedly with age, which to the ocular dominance bands are shown in the fifth and sixth is surprising since the spacing of ocular dominance bands and columns of Table 2. Even in the youngest animal (BF-3.5), which even the cortical surface area increase steadily in an age- was only 3^ weeks old at the time of study, there is a strong dependent fashion (Hubel et al., 1977a; Purves & LaMantia, orthogonality between regions of constant orientation prefer- 1993). ence and regions of constant ocular dominance, which cause the iso-orientation regions to appear stretched and slab-like along axes perpendicular to the ocular dominance bands. Sim- Angles of intersection between bands of orientation ilar findings were also reported previously for adults (Obermayer and ocular dominance & Blasdel, 1993). Another observation made in adults was that singularities and fractures running parallel to ocular dominance columns tend to lie in their centers while fractures and iso-orientation bands Local organizations of orientation preference crossing the borders of ocular dominance columns tend to do As mentioned above, the four different arrangements of ori- so at right angles (Blasdel & Salama, 1986; Swindale, 1992; entation preference that are distinguished in striate cortex of Obermayer et al., 1992a; Blasdel, 1992ft; Obermayer & Blas- adult animals are present at even the earliest age. Hence, linear del, 1993). From Fig. 8, where the borders of ocular dominance

columns (indicated in black) are superimposed on maps of iso- ear zones (defined here as the shaded regions where our mea- orientation contours from monkeys BN-5.5 and BF-3.5, it sure of alignment, POP, produces values greater than 0.6). appears that this perpendicularity is present by 3^ weeks of Fig. 9 shows histograms of angles between iso-orientation and age. From visual inspection of this figure, it furthermore ocular dominance slabs in the linear zones where, as one can becomes obvious that singularities tend to lie in the centers of see, the largest concentration occurs at 90 deg. Alignments ocular dominance bands, and that where iso-orientation con- within ±18 deg of perpendicular are 1.8-3.9 times as abundant tours run parallel to one another, in linear zones, they intersect as those within ±18 deg of parallel. the borders of ocular dominance slabs at angles of approximately 90 deg. Discussion These arrangements can also be demonstrated quantitatively, by determining the angles of intersection between contours of These results confirm and extend previous observations (Wiesel constant orientation and constant ocular dominance in the lin- & Hubel, 1974) that neurons in monkey striate cortex become orientationally selective and that linear organizations of orientation preference are present shortly after birth if not before. The regularity of linear zones, saddle points, fractures, and singularities, moreover, indicates that patterns of orientation pref- SELECTIVITY PREFERENCE erence have many, if not all, of their adult features by 3| weeks of age. In agreement with Des Rosier et al. (1978) and LeVay et al. (1980), we find banded patterns of ocular dominance in our youngest animals. But since the LGN afferents are BF-3.S unlikely to have segregated in layer 4c by 35, or even 5\ weeks of age, it makes sense that the optical signals associated with ocular dominance are less pronounced in younger animals, and that they increase steadily with age. The relationships between ocular dominance and orientation patterns that were described recently in adult animals (Blasdel, 1992a, 6; Obermayer & Blasdel, 1993) also appear to be present BN-5.5 from very early on. The inverse correlation between orientation selectivity and ocular dominance, for example, that causes differential images of orientation to appear paler in the centers of ocular dominance columns, is pronounced by 3| weeks of age. And the proportions of iso-orientation contours oriented perpendicular to ocular dominance columns are indistinguishable from the proportions observed in adults (see Table 2). The for- BN-7.5 mer effect can be seen clearly in Fig. 3, where the centers of ocular dominance columns appear superimposed on differential images of orientation. And the latter is apparent from the Fourier spectra of ocular dominance and orientation patterns in Fig. 6 as well as from the histograms in Fig. 9 where the angles between ocular dominance and orientation slabs in the linear zones are represented in histograms. From the generality of these correlations, as well as from the ease with which they were BN-9

Fig. 5. This figure depicts the distribution of orientation selectivities and preferences in each animal, in polar coordinates. Gray scale images on the left indicate the degree of orientation selectivity (tuning) at every BN-14 point, with light and dark areas representing more and less selectivity. Color images on the right represent preferred orientations, with opponent colors used to indicate preferences for orthogonal orientations. Hence, red and green indicate preferences for horizontal and vertical while blue and yellow indicate preferences for left and right oblique. As one can see by comparing images in different rows, patterns of orientation selectivity and preference do not change dramatically with age. Linear zones, saddle points, fractures, and singularities are present in all the images, in approximately the same proportions. By comparing NM-1 representations at the top and bottom, moreover, it becomes apparent that age-dependent changes in the periodicity of orientation preferences are minimal, and much less striking than age-dependent changes in the periodicity of ocular dominance which are clearly evident in the right- hand column of Fig. 4, and demonstrated quantitatively in Fig. 10 and 1 mm Table 2.

2/mm BF-3.5 I •*•*•*• Fig. 6. Fourier spectra of orientation and ocular dominance maps from each animal, arranged according to age. Left: Fourier spectra of the orientation map, with each pixel corresponding to one mode and its gray value indicating its amplitude squared. As one can easily appreciate, each of these spectra is characterized by a BN-5.5 slightly elliptical band of values; where this ellipticity is significant we have indicated the axis of elongation with a line. Right: Fourier spectra of ocular dominance columns from the same region of cortex. As described previously (Obermayer & Blasdel, 1993), and as one can verify by comparing these spectra with the ocular dominance patterns used as input (see Fig. 4), periodic bands aligned along one axis produce two clusters of modes BN-7.5 that extend along an orthogonal axis and that are sep- arated by an amount proportional to the frequency of repetition. In other words, more widely spaced modes indicate a higher frequency of repetition. The axis of highest frequency repetition can therefore be calculated directly, and represented by a solid line that intersects the origin and both modes. As one can see, the resulting lines on the left and right, indicating axes of high- BN-9 est frequency repetition for orientation and ocular dominance, tend to run at right angles to one another in each row. This is similar to the orthogonal relationship between orientation and ocular dominance shown previously for adults (see Obermayer and Blasdel, 1993). Because this phenomenon results from the local intersection of ocular dominance bands and orientation bandlets at right angles, the global relationship becomes BN-14 most pronounced when ocular dominance bands are aligned significantly over a large region.

acquired in young animals, we conclude that gross topographies of data points, and that the resulting trends towards expansion of orientation selectivity and ocular dominance are well estab- are consistent and clear. As one can see in Fig. 10a, where col- lished by 3 5 weeks of age. umn widths are plotted as a function of age, this increase shows with a linear correlation coefficient of r = 0.94. The observed increase also agrees with previous observations (Hubel et al., 1977a) that the ocular dominance bands observed in layer 4c Ocular dominance periods and cortical surface area of infants appear 20% narrower than they do in adults. While the actual mechanisms that generate orientation prefer- By measuring the area of striate cortex carefully, in 11 infant ence and ocular dominance remain to be elucidated, it is clear macaques at various ages, Purves and LaMantia (1993) esti- (for Old World primates) that ocular dominance columns align mated that the area of striate cortex expands by approximately with segregated bands of right and left eye afferents in layer 16% after birth. Since this approximates the currently observed 4c (Rakic, 1976; Hubel & Wiesel, 1972; LeVay et al., 1975). increase in ocular dominance width, however, it can be ac- Since these anatomically segregated bands of axons are unlikely counted for entirely by one-dimensional expansion. A more sym- to rearrange themselves in any meaningful way, other than to metrical, two-dimensional expansion (e.g. perpendicular and segregate more completely, their alignment with ocular domi- parallel to the ocular dominance columns) is in fact ruled out, nance slabs means that the positions of the latter should be rel- since this would require an increase of at least 40% — more than atively stable, anchored in place, as it were, by these afferents. twice that actually observed. The fact that a two-dimensional Consequently, it is no more surprising that the spacing of ocu- expansion is at some level required can be reconciled with these lar dominance columns increases with age than it is that the cor- findings by noting that ocular dominance columns assume a tical surface expands. And such an increase is apparent in the number of different orientations that would allow for growth right-hand column of Fig. 4, where the ocular dominance pat- in all directions. Given the orthogonality between ocular dom- terns from different animals appear above one another, accord- inance columns and the 17/18 border (LeVay et al., 1975), a ing to age. While there are only five patterns in this sample, it region that must undergo particularly rapid elongation with cor- is important to realize that each is defined by tens of thousands tical expansion, the ocular dominance columns seen in adult ani-

BN-5.5

L = Linear zone S = Singularity F = Fracture H = Saddle-point

BF-3.5

1 mm

Fig. 7. This figure shows enlarged sections of the orientation preference maps obtained from the two youngest animals (BN- 5.5 and BF-3.5). The representations on the left depict orientation preferences in opponent colors for orthogonal orientations (with red and green indicating horizontal and vertical and blue and yellow indicating left and right oblique, respectively) — similar to the example in Fig. la. The representations on the right depict contours of iso-orientation, similar to those in Fig. lb, where the indicated lines represent paths along which orientation preferences remain constant. Accordingly, these representations facilitate the detection of linear zones (which appear as evenly spaced sets of parallel lines), saddle points (which appear as blank zones), fractures (indicated by clumps of lines), and singularities (where multiple lines converge), and make it clear that all four of these organizations are present in approximately adult proportions, with approximately adult spacing, in even the youngest animals.

mals might be little more than fiduciary "stretch marks," left that patchy lateral connections develop in the upper layers before by an earlier expansion. their inputs (from layer 4) arrive (Callaway & Katz, 1992). Differential growth, perpendicular to the ocular dominance Differential growth might also underlie the anisotropic dis- columns, is also suggested by the anisotropic magnification fac- tribution of orientation preferences, which repeat at higher fre- tors seen in adults, where the magnification across ocular dom- quencies along the ocular dominance columns than they do inance columns exceeds that along them by approximately 60% across them. In this instance, it could explain why regions pre- (Van Essen et al., 1984; Tootell et al., 1988). Given the likeli- ferring the same orientation tend to elongate into slabs in the hood that this arises from duplicate, side-by-side representations first place, and why these slabs cross the borders of ocular dom- of the visual field in layer 4c, as commonly argued (LeVay et al., inance columns at angles of approximately 90 deg. Regions pre- 1975; Blasdel & Fitzpatrick, 1984), it should not be present, or ferring the same orientation might literally become stretched, even necessary, at birth, before the ocular dominance bands in perpendicular to the ocular dominance columns and into their layer 4c emerge. In fact, an isotropic representation would make adult form, as the cortical surface expands. If so, much of this more sense, in the context of plausible developmental mecha- stretching would have to occur early, since the orthogonality nisms. And a subsequent expansion, perpendicular to emerg- between orientation and ocular dominance slabs is well estab- ing ocular dominance columns, could explain the resulting lished by 3| weeks of age (see Figs. 6, 8, and 9). anisotropy nicely. It could also explain why the patchy horizontal connections, seen in the upper layers of striate cortex following microinjections of biocytin (Yoshioka et al., 1992; Patterns of orientation preference Blasdel et al., 1992; Yoshioka et al., 1995), extend almost twice If adult patterns of orientation and ocular dominance really are as far across the ocular dominance columns as they do along as correlated as our recordings suggest (Blasdel, \992b; Ober- them. And it also fits with recent findings, in cat striate cortex, mayer & Blasdel, 1993), the periodicity of orientation preferences

BN-5.5

22° 45° 67° 90°

BN - 5.5

BF-3.5

22° 45° 67° 90°

BN - 7.5

1 mm 22° 45° 67° 90° Fig. 8. When the borders of ocular dominance columns are added to the iso-orientation contours from Figure 7, it becomes clear: (1) that 16% singularities (indicated by lines converging to points) lie mostly in the centers of ocular dominance columns; and (2) that where iso-orientation 12%. contours cross the borders of ocular dominance columns, they do so at angles of approximately 90 deg, as observed previously in adults BN-9 (Obermayer & Blasdel, 1993). Both of these tendencies are present and robust by 3j weeks of age.

0° 22° 45° 67° 90° also should expand with age. Yet it does not. From Table 2 and Fig. 10b, one can see that even along axes perpendicular to the ocular dominance columns, the periodicity of orientation pref- 16% erences increases only half as fast as the periodicity of ocular dominance. Since orientation and ocular dominance patterns 12%. were imaged from the same cortical regions at the same times, BN-14 moreover, this discrepancy cannot be attributed to individual variation or other experimental variables (e.g. swelling). One possibility is that the periodicity of orientation preferences does not correlate with the periodicity of ocular domi- o% nance. But other data (Obermayer & Blasdel, 1993) suggest that 22° 45° 67° 90° it does —at least in adult animals. The fact that this relation appears not to hold across infants, recorded at different ages, Fig. 9. When the angles of intersection, for iso-orientation contours crossing lines of iso-ocular dominance, are tabulated in the linear zones might therefore reflect a changing relationship between orien- OP tation and ocular dominance as infant animals mature. This is (i.e. where the calculated value of alignment, P , exceeds 0.6), a strong preponderance of 90-deg intersections becomes apparent, and also the implication of Table 3, where it is clear that the den- is especially pronounced in the youngest animal, whose patterns of ocu- sity of singularities (points where preferences for all orienta- lar dominance and orientation were determined at 3j weeks of age. tions converge) tends to remain the same as the cortical surface Hence, the orthogonality between ocular dominance and iso-orientation expands. The only obvious way in which this can happen is if contours described previously in adults (see Obermayer & Blasdel, 1993), new singularities are added in pairs (e.g. one "+" singularity is well entrenched by 3j weeks of age.

Table 3. Percentage of cortical area covered by linear zones, where our measure of alignment POP' exceeded 0.6, and the densities of different types of singularities0

Density of Linear zones Density of +180 deg Density of -180 deg singularities Animal Age (weeks) (% of area) singularities (mm~2) singularities (mm~2) (mm-2)

BF-3.5 3.5 47 3.9 3.9 7.8 BN-5.5 5.5 49 3.7 3.7 7.4 BN-7.5 7.5 45 4.5 4.5 9.0 BN-9 9 54 3.7 3.7 7.4 BN-14 14 36 3.9 3.8 7.7 Adult (nemestrina) 48 ± 7 4.0 ± 0.4 3.9 ± 0.5 7.9 ± 0.9 Adult (fascicularis) 49 4.1 4.2 8.3

aThe values for adult Macaca nemestrina reflect averages of four animals (with error margins derived from standard deviations). The values for adult Macaca fascicularis are averages from two recording sites in the same animal. Note (right-hand column) that the density of singularities (the number per unit area) stays approximately the same as infant animals mature. This contrasts with the spacing of ocular dominance columns, which increases by at least 20%, and even the area of striate cortex, which is thought to increase by at least 16% (Purves & LaMantia, 1993), during the same period of time. Data for the adults were taken from Obermayer and Blasdel (1993).

and one "—" singularity; see Blasdel, 1992ft, for details), as the are not unprecedented (Hirsch & Spinelli, 1970, 1971; Blake- cortical surface expands. more & Cooper, 1972; Blasdel et al., 1977). And while some These findings suggest the possibility that patterns of orien- of these reports have been contested (see Stryker & Sherk, 1975, tation preference may not be as stable or as hard wired as pat- for example), recent descriptions of dynamic change in the terns of ocular dominance. And since they are not anchored in response properties of adult neurons, including orientation pref- place by anatomically segregated bands of afferents, some erences, receptive-field sizes, and positions (Gilbert & Wiesel, degree of rearrangement may actually be possible. As startling 1992), suggest that developmental changes in these properties as this possibility may seem, reports of orientational plasticity are not out of the question.

a) b) Correlation coefficient (r) = 0.939 Correlation coefficient (r) = 0.576 Slope = 13.031 Slope = 6.85 950 950 O OP perpendicular 900

600 600 8 10 12 14 16 Age Age

Fig. 10. The spacing of ocular dominance columns and orientation preferences, as a function of age (indicated in weeks), (a) The strong correlation between age and the spacing of ocular dominance columns becomes especially apparent when one is plotted as a function of the other. As one can see, the average spacing of ocular dominance columns increases by approximately 12 pm per week between 3| and 14 weeks of age, for a total increment of 20%, which agrees approximately with estimates derived from the earlier work of Hubel et al. (1977a) that seem to suggest a 25% increase during the first few months of life. If the calculations of Purves and LaMantia (1993) are correct, therefore, and the total surface area of striate cortex increases only by 16% or so after birth, the observed 20% increase in the ocular dominance spacing suggests strongly that most of this growth occurs anisotropically, along one axis — perpendicular to the ocular dominance columns —as the neonatal cortex matures, (b) In this figure, where the average spacing of orientation preferences perpendicular to the ocular dominance columns is plotted as a function of age, it becomes apparent that the spacing of orientation preferences increases only half as fast as the spacing of ocular dominance columns, measured along similar axes. This implies that the relationships between ocular dominance and orientation preference that appear so prominent in adult cortex may not be rigid in young animals, and indeed may undergo some re-arrangement as infant animals mature. These studies clearly need to be replicated in greater detail, however, preferably by following the development of identified patterns of ocular dominance and orientation preference in single animals over time.

Conclusion HUBEL, D.H. & WIESEL, T.N. (19746). Uniformity of monkey striate cortex: A parallel relationship between field size, scatter, and mag- These results demonstrate that many, if not all, of the organi- nification factor. Journal of Comparative Neurology 158, 295-306. zational features of adult striate cortex are well established by KIORPES, L. & BLASDEL, G.G. (1987). Optical studies of macaque stri- 35 weeks of age. They also raise new questions about the sym- ate cortex during development. Neuroscience Abstract 13, 11243. LEVAY, S., HUBEL, D.H. & WIESEL, T.N. (1975). The pattern of ocular metry of cortical development, whether it expands uniformly dominance columns in macaque striate cortex revealed by a reduced or perpendicular to the ocular dominance columns, and the sta- silver stain. Journal of Comparative Neurology 159, 559-576. bility of relationships between orientation and ocular dominance LEVAY, S. WIESEL, T.N. & HUBEL, D.H. (1980). The development of patterns, as infant animals mature. These questions may best ocular dominance columns in normal and visually deprived mon- be answered, it seems, by using similar methods to follow the keys. Journal of Comparative Neurology 191, 1-51. OBERMAYER, K. & BLASDEL, G.G. (1993). Geometry of orientation and development of ocular dominance and orientation patterns, as ocular dominance columns in monkey striate cortex. Journal of Neu- well as cortical magnification, in single infant animals as they roscience 13, 4114-4129. mature. OBERMAYER, K., BLASDEL, G.G. & SCHULTEN, K. (1991). A neural network model for the formation and for the spatial structure of ret- inotopic maps, orientation and ocular dominance columns. In References Artificial Neural Networks, ed. KOHONEN, T., et al., pp. 505-511. Amsterdam, The Netherlands: Elsevier (North Holland). BARTFELD, E. & GRINVALD, A. (1992). Relationships between OBERMAYER, K., BLASDEL, G.G. & SCHULTEN, K. (1992a). A statistical orientation-preference pin-wheels, cytochrome oxidase blobs, and mechanical analysis of self-organization and pattern formation dur- ocular-dominance columns in primate striate cortex. Proceedings ing the development of visual maps. Physiological Reviews A 45, of the National Academy of the Sciences of the U.S.A. 89, 11905- 7568-7589. 11909. OBERMAYER, K., SCHULTEN, K. & BLASDEL, G.G. (19926). A compari- BLAKEMORE, C. & COOPER, G.F. (1972). Development of the brain son of a neural network model for the formation of brain maps with depends on the visual environment. Nature (London) 228, 477-478. experimental data. In Advance in Neural Information Processing BLASDEL, G.G., MITCHELL, D.W., MUIR, D.W. & PETTIGREW, J.D. Systems, ed. TOURETZKY, D.S. & LIPPMAN, R., pp. 83-90. San (1977). A physiological and behavioral study in cats of the effect Mateo, CA: Morgan Kaufmann Publishers. of early visual experience with contours of a single orientation. Jour- OBERMAYER, K., KIORPES, L. & BLASDEL, G.G. (1994). Development nal of Physiology (London) 265, 615-635. of orientation and ocular dominance columns in infant macaques. BLASDEL, G.G. (1989). Visualization of neuronal activity in monkey stri- In Advances in Neural Information Processing Systems 6, ed. ate cortex. Annual Review of Physiology 51, 561-581. COWAN, J., TESAURO, G., & ALSPECTOR, J., pp. 543-550. San BLASDEL, G.G. (1992a). Differential imaging of ocular dominance and Mateo: CA: Morfan Kaufmann Publishers. orientation selectivity in monkey striate cortex. Journal of Neuro- PETTET, M.W. & GILBERT, CD. (1992). Dynamic changes in receptive science 12, 3117-3138. field size in cat primary visual cortex. Proceedings of the National BLASDEL, G.G. (19926). Orientation selectivity, preference, and conti- Academy of the Sciences of the U.S.A. 89, 8366-8370. nuity in monkey striate cortex. Journal of Neuroscience 12, 3139— PETTIGREW, J.D., COOPER, M.L. & BLASDEL, G.G. (1979). Improved 3161. use of tapetal reflection for eye position monitoring. Investigative BLASDEL, G.G. &.FITZPATRICK, D. (1984). Physiological organization Ophthalmology and Visual Science 18, 490-495. of layer 4 in macaque striate cortex. Journal of Neuroscience 4, PRESS, W.H., FLANNERY, B.P., TEUKOLSKY, S.A. & VETTERLING, W.T. 880-895. (1988). Numerical Recipes in C. Cambridge, MA: Cambridge Uni- BLASDEL, G.G. & SALAMA, G. (1986). Voltage sensitive dyes reveal a versity Press. modular organization in monkey striate cortex. Nature 321, 579-585. PURVES, D. & LAMANTIA, A. (1993). Development of blobs in the visual BLASDEL, G.G., YOSHIOKA, T., LEVITT, J.B. & LUND, J.S. (1992). Cor- cortex of macaques. Journal of Comparative Neurology 332, 1-7. relation between patterns of lateral connectivity and patterns of RAKIC, P. (1976). Prenatal genesis of connections subserving ocular orientation preference in monkey striate cortex. Society for Neuro- dominance in the rhesus monkey. Nature (London) 261, 467-471. science Abstracts 18, 389. STRYKER, M.P. & SHERK, H. (1975). Modification of cortical orientation- BONHOEFFER, T. & GRINVALD, A. (1991). Orientation columns in cat selectivity in the cat by restricted visual experience: A reexamina- are organized in pin-wheel like patterns. Nature 353, 429-431. tion. Science 190, 904-906. CALLAWAY, E. & KATZ, L. (1992). Development of axonal arbors of layer SWINDALE, N.V. (1982). A model for the formation of orientation col- 4 spiny neurons in cat striate cortex. Journal of Neuroscience 12, umns. Proceedings Royal Society B (London) 215, 211-230. 570-582. SWINDALE, N.V. (1992). A model for the coordinated development of DESROSIERS, M.H., SAKURADA, O., JEHLE, J., SHINOHARA, M., KEN- columnar systems in primate striate cortex. Biological Cybernetics NEDY, C. & SOKOLOFF, L. (1978). Functional plasticity in the imma- 66, 217-230. I4 ture striate cortex of the monkey shown by the [ C) deoxyglucose TOOTELL, R.H., SWITKES, E., SlLVERMAN, M.S. & HAMILTON, S.L. method. Science 200, 447-449. (1988). Functional anatomy of macaque striate cortex. II. Retino- GILBERT, CD. & WIESEL, T.N. (1992). Receptive field dynamics in adult topic organization. Journal of Neuroscience 8(5), 1531-1568. primary visual cortex. Nature 356, 150-152. T'so, D.Y., FROSTIG, R.D., LIEKE, E.E. & GRJNVALD, A. (1990). Func- HIRSCH, H.V.B. & SPINELLI, D.N. (1970). Visual experience modifies tional organization of primate visual cortex revealed by high reso- distribution of horizontal and vertical oriented receptive fields in lution optical imaging. Science 249, 417-420. cats. Science (N.Y.) 168, 869-871. VAN ESSEN, D.C, NEWSOME, W.T. & MAUNSELL, H.R. (1984). The visual HIRSCH, H.V.B. & SPINELLI, D.N. (1971). Modification of the distri- field representation in striate cortex of the macaque monkey: Asym- bution of receptive-field orientation in cats by selected visual expe- metries, anisotropies, and individual variability. Vision Research 24, riences during development. Experimental Brain Research 13, 429-448. 509-527. WIESEL, T.N. & HUBEL, D.H. (1974). Ordered arrangement of orienta- HUBEL, D.H. & WIESEL, T. (1962). Receptive fields, binocular inter- tion columns in monkeys lacking visual experience. Journal of Com- action, and functional architecture of cat striate cortex. Journal of parative Neurology 158, 307-318. Physiology (London) 160, 106-154. WOLBARSHT, M.L., MACNICHOL, E.F. & WAGNER, H.G. (1960). Glass HUBEL, D.H. & WIESEL, T. (1968). Receptive fieldsan d functional archi- insulated platinum microelectrode. Science 132, 1309-1310. tecture of monkey striate cortex. Journal of Physiology (London) YOSHIOKA, T, BLASDEL, G., LEVITT, J. & LUND, J.S. (1992). Patterns 195, 215-243. of lateral connections in macaque visual area VI revealed by biocy- HUBEL, D.H. & WIESEL, T.N. (1972). Laminar and columnar distribu- tin histochemistry and functional imaging. Society for Neuroscience tion of geniculo-cortical fibers in the macaque monkey. Journal of Abstracts 18, 299. Comparative Neurology 146, 421-450. YOSHIOKA, T., BLASDEL, G.G., LEVITT, J.B. & LUND, J.S. (1995). Rela- HUBEL, D.H. & WIESEL, T.N. (1974a). Sequence regularity and geom- tion between patterns of intrinsic lateral connectivity, ocular domi- etry of orientation columns in the monkey striate cortex. Journal nance, and cytochrome oxidase-reactive regions in macaque monkey of Comparative Neurology 158, 267-294. striate cortex. Cerebral cortex (in press).