Multiple loci for foveolar vision in macaque monkey visual cortex

Abstract

In humans and nonhuman primates, the central 1° of vision is processed by the foveola, a retinal structure that comprises a high density of photoreceptors and is crucial for primate-specific high-acuity vision, color vision and gaze-directed visual attention. Here, we developed high-spatial-resolution ultrahigh-field 7T functional magnetic resonance imaging methods for functional mapping of the foveolar visual cortex in awake monkeys. In the ventral pathway (visual areas V1-V4 and the posterior inferior temporal cortex), viewing of a small foveolar spot elicits a ring of multiple (eight) foveolar representations per hemisphere. This ring surrounds an area called the 'foveolar core', which is populated by millimeter-scale functional domains sensitive to fine stimuli and high spatial frequencies, consistent with foveolar visual acuity, color and achromatic information and motion. Thus, this elaborate rerepresentation of central vision coupled with a previously unknown foveolar core area signifies a cortical specialization for primate foveation behaviors.

Fig. 1. Acquisition stability.

a, Left, head motion in six dimensions: roll, pitch, yaw, A–P (anterior–posterior), R–L (right–left) and I–S (inferior–superior). Right, head displacement: disp, displacement relative to the initial frame (mean = 120 µm); Δdisp, difference between two sequential frames (mean = 80 µm). Data were averaged across eight sessions in monkey E. b, Eye fixation behavior. Left, eye position trace in one run. Red and blue lines represent the x and y eye position coordinates, respectively. Two red arrows denote blinks. Note: eye blinks did not cause disruptions of the BOLD signal. Right, all eye positions in four runs (250 Hz, 4 × 280 s). Eye fixations were within a 0.5° radius 94% of the time. Right, despite occasional saccades, across four runs (each trial: 20 s blank and 20 s visual stimulus; seven trials per run; a total of 280 s), the center of mass of the eye trace was within a 0.5° radius of the fixation point (central dotted red circle). With sufficient training, the monkeys maintained stable performance with little body motion and few saccades during scanning. c, BOLD signal time course. Left, time course of activated voxels during eye fixation in b (n = 348 voxels). Right, average BOLD time course. Yellow shading denotes the stimulus on period. d, Map stability. Maps obtained from 3, 7 and 28 trials. Left inset, visual stimulus: 1° and 5° eccentricity rings. Activation clusters remained stable across trials, indicating stable eye fixation. Yellow arrows, V1; red arrows, V2. Right, each dot is a measurement of the shift in center of mass between activations from 7 and 28 trials. For each of the activation clusters, six measurements were taken from six sequential slices. e, Increased trial number, from 7 (middle panel) to 28 (right panel), reveals additional visual areas. Left, section from D99 atlas. Activations in V1 (1° and 5°), V2 (yellow circles), V3 (green arrow), V4 (blue arrow) and TEO (red arrows). Same as in d. Map thresholds were significant at P < 10⁻³ according to a one-sided t-test. All data are presented as the mean values ± s.e.m. Source data

Fig. 2. Mapping visual cortical borders by mapping HM and VM and isoeccentricity mapping.

a,b, Stimuli for imaging VM and HM, in monkey E (a) and monkey J (b), comprised horizontal and vertical 0.15°-wide bands containing alternating saturated red and blue checkerboards (insets). All images were acquired within the same sessions but illustrated separately for each area (Methods). For each monkey, the four panels on the left were overlain in the rightmost images. Black and white dashed lines denote the VM and HM, respectively. LS, lunate sulcus; IOS, inferior occipital sulcus; STS, superior temporal sulcus. ECS, ectocalcarine sulcus. Monkey E, P < 10⁻³; monkey J, P < 10⁻⁹ (one-sided t-test). c, Isoeccentricity maps from monkey E. A paired stimulus paradigm (six sets of paired rings: 0.4° and 2°, 0.4° and 4°, 0.6° and 3°, 0.8° and 4°, 1° and 5° and 1.2° and 6°) reduced the number of stimulus conditions needed by half and further confirmed that the maps were stable across sessions (for example, similar 4° activations were obtained in 0° and 4°, 0.4° and 4° and 0.8° and 4° maps). White circle, intersection location between isoeccentricity and meridians. d, Isoeccentricity (0.8° and 4°, 1° and 5° and 1.4° and 7°) maps from monkey J. e,f, Summary of findings in c (e) and d (f).

Fig. 3. Determining locations of foveolar representation.

a, Activations to foveated small spot stimuli (monkey E: 0.4°, 0.6° and 0.8° flashing saturated red and blue squares alternating at 3 Hz, shown in the first, second and third row, respectively). The activation maps to each stimulus are shown in each column (TEO/FST, V4/TEO, V3/V4, V2/V3 and V1/V2). Each panel presents significant voxels at each of three thresholds (P values according to a one-sided t-test are color-coded: orange, lowest; green, middle; blue, highest). For each visual area, the center of highest significance is consistent across spot sizes (colored dashed lines). The last column consists of the set of all foveola (colored dashed circles, overlay from four leftmost columns). Black and white dotted lines denote the VM and HM, respectively. Gyri, light gray; sulci, dark gray. b, Foveolar activations in monkey E, right hemisphere (same as response to 0.6° in a). c, Foveolar activations in monkey E, left hemisphere. d,e, Foveolar activations in monkey J’s left hemisphere (d) and right hemisphere (e) (Extended Data Fig. 1). In b–e, small colored circles denote the foveolar activations in each cortical area. Lavender dotted oval, foveolar core.

Fig. 4. Phase encoding using fine isoeccentricity and isopolarity stimuli in visuotopic V1–V4 and in foveolar core.

a, Isoeccentricity map. Inset, isoeccentricity stimulus: continuous increasing size of isoeccentric rings (each ring: 0.15° wide; one trial consisted of 50 rings over 3°, 0.06° per ring, lasting 50 s with a 10-s ISI; fixation point: 0.05° white dot, constant over a total of 21 trials). b, Isopolarity map. Inset, isopolarity stimulus: continuous rotating checkerboard wedge (5° wedge, rotated over 50 s from 90° to −90° in the contralateral visual field; size: 3°, over a total of 21 trials). Black circles in a and b denote the locations of voxel time courses shown in d,e. White circles in a and b denote the foveolar loci identified in Fig. 3d,e. c, Histogram showing that FFT amplitude at the stimulus frequency is weaker in ‘inside core’ compared to ‘outside core’ samples. FFT amplitude: inside core, mean = 1.2; outside core, mean = 3.0; P < 0.0001 (two-sided chi-square test). d, Isoeccentricity time courses (BOLD%). ROIs 1–6 are in V1 along the VM. Red arrowheads (center of mass of peak) indicate a peak shift as ROIs approach the center. ROIs 7–15 are in the foveolar core along an anteroposterior axis through the core. ROIs 16 and 17 are along a posterosuperior axis. ROIs 18 and 19 are along a posteroinferior axis. ROIs 20–22 are along an anteroinferior axis. e, Isopolarity time courses (BOLD%). ROIs 1–9 are along three isoeccentricity rings in V1. ROIs 10–12 are in foveolar core parallel to isoeccentricity rings in V1. Note: because the dorsal cortex represents ventral fields, the phase shift because of the wedge approaching the VM in the ventral field results in a shortening latency to response (red arrowheads). Red boxes represent time courses from ROIs in the core region. Numbers in the right corner of each graph denote FFT amplitude indices (Extended Data Fig. 6). All data in d,e are presented as mean values ± s.e.m. Source data

Fig. 5. Mesoscale functional domains within the foveolar core.

Color-coded responses to different visual stimuli. a, Activations to achromatic high-SF gratings (in cycles per degree). Yellow, SF11; purple, SF15; orange, SF18. b, Overlay of high-SF (yellow, SF11 + SF15 + SF18) and low-SF (red, SF0.2) activations. c, Overlay of low-SF (red, SF0.2), motion (cyan, clockwise moving dots) and color (blue, 0.8° flashing spot) activations. d, Overlay of all domains. e, Four fields of view from d showing multiple types of functional domains within a small region. f, Averaged time course of each domain type selected from 3–5 clusters within the core. All data are presented as the mean values ± s.e.m. The numbers of voxels from top to bottom are 588, 420, 387 and 385, respectively. g, Overlay indices between different populations of domains show little overlap between domains (overlap indices: 0.09, range 0–0.25). h, Domain size distribution (total number: 111; Methods and Extended Data Fig. 8). Source data

Fig. 6. Multiple foveolar representations revealed by optical imaging.

a, Ocular dominance (OD) map. Scale bar, 1 mm. b, Orientation map (45° versus 135°). Green outlines, V4 orientation bands; red outlines, V4 color bands. c, Color map (color versus achromatic). Yellow arrows, color stripes in V2 (same locations as stripes shown in d). d, Cytochrome oxidase stained stripes in V2 align well with color stripes in c (yellow arrows). e, Blood vessel map over V1, V2, V4 and TEO. M, medial; L, lateral; A, anterior; P, posterior. f, Color-coded eccentricity map (red, 0.01°; yellow, 0.10°; green, 0.20°; cyan, 0.30°; blue, 0.40°; purple, 0.50°). Lavender dashed circle, foveolar core; colored arrows, shifted eccentricities. White arrow: V1/V2 border. g, Color-coded isopolarity map (red, 0°; yellow, 15°; light green, 30°; dark green, 45°; cyan, 60°; blue, 75°; purple, 90°). Numbered white dotted circles, foveolar locations at V1/V2 (1 and 2), V2v/V3v (4), V3v/V4v (6), V4v/TEO (7) and TEO/FST (8). Locations V2d/V3d (3) and V3d/V4d (5) are within the LUS and not visible in optical imaging. Colored arrows, shifted polar angles. h, For comparison, 7T MRI of all foveola in monkey E (left hemisphere, from Fig. 3b) numbered with the corresponding locations in g. The orientation of all optical imaging maps was rotated to correspond with the MRI results shown in h.

Fig. 7. The foveolar core has a large CMF.

a,b, Linear CMF in central 1° of V1 as a function of eccentricity for monkey E (a) and monkey J (b) (Extended Data Fig. 1). Colored lines were taken from a previous study. c, Linear CMF in central 0.5° of V1 (from optical imaging data in Extended Data Fig. 9). d,e, Foveolar core bounded by eight foveolar loci in monkey E (d) and monkey J (e). Yellow outlines denote the boundaries for area size calculation, approximated by a pink dotted oval. f, Optical image of central 0.01° and 0.1° (yellow circles). The CMF of the central 0.1° (yellow pixels) is 1,500 mm² per degree and that of the central 0.01° (half of the red pixels in V1/V2d) is 22,000 mm² per degree. g, A summary schematic of topographic (light gray) and nontopographic (dark gray) foveolar representation. Numbers denote the estimated area (mm²) per degree. Source data

Fig. 8. Revised view of visual foveolar representation.

a, Previous classical view of topographic representation (light blue). The foveolar confluence (red stars) comprises a single foveolar locus (top) or one locus per area (bottom; in humans and macaques)^,. b, Our study shows that (1) the foveolar center is represented eight times (eight red stars), one at each of the dorsal and ventral representations of V1/V2, V2/V3, V3/V4 and V4/TEO; (2) the foveolar core (orange) is an area within the ring of stars and outside the area of visuotopic representation; and (3) there are functional domains (colored dots) within the core that are responsive to large but not small foveolar stimuli (including color, orientation and motion stimuli). Red lines, HM; green lines, VM.

Extended Data Fig. 1. Determining locations of foveolar representation on Monkey J.

Shown in each row (V1/V2, V2/V3, V3/V4, V4/TEO, and TEO/FST) are the activation maps to each of the 3 spot stimuli (0.6°, 0.8°, 1.0°). Conventions same as Fig. 3a.

Extended Data Fig. 2. Slice views of bilateral foveolar representation on Monkey E. (A, B).

Surface view of multiple foveolar representations on the left (A) and right (B) hemispheres, methods same as in Fig. 3. [p-values indicated by color code: orange (lowest), green (middle), and blue (highest)]. (C) Slice view of bilateral foveolar activations to spot stimulation of 0.8 deg in transverse sections. Color circles refer to the foveolar loci in A&B. Numbers: slice number. (D) Corresponding sections to C in D99 atlas. Color circles: foveolar activations corresponded with visual area borders.

Extended Data Fig. 3. Stability and separability of foveolar loci despite different distributions of eye movements.

With (A) poorer (85% within 1°) and (B) better (95% within 1°) eye fixation behavior, activations on the cortex remain very similar (see circles). We considered the possibility that the spatial distribution of eye movements leads to stimulation by the edges of the spot that activate the loci that we interpret as the foveolar centers. We show here that our data preclude this possibility. First, this scenario predicts that if, the eye movement distribution is larger, then the edges should activate more peripheral loci on the cortex. However, in two sessions in which the monkey exhibited (A) poorer (85% within 1°) and (B) better (95% within 1°) eye fixation behavior, the activations on the cortex remain very similar (see circles marking foveolar loci). Second, it predicts that a larger (for example 0.8°) would activate more peripheral locations on the cortex than a smaller (for example 0.4°) stimulus. However, as shown in Fig. 3 and Extended Data Fig. 1 of the manuscript, the foveolar loci are stable across 3 spots sizes, in each of 2 monkeys. Finally, if the responses were due to edge effects from eye movements, the result would not be two focal and punctate activations, but a continuous activation zone elicited by the many directions of eye movements. Logically, blurring would only serve to make the loci larger and less separable, thereby indicating our results are an overestimate of foveolar activation size, which would mean the actual size of foveolar loci are even smaller and even more distinguishable. Thus, our results are not an artifact of eye movements. Source data

Extended Data Fig. 4. Retinotopic stimuli do not activate foveolar core.

Comparison of (A) Color alone (0.6° red/blue square), p < 10⁻⁵ evokes multiple foveolar activations (yellow circles) (p < 10⁻⁵), revealing activation-free foveolar core (dotted pink oval). (B) Small fixation cross (lines 0.1° wide, 0.3° long) evokes some weak response at foveolar loci (threshold lowered to p < 10⁻³ for visibility), and no activation in core. (C) Map of 0.15° wide lines in HM (red) minus VM (blue) does not invade core area (same as manuscript Fig. 2b). We considered whether the absence of activation in the foveolar core was due to the subtraction of fixation cross. Single condition maps (no subtraction) to small foveal color spots (0.6° red/blue square) p < 10⁻⁵) evoke foveolar activations (A, yellow circles) revealing activation-free foveolar core area. The small fixation cross (lines 0.1° wide, 0.3° long) alone evokes little activation (B); thus, the absence of signal in the core remains, and is not due to fixation cross subtraction. We also note that the HM and VM stimuli robustly delineate the visuotopic regions of V1-V4 but fail to invade the core; in fact, each of these activations terminate near their respective foveolar loci (C).

Extended Data Fig. 5. FFT analysis of the time course.

Upper panel: Original time-course. All data are presented as mean values ± SEM. Lower panel: FFT analysis of upper time-course. (A) Example form ROI #15 (Fig. 4) within the core, n = 85 voxels. Red arrow: one of multiple peaks in FFT, indicating lack of clear phase encoding. (B) Example from ROI #6 (Fig. 4) outside the core, n = 63 voxels. Red arrow: clear single peak FFT at frequency of phase encoding stimulus. Source data

Extended Data Fig. 6. Similarities of human foveolar maps with this study.

(A) Phase encoding retinotopic mapping adapted from Schira 2009 Fig. 5 reveals nonhomogeneous foveolar region (inside dashed ovals). (B) pRF analysis of retinotopic mapping adapted from Doumulin 2008 Fig. 5. Dashed oval: non-homogenous central area. Note several red foveolar loci at the ends of VM and HM (black arrows) on the dashed ring, an organization similar to what we find. These maps reveal strong similarities with our phase encoding maps, though this organization was not previously recognized. Close perusal of these images reveals the presence of similar foveolar loci along a ring surrounding a central nonhomogeneous region (A). Although statistical thresholding was used, it would be important to see the strength of correlation of phase encoding for the central field voxels. We also found potentially similar published results using retinotopic mapping by pRF analysis (B). The central area in these maps was not homogeneous and exhibited focal red loci which fell at the end of the VM and HMs and on a ring that encircled what may be the foveolar core. Other published phase encoded maps also illustrate VM and HMs that stop short of a common center. We suggest that the core region does respond to foveolar stimuli, but their receptive fields may not be topographically ‘foveola only’ (for example the lack of response to small focal stimuli and, in some cases, preference for large stimuli, see Fig. 5 on functional domains).

Extended Data Fig. 7. Little activation to standard visual stimuli in the foveolar core.

Color stripe mapping in V2. Color grating vs achromatic 1 cyc/deg grating. Green outline: four color stripes in V2.

Extended Data Fig. 8. Domain size calculation and distribution.

Long axis of domain measured in the surface mesh. Two examples (1 mm and 2 mm size) shown.

Extended Data Fig. 9. Optical imaging single condition maps.

A-F. Images (original raw data) of cortical activation to iso-eccentric arcs of radius r in degrees. A: r = 0.01°, B: r = 0.10°, C: r = 0.20°, D: r = 0.30°, E: r = 0.40°, F: r = 0.50°. Scalebar = 1 mm. M: medial, A: anterior, LUS: lunate sulcus, STS: superior temporal sulcus, ISO: inferior occipital sulcus. G. Color-coded eccentricity vector map (vector summation of 6 iso-eccentricity responses). Inset: iso-eccentricity color code. H. Central (0.01°) stimulus activates a surprisingly large cortical region (red pixels) measuring over 7 mm in extent along V1/V2 border. I-O. Images of cortical activation to iso-polar bars in angular rotation Δ in degrees. H: Δ = 0°, I: Δ = 15°, J: Δ = 30°, K: Δ = 45 °, L: Δ = 60°, M: Δ = 75 °, N: Δ = 90°. P. Color-coded angularity vector map (vector summation of 7 iso-polarity responses). (Same case shown in Fig. 4c–g).