Receptive-Field Properties of Neurons in Middle Temporal Visual Area

JOURNALOF NEUROPHYSIOLOGY Vol. 52, No. 3, September 1984. Printed in U.S.A.

Receptive-Field Properties of Neurons in Middle Temporal Visual Area (MT) of Owl Monkeys

DANIEL J. FELLEMAN AND JON H. KAAS

Departments of Psychology and Anatomy, Vanderbilt University, Nashville, Tennessee 3 7240

SUMMARY AND CONCLUSIONS

1. Response properties of single neurons in the middle temporal visual area (MT) of anesthetized owl monkeys were determined and quantified for flashed and moving bars of light under computer control for position, orientation, direction of movement, and speed.

2. Receptive-field sizes, ranging from 4 to 25” in width, were considerably larger than receptive fields with corresponding eccentricities in the striate cortex. Neurons were highly binocular with most cells equally or nearly equally activated by either eye.

3. Neurons varied in selectivity for axis and direction of moving bars. Some neurons demonstrated little or no selectivity, others were bidirectional on a single axis, while the largest group was highly selective for direction with little or no response to bar movement opposite to the preferred direction. Over 70% of neurons were classified as highly selective and 90% showed some preference for direction and/or axis of stimulus movement.

4. Neurons typically responded to bar movement only over a restricted range of velocities. The majority of neurons responded best to a particular velocity within the 5-6O”/s range, with marked attenuation of the response for velocities greater or less than the preferred. Some neurons failed to show significant response attenuation even at the lowest tested velocity, while other neurons preferred velocities of 1 OO”/s or more and failed to attenuate to the highest velocities.

5. Response magnitude varied with stimulus dimensions. Increasing the length of the moving bar typically increased the magnitude

of the response slightly until the stimulus exceeded the receptive-field borders. Other neurons responded less to increases in bar length within the excitatory receptive field. Neurons preferred narrow bars less than 1” in width, and marked reductions in responses characteristically occurred with wider stimuli.

6. Moving patterns of randomly placed small dots were often as effective as or more effective than single bars in activating neurons. Selectivity for direction of movement re- mained for the dot pattern.

7. Poststimulus time (PST) histograms of responses to bars flashed at a series of 2 1 different positions across the receptive field, in the “response-plane” format, indicated a spatially and temporally homogeneous receptive- field structure for nearly all neurons. Cells characteristically showed transient excitation at both stimulus onset and offset for all effective stimulus locations. Some cells responded mainly at bright stimulus onset or offset.

INTRODUCTION

The middle temporal visual area, MT, is a representation of the contralateral visual hemifield that was first discovered by electrophysiological mapping methods in the temporal lobe of New World owl monkeys (3) and subsequently revealed in prosimian primates (6) and Old World monkeys (14, 36). MT is also distinguished by dense myelination in appropriately stained brain sections and by direct projections from striate cortex (see Ref. 38). With the possible exception of the second visual area, V-II, MT is the only extrastriate

488 0022-3077/84 $1 SO Copyright 0 1984 The American Physiological Society

PROPERTIES OF NEURONS IN AREA MT 489

area that has been described in terms of architecture, retinotopic organization, connections, and precise boundaries in a range of primate species. Because MT is a widely accepted subdivision of visual cortex and because it can be identified easily and clearly by a number of criteria, MT is an exceedingly appropriate area for electrophysiological studies of the properties of single units. Determining these properties is an important step in un- derstanding the functional role of MT in the visual system.

The first reports on neurons in MT involved recordings from the striate cortex projection zone in the superior temporal sulcus of macaque monkeys at a time when the homology with MT of New World monkeys was not yet recognized ( 10, 4 1). The most important observation from these pioneering investigations was that MT neurons respond best to stimuli moving in particular directions. Subsequent studies have consistently demonstrated that a high percentage of MT neurons are selective for the direction of moving stimuli (1, 2, 7, 22,23,28, 36,42). Other features of MT neurons have also been described in the studies, and it is generally concluded that MT neurons respond very poorly to stationary stimuli, are nonselective for color, have stimulus velocity preferences, and are typically binocular. From the first studies, stimulus shape has been described as unimportant, but the evidence is somewhat inconsistent (see DISCUSSION). In the present study, we confirm previous observations that the axis and direction of moving stimuli are critical variables for MT neurons and find that the overall directional tuning of these neurons in owl monkeys is remarkably similar to that reported for macaque monkeys (2). As others have reported (7,22,4 l), we find MT neurons to be selective for stimulus velocity and also that neurons vary in preferred velocity over a wide range. In contrast to most previous reports, we find MT neurons to be extremely selective for stimulus width, with all neurons tested preferring narrow stimuli. Finally, receptive-field structure has been studied quantitatively for the first time, and the results indicate that MT neurons have spatially homogeneous receptive fields with coextensive on- and off-discharge regions. Some of these results have been presented briefly elsewhere (11).

METHODS

Properties of neurons in the middle temporal visual area, MT, were determined using computer- controlled stimulation and data-collection procedures and microelectrode recording techniques in 10 adult owl monkeys, Aotus trivirgatus. These animals were commercially purchased, wild-caught juveniles or young adults of both sexes. Anesthesia was maintained throughout all surgical and recording procedures.

Surgical preparation Each monkey was initially anesthetized with

urethan injected intraperitoneally at a dosage of 100 mg/ 100 g body wt. Smaller supplemental doses were given as needed to achieve and maintain surgical levels of anesthesia. Urethan was the anesthetic of choice because in previous experiments, urethan provided effective, long-lasting, and stable levels of surgical anesthesia without depressing the evoked activity of neurons in extrastriate cortex, including the MT (e.g., Refs. 3-5). Each monkey was placed in a visual stereotaxic frame and area MT was ex- posed for recording. Two different surgical procedures were used. For some monkeys, a small craniotomy was performed overlying MT and the opening was surrounded by a small stainless steel chamber. After the dura was opened and retracted, the chamber was filled with silicone fluid or 4% agar in saline. The chamber was then hydraulically sealed with liquefied dental wax. The second procedure involved making a large craniotomy and surrounding it with a well formed of dental acrylic, which was then filled with silicone fluid. This second procedure allowed for clear visualization of electrode placement and the large opening reduced brain movements and provided excellent recording stability. For both conditions, surgical wounds were locally anesthetized with 0.5% procaine HCl.

Ophthalmic preparation and physiological maintenance

In order to prepare the eyes for stimulation, the corneas were locally anesthetized with 0.5% procaine HCl and the lids retracted with suture. Cy- cloplegia and mydriasis were achieved by applying 1% cyclopentolate HCl or 1% atropine sulfate. Eye movements were prevented by either mechanical or pharmacological procedures. The eyes were either rigidly fixed to immobile eye rings using suture and a histological adhesive (Histoacryl blue) or ocular immobilization was achieved through general neu- romuscular blockade induced with 2-5 mg galla- mine triethiodide and maintained by a continuous intravenous infusion of 7.5 mg l kg-’ l h-’ of gal- lamine in 0.45% saline with 5% dextrose. Paralyzed animals were mechanically ventilated with humid- ified room air to maintain an end-tidal CO* con-

490 D. J. FELLEMAN AND J. H. KAAS

centration of 3.5-4.0%, as measured by a Beckman LB-2 medical gas analyzer. In all experiments, the animal’s body temperature was maintained at 37.5 “C by a thermostatically controlled warm-water heating pad, and the electrocardiogram (ECG) was monitored continuously. In all experiments, general anesthesia was maintained during recording. The ECG was monitored as the primary indicator of anesthetic state in the paralyzed preparations, while nonparalyzed monkeys were often checked for no- ciceptive reflexes.

Contact lenses of appropriate curvature and cor- rective power were applied to the corneas with a cushioning solution of hydroxypropyl methylcel- lulose. The refractive states of the eyes were estab- lished by slit retinoscopy and were adjusted by contact lenses and trial case lenses to bring a tangent screen at 115 cm into focus on the retina. The locations of the optic disks were projected onto the tangent screen using the method of Fernald and Chase (13). The location of the area centralis for each eye was judged to be 20’ nasal to the optic disk along the horizontal meridian (3).

Visual stimuli Stimuli were projected onto a tangent screen 115

cm in front of the animal. A tungsten-filament bulb diffusely illuminated the screen at levels up to 10 cd/m2. Bright stimuli 0.3- 1.5 log units above background were produced by projecting the image of a 35-mm transparency or of a slit formed with an adjustable rectangular aperture. The orientation and axis of movement of stimuli were determined by the angle of rotation of a dove prism controlled by a stepping motor. Stimuli were statically positioned or smoothly moved by reflecting the image off the front surface of a mirror rotated with a galvanom- eter. The duration of flashed stimuli were controlled by an electromechanical shutter with a 10-90s transition time of 5 ms. Stimulus movement, position, rotation, and duration were under computer control (PDP 8/A). Whenever possible the multi- histogram technique first introduced by Bishop et al. (8) was employed so as to minimize the con- sequences of slow variations in cortical excitability. Thus, individual stimulus parameters were pseudorandomly varied and repeated on several occa- sions in order to produce the best possible estimate of each cell’s response to those stimulus configu- rations.

Electrodes and signal processing Single-unit recordings were made with Paralyene

C-coated tungsten microelectrodes with tip imped- ances of 1 .O-3.0 MQ (at 1,000 Hz). Extracellular potentials were conventionally filtered and ampli- fied to provide input to an audiomonitor and os- cilloscope. In addition, the signal provided input to a window discriminator that segregated unit waveforms by amplitude and time-course differ-

ences. Each isolated action potential was converted to a digital pulse that provided input to the computer.

Recording procedure and neuron classijication

The microelectrode was advanced by a hydraulic microdrive while a search stimulus swept through the visual field. The search stimulus was either a grating or a bar swept slowly back and forth while being rotated slowly under computer control or a hand-held projector image swept through visual space. Once a unit was isolated, its receptive-field location and dimensions were determined and stimulus requirements of orientation, direction of movement, velocity, and size were studied qualitatively. At this stage, the responsiveness of the neuron to stimuli presented to each eye separately was estimated and the contribution of each eye to the responsiveness was qualitatively expressed as a number from 1 to 7 according to the scheme introduced by Hubel and Wiesel (17). Thus, strictly monocular responses were assigned to class 1 for the contralateral eye and class 7 for the ipsilateral eye. Cells showing equal contributions by the two eyes were assigned to class 4, with categories to either side indicating decreasing amounts of binocular input. The eye mediating the weaker response was covered with a black opaque occluder during further tests.

The receptive-field sizes for neurons in area MT were determined during the hand-plotting procedures. Either a small spot or a narrow slit of light was moved or flashed to define the area of visual space or “minimum discharge field” (8) where the stimuli excited the cell. Neurons typically preferred stimuli moving in a particular direction or along a particular axis. Measurements of the excitatory receptive field were most easily and accurately determined along the axis of the preferred stimulus movement, which was used to determine the receptive-field width. The receptive-field length was defined as the distance along the orthogonal axis over which the moving stimulus continued to yield a response. Widths and lengths were arbitrarily assigned for neurons that were not selective for movement axis. Receptive-field sizes were obtained by multiplying length by width. This assumption of a rectangular shape is probably not quite accurate, since detailed plots of excitatory receptive fields resulted in ovoid shapes. Another limit of the ac- curacy of the hand-plotted receptive fields is that fringe regions of reduced or marginal responsiveness are not included in the excitatory receptive field. When we compared our hand plots of receptive fields with objectively determined measures of receptive-field width using the response-plane technique (Fig. 14), the inclusion of regions of reduced responsiveness increased the widths of the receptive


fields by some 1 O-20%. Receptive-field locations were plotted in visual space relative to the optic disk and the estimated location of the area centralis (3).

The properties of directional and axis selectivity were evaluated by collecting a family of poststimulus time (PST) histograms of responses to stimulus movements for each of 24 directions (Fig. 3). The mean firing rate during stimulus movement for each direction was illustrated for each neuron in polar coordinates (Fig. 3C) and in rectilinear coordinates (Fig. 30).

In order to describe further the sample of neurons, a directional selectivity index was defined as DSI = 1 - (response to antipreferred movement/ response to preferred direction of movement). Since cells in these anesthetized preparations had little if any appreciable spontaneous activity and suppression of spontaneous activity was not measured, a DSI value of 0 indicates complete nondirectionality and 1 .O indicates complete directionality for the neurons in the sample. The present DSI differs from those used previously for MT neurons (7,22) where the suppression of spontaneous firing was included in calculations so that DSI values greater than 1 .O could occur. Thus, the present DSI values would tend to be somewhat less than previous estimates, depending on the amount of spontaneous activity and the amount of suppression of such activity. A measure of axis selectivity was obtained from the rectangular plots of response amplitude by determining the angle for which deviation of the stimulus bar from the preferred orientation reduced the cell’s response by 50%. For some neurons, further tests were made of their responsiveness and tuning to a moving 30’ “visual noise” field constructed of 0.75 O random dots at a density of 10%. This field extended well beyond excitatory receptive-field boundaries for all tested neurons. As for single oriented slits, responses were determined for movement of the visual noise field along 12 equally spaced movement axes and for the 2 directions for each axis. Responses to moving bars or random-dot fields were displayed as polar or rectangular plots of response magnitude for each direction of movement. Initially, peak firing rate, mean firing rate, and total number of spikes were all used as measures of response magnitude. Since all of these measures produced similar plots, only mean firing rate was used in subsequent eval- uations.

For many neurons, the effectiveness of bars of light of different lengths and widths was determined. For these tests, a narrow bar of a qualitatively optimal width was swept through the receptive field in the most effective direction and returned through the receptive field for 20 trials at a given bar length. By presenting other sets of trials with different bar lengths, the most effective bar length could be determined for movements both in the preferred and

antipreferred directions. Similarly, bar widths were varied for sets of trials with a bar of nearly optimal length.

The velocity-tuning characteristics of neurons were determined by quantifying responses for different velocities of bars moved smoothly over a total range of l-25O”/s. Stimulus velocities within each of two or three limited velocity ranges were pseudorandomly mixed for 1 O-50 repetitions, and up to 16 velocities were tested. A PST histogram was generated for each velocity, and the mean firing rate during the stimulus movement through the receptive field was calculated. For the fastest velocities, an integration time of 250 ms was used to avoid overestimates of response strength (see Refs. 22, 26). While velocity tuning can be described in terms of mean firing rate, peak firing rate, or total number of spikes, measures of mean firing rate are preferred, since peak firing rate measures tend to favor higher stimulus velocities and total spike estimates tend to overestimate the low-velocity responses of cells (7, 26).

For most neurons, the spatiotemporal organization of the receptive field was evaluated using flashed bars with the response-plane technique (32). Typically, a narrow slit was flashed at a series of 21 positions across the receptive field. The flash duration was typically either 500 or 1,000 ms, followed by an off period of 500-5,000 ms. The entire sequence was repeated lo-25 times and the evoked spike activity was analyzed and presented as a sequence of PST histograms, one for each stimulus position (see Fig. 15). It is also possible to simplify the information in the response plane and only display those portions of the PST histograms in which responses occur above some specified criterion firing rate. These “contour planes” provided an easy method for identifying regions of excitation and inhibition in this time-space coordinate system (Fig. 15).

Histological con$rmation of recording sites

In order to avoid sampling other visual areas, an effort was made to place electrode penetrations well within the borders of the MT. Electrode penetrations were made approximately normal to the cortical surface and were marked with small elec- trolytic lesions (10 PA for 10 s). At the completion of the experiment the animal was deeply anesthetized with ketamine HCl (25 mg/kg) or sodium thiopental (35 mg/kg) and perfused intracardially with normal saline followed by a buffered glutar- aldehyde-paraformaldehyde fixative. The brain was then removed, photographed, frozen, and sectioned in the frontal plane at 50 pm. Sections through the area of interest were stained with cresyl violet for cytoarchitecture or a hemotoxylin procedure (20) for myeloarchitecture. The boundaries of area MT

492 D. J. FELLEMAN

were reconstructed from these sections and electrode tracks were identified. All recorded neurons in the present sample were judged to be within MT.

RESULTS

The responses of 65 neurons in the middle temporal visual area (MT) of owl monkeys were studied using a variety of stimuli, including flashed and moving bars of various dimensions and moving random-dot fields. Since the complete battery of tests took several hours, not all cells were tested on all stimulus parameters. Overall, neurons in area MT were highly responsive to moving bars, usually over restricted ranges of movement axis, direction, velocity, and bar width. In the following sec- tion we describe the responses of MT neurons to different stimulus parameters. We first present examples of the responses of individual cells, showing both PST histograms and individual tuning curves, and then show the population distribution of the metrics generated from the tuning curves.

Receptive-field size and eccentricity In characterizing each neuron, the width

and length of the excitatory receptive field was determined with a hand-held projector using

>300 1

4 300 - w a a 250 - P Jr* ;“y 200-

W ma 2 (3150-

:: - _

0

AND J. H. KAAS

spots and slits of light, and the receptive-field area was calculated from length times width. Receptive-field sizes varied considerably, ranging from 16 to over 300 deg2 (Fig. 1). There was also considerable scatter in receptive-field size for any given eccentricity. How- ever, receptive-field size was significantly cor- related with eccentricity (r = 0.3 1; P < 0.002). A similar relation held for receptive-field width and eccentricity (r = 0.42; P < 0.002).

Ocular dominance The vast majority of MT neurons were

found to be binocular, with few cells showing only monocular or weak binocular influences (Fig. 2). Thus, 93% of the sampled neurons were strongly binocular, showing equal driving from each eye or only slight asymmetries of ocular dominance. There was a slight tendency for neurons to be more strongly driven by the contralateral eye. Since the two eyes were not brought into precise binocular alignment and independent simultaneous stimulation of both eyes was not employed, no systematic attempt was made to quantitatively investigate binocular facilitatory or inhibitory interactions. However, there was some indication from hand-plotting observations that one eye oc- casionally exerts a suppressive influence on

. .

w-loo 0 W a

50

5 10 15 20 25 30 35

ECCENTRICITY (DEGREES)

FIG. 1. The relation of receptive-field location to receptive-field area for neurons in MT. Each dot indicates the hand-plotted area -of a receptive field and the distance of the receptive equation: area = 0.64 (eccentricity) + 18. ,93 deg2. See text for details.

field from the center of gaze. Regression


50-

cn40- 5 y,,_

:

$20-

I"

LO- r t-

2 3

I

4 5

I

7

OCULAR DOMINANCE CLASS

FIG. 2. The distribution of ocular dominance for 63 neurons in area MT based on qualitative observations during the hand-plotting procedure. The vast majority of neurons received balanced input from both eyes (class 4), while several cells were slightly biased toward the contralateral (class 3) or ipsilateral (class 5) eye.

the excitatory responses from the other eye. It sometimes appeared that binocular stimulation of the nonaligned eyes produced less of a response than monocular stimulation of either alone. We failed to find any cells showing opposite preferred directions of movement, as described by Zeki (4 1) for macaque MT and Cynader and Regan (9) for cat parastriate “looming” detectors. Movement axis tuning directional selectivity

and

In general, moving bars or spots of light were highly effective stimuli. However, the effective axes and directions of stimulus movement for individual neurons were usually quite restricted. Most neurons responded best to movement in a particular direction and less well or not at all to movements in other directions (directionally selective neurons; cells l-9, Fig. 4). Other neurons responded well to movement in both directions along particular axes but less well or not at all along other axes (bidirectional or axis-selective neurons; cells lo- 12, Fig. 4). A few neurons responded equally or nearly equally to stimulus movement in both directions along all axes (pandirectional ‘or nondirectional neurons; cells 13-15, Fig. 4).

The distribution of directional selectivity values for 6 1 MT neurons is shown in Fig. 5. These MT neurons varied continuously from being nonselective to highly selective. The distribution of sampled neurons has a mean of 0.62 (SD, 0.30). If a DSI criteria of 0.50 (100% or more greater response in the preferred direction) is used, then 70% of the sampled neurons were directionally selective, an additional 13% were directionally asymmetric (DSI, 0.33-0.50; 50-100% greater response in the preferred direction), while 17% were nonselective or nearly so (including bidirectional neurons). If directionally selective, directionally asymmetric, and bidirectional responses are included, then 90% of MT neurons dis- criminate movement direction or axis. DSI values showed a slight positive correlation with increases in receptive-field eccentricity (r = 0.25; P < 0.03). Thus, directionally selective neurons appear to be slightly more prevalent at increasing eccentricities in MT.

The DSI fails to indicate the tuning properties of neurons and describes bidirectional neurons as relatively nonselective. Therefore, the width of the tuning curve is a useful additional measure of directional selectivity. Since the tuning curves for MT neurons are reasonably symmetrical on both sides of a peak response (Figs. 30 and 6C and D), tuning curves were expressed in terms of the standard measure of half-width at half-amplitude. This index simply indicates the value in degrees of the change in movement axis from the optimum that will reduce the response magnitude by half. For a bidirectional neuron, the tuning curve was calculated for the greatest peak, but values for either peak were typically about the same.

The tuning widths of MT neurons varied considerably. In a sample of 5 1 neurons (Fig. 6E), half-width at half-amplitude values ranged from just over 1 O”, indicating a very rapid drop in responsiveness with change in stimulus direction, to values over 60° for nonselective and weakly tuned neurons. Forty- two neurons (82%) were strongly selective for stimulus direction or axis so that deviations of 60° or less from the preferred direction reduced the response to half the maximum. This subgroup of selective neurons had a mean half-width of 34.6’ and values ranged from 12 to 56’. Nine neurons in the sample were


1 I

B

108 s/s

A-

C

-

D WlOO-, cn - 2 - 0 - IL - co 50 w - A

AXIS / DIRECTION

FIG. 3. Experimental technique employed to determine the movement axis selectivity and directional selectivity of neurons in area MT. A: hand-plotted receptive field with 3 of the 12 tested movement axes indicated by arrows. The stimulus movement-evoked activity along each axis is compiled as a PST histogram. Each histogram indicates responses for both directions of movement with the point of reversal indicated by the reversal in slope of the stimulus marker below each histogram. B: family of PST histograms generated from slit movement along 12 axes ( 15” increments) and two directions at each axis. C polar plot of mean firing rate during stimulus movement as a function of movement axis and direction. D: rectilinear plot of same data, with normalized responses.

nonselective or- poorly selective for direction When these nonselective neurons are included of movement and were therefore very broadly in the distribution, the mean half-width tuning tuned with half-widths of more than 60°. becomes 42 O.

PROPERTIES OF NEURONS IN AREA MT

DSI 0.98 DSI 0.90 DSI 0.98

1 3 2O 2 3 2” 3 26’

DSI 0.99 DSI 0.95 DSI 0.98

24’ 5 36O 6 34O

DSI 0.75

34O

DSI 0.17

10 2o”

DSI 0.11

13 U

DSI 0.78 DSI 0.88

DSI 0.03 DSI 0.20

U 15 U

9 28O I

DSI 0.45

12 16’

FIG. 4. Types of movement axis tuning and direction selectivity for neurons in MT. Polar plots of mean firing rate as a function of stimulus direction are shown along with a directional selectivity index (DSI, see text) and an axis-tuning value (half-width at half-amplitude). Neurons l-9 are typical MT neurons with obvious directional selectivity and narrow movement axis tuning. Neurons lo- 12 are bidirectional and axis selective. Neurons 13- 15 are not strongly selective for direction or axis. U, undefined axis tuning. All polar plots are drawn relative to the maximum mean firing rate for each neuron. Mean firing rates for movement along the best direction in spikes per second. 1, 28.4; 2, 22.8; 3, 22.7; 4, 17.8; 5, 19.1; 6, 17.2; 7, 21.4; 8, 16.7; 9, 28.7; 10, 27.6; 11, 14.7; 12, 31.5; 13, 20.8; 14, 22.1; and 15, 23.5.


FIG. 5.

.6 .8

DIRECTIONAL SELECTIVITY INDEX

Distribution of directional selectivity index values for a sample of 6 1 neurons. See text.

There was evidence that neurons were grouped in MT according to preferred stimulus movement direction or axis. Figure 7 indicates the polar response plots for seven neurons successively recorded in a vertical electrode penetration. Note that all seven neurons preferred bar movement close to the horizontal plane although the first neuron was nearly pandirectional and the last two neurons were bidirectional. Results from other electrode penetrations normal to the cortical surface were similar in that selective neurons shared the same preferred movement axis or direction.

Directional and axis selectivity to moving texture

dot field was an effective stimulus over a broader range of axes than the moving bar. Several cells, however, showed an increase in axis selectivity with texture as compared to bar movement. Typically, the moving random-dot field produced a longer response. This was the consequence of the difference in the two stimuli. The moving bar started and ended outside the receptive field, while some of the moving elements of the random-dot field were always within the receptive field. Thus, the texture field was an effective stimulus for the duration of any movement. The moving random-dot field also produced more spikes per second. This suggests that there is a modest summation of excitatory effects of the individual elements of the moving texture

During hand plots of receptive fields, neu- within the receptive field that is not completely rons in MT demonstrated direction and axis countered by inhibitory effects, if any, of these selectivity to moving spots as well as bars of elements within or outside the excitatory re- light. In addition, selectivity of MT neurons ceptive field (24). for nonoriented stimuli was demonstrated for When the properties of directional selectiv- a sample of 24 neurons by using a moving ity and movement axis tuning were compared texture field of random dots. Results from one for moving bars and moving texture fields, 8 cell are shown in Fig. 8, where responses to of 15 neurons had broader axis tuning with a moving bar and a moving random-dot field little change in the magnitude of the direc- are compared. For both stimuli, the response tional selectivity index for the texture field was highly dependent on direction of stimulus (Fig. 9A). However, several neurons had a movement. Thus, the array of PST histograms, narrower axis-tuning curve and a few neurons the individual histogram with the greatest re- showed enhanced directional selectivity with sponse, the polar plot, and the rectilinear plot texture as compared to single bar movement. all indicate that movements to the left were highly effective; while movements to the right

Figure 9B compares the magnitude of the response for 15 neurons for movement of bars

were not. For some cells, the moving random- and texture in the preferred direction. Since


A

C

B

100.

50.

D 100.

50-

180” 360”

AXIS/DIRECTION

J% : : : +-- i ,I : : : : . I 1803 360°

AXIS/DIRECTION

10 20 30 40 50 60 >60

MOVEMENT AXIS

HALF WIDTH AT HALF AMPLITUDE

FIG. 6. Movement-axis selectivity of MT neurons. Polar (A) and rectilinear plots (B) for a narrowly tuned and a broadly tuned (C and D) neuron are shown. Dotted lines indicate the full width of curves at half-amplitude (E). Distribution of half-widths at half-amplitude values in degrees for 61 neurons. Mean half-width of tuned neurons, 34.6”. In B and D, ordinate is percent maximum response.

the two stimuli differed greatly in regional flux . in the effectiveness of the two stimuli might and contour, both within and outside the ex- be expected. Interestingly, the most effective citatory receptive field, a systematic difference stimulus differed for individual neurons.


F

PENETRATION \

81-59-L

+157

ti - .5 mm

FIG. 7. Directional selectivities for neurons along a typical electrode penetration in area MT. This penetration started with a superficially located pandirectional cell, continued with several directionally selective neurons sharing similar preferred directions, and ended in layer IV with two movement axis-selective, bidirectional cells. Polar plot of axis tuning and directional selectivity as in Fig. 4.

Velocity selectivity ulus velocity is shown for two typical neurons The magnitude of the response of MT neu- in Fig. 10. One cell responded over a broad

rons to moving stimuli in the preferred di- range of velocities (5-2OO”/s) with the greatest rection depended on the rate of movement, mean firing rate to an intermediate velocity and the most effective rates of movement var- of approximately 2O”/s (Fig. 1OA and B). Since ied from neuron to neuron. The role of stim- the stimulus was in the receptive field for lon-


CELL +8 l-50-1-5

1°x120BAR

RANDOM DOTS

E 160

AXIS / DIRECTION

FIG. 8. Responses elicited by bar (A-D) and random dot (E-H) stimuli. Families of PST histograms were generated by movement in two directions along 12 axes (A and E). Responses at the preferred axis are in B and F. Corresponding polar and rectilinear plots of axis tuning and directional selectivity are in d and D and in G and H. Conventions as in Fig. 3. Triangles indicate the point of stimulus direction reversal.

D. J. FELLEMAN

A TEXTURE - BAR

.8

.6

i -4

1 2 0 0

z -.2

-.4

-.6

-.8

50 40 30 20 10 0 10 20 30 40 50

neg. pos.

AXIS TUNING CHANGE

HALF wIDTH (Degrees)

B TEXTURE

5 1

BAR

NiIclIn-n .5 1.0 1.5 >2.0

RESPONSE RATIO

FIG. 9. Differences in responses to bar or random-dot field movement. A: directional selectivity index (DSI) and movement axis-tuning values for moving bars were subtracted from values for-moving random-dot fields for 15 neurons. The plotted changes (dots) indicate that most of these neurons were more broadly tuned (positive change in movement axis values) to the texture field with little change in the magnitude of the directional selectivity. See text. B: comparison of mean firing rates during stimulus movement for bar and texture stimuli. Ratios greater than 1 .O indicate that the greater response was to texture.

ger times at these more effective slower velocities, the number of spikes produced by the bar was also greater. The other cell also responded over a wide range of stimulus velocities, but the most effective presented velocity was over lOO”/s and the preferred velocity may have been even higher (Fig. 1OC and D). Slower stimulus rates of 2O”/s or less produced

4ND J. H. KAAS

weaker responses of durations that did not noticeably lengthen with the increase in the time that the stimulus was in the receptive field. In addition, the response to movement in the antipreferred direction was greatly reduced or absent for slower velocities.

The two neurons characterized in Fig. 10 are examples of how cells in MT differ in selectivity to stimulus velocity. The velocity- tuning curves from these and other neurons in a sample of 40 neurons in MT are shown in Fig. 11. For convenience of description, these cells are divided into three groups. Most neurons in the sample preferred stimulus velocities of less than 6O”/s and showed marked attenuation of the mean response rate to below 50% of maximum for stimulus velocities greater or less than the preferred (26/40,65%; Fig. 11B). Each of these neurons had a preferred stimulus velocity and was relatively narrowly tuned. However, the preferred stimulus velocity varied in the group in the range of 5-6O”/s. Other neurons (8/40, 20%) responded best to velocities of loo/s or less, showed marked attenuation of the response for higher velocities, but marked attenuation at the lower velocities was either not present or not determined because low enough velocities were not tested (Fig. 11A). Of course, these neurons might have shown marked response attenuation for extremely low velocities, but for descriptive purposes we refer to them as low-pass velocity neurons in contrast to the velocity-tuned neurons in Fig. 11B. A third group of high-pass velocity neurons (6/40, 15%) responded best at the highest tested velocities of over lOO”/s and no high- velocity attenuation was detected, although attenuation must occur at extremely high velocities (Fig. 11 C). Thus, the sample revealed that MT neurons differ in velocity preference. However, the distribution of preferred velocities (Fig. 12A) for the sample of neurons was uneven, with a suggestion of three groups of neurons, one preferring low velocities, one preferring middle velocities of around 4O”/s, and a third preferring velocities higher than lOO”/s.

There was no apparent relation between preferred velocity and velocity selectivity for MT neurons. The bandwidth of the tuning curve at half-maximum amplitude was determined and related to the preferred velocity


A

B

( i\, lb

VELOCI

r 1

100 500

TV

C L 20%.

WA-

h!!f- 250 k & s’sL 200%.

‘T

::-./

1 10 100 500 VELOCITY

FIG. 10. Velocity tuning for two neurons in MT. A: poststimulus time histograms for movement of a bright bar in the preferred and antipreferred directions for six different velocities for a velocity-tuned neuron. B: tuning curve for stimulus velocity for the neuron above. The mean firing rate was determined for the response in the preferred direction for each of the velocities in A as well as for additional velocities over a wider range and plotted as a percent of the maximal response. C and D: results from a neuron preferring high velocities. Conventions as in A and B.

for each velocity-tuned neuron (Fig. 12B). Bandwidths ranged from 1.20 to 3.75 octaves (mean, 2.34 octaves). The relation of bandwidth tuning to best velocity was nonsignifi- cant (r = 0.25; P < 0.13). Thus, neurons selective for either low or moderate velocities were, overall, equally selective for their preferred velocities. Of course, neurons without complete tuning curves were not included, and it remains to be seen if comparable selectivity is observed for low-pass velocity neurons when sufficiently low velocities are tested.

In an effort to gain some insight into mechanisms subserving directional selectivity of neurons in area MT, we examined the relationship of directional selectivity to stimulus velocity. Because we have only considered a few neurons, our data are still preliminary. However, most neurons examined had the highest index of directional selectivity when stimulated at the preferred velocity, and deviations from this optimum velocity led to small decreases in the magnitude of directional selectivity. Typically the magnitude of response was reduced considerably by the change in stimulus velocity before a large

change in directional selectivity was apparent. Some cells, however, did not exhibit peak directional selectivity at the preferred stimulus velocity. The cell in Fig. 1OC is such an ex- ample. This cell preferred stimuli moving at 2OO”/s, but the peak directional selectivity was not seen until much lower velocities (20-30”/ s). Here the velocity-tuning curve for movement in the null direction (not shown) dropped off more rapidly at low velocities than in the preferred direction. This difference un- derlies the increase in directional selectivity seen at the lower velocities. In other cells, strong directional selectivity, which was observed at relatively low velocities, was reduced or lost at high stimulus velocities. In such cases, it can be suggested that very fast stimulus movement exceeds the spatiotemporal limits of an underlying inhibitory process, and thus the process in the null direction is expressed (see Ref. 27). In this case, it can be hypoth- esized that a null direction inhibitory mechanism is maximally activated by a relatively slow stimulus movement, while the main excitatory process appears to prefer significantly higher velocities.


80-

60-

40-

20-

0’ 1 I \

1 10 100 500

ii B $lOO-

.

z 80-

: 60-

5 40-

z - 20-

2 -yb

z O’

WA

a I 1 I

1 10 100 500 i+

80-

60-

40-

1 10 100 500

VELOCITY

FIG. 11. Velocity-tuning curves for MT neurons. Rep- resentative tuning curves are grouped into low-pass neurons (A), velocity-tuned neurons (B), and high-pass neurons (C). See text for details.

Stimulus dimension tuning In hand plots of receptive fields, narrow

bars of light often appeared to be more effective stimuli than broad bars or edges. To ex- amine further the relation of stimulus size to response magnitude, the preferred stimulus dimensions for moving bright bars were determined quantitatively for 20 neurons in the MT. Results from a typical cell are shown in Fig. 13. As the length of the bar increased toward the receptive-field length (6”), the response magnitude increased, especially for bar movement in the preferred as compared to the antipreferred directions (since 6O was not tested, it is uncertain if the maximal response

would have been at 6’ or if the hand plot slightly underestimated the length of the excitatory receptive field). In contrast, when bars of different widths were tested, the effectiveness of the stimulus dropped sharply with increasing widths even though all widths tested were a fraction of the total receptive-field width (4”). In fact, the most narrow width tested (0.25”) was the most effective, and a width of one- half the receptive field width produced no clear response. In addition, a clear response to the stimulus bar moved in the antipreferred direction occurred only for the most narrow width tested. Thus, this neuron was quite responsive to changes in stimulus dimensions. It preferred a stimulus length at or close to the length of the receptive field and a stimulus width & or less that of the receptive field. These stimulus dimensions were still preferred when the bar was moved in the antipreferred direction.

Examples of length and width tuning curves for MT neurons are shown in Fig. 14. A typical pattern of response to increasing stimulus length was an increase in the magnitude of the response until the receptive-field length was reached or exceeded. However, a few neurons appeared to respond better to shorter than longer stimuli. All but one of the tested neurons produced maximal responses to bar widths close to or at the minimal width tested, which was typically less than 1/]0 the width of the receptive field and usually 0.25O (it was impractical to test widths smaller than 0.125 “). Thus, almost all MT neurons appear to prefer very narrow stimuli in comparison to the width of the receptive fields. In summary, MT neurons were selective for stimulus dimension. Increases of stimulus length within the excitatory receptive field typically increased the response and increases of stimulus width within the receptive field often greatly decreased the response.

Color In our first experiments, we tested some

neurons with bars of different wavelengths matched for behaviorally determined lumi- nosity functions for owl monkeys (19). No strong color preferences were detected and we quickly discontinued these lengthy tests. Our impression that wavelength is not an important variable for MT neurons is consistent with


VELOCITY (DEG./SEC.)

ml LOW PASS

El HIGH PASS

cl TUNED

10 20 30 40 50 60 ~60

VELOCITY (DEG./SEC.)

FIG. 12. A: distribution of preferred velocity for 40 MT neurons of three types (see Fig. 11). B: velocity-tuning curve bandwidths as a function of preferred velocity for 2 1 velocity-tuned neurons.

similar conclusions of Maunsell and Van Es- the response-plane technique (32). Typically, sen (22) and Zeki (41) that MT neurons in a response plane was generated by flashing a macaque monkeys are without any obvious 1 O x 12” bright bar in the preferred orientation color selectivity. at 2 1 positions across the receptive field. This

Receptive-field structure: responses to procedure revealed temporal characteristics of

flashed stationary stimuli responses for different positions of the stimulus in the receptive field. The typical cell illus-

Receptive-field structure of neurons in MT trated in Fig. 15 had a hand-plotted receptive was evaluated in a systematic manner by using field of approximately 10” in width. This neu-

504

A

D. J. FELLEMAN AND J. H. KAAS

PST HISTOGRAMS TUNING CURVES

C LENGTH (degrees)

WIDTH (degrees)

FIG. 13. Stimulus dimension tuning for an MT neuron. Poststimulus time histograms for movements in the preferred and antipreferred directions are shown for increasing bar lengths (A) and widths (C). The triangle indicates the point of stimulus reversal. The resulting tuning curves for the preferred and antipreferred movements are shown for bar length (B) and width (D). An arrow indicates the length of the receptive field estimated from the hand plot. Receptive-field width, 4”. Velocity, 5 O/s.

ron responded to both the onset and offset of position was a decrease iti response magnitude the flash in a transient manner regardless of as the receptive-field margins were reached. stimulus position. The only effect of stimulus Thus, the receptive field of this neuron can


;.; .; .i .5. .8 .i .Q .fi 1.b I:, I:* 1:3 114 1.5 .l .2 .3 .4 .5 .6 .7

l 59-1-2 l 59-1-3 v 59-l-4 A 59-1-5 o 59-l-7

LENGTH WIDTH

proport ion field dimension

100

3

or

5:

! !

. 2

50

E

P 0

LENGTH WIDTH

100,

50

0

n 74-l-5 l 66-l-3 A 66-4-2 0 74-l-4

I I I I I I 1-71

-1 .i .i .i .5 .6 .7

FIG. 14. A: length (left) and width (right) tuning curves for preferred directions of movement for nine MT neurons. Responses are expressed as a function of percent of receptive-field length or width. Most MT neurons show increasing responses for increasing bar length within the receptive field, and most are selective for relatively narrow bar widths.

be described as spatially homogeneous. For this neuron, the magnitude of the on-response was slightly greater than the off-response.

The spatial and temporal distributions and amplitude differences of the on- and off-responses are shown in more detail for one neuron in two contour planes (Fig. 15). A contour plane consists of a two-dimensional space versus time plot of responses where each bin of the PST histogram is compared to some criterion value and a dot displayed when that criterion is met. In Fig. 15, the contour plane

on the left has the higher criterion value, and it is apparent that the on-response has a greater spatial extent at this value. The contour plane on the right more accurately reflects the spatial overlap of the on- and off-responses. Thus, by varying the levels of the contour plane, various features of the receptive-field structure are revealed.

Response planes and contour planes were derived for a total of 26 MT neurons. Most were of the type shown in Fig. 15. Receptive fields were spatially homogeneous, and the


. : : : : : : : : : : : : : : : : : : i ; x : . " " ' . " : . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . : : . . . . . . . . . . . . . . . . . . . . . . . . . .

~~~~~~ - : - - -

. " : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . " " " . ' : : : . . . . . . . . . . . . . . . . . . . : : : : : : : : : , " " " " : : .

. : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : , : : : : : : : : : : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

" " " " ' : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : : : : :

. . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . . . . . . . . . : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : : : . . . . . . . . . . . : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . " " " " : : : : : : : : , . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . : : : : : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

: : : : : : : : : : : : : : : : : : : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . : : : : : : . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : : : : . . . . . . . . . . . . . . . . . .

" " ' : : : : : : : : : : . . . . . . . . . . . . . . : : : : : : : "."""""::::::i~:~~~~~~~~~~~~~~~~~:: . . . . . . . . . . . . . . . . . : : : : : : : : : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . e : : : : : : : : : : : : : : . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : : : : : : : : : : : . . . . . c : : : : : : : : : : : : : : : : : : : : : . . . . .

. . . . . . . . . . . . . . . . . . . . . . . : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : : : : . . . . . . . . . . . . . . . . . . : : : : : : : : : : : : : : : : : : : : . . : : : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . : : : : : : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . : : : : : : : : : : : : : : : : : : : : : : : i i : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : : : : : : :

. . . . . : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : : : : : : .

. . . . . s . . . : : : : : : : : : : : : : . ..~~ t : : : : : : : : : : : : : : : : : . . . . . . . . . . . . . . . . : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : : : : : . . . . . . . . . . . . . . . . . . . . . z : : : : : : : : : . . : .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. : : : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : : : : : : : : . . . . . . . . . . : : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . : : : : : : . . . : : : : : : : . . . . . . . . . . . " " " : : : : . " " " : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : : : : : : ; : : : : : ' . . . . . . . . . : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : : : : : : : : : : : : : : : : : : : : : . . . . . . . . . . . . . . . . . . . : : .

. : : : : : : : . . . . . . . . . . . . . r : : : : : : : : : : : : : : : : : : : : : : : .

. . . . . . . . : : : : : : : : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. : : : : : : : : : : : : : : : : : : : : : : : : : : : : : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . : : : : : . . . . . . . . .

. . . . . . . . . . . . . . : : : : : : : : : : : : : : : :

\

. . . . . . . . . : : : : : : : : : : : : .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . : : . . . . . : : : : : : : : .

. : : : : : : : : : : : : : . . . .

. . . . . . . . . . . . . . . : : : : : : . . . . . . . . . . . . . . . . . .

.:~~~~~~~ ~~

t

RESPONSE PLANE CONTOUR PLANES

-- - - -- - - - - - - - - - - - - - - _-

I 1

FIG. 15. A response plane and contour plane for an MT neuron. Above left: a hand plot of the receptive field (shaded) with 21 stimulus positions (dots) for a flashed 1 O x 12” bright bar (bar indicated at positions 1 and 2 1). Above right: poststimulus time histograms for each spatial position. Key indicates the three axes of the response plane: time, space, and response (firing rate). This response plane was generated from 20 presentations at each position flashed on for 500 ms and off for 4,500 ms. See METHODS for further details. Below: response plane from above and contour planes generated at 50 and 25% of the peak firing rate.

neurons responded in a transient manner to both stimulus onset and offset. Often the on- response was more pronounced, as shown in Fig. 15, but sometimes the off-response was slightly favored. Figure 16 illustrates response and contour planes for another typical on/off cell, as in Fig. 15, as well as for three of the less commonly encountered types. Two of the unusual neuron types in Fig. 16 appear to be no more than variations of the on/off type. Cell B was almost completely on, but a weak off-response (contour plane on the right) could be detected. This cell might be commonly classified as an on-cell because the weak off- response was difficult to reveal. Similarly, cell C has a strong off-response, but a weak on-

response could be detected (contour plane on the right). Thus, this off-cell appears to be a variety of the on/off type. The third type of cell (Fig. 16D) was only identified twice so that it appears to be quite rare in the MT. The response of greatest magnitude is a transient excitatory off-response so that this cell could be classified as an off type. However, the contour planes clearly revealed a weak sustained on-response. The on-response had a much greater latency than the off-response, suggesting that it may relate to a different mechanism.

In summary, spatially overlapping transient on- and off-excitatory responses to flashed bars were typical for MT neurons. In none of the

PROPERTIES OF NEURONS IN AREA MT

50% 25%

A ON/OFF

ON

4000 msec.

OFF

D MIXED

. . . -.- - . . . . . - .-.. - -... . . - . . . - .- -.. - . . - . - . -

. w

. .

. -

. -

. -

. - s -

- - - - . - . - - - - - - - - - - - - - - - - - - -

. . . . . .

. . . .-..c . . .

. . . -. . . . . - . .

-- - -.--. - .

--.... - . . .---.. -

. --. - . . ---- . ..--

. .--.a - . - ..----. .

.----- . - . s-e-.. -

- --. -m . . . . . . -. . . .

- . ̂ -.. . -.

. .

507

t I 12O

t I 18O

t I 10”

RESPONSE PLANE CONTOUR PLANES

FIG. 16. Examples of response planes for neurons in MT. Contour planes are at 50 and 25% peak response. Conventions as in Fig. 15. See text for details.

response planes was there any indication of DISCUSSION spatial asymmetry or subunit structure that The present results indicate that most neu- was unrelated to magnitude differences in the rons in MT of owl monkeys are highly selective on- and off-responses. for axis and direction of stimulus movement.


Neurons respond best for certain stimulus velocities, with some neurons preferring low; others, middle; and others, high velocities. MT neurons appear to prefer stimulus widths that are relatively narrow in comparison to receptive-field widths. Neurons respond transiently to the onset and offset of a flashed bar of light throughout the receptive field. These conclusions are discussed and related to relevant previous findings.

Direction and axis selectivity Perhaps the most obvious characteristic of

neurons in MT is the tendency to respond best to stimuli moving in a particular direction or along a particular axis. In the original study of MT neurons (neurons in the projection field of striate cortex in the superior temporal sulcus) in macaque monkeys, Dubner and Zeki (10) reported that 5 1 of 70 neurons responded best to movement in a particular direction and little or not at all to movement greater than 30-45’ from the preferred direction. Only seven neurons responded well to stimulus movement in all directions. Results from subsequent studies have supported the conclusion that most neurons in MT are directionally selective for stimulus movement (Ta- ble 1). Thus, 90% or more of the neurons show at least some degree of axis or direction selectivity, and 70-80% of the neurons in MT are highly selective to direction of stimulus movement, responding at least twice as well in the preferred as in the antipreferred direction. An additional 5- 10% of the neurons are axis selective or bidirectional. These estimates apply almost equally to owl and macaque monkeys, suggesting that there are no major species differences in the property of direction selectivity of MT neurons. Furthermore, the estimates are derived from data gathered under different conditions of stimulation and anesthesia, so that it appears that directional selectivity is a robust property of MT neurons that is not easily altered by different recording conditions.

While the majority of neurons in MT can be classified as highly selective for direction of stimulus movement, it also appears that the population of neurons varies continuously from unselective to highly selective types. Thus, when individual neurons in the present study were assigned a directional index, the

TABLE 1. Directionally selective neurons in MT

Reference No. Conclusions

Owl monkey Present study

7

42

70% highly selective (preferred 2 X opposite)

13% asymmetric 7% bidirectional 10% pandirectional 84%’ highly selective (preferred

2 X opposite) 76% highly selective (subjective) 11% bidirectional 13% pandirectional

Macaque monkey 22, 23

1

36

41

10

8 5%” highly selective (preferred 2 X opposite)

84% selective (subjective) 8% bidirectional 8% pandirectional 79%* highly selective

(subjective) 18% asymmetric 3% pandirectional Overwhelming majority highly

selective 73%’ highly selective

(subjective) 10%” pandirectional

* Our calculations from presented data.

resulting values formed a continuous distribution skewed toward the maximum value of direction selectivity. Similar distributions of directional selectivity indices have been reported for owl monkeys (7) and macaque monkeys (2, 22), and these distributions also were continuous and skewed toward the highly selective end.

Another measure of direction selectivity is reflected in the tuning curves for MT neurons. Using a fairly conventional metric, we calculated the half-widths of direction tuning curves at the level where the response was half the maximum value and found a unimodal sample of directionally selective cells to be distributed over a range of 12-56” from a mean of 34.6” (Fig. 6). A second mode of poorly or unselective cells had half-bandwidths of over 60”, shifting the total sample mean to 42O. In macaque monkeys, Albright and Gross (2) also determined bandwidths for direction-


selectivity tuning curves for MT neurons, and the total sample included a range of 1% 130’ with a mean of 42” for half-bandwidths at half-maximum. Thus, the tuning curves for MT neurons sampled had similar distributions and identical means for owl and macaque monkeys. Likewise, for both macaque (22) and owl monkeys (7), distributions of tuning values were quite similar for another measure of direction tuning, a tuning index that in- tegrates the area under the tuning curve.

While direction tuning for MT neurons has generally been studied with moving bars, Baker et al. (7) noted that a moving array of random dots is an effective stimulus that also can be used to determine direction preferences. In the present experiments, a few neurons were tested with both moving bars and moving random-dot fields. For random dots, most neurons maintained their previous direction preference. Some of these neurons broadened their direction tuning (Fig. 8), while a few changed somewhat in direction preference. Thus, the tuning values for MT neurons may depend in part on stimulus features (see Stimulus dimension selectivity).

There is considerable evidence that neurons are distributed according to axis or direction preference in some systematic manner in MT. Present observations were from electrode penetrations orthogonal to the cortical surface, and it was noted that neurons along each penetration tended to have similar axis preferences. Typically, they also have the same direction preference, but nearby neurons may prefer opposite directions. Baker et al. (7) previously found in owl monkeys that with electrode penetrations perpendicular to the cortical surface area MT neurons tended to prefer stimuli of the same orientation, and that neurons with opposite preferred directions may lie juxtaposed. In macaque monkeys (1, 22, 36, 41), systematic changes tend to occur in the direction and axis selectivity preferences of MT neurons recorded along electrode penetrations oblique to the surface. Furthermore, abrupt 180’ reversals of direction preference for adjacent neurons have been observed (2). After investigating rates of change of axis preference for neurons along oblique electrode penetrations, Albright et al. ( 1) postulated that slablike rather than cylindrical axis of motion “columns” exist in MT. Such a suggestion is

consistent with the modular projection pattern of striate cortex to MT (2 1,25) and the band- ing of intrinsic connections in MT (34). Indeed specific axis of motion (or orientation) bands crossing MT may receive input from matched orientation-preference bands in striate cortex.

Stimulus dimension selectivity

MT neurons have been commonly described as unselective for aspects of stimulus form ( 1, 10, 22, 4 1). More specifically, MT has been called a motion area, where neurons are uninterested in such features as form. However, in the present experiments, the width and, to a lesser extent, the length of stimulus bars were important variables for MT neurons. For almost all neurons tested with moving bars of different widths (changes in stimulus dimension parallel to the axis of preferred motion), narrow bars were highly effective stimuli, while wide bars were much less effective stimuli. The preferred stimulus width was usually the narrowest tested, 0.25 O, and stimuli as wide as 1.0’ often produced little or no response. Considering that the most effective stimuli were only a fraction of the total width of the large receptive fields, MT neurons showed remarkable selectivity for stimulus width. Although our sample was small, the results were so consistent that it seems likely that most MT neurons are highly selective for stimulus width. However, this conclusion is not strongly supported by previous findings (7, 27, 36, 41, 42).

Perhaps the most relevant related findings are those made for MT neurons in owl monkeys. Baker et al. (7) studied width selectivity with moving bars of various widths for neurons in MT and presented results in terms of a dimension-selectivity index. Although the response of most of the neurons clearly decreased with changes from the optimal stimulus width, reductions of 50% or more in the response were achieved in only 13 of 65 (20%) neurons. On the other hand, Zeki (42), in his study of MT in owl monkeys, noted that neurons were “often highly exigent in their preference for a particular slit width.” As an ex- ample, he described a cell that was very responsive to a slit of 0.5” width but unresponsive to a slit 1.5 O in width. The evidence for stimulus width selectivity by MT neurons in macaque monkeys also seems


somewhat inconsistent. Recently, Maunsell and Van Essen (22) reported “no significant effects of width” of a bar moving in the preferred direction for 18 neurons. However, Van Essen et al. (36) did describe a few cells as responding best to very small spots, and Zeki (4 1) stated that for some cells the width of the stimulus slit was “critical.” Finally, Newsome et al. (27) report that neurons in MT are tuned for different spatial frequencies in the range of 0.1-4.0 cycles/deg, thus indicating some form of spatial selectivity.

While the direct evidence for stimulus width selectivity for MT neurons is somewhat inconsistent, there are reasons to suspect that such selectivity is present. One of the features of MT neurons that we did not systematically study was the characteristic of orientation selectivity. However, it was apparent that flashed bars were most effective when elongated in the direction orthogonal to the preferred direction of motion, and this observation is consistent with selectivity for stimulus orientation. In addition, Baker et al. (7) found MT neurons in owl monkeys to be highly tuned to flashed- bar orientation, with the cells actually being more sharply tuned to the orientation of stationary flashed bars than to direction of movement. Likewise, Zeki (42) concluded that almost as many neurons in MT of owl monkeys are orientation selective as direction selective (7 1 versus 76%). In the striate cortex, orientation-selective cells, both simple and complex, are extremely sensitive to stimulus width (16, 17), and given current models of how orientation selectivity is determined (e.g., Ref. 15), it would be puzzling to find orientation-selective cells that are not sensitive to stimulus width.

In macaque monkeys, Zeki (41) concluded that very few MT neurons are orientation selective and, therefore, these neurons appeared to differ considerably from MT neurons in owl monkeys. This claim by Zeki (41), however, was not supported by the recent study of Maunsell and Van Essen (22; also see Ref. 2), where objective and quantified criteria were used to classify 77% of the neurons in MT of macaque monkeys as orientation selective. These orientation-selective neurons may also be sensitive to changes in stimulus width.

Another aspect of stimulus form is length, the distance orthogonal to the preferred di-

rection of stimulus movement for the neuron. Our results suggest that MT neurons are sensitive to stimulus length. There generally was some increase in the response with increase in bar length until the length of the receptive field was reached. With further increases in bar length, some neurons showed a slight decrease in response. Similar findings have been reported previously. In owl monkeys, Petersen et al. (30) found that increasing the length of a stimulus over the optimum length generally produced slight decrements in response strength but, for some neurons, the response could be reduced to one-half or less. In addition, in a recent study of MT neurons in owl monkeys (24), moving textured back- grounds outside the excitatory receptive field of neurons greatly reduced the response to a moving bar when the bar and the textured background were moving in the same direction. Thus, these experiments revealed an an- tagonistic surround mechanism that also would be activated by increasing stimulus bar length into the surround. In macaque monkeys, Zeki (4 1) described MT neurons that became more responsive as a slit was increased in length up to the edges of the receptive field but did not change in responsiveness with greater increases in length. For other neurons, increasing the length of the stimulus beyond the excitatory receptive field abolished the response. More recently, Maunsell and Van Es- sen (22) studied the effects of changing stimulus length on 18 MT neurons in macaque monkeys and concluded that most neurons were only slightly influenced. However, two neurons showed substantial end stopping and one neuron showed considerable response en- hancement with increases in stimulus length. Thus, it seems reasonable to conclude that stimulus length may be a significant variable for many or most MT neurons in both owl monkeys and macaque monkeys.

Stimulus velocity

The rate at which the stimulus bar moves across the receptive field was an important parameter for neurons in MT. Most cells were tuned to a particular velocity in the range of 5-6O”/s. Other cells preferred velocities of less than 15”/s and responded well at even the slowest tested velocities. A few cells responded well to velocities of lOO”/s and higher and


responded poorly to lower velocities. Similarly, Baker et al. (7) found that most neurons in the MT of owl monkeys were tuned to velocities in the range of 5-6O”/s, with a distribution of preferred velocities that closely matches that determined in the present study. As in the present study, some neurons preferred a stimulus velocity of lOO”/s or more. In macaque monkeys, MT neurons are also tuned for stimulus velocity and preferred velocities are distributed over a range of two orders of magnitude (22). This is in marked contrast to the distribution of preferred velocity in macaque striate cortex where a much more restricted range of preferred velocities is represented (D. C. Van Essen, personal communication). Preliminary evidence from owl monkey striate cortex also suggests that a limited range of lower velocities is represented (12). Thus, neurons in area MT appear to have a broader range of velocity selectivity than neurons in striate cortex for both owl and macaque monkeys.

Receptive-field structure The present report quantitatively describes

the receptive-field structure of MT neurons for the first time. Almost all the neurons in the MT had the same type of receptive-field structure. The response planes indicated that MT neurons had spatially homogeneous receptive fields with coextensive on- and off- discharge regions or domains. Throughout the excitatory receptive field, neurons responded transiently for 100-200 ms to both stimulus onset and offset with latencies of about 50 ms. Possibly because of the low levels of spontaneous activity, inhibitory domains were not obvious in any of the response planes.

The type of receptive-field structure revealed for MT neurons is very similar to that of the most commonly encountered neuron in striate cortex of cats, the “C-2” complex cell (29). These cells also have two overlapping transient (about 40 ms) excitatory domains, one at light on and one at light off. Perhaps more significantly, MT neurons are very similar to the C-type of complex cell in the striate cortex of owl monkeys (12). These C-type neurons also have spatially homogeneous receptive fields with transient on- and off-responses. Given the projections from striate cortex to MT (21, 25, 35, 40), these C-type cells could provide

much of the driving activity for MT neurons. Because Y-cells in the lateral geniculate nucleus of cats (32) and owl monkeys (unpub- lished studies) also respond transiently in a relatively spatially homogeneous way to light onset and offset, we have suggested a Y-cell influence on owl monkey C-cells (12). It fol- lows that MT may be a further station in a Y-cell-dominated pathway. Such a possibility would be consistent with other properties of MT neurons, such as an apparent lack of color coding (1, 22, 41).

The response characteristics of MT neurons to flashed bars of light have been previously reported and they largely correspond to the present observations. Thus, Baker et al. (7) reported that cells in MT of owl monkeys generally responded transiently to the onset of a flashed bar and somewhat less often to the offset of the bar. In macaque monkeys, Maun- sell and Van Essen (22) found that responses of MT neurons were almost always transient. While some neurons responded either to stimulus onset or offset, most neurons responded to both.

Receptive-field size There is general agreement that the recep-

tive fields for MT neurons are larger than those for neurons in earlier stages of visual processing. This is hardly surprising, since there is a clear general tendency across species for receptive-field size for matched eccentricities to increase with each step in the processing chain. However, Gattass and Gross (14) note that receptive fields for neurons in MT are considerably larger than for neurons in V-II, which, like MT, receives its major input from V-I and therefore may be considered to be at a similar processing level. Receptive-field sizes in our sample ranged from 4 or 5” in width near the center of gaze to 10-25” in width some 25-30” from area centralis. These values are not notably different from those illustrated for multineuron recording sites for mapping studies of MT in owl monkeys (3), and Baker et al. (7) reported a mean receptive-field size for MT neurons in owl monkeys that, when related to mean receptive-field eccentricities, would indicate the sampling of neurons with only slightly larger receptive fields than in the present sample. In macaque monkeys, receptive-field sizes for multiunit recording sites ( 14)

512 D. J. FELLEMAN AND J. H. ISAAS

were only somewhat larger than for neurons at matched eccentricities in owl monkeys. The slight differences in receptive-field sizes that have been described in these studies could relate to differences in the ways receptive fields were defined, multineuron versus single-neuron recording, species differences, or sampling errors.

The larger receptive fields for MT neurons roughly correspond to the proposed relation between receptive-field size and cortical magnification (18, 33, 46). MT, as a small area about r/l0 the size of striate cortex (4), should have receptive fields about 10 times as large. In macaque monkeys, paracentral and pe- ripheral receptive fields appear to be in the range of 5- 10 times larger for MT than area 17 neurons (14), and the relative sizes of receptive fields in the striate cortex of owl monkeys (4, 12) indicate a similar relation in owl monkeys. If MT receives input from the C- type complex cell in the striate cortex (12), it may be relevant that these cells generally have larger receptive fields than other complex cells in the striate cortex. However, most of the increased size of MT neurons undoubtedly reflects the considerable convergence and di- vergence of striate cortex inputs into the MT (e.g., Refs. 21, 25, 34, 35, 40).

Ocular dominance In the present experiments, most neurons

in MT were equally activated by either eye. Cells only activated by simultaneous stimulation of the two eyes were not detected. Zeki

REFERENCES

1.

2.

3.

4.

5.

ALBRIGHT, T.D., DESIMONE, R., ANDGROSS,C.G. Columnar organization of directionally selective cells in visual area MT of the macaque. J. Neurophysiol. 51: 16-31, 1984. ALBRIGHT, T. D. AND GROSS, C. G. Direction and orientation selectivity of neurons in area MT of ma- caques. Sot. Neurosci. Abstr. 8: 8 11, 1982. ALLMAN, J. M. AND KAAS, J. H. A representation of the visual field in the caudal third of the middle temporal gyrus of the owl monkey (Aotus trivirgatus). Brain Res. 3 1: 84-105, 197 1. ALLMAN, J. M. AND KAAS, J. H. Representation of the visual field in striate and adjoining cortex of the owl monkey (Aotus trivirgatus). Brain Res. 35: 89- 106, 1971. ALLMAN, J. M. AND BAAS, J. H. The organization of the secondvisual area (V-II) in the owl monkey: a second order transformation of the visual hemifield. Brain Res. 76: 246-265, 1974.

(41) also found that the majority of neurons in MT of owl monkeys were equally driven by either eye and that most cells were neither facilitated nor inhibited by simultaneous binocular stimulation. Yet, some 20% of the cells could be activated only by simultaneous binocular stimulation, and other cells showed some lesser degree of binocular interactions. However, in macaque monkeys, Maunsell and Van Essen (23) reported that most neurons in MT receive approximately balanced inputs from the two eyes and could be driven by either eye alone. The distribution of neurons according to degree of contralateral or ipsilateral eye dominance, using the ocular dominance histogram scheme of Hubel and Wiesel (17) was very similar for macaque (23) and owl monkeys (present study).

ACKNOWLEDGMENTS

We thank J. P. Jones and D. Bloom for their excellent computer programming. We also thank J. P. Jones for participating in some of the earlier experiments, and J. Allman, D. Van Essen, C. Cusick, M. McAvoy, M. Sesma, and R. Weller for comments on the manuscript.

This study was supported by National Institute of Mental , Health Fellowship IF3 1 MH08 179 to D. J. Felleman and National Institutes of Health Grant EY02686 to J. H. Kaas.

Present address of D. J. Fellernan: Division of Biology 2 16-76, California Institute of Technology, Pasadena, CA 91125.

Received 10 June 1983; 1984.

accepted in final form 3 April

6. ALLMAN, J. M., KAAS, J. H., AND LANE, R. H.The middle temporal visual area (MT) in the bushbaby, Galago senegalensis. Brain Res. 57: 197-202, 1973.

7. BAKER, J. F., PETERSEN, S. E., NEWSOME, W. T., AND ALLMAN, J. M. Visual response properties of neurons in four extrastriate visual areas of the owl monkey (Aotus trivirgatus): a quantitative comparison of medial, dorsomedial, dorsolateral, and middle temporal areas. J. Neurophysiol. 45: 397-4 16, 198 1.

8. BISHOP, P. O., COOMBS, J. S., AND HENRY, G. H. Interaction effects of visual contours on the discharge frequency of simple striate neurones. J. Physiol. Lon- don 219: 659-687, 1971.

9. CYNADER, M. AND REGAN, D. Neurones in cat parastriate cortex sensitive to the direction of motion in three dimensional space. J. Physiol. London 274: 549-569, 1978.

10. DUBNER, R. AND ZEKI, S. M. Response properties and receptive fields of cells in an anatomically defined


12.

13.

14.

16.

20.

21.

22.

23.

24.

25.

26.

27.

region of the superior temporal sulcus in the monkey. Brain Res. 35: 528-532, 197 1. FELLEMAN, D. J., JONES, J. P., AND KAAS, J. H. Some response properties of neurons in the middle temporal visual area (MT) of owl monkeys. Assoc. Res. Vision Ophthalmol. 3: 148, 198 1. FELLEMAN, D. J. AND KAAS, J. H. Receptive field properties of neurons in the striate cortex of the owl monkey, Aotus trivirgatus (Abstract). Invest. Ophthalmol. Visual Sci. 22, Suppl. 118, 1982. FERNALD, R. AND CHASE, R. An improved method of plotting retinal landmarks and focusing the eyes. Vision Res. 11: 95-96, 197 1. GATTASS, R. AND GROSS, C. G. Visual topography of striate projection zone (MT) in posterior superior temporal sulcus of the macaque. J. Neurophysiol. 46: 621-638, 1981. HEGGELUND, P. Receptive field organization of complex cells in cat striate cortex. Exp. Brain Res. 42: 99-107, 1981. HENRY, G. H., BISHOP, P. O., AND COOMBS, J. S. Inhibitory and subliminal excitatory receptive fields of simple units in cat striate cortex. Vision Res. 9: 1289-1296, 1969. HUBEL, D. H. AND WIESEL, T. N. Receptive fields, binocular interaction, and functional architecture in the cat’s visual cortex. J. Physiol. London 160: 106- 154, 1962. HUBEL, D. H. AND WIESEL, T. N. Uniformity of monkey striate cortex: a parallel relationship between receptive field size, scatter and magnification factor. J. Comp. Neural. 158: 295-302, 1974. JACOBS, G. H. Visual capacities of the owl monkey (Aotus trivirgatus). I. Spectral sensitivity and color vision. Vision Res. 17: 8 1 l-820, 1977. LIN, C.-S. AND BAAS, J. H. Projections from cortical areas 17, 18, and MT onto the dorsal lateral geniculate nucleus in owl monkeys. J. Comp. Neurol. 173: 457- 474, 1977. LIN, C.-S., WELLER, R. E., AND KAAS, J. H. Cortical connections of striate cortex in the owl monkey. J. Comp. Neurol. 2 11: 165-176, 1982. MAUNSELL, J. H. R. AND VAN ESSEN, D. C. Single unit responses in the middle temporal area of the macaque. I. Direction, speed, and orientation. J. Neurophysiol. 49: 1127- 1147, 1983. MAUNSELL, J. H. R. AND VAN ESSEN, D. C. Single unit responses in the middle temporal area of the macaque. II. Binocular interactions and disparity selectivity. J. Neurophysiol. 49: 1148- 1167, 1983. MIEZEN, F., MCGUINNES, E., AND ALLMAN, J. An- tagonistic direction specific mechanisms in area MT in the owl monkey. Sot. Neurosci. Abstr. 8: 68 1, 1982. MONTERO, V. M. Patterns of connections from the striate cortex to cortical visual areas in superior temporal sulcus of macaque and middle temporal gyrus of owl monkey. J. Comp. Neurol. 189: 45-59, 1980. MOVSHON, J. A. The velocity tuning of single units in cats striate cortex. J. Physiol. London 249: 445- 468, 1975. NEWSOME, W. T., GIZZI, M. S., AND MOVSHON, J. A. Spatial and temporal properties of neurons in macaque MT (Abstract). Invest. Ophthalmol. Visual .

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

NEWSOME, W. T., MIKAMI, A., AND WURTZ, R. H. Direction selective responses to sequentially flashed stimuli in extrastriate area MT in the awake macaque monkey. Sot. Neurosci. Abstr. 8: 8 12, 1982. PALMER, L. A. AND DAVIS, T. L. Receptive-field structure in cat striate cortex. J. Neurophysiol. 46: 260-276, 1981. PETERSEN, S. E., BAKER, J. F., AND ALLMAN, J. M. Dimensional selectivity of neurons in the dorsolateral visual area of the owl monkey. Brain Res. 197: 507- 51 1, 1980. SCHILLER, P. H., FINLAY, B. L., AND VOLMAN, S. F. Quantitative studies of single-cell properties in monkey striate cortex. I. Spatiotemporal organization of receptive fields. J. Neurophysiol. 39: 1288-13 19, 1976. STEVENS, J. K. AND GERSTEIN, G. L. Spatiotemporal organization of cat lateral geniculate receptive fields. J. Neurophysiol. 39: 2 13-238, 1976. SUR, M., MERZENICH, M. M., AND KAAS, J. H. Mag- nification, receptive-field area, and “hypercolumn” size in areas 3b and 1 of somatosensory cortex in owl monkeys. J. Neurophysiol. 44: 295-3 11, 1980. TIGGES, J., TIGGES, M., ANSCHEL, S., CROSS, A., LEDBETTER, W. D., AND MCBRIDE, R. L. Area1 and laminar distribution of neurons interconnecting the central visual cortical areas 17, 18, 19, and MT in the squirrel monkey (Saimiri). J. Comp. Neurol. 202: 539-560, 1981. UNGERLEIDER, L. AND MISHKIN, M. The striate projection zone in the superior temporal sulcus of Macaca mulatta: location and topographic organization. J. Comp. Neural. 188: 347-366, 1979. VAN ESSEN, D. C., MAUNSELL, J. H. R., AND BIXBY, J. L. The middle temporal visual area in the macaque: myeloarchitecture, connections, functional properties and topographic organization. J. Comp. Neurol. 199: 293-326, 1981. WELLER, R. E. AND KAA~, J. H. Connections of the dorsolateral visual area (DL) of extrastriate visual cortex of the owl monkey (Aotus trivirgatus). Sot. Neurosci. Abstr. 6: 579, 1980. WELLER, R. E. AND KAAS, J. H. Cortical and sub- cortical connections of visual cortex in primates. In: Cortical Sensory Organization. Multiple Visual Areas, edited by C. N. Woolsey. Clifton, NJ: Humana, 198 1, vol. 2, p. 121-155. WELLER, R. E. AND KAAS, J. H. Connections of visual cortex rostra1 to the Middle Temporal Area (MT) in owl monkeys: evidence for a visual region in superior temporal cortex. Sot. Neurosci. Abstr. 9: 958, 1983. WELLER, R. E. AND KAAS, J. H. Retinotopic patterns of connections of area 17 with visual areas V-II and MT in macaque monkeys. J. Comp. Neurol. 220: 253-279, 1983. ZEKI, S. M. Functional organization of a visual area in the posterior bank of the superior temporal sulcus of the rhesus monkey. J. Physiol. London 236: 549- 573, 1974. ZEKI, S. M. The response properties of cells in the middle temporal area (area MT) of owl monkey visual cortex. Proc. R. Sot. London Ser. B 207: 239-248,

Sci. 24, Suppl. 107, 1983. 1980.

Documents

Receptive-Field Properties of Neurons in Middle Temporal Visual Area