Heading encoding in the macaque ventral intraparietal area (VIP)

Heading encoding in the macaque ventral intraparietalarea (VIP)

Frank Bremmer,1,2 Jean-ReneÂ Duhamel,1,3 Suliann Ben Hamed1,3 and Werner Graf11Laboratoire de Physiologie de la Perception et de l'Action, CNRS±ColleÁge de France, 11 place Marcelin Berthelot, F-75231

Paris Cedex 05, France2Department of Neurophysics, Philipps University, Renthof 7, D-35032 Marburg, Germany3Institut des Sciences Cognitives, CNRS UPR 9075, 67 bvd Pinel, F-69675 Bron, France

Keywords: area MST, area VIP, heading, monkey, posterior parietal cortex, self-motion

Abstract

We recorded neuronal responses to optic ¯ow stimuli in the ventral intraparietal area (VIP) of two awake macaque monkeys.According to previous studies on optic ¯ow responses in monkey extrastriate cortex we used different stimulus classes:

frontoparallel motion, radial stimuli (expansion and contraction) and rotational stimuli (clockwise and counter-clockwise). Seventy-

®ve percent of the cells showed statistically signi®cant responses to one or more of these optic ¯ow stimuli. Shifting the locationof the singularity of the optic ¯ow stimuli within the visual ®eld led to a response modulation in almost all cases. For the majority

of neurons, this modulatory in¯uence could be approximated in a statistically signi®cant manner by a two-dimensional linear

regression. Gradient directions, derived from the regression parameters and indicating the direction of the steepest increase inthe responses, were uniformly distributed. At the population level, an unbiased average response for the stimuli with different

focus locations was observed. By applying a population code, termed `isofrequency encoding', we demonstrate the capability of

the recorded neuronal ensemble to retrieve the focus location from its population discharge. Responses to expansion and

contraction stimuli cannot be predicted based on quantitative data on a neuron's frontoparallel preferred stimulus direction andthe location and size of its receptive ®eld. These results, taken together with data on polymodal motion responses in this area,

suggest an involvement of area VIP in the analysis and the encoding of heading.

Introduction

Motion through the environment (self-motion) requires the process-

ing of a variety of incoming sensory signals. If forward self-motion

occurs on a straight path, the visual optic ¯ow arriving at the retina is

radially symmetric and the singularity within this ¯ow ®eld indicates

the direction of heading (Gibson, 1950). In addition, tactile, auditory

and vestibular signals can be used to deduce self-motion information.

The macaque medial superior temporal area (MST) is assumed to

play a prominent role in the encoding of visual self-motion

information (for reviews see, e.g., Bremmer et al., 2000; Duffy,

2000). Most neurons in area MST are selective for the direction of

stimulus motion in the frontoparallel plane (Tanaka et al., 1986) but

also respond to stimuli mimicking either forward (expansion),

backward (contraction), or rotatory (clockwise, counterclockwise)

motion (Saito et al., 1986; Tanaka et al., 1986; Tanaka & Saito, 1989;

Tanaka et al., 1989; Duffy & Wurtz, 1991a; Duffy & Wurtz, 1991b;

Graziano et al., 1994; Lagae et al., 1994). Response strength to such

optic ¯ow stimuli is often in¯uenced by the location of the singularity

of the optic ¯ow corresponding to different directions of heading

given that the direction of gaze is constant (Duffy & Wurtz, 1995;

Lappe et al., 1996; Duffy, 1998). Further evidence for the involve-

ment of area MST in the process of self-motion encoding comes from

studies indicating responses to real physical movement (Duffy, 1998;

Bremmer et al., 1999).

Area MST, however, is not the only motion-sensitive area in the

macaque posterior parietal cortex. Based on anatomical work, the

ventral intraparietal area (VIP) was originally de®ned as the medial

temporal (MT) projection zone in the intraparietal sulcus (Maunsell

& Van Essen, 1983). Later physiological studies revealed that the

majority of VIP neurons display a strong preference for the direction

of visual stimulus motion (Duhamel et al., 1991; Colby et al., 1993)

and also respond to basic optic ¯ow stimuli (Schaafsma & Duysens,

1996; Schaafsma et al., 1997). In addition, many neurons in area VIP

respond in a directionally selective manner for tactile, auditory and

vestibular stimulation (Bremmer et al., 1995; Graf et al., 1996;

Bremmer et al., 1997a; Duhamel et al., 1998; Klam et al., 1998;

Schlack et al., 2000), strongly suggesting an involvement of area VIP

in the processing of self-motion information.

The emphasis of our present study, compared to the two previous

studies on basic optic ¯ow properties in area VIP (Schaafsma &

Duysens, 1996; Schaafsma et al., 1997), was to address the issue of

heading and the role of area VIP in self-motion perception.

Furthermore, the extent to which the responsiveness to expansion

and contraction stimuli re¯ects the sensitivity of these neurons to

frontoparallel motion needed to be investigated. Discharge of the vast

majority of neurons was strongly modulated by the simulated heading

direction, i.e. the location of the singularity of the optic ¯ow (SOF).

Tuning for this stimulus parameter mostly was broad and linearly

dependent on horizontal and vertical SOF location. We demonstrate

Correspondence: Dr Frank Bremmer, 2Department of Neurophysics as above.E-mail: [email protected]

Received 13 March 2002, revised 15 July 2002, accepted 25 July 2002

doi:10.1046/j.1460-9568.2002.02207.x

European Journal of Neuroscience, Vol. 16, pp. 1554±1568, 2002 ã Federation of European Neuroscience Societies

the capability of the recorded neurons to encode heading direction by

applying a population code termed `isofrequency encoding'. Finally,

predictability of a preference for either expansion or contraction

stimuli based on information about the location and structure of a

neuron's receptive ®eld and its frontoparallel preferred direction was

at chance level. Preliminary reports of these results have been

presented earlier (Bremmer et al., 1995; Duhamel et al., 1997a).

Materials and methods

Experiments were performed in two female macaque monkeys, one

rhesus monkey (M. mulatta, 4.6 kg) and one fascicularis monkey (M.

fascicularis, 3.8 kg). All animal care, housing and surgical proced-

ures were in accordance with national French and international

published guidelines on the use of animals in research (European

Communities Council Directive 86/609/ECC).

Animal preparation

All surgical procedures and interventions were carried out under

deep general anaesthesia (Propofol: 10 mg/kg for induction;

25 mg/kg/h for maintenance) and sterile conditions. Monkeys

were prepared for recordings by implanting a head holding device

and two scleral search coils to monitor eye movements according

to the method published by Judge et al. (1980). The leads were

connected to a plug on top of the skull. A recording chamber for

microelectrode penetrations through the intact dura was anchored

¯at to the skull centred at P 3.5, L 12 mm. This semistereotaxic

approach allowed long electrode penetrations parallel to the

intraparietal sulcus. Recording chamber, eye coil plug and head

holder were embedded in dental acrylic which itself was

connected to the skull by self-tapping screws. Analgesics and

antibiotics were applied postoperatively and recording started one

week after surgery.

Behavioural paradigm and recordings

During training and recording sessions, the animals were restrained in

a primate chair with the head ®xed during recordings, while they were

performing ®xation tasks for liquid rewards. Rewards were given for

keeping the eyes within an electronically de®ned window of 2 3 2°centred on the ®xation target. A PC running the REX software

package (NIH) controlled behavioural paradigms and data acquisi-

tion. At the end of a training or an experimental session, the monkeys

were returned to their cages. The animals' weights were monitored

daily and supplementary fruit or water were provided if necessary.

For extracellular recordings, tungsten-in-glass electrodes

(Frederick Haer, Inc.; impedance 1±2 MW at 1 kHz) were advanced

using a hydraulic microdrive (Narishige) which was mounted on the

recording chamber. Neuronal activity and electrode depth were

recorded to establish the relative position of landmarks such as grey

and white matter, and neuronal response characteristics. During

recording sessions, area VIP was identi®ed by its location within the

intraparietal sulcus and its typical physiological response character-

istics: selectivity for the direction of visual stimulus motion and often

also directionally selective tactile responses to stimulation of the face

or head area (Colby et al., 1993; Duhamel et al., 1997b). These are

different response characteristics from those found in the neighbour-

ing lateral intraparietal area (LIP) and medial intraparietal area

(MIP). Whereas one animal is still being used in experiments,

histological analysis of the ®rst animal showed that recording sites

were located in area VIP (see below).

Visual stimulation

During the experiments the animals viewed a translucent screen

subtending the central 70 3 70° of the visual ®eld. Computer-

generated visual stimuli as well as the ®xation target were back-

projected onto this screen by a liquid crystal display system. During

visual stimulation, the monkeys had to keep their eyes within the

tolerance window of a central ®xation spot at straight-ahead position

[(x, y) = (0°, 0°)] for 3500 ms to receive the liquid reward. Visual

stimuli were random dot patterns, consisting of 240 dots. Each

individual dot was 0.15° in size with a luminance of 0.5 cd/m2. A

single dot leaving the display area at its outer borders was replaced at

a new location. A speci®cally adapted replacement algorithm (Picto

2-D, EPITA, Paris) guaranteed a constant density of dots (one per

20.4°2) across the entire screen throughout the stimulation period.

The stimulus background was always dark, i.e. luminance was

< 0.01 cd/m2.

Unless mentioned otherwise, visual stimuli covered the full tangent

screen, i.e. the central 70 3 70° portion of the visual ®eld. In order to

determine a neuron's preferred direction for movement in the

frontoparallel plane, a random dot pattern was moved along a circular

pathway (continuous mapping of directional selectivity; see also, e.g.,

Bremmer et al., 1997b). In this paradigm the speed of the stimulus is

constant throughout a stimulus trial (cycle), but stimulus direction

changes continuously (0±360°) within a complete stimulus cycle.

Thus, each pattern element moves with the same speed (typically 27 or

40°/s) around its own centre of motion (the radius being typically 5 or

10°). This kind of stimulation is different from traditional mapping

procedures to determine a neuron's preferred direction (PD), where

responses to several (typically eight) uni-directional pattern move-

ments are compared. The continuous mapping of directional select-

ivity has the experimental advantage that all possible stimulus

directions within the frontoparallel plane are presented during a single

trial without the need to test a critical number of unidirectional pattern

movements. Experiments in cat visual cortex, monkey pretectal nuclei

and monkey area MST have shown that directional tunings obtained

using the continuous mapping procedure are equivalent to the tunings

obtained by unidirectional pattern movements (Schoppmann &

Hoffmann, 1976; Hoffmann & Distler, 1989; Bremmer et al.,

1997b). In order to verify the equality of the two experimental

paradigms for area VIP, some neurons were tested with both

approaches and the results for determining the neuron's preferred

directions were compared off-line.

Usually, neurons were also tested with radial and rotational optic

¯ow pattern, with the singularity of this optic ¯ow (SOF) at the screen

centre. In the case of radial optic ¯ow pattern, average speed was 40°/

s; in the case of rotational pattern the rotation rate was 90°/s. If a

neuron responded to one stimulus or the other, optic ¯ow stimuli with

the singularity located at one of nine possible locations

[(x, y) = (0°, 0°), (0°, 625°), (625°, 0°), (617.67°, 617.67°)]

were presented. We applied this paradigm with radial (expansion

and contraction) stimuli in both monkeys. In such cases, expansion

and contraction stimuli were presented interleaved in pseudo-

randomized order.

A neuron's visual receptive ®eld (RF) was initially located manually

by means of a hand-held projector and optimal stimulus parameters

were identi®ed. For a subpopulation of cells, the RF was subsequently

measured quantitatively while the monkey ®xated on the screen centre.

The visual stimulation area covered the central 70 3 70° portion of the

visual ®eld divided into a virtual square grid of 49 nonoverlapping

patches, each one subtending 10 3 10°. A single stimulus consisted of

a 10 3 1° white bar appearing at one edge of a given patch, moving in

Heading encoding in area VIP 1555

ã 2002 Federation of European Neuroscience Societies, European Journal of Neuroscience, 16, 1554±1568

the optimal direction perpendicular to the orientation of the bar at a

constant velocity of 100°/s for 100 ms and then disappearing, thus

covering exactly 10° during a single sweep. The stimulus reappeared at

a different location 300 ms later, moved for 100 ms, disappeared, and

so on. Six to eight stimuli were presented in this rapid sequence at

randomly selected locations in the course of a single successful

®xation trial. We recorded a complete stimulation grid with 6±10

stimulus presentations per grid position. RF maps were constructed

off-line by counting the total number of spikes evoked by stimulating a

given grid patch using a shifted temporal window adjusted to the cell's

response latency. For further details concerning the mapping proced-

ure see Duhamel et al. (1997b).

Data analysis and histology

Preferred directions of visual stimulus motion were determined

utilizing the weighted average method. The circular pathway stimulus

changes stimulus direction continuously, i.e. the neuronal response

observed at time t = x ms is related to a stimulus direction presented

at time t = (x ± latency) ms. Thus, the computation of a cell's

preferred direction included latency correction. Average ®ring rates

of a cell's response to a circular pathway stimulus usually were

determined over a 500-ms period centred in the temporal domain on a

point corresponding to the cell's PD. Average ®ring rates of a cell's

response to expansion, contraction or rotational stimuli were

computed from the full response period; for some cells this included

phasic responses to motion onset.

Differences in activity were tested for statistical signi®cance with a

Mann±Whitney rank test (expansion vs. contraction responses) or a

distribution-free ANOVA [circular pathway, expansion, contraction,

clockwise (cw) and counterclockwise (ccw); expansion or contraction

responses for stimuli with nine SOFs]. Two-dimensional linear

regression analysis was applied to quantify the in¯uence of the focus

location on the neuronal response. An F-statistic was employed to

validate the planar model as a ®t of the observed data.

After recording sessions were terminated in the ®rst monkey (M.

mulatta), microlesions (50 mA for 15 s) were made at speci®c

locations at the fundus and in the lateral bank of each intraparietal

sulcus of both hemispheres. After a 24-hour survival period,

anaesthesia was induced by ketamine injection (10 mg/kg, i.m.) and

then the animal was administered a lethal dose (60 mg/kg, i.p.) of

sodium pentobarbital. After respiratory block and cessation of all

re¯exes, the animal was transcardially perfused (0.9% NaCl in 0.1 M

phosphate buffer, 5000 U of heparin at 36 °C). This was followed by

®xative (3 L of 4% paraformaldehyde in 0.1 M phosphate buffer and

10% sucrose in 2 L of 4% paraformaldehyde in 0.1 M phosphate

buffer, pH 7.4, at 4 °C). By use of a stereotaxic apparatus, marking

pins were inserted through the periphery of the recording grid with a

microdrive to document the extent of the chamber as well as the

orientation of the recording grid. Subsequently, after post®xation, the

brain was extracted and a block of tissue containing the intraparietal

sulci and neighbouring regions were cut from both hemispheres.

These tissue blocks were immersed in 0.4 M phosphate buffer/10%

sucrose solution for 2 days. The tissue was then cut (50 mm) on a

freezing microtome. To be able to follow the entire length of a given

electrode penetration within single brain sections, the tissue was cut,

not in stereotaxic coordinates, but along planes parallel to the

marking pins. The majority of sections were counter-stained with

thionine. One section in 10 was stained for myelin using the Schmued

method (Schmued, 1990). Sections were analysed by light micro-

scopy and archived with the Micro-Bright®eld Neurolucida system.

The archived sections were used for a traditional display of the

recording sites (Fig. 1), but they could also be reassembled for a

three-dimensional representation of the areas and recording sites

around the intraparietal sulcus.

Results

A total of 181 cells were recorded from two left hemispheres of two

monkeys: 94 cells from the ®rst monkey (Macaca mulatta: monkey

V) and 87 cells from the second monkey (Macaca fascicularis:

monkey R); (see also Table 1).

Recording sites

At the beginning of the experimental sessions, electrode penetrations

were made into the general area of the grey matter about the

intraparietal sulcus with the goal of determining the extent of areas

LIP and VIP, and to map the intraparietal sulcus (Fig. 1). While area

FIG. 1. Location and reconstruction of recording sites. (A) Overall lateralview of left hemisphere indicating the topographical relationship of corticallandmarks, i.e. the superior temporal sulcus (st), and the unfolded lunar (lu)and intraparietal (ip) sulci (frame). (B) Higher magni®cation display ofunfolded lunar and intraparietal sulci of the framed area in A, indicatingantero-posterior extent of the lateral intraparietal (LIP) area and the ventralintraparietal (VIP) area. LIP is shown in dark grey, VIP in medium grey, asdetermined by functional and histological properties (for details see maintext). Vertical lines indicate the anterior and posterior borders of a virtualcylinder as projecting from the recording chamber, inside of which therecordings were made. sl1, sl2 and sl3, respectively, identify three coronalsections and their position in the intraparietal sulcus. Area VIP wasidenti®ed in two slices (sl1 and sl2, grey). The trajectory and endpoint of atypical electrode track has been reconstructed and displayed in sl2. Notethat, in order to reach VIP, electrode penetrations in most cases had to crossthe intraparietal sulcus and area LIP. This topographical relationship servedas an identi®cation criterion for area VIP. Typically, there was a noticeabledifference in neuronal discharge characteristics when entering area VIPfrom area LIP.

1556 F. Bremmer et al.


LIP was systematically studied simultaneously in the same animals

for different research projects (Ben Hamed et al., 2001; Ben Hamed

et al., 2002), the medial bank of the sulcus with MIP and area 5 was

not further explored.

Area VIP is located in the fundus of the intraparietal sulcus,

extending dorsally into its anterior and posterior banks.

Histologically, the border to area LIP could be distinguished based

on differences in myelination (Blatt et al., 1990). Functionally there

was a clear difference in neuronal discharge characteristics when the

electrode passed from the saccade-related area LIP to the polysensory

area VIP (Colby et al., 1993). Based on these histological and

functional criteria, area VIP was encountered at a depth of 8±12 mm

from the dural surface (Fig. 1) and covered an area of approximately

2±4 mm within the extent of our recording chambers.

Optic ¯ow responses: stimuli mimicking frontoparallel motion

One hundred and ®fty-eight neurons were tested for their

responsiveness for optic ¯ow stimuli mimicking movement in

the frontoparallel plane. Seventy-nine percent of the cells (125/

158) responded in a statistically signi®cant manner (Mann±

Whitney rank test, P < 0.05); 64/80 (80%) from monkey V and

61/78 (78%) from monkey R. For n = 44 of these cells (n1 = 38,

FIG. 2. Tuning to frontoparallel visual stimulus motion. (A) Responses of a VIP neuron to linear stimulus motion in eight different directions. The PSTHs arepositioned at the respective stimulus directions; the PSTH on the right shows the responses of the neuron to stimulus motion to the right, etc. The centralpolar plot depicts the directional tuning of the cell (solid thick line) as obtained from vector-averaging the mean discharges of the cell during stimulation ineach direction. This cell prefers rightward and downward motion (PD 342.9°). Motion onset, as indicated by the ®rst dotted line within each responsehistogram, is accompanied by a phasic burst for the stimulus directions containing components into the preferred stimulus direction (PD). Motion end, asindicated by the second dotted line within each PSTH, is accompanied by a phasic burst for stimulus motion containing components directed into the cell'snonpreferred directions (NPD, i.e. leftward and upward for this neuron). (B) Response histogram and polar plot for the same cell during frontoparallel motionalong a circular pathway. The PSTH shows the stimulus response as plotted against stimulus direction, while the polar plot depicts the same data plotted inthe frontoparallel stimulus plane. The preferred direction was obtained by vector-averaging the response bins for a full stimulus cycle (PD 341.1°). Thenumerical values for the cell's PD obtained by the two different approaches provide essentially the same estimate of the cell's preferred direction forfrontoparallel visual stimulus motion.

TABLE 1. Numbers of neurons tested in different paradigms

ParadigmMonkey Vneurons (n)

Monkey Rneurons (n)

Totalneurons (n)

Any 94 87 181Circular pathway 80 78 158Circular pathway and translation 38 6 44Velocity tuning 16 51 67Optic Flow I (circular pathway, exp, cont) 65 25 90Optic Flow II (circular pathway, exp, cont, cw, ccw) 0 36 36Optic Flow ± Shifted SOF 23 31 54Receptive ®eld (RF) map 6 49 55RF map and Optic Flow 4 22 26



monkey V; n2 = 6, monkey R), we compared the PD determined

by the circular pathway stimulation with the PD obtained using

eight separate frontoparallel stimulus directions (`classical linear

planar motion', Fig. 2). For the classical testing the PD was

computed as the average of the eight mean responses, each

individual response weighted with its respective stimulus direc-

tion. An analogue computational process was used to determine

the PD derived from the circular planar motion. In this case,

however, the weighted average of individual bins of the response

histogram, representing different stimulus directions, was con-

sidered. In the example given in Fig. 2, both experimental

paradigms resulted in essentially the same values for the PD:

342.9° (linear planar motion) vs. 341.1° (circular pathway

stimulation). For the population of neurons tested, the average

difference in PD values resulting from the two paradigms was ±

0.6°. For area VIP, the circular pathway stimulation can thus be

considered equally as effective in determining a neuron's PD of

visual stimulus motion as the linear planar motion approach.

Preferred directions from the population of cells were distributed

continuously across the frontoparallel stimulus space (Fig. 3).

However, although recording sites were evenly distributed across

area VIP, PD values were not uniformly distributed but rather biased

towards contralateral downward motion with respect to the recorded

hemisphere (n = 125, c2 = 23.7, 3 DF, P < 0.0001). This bias of

direction selectivity was observed for each monkey individually

(Fig. 3B): monkey V, n = 64, c2 = 16.41, 3 DF, P < 0.001; monkey

R, n = 61, c2 = 8.84, 3 DF, P < 0.03.

Optic ¯ow responses: stimuli mimicking forward and backwardmotion

In addition to the large-®eld optic ¯ow stimuli mimicking movement

in the frontoparallel plane, responses to radial and rotational optic

¯ow stimuli were tested. In 90 neurons (n1 = 65, monkey V; n2 = 25,

monkey R) we recorded the responses to radial stimuli [expansion

(exp) and contraction (cont)] with the singularity located at the centre

of the screen. This mimicked a situation with coinciding gaze and

heading both directed straight ahead. Stimuli always covered the

central 70 3 70° portion of the visual ®eld and were not adapted to

the RF properties of individual neurons. Examples for two neurons

tested with expansion or contraction stimuli are shown in Fig. 4.

Neurons were classi®ed according to whether they increased their

®ring rate during expansion (expansion cell; Fig. 4A), or during

contraction (contraction cell; Fig. 4D). Optic ¯ow patterns in the

opposite direction, i.e. a contraction stimulus, often resulted in

inhibiting an expansion cell (Fig. 4B), and an expansion stimulus

resulted in inhibiting a contraction cell (Fig. 4C). In addition, many

neurons also responded with phasic bursts to the different `visual

events' (stimulus onset, motion onset etc.) during presentation of the

stimulus pro®les. There was a statistically signi®cant response by 67/

90 cells (74.4%) to at least one of the stimuli (circular pathway,

expansion and/or contraction; P < 0.05, distribution-free ANOVA). We

observed the following distribution of strongest responses: 21/67

(31.3%) for the circular pathway stimulus, 30/67 (44.8%) for the

expansion stimulus and 16/67 (23.9%) for the contraction stimulus

(Fig. 5). Considering only expansion and contraction responses,

about two thirds of the neurons gave a stronger response for

expansion (30/46 = 65%) than for contraction (16/46 = 35%) stimuli.

In n = 36 neurons (monkey R) we recorded in addition responses

to rotational, cw and ccw stimuli. Three-quarters (27/36 = 75.0%) of

these cells responded statistically signi®cantly to at least one of the

®ve different optic ¯ow patterns. The distribution of strongest

responses was as follows: 5/27 (18.5%) for the circular pathway

stimulus, 9/27 (33.3%) for the expansion stimulus, 4/27 (14.8%) for

the contraction stimulus, 2/27 (7.4%) for the cw stimulus and 7/27

(25.9%) for the ccw stimulus (Fig. 5B). The percentage of strongest

responses might be underestimated for circular pathway stimuli.

Concerning this analysis, responses were quanti®ed within a 500-ms

response window in order to get a reliable and stable estimate of the

neuronal discharge. Within this time interval, however, the stimulus

pattern had changed movement direction within the frontoparallel

plane by either 60 or 90° (corresponding to a cycle length of either

3000 or 2000 ms). As a consequence, neuronal discharge was not

only determined for the optimal stimulus direction but also for

directions within 630° or even 645° with respect to the preferred

direction. While increasing the signal-to-noise ratio of the response

estimate, this procedure in turn leads to a decrease of the estimated

®ring rate.

In¯uence of the location of the SOF on neuronalresponsiveness

In our study we were particularly interested in the role of area VIP in

the process of encoding the direction of self-motion (heading). In the

FIG. 3. Distribution of preferred directions. (A) Each individual line represents the preferred direction for visual stimulus motion in the frontoparallel plane ofa single cell. (B) Preferred directions as classi®ed in 90° bins. Light and dark grey bars indicate values from the ®rst (monkey V) and second (monkey R)animal used in this study. Statistical analysis indicates a signi®cant (P < 0.05) bias for downward and contralateral directions.



®rst stage we had tested the responses of neurons to stimuli

simulating straight self-motion either being forward (exp) or back-

ward (cont). In the next step we extended our tests to optic ¯ow

stimuli simulating heading into different directions during stable gaze

directions. To this end, we tested the in¯uence of changing the

location of the SOF of either expansion or contraction stimuli on the

neuronal discharge.

The response pattern of the majority of neurons was in¯uenced by

changing the position of the SOF within the visual ®eld (Fig. 6, same

neuron as in Fig. 4). In the illustrated case, responses increased

FIG. 5. Optic ¯ow preferences of neurons in area VIP. (A and B) Distribution of optic ¯ow preferences when cells were tested with either three (A: circularpathway, expansion, and contraction) or ®ve (B: former three plus cw and ccw rotation) optic ¯ow stimuli. Most neurons had a maximum discharge forexpansion stimuli. As indicated in more detail in the main text, the percentage of cells with peak discharge during circular pathway stimulation might beunderestimated due to temporal averaging.

FIG. 4. Responses to expansion and contraction stimuli in area VIP. Responses of two different cells (found in the same recording session) to (A and C)expansion and (B and D) contraction stimuli. The tickmarks in the spike rasters of each panel indicate (®rst) onset of the stationary stimulus, (second) onsetof stimulus motion, (third) end of stimulus motion and (fourth) end of the stimulus. (A and B) Neuron preferring the expansion stimulus over the contractionstimulus. In addition, this cell responded with a strong phasic discharge to expansion onset. (C and D) Neuron preferring contraction over expansion. Each`visual event' (stimulus onset, motion onset, etc.) caused a phasic discharge. Insets indicate the location of the focus of expansion/contraction stimulus, in thiscase in the centre of the screen.



monotonically for singularities shifted up and to the left, and

decreased for singularities shifted down and to the right. This was

true for the early, phasic component of the cell's response as well as

for its tonic component (ANOVA, P < 0.0001).

In order to quantify the modulatory in¯uence of the SOF location

on the neuronal responses we compared qualitatively different

statistical models. Previous studies have suggested the use of

sigmoidal tunings (Lappe et al., 1996) or Gaussian tunings (Perrone

& Stone, 1994) to describe similar response characteristics of neurons

from area MST. From pure visual inspection, most neurons did not

reveal a single response peak, as would be the case for a Gaussian

tuning. Moreover, the relatively small number of sample points

(n = 9) may also not be suited to test quantitatively this type of

statistical model. Sigmoidal tunings on the other hand need speci®ed

saturation values. These were obviously not available from our data.

We thus decided to employ a different statistical model, namely the

two-dimensional linear regression analysis, to quantify the modula-

tory in¯uence. For the responses of the neuron shown in Fig. 6, a

two-dimensional regression plane could be signi®cantly ®tted to the

discharges (mean discharges, Fig. 7A; multiple linear regression

analysis, P < 0.0001, Fig. 7C). The two-dimensional regression

analysis was also used to quantify the responses of the same cell to

the nine different contraction stimuli. Mean discharges and the

approximated regression plane are shown in Fig. 7B and D. The

regression plane could be approximated signi®cantly (P < 0.01) to

the contraction responses of this neuron.

Population level

A subpopulation of n = 54 cells (n1 = 23 from monkey V; n2 = 31

from monkey R) were tested for their sensitivity to the position of the

SOF within the visual ®eld. An in¯uence of the focus location on the

responses to both expansion and contraction stimuli was found for

83% of the neurons (45/54). Another 11% (6/54) revealed a

modulatory effect for one of the two optic ¯ow stimuli (either

expansion or contraction). Only 6% (3/54) of the neurons showed no

in¯uence of the focus location on their discharge. A regression plane

was approximated to the discharges of all neurons revealing a

modulatory in¯uence of the focus location in the one (exp) and/or

other (cont) stimulus condition (n = 45 3 2 + 6 3 1 = 96). This ®t

was statistically signi®cant for 74/96 stimulus conditions (77%). The

distribution of the gradients of these planes is shown in Fig. 8. The

gradient direction represents the direction of the steepest slope of the

regression plane and can be computed as a = arc tan (b/a), with a and

b the amounts of slope in horizontal and vertical directions. In the

central scatter plot, each individual dot represents the gradient of an

individual regression plane. The histograms on top of and beside each

scatter plot indicate the distribution of slopes along a single axis; the

histogram on top indicates the distribution along the horizontal axis,

the histogram beside indicates the slope distribution along the vertical

axis.

The observed uniform distribution of gradient directions does not

predict the average response of the population of neurons for the

different expansion and contraction patterns. If, as an example, all

FIG. 6. Response of a VIP neuron to expansion stimuli with shifted SOFs. The nine PSTHs depict the response of a neuron (same neuron as in Fig. 4A andB) to expansion stimuli with nine different SOF locations. The location of each PSTH represents the location of the SOF in the visual ®eld (also indicated bythe placement of the solid square inside the insets in each panel): the PSTH in the upper left shows the responses for the trials with the singularity in theupper left, etc. For this neuron, responsiveness was profoundly in¯uenced by the focus location. Responses increased for focus locations shifted upward andleftward. Responses decreased for SOF locations shifted downward and rightward.



neurons with a positive slope in the horizontal direction would have a

stronger discharge than those with a negative slope, the average

response for rightward singularities would be much stronger than the

one for leftward singularities. We thus were also interested in the

average response of the population of neurons for the different focus

locations. Indeed, at the population level, the modulatory in¯uence

observable for the individual neurons was balanced out (Fig. 9).

Average responses for the different SOF locations of the expansion

and contraction stimuli are shown in Fig. 9A (exp) and 9B (cont).

The average response planes are shown in Fig. 9C (exp) and 9D

(cont). A statistical test (distribution-free ANOVA) revealed no

signi®cant difference for the different SOFs within the group of

expansion or contraction responses. Pair-wise comparison between

expansion and contraction responses for identical SOF locations

revealed a signi®cant difference for the following four SOF locations:

Left Up, Up, Right-Up, and Centre. (Mann±Whitney rank test,

P < 0.01). A distribution of the r-values of the linear regression

indicating the goodness-of-®t of this statistical model is shown for

expansion (Fig. 9E) and contraction (Fig. 9F) responses. Statistically

signi®cant ®ts were obtained for 81% (39/48) of the expansion

responses and 73% (35/48) of the contraction responses.

Retrieval of heading direction

While a modulatory in¯uence of the focus location on the neuronal

discharges was observed at the single cell level, this in¯uence was

balanced out at the population level. Furthermore, gradient directions

were uniformly distributed. From a theoretical point of view, these

data are prerequisites for application of a population encoding of a

given parameter, in this case the focus location. We thus applied a

recently developed population code termed isofrequency encoding

FIG. 7. Quanti®cation of the in¯uence of the SOF location on neuronal responsiveness. (A and C) Responses to expansion stimuli. (A) The histogram showsthe average response (6 SD) of the cell for stimuli with the SOF at different locations in the visual ®eld: (C, centre; LU, left up; U, up; RU, right up; R,right; RD, right down; D, down; LD, left down; L, left). (C) The shaded plane represents the two-dimensional linear regression to the mean discharge(r2 = 0.964, P < 0.0001). The x-y plane in this plot represents the central 80 3 80° of the visual ®eld. The base point of each drop line depicts the SOFlocation on the screen, and the height of each line depicts the mean activity value for stimulation with the SOF at this location. (B and D) Responses of thesame neuron to contraction stimuli. (B) Responsiveness was highest for contraction stimuli with SOF locations down and to the right. (D) Approximation of alinear regression plane was statistically signi®cant (P = 0.0058).



(Boussaoud & Bremmer, 1999). For each individual cell we took its

discharge value observed for a speci®c SOF location and determined

the isofrequency line of this discharge value on the regression plane.

In a next computational step, all points of intersection between

individual isofrequency lines for a speci®c SOF location were

computed and sorted in 5°-wide bins. The estimate of the focus

location was computed as the centre of mass of the distribution of the

points of intersection of these isofrequency lines. If the tunings of the

cells had been ideally planar, i.e. if the regression planes would

perfectly ®t the neuronal responses, there would be no scatter of the

distribution of points of intersection but rather a single peak. Thus,

the width of the distribution of the points of intersection is a

qualitative measure of the goodness of the regression plane as ®t to

the neuronal responses.

We applied this computation to data from all nine SOF locations.

Numerical data for all nine SOF locations are given in Table 2;

examples for two of them (LU, left up, and RD, right down) are

shown in Fig. 10. The error for the retrieval of the upper left focus

location was 4.7°, that for the lower right position was 2.7°. The mean

mismatch between real and computed focus location was 4.6°. This

value is quite good considering the relatively small number of

neurons (n = 54).

RF location and optic ¯ow responses

Receptive ®elds were mapped quantitatively for n = 55 cells (n1 = 6

from monkey V, n2 = 49 from monkey R). Examples for three

individual cells are given in Fig. 11A±C. The majority of cells had a

strong contralateral (i.e. right) component but often also extended

into the ipsilateral visual ®eld (Fig. 11A). Some of the neurons were

purely contralateral (Fig. 11B) or ipsilateral (Fig. 11C). Most

neurons' RF form was complex and often different from the classical

Gaussian-shaped pro®le. Finally, many neurons' RF extended beyond

the mapping range, i.e. beyond the central 70 3 70° of the visual

®eld.

We de®ned the `relative receptive ®eld size' (RRFS) with respect

to the mapping range (70 3 70°) and considered it to be that part of

the visual ®eld within which the neuron's discharge was > 50% of the

peak discharge. Across the population of cells, minimum RRFS was

6%, maximum RRFS was 94% while the average RRFS was 32%,

corresponding to 0.32 3 70 3 70°2 = 1568°2 (Fig. 11D). However,

as many neurons' RF extended beyond the mapping range, this

de®nition of the RRFS probably underestimates the full RF size.

For 26 neurons we could keep stable recording conditions long

enough to measure optic ¯ow responses (circular pathway, expansion

and contractions) and map the visual receptive ®eld. Two examples

are shown in Fig. 12. In one case (Fig. 12A) the neuron's PD was

almost exactly to the right. The RF of this neuron (Fig. 12B) covered

a large part of the right, i.e. contralateral, visual hemi®eld. A `hot

zone', i.e. an area of higher excitability within the RF, was con®ned

to a smaller region in the upper right quadrant. In the second case

(Fig. 12C), the neuron preferred visual stimulus motion to the left.

The RF of this neuron was also located predominantly in the right,

contralateral, visual ®eld, revealing a hot zone in the peripheral

portion of the upper right quadrant (Fig. 12D).

The question arises whether responses to expansion and contrac-

tion stimuli can be deduced by considering a neuron's frontoparallel

PD and its RF location. Such an explanation would be based on a

linear summation of responses across the visual ®eld, dependent on

the directional components of the optic ¯ow stimulus weighted by the

receptive ®eld strength at that speci®c part of the visual ®eld. To test

for such a predictability of responses we subdivided the stimulation

area into a virtual grid of 13 3 13 patches. For each patch location,

we computed the average directional component of the local optic

¯ow vector and assigned the neuron's response to this stimulus

direction as measured by the circular pathway paradigm. This

discharge value was weighted by the relative strength of the visual RF

at that patch location. To obtain these weights we constructed RF

maps off-line as indicated in Materials and Methods section. The

resulting 7 3 7 raw matrix was subsequently converted into a

13 3 13 matrix by linear interpolation of one new point between

each original data point. Each discharge value was divided by the

maximum discharge value within this RF map to obtain normalized

FIG. 8. Distribution of the gradients of the regression planes. In the scatter plots each single dot represents the gradient of an individual linear regressionplane approximated to (A) the expansion responses and (B) the contraction responses. The directions of the gradients in both conditions are uniformlydistributed. Histograms beside and on top of the scatter plots indicate the distribution of slopes along the horizontal and the vertical axes.



discharge values ranging between 0 (zero discharge) and 1 (peak

discharge). In order to test for predictable response preferences for

either expansion or contraction stimuli, we determined the hypothet-

ical response pattern across the 13 3 13 sub®elds for the two optic

¯ow stimuli (expansion and contraction) and tested it for signi®cant

differences (Mann±Whitney rank test). In the illustrated examples,

the neuronal response could be explained by the described linear

summation technique in the ®rst case (Fig. 13A and B), but it could

not be explained in the other case (Fig. 13C and D). Both cells clearly

preferred an expansion stimulus (Fig. 13A and C) over a contraction

stimulus (Fig. 13B and D). All local motion vectors of the expansion

stimulus in the right visual hemi®eld have a rightward component, i.e.

a directional component that coincides with the ®rst cell's PD (see

Fig. 13A). In this case, linear summation predicted a signi®cantly

larger response to expansion than to contraction stimuli (P < 0.002).

The opposite is true for the responses of the second cell, for which the

expected contraction response was signi®cantly stronger than the

expansion response (P < 0.0001). However, its `real' preference for

FIG. 9. Population response for expansion and contraction stimuli with shifted SOFs. (A and B) Mean and SD of the discharge rate averaged across theneuronal samples for each focus location. Abbreviations are as in Fig. 7. (C and D) Mean regression planes for the total sample of cells. Note that for bothstimuli, the resulting regression planes are almost ¯at. (E and F) Distribution of the r-values as indication of the goodness-of-®t of the linear regression. In81% of the cases expansion responses could be approximated signi®cantly by a 2-D linear regression. The same was true for 73% of the contractionresponses. The dashed lines in each panel indicate the r-value (r = 0.795) above which ®ts are signi®cant at P < 0.05.



expansion stimuli was similar to that of the ®rst neuron (Fig. 13C and

D). Such an incompatibility between real response and predicted

response based on linear summation was observed in 14 of the 26

tested neurons (54%) while a compatibility between real and

predicted response pattern was observed in the remaining 12 cells

(46%). Taken together, predictability for expansion and contraction

responses based on RF characteristics and directional preference for

frontoparallel motion was at chance level.

Discussion

General ®ndings

The majority of neurons in area VIP responded signi®cantly to optic

¯ow stimuli (translation in the frontoparallel plane, forward or

TABLE 2. Experimentally determined and computed SOF locations

Position Horexp Horcomp Vertexp Vertcomp Error

C 0.0 ±5.1 0.0 ±7.2 8.8L ±25.0 ±21.2 0.0 ±2.7 4.7LU ±17.6 ±17.2 17.6 13.0 4.7U 0.0 ±0.4 25.0 25.5 0.6RU 17.6 18.6 17.6 18.9 1.5R 25.0 24.6 0.0 ±0.7 0.8RD 17.6 15.3 ±17.6 ±18.8 2.7D 0.0 2.0 ±25.0 ±17.5 7.8LD ±17.6 ±15.0 ±17.6 ±8.3 9.8

All values are in degrees. Hor, horizontal; Vert, vertical; exp, experimentallydetermined; comp, computed; C, centre; L, left; LU, left up; U, up; RU, rightup; R, right; RD, right down; D, down; LD, left down.

FIG. 10. Isofrequency encoding of SOF location. The four panels show the results of applying the isofrequency encoding to the data. For each individual cell,discharge values observed for a speci®c SOF location were identi®ed and the isofrequency line was determined from these discharge values. In a nextcomputational step, all points of intersection between individual isofrequency lines for a speci®c SOF location were computed and sorted in 5°-wide bins. Theestimate of the focus location was computed as the centre of mass of the distribution of the PIs of these isofrequency lines. (A and B) The result for theupper left SOF location [(x, y) = (±17.6°, 17.6°)] as (A) 2-D and (B) 3-D views. The error, i.e. the Euclidean distance between computed and real SOFlocations, for the retrieval of the focus location was 4.7°. (C and D) The result for the lower right SOF location [(x, y) = (17.6°, ±17.6°)] as (C) 2-D and (D)3-D views. The error for the retrieval of the focus location was 2.7°. Crosses in A and C indicate the experimentally given SOF locations.



backward motion, or rotation). The vast majority of neurons revealed

an in¯uence of the location of the focus of either expansion or

contraction on their responsiveness. By applying a recently intro-

duced population code, termed the isofrequency encoding, the

location of the focus of expansion could be retrieved from the

population discharge. Tuning for expansion or contraction stimuli

was often not compatible with a neuron's frontoparallel stimulus

preference and its RF properties. All results taken together strongly

suggest an involvement of area VIP in the analysis and encoding of

heading.

FIG. 12. Determining preferred stimulusdirection and receptive ®eld location. (A andC) Preferred directions of visual stimulusmotion and (B and D) mapping of thereceptive ®eld (RF) in two different cells(upper and lower row). The ®rst cell (A)preferred rightward motion and (B) had a RFlocated mainly in the right part of the visual®eld. The second cell (C) preferred leftwardmotion and (D) had its RF hot zone locatedmainly in the right part of the visual ®eld.

FIG. 11. Visual receptive ®eld maps. (A±C)Visual receptive ®elds from individual cells inarea VIP. The relative receptive ®eld size(RRFS) as compared to the total mappingrange is indicated on top of each panel. (D)The average receptive ®eld is locatedpredominantly in the contralateral visual ®eld.For details of the mapping procedure and thecomputation of the RRFS see main text.



Retrieval of heading direction

Responsiveness of most neurons was in¯uenced by the location of the

singularity of the optic ¯ow (SOF). In our experiments we varied the

SOF location within the visual ®eld rather than with respect to an

individual neuron's visual RF. This experimental approach is

different from earlier studies on optic ¯ow sensitivity in area MST

(e.g. Graziano et al., 1994; Lagae et al., 1994) but very similar to

more recent studies on optic ¯ow responses and their dependency on

the SOF location (e.g. Lappe et al., 1996; Duffy, 1998). The reason

for our experimental approach was threefold: (i) it allows us to

compare directly results from area VIP with those obtained recently

in area MST; (ii) employment of always the identical set of stimuli,

not adapted to individual cell properties such as RF location, allows

computation of a population response for optic ¯ow stimuli; (iii) for a

system's approach, considering optic ¯ow and its role in self-motion

encoding presentation of (identical) large-®eld visual stimuli appears

biologically and ecologically more plausible than smaller stimuli

adapted to individual RF properties. Our data on the in¯uences of the

SOF locations on the neuronal responses in area VIP are in good

agreement with results obtained in recent studies on optic ¯ow

responses in area MST (Duffy & Wurtz, 1995; Lappe et al., 1996). In

our study we applied a two-dimensional regression analysis to

quantify this modulatory in¯uence. The primary reason for using this

kind of statistical analysis was to apply a straightforward mathemat-

ical tool that provides a directional component (gradient direction)

and a numerical value for characterizing the strength of the effect

(slope along horizontal and vertical axes). We also needed to have a

statistical model which could be used with a fairly small number of

sample points (n = 9). Although we could have used sigmoids as a

statistical model rather than a plane (Lappe et al., 1996), this would

have required de®nition of appropriate saturation values. These

saturation values, however, could not be deduced from our set of data

points. Moreover, planar and sigmoidal tunings are very similar in the

working range of the sigmoid. We thus decided to use the plane as a

statistical model without claiming that this is the only valid one for

the data treatment.

From a relatively small sample of neurons (n = 54), we were able

to retrieve the direction of heading with an average error of 4.6°. Our

results are thus in good agreement with data on the retrieval of

heading direction from neuronal discharges in area MST (Lappe et al.,

1996). While the modulatory in¯uence could be observed at the

single-cell level, it was in equilibrium at the population level. Thus,

all heading directions can be detected equally well, i.e. there exists no

bias for any focus location in the visual ®eld. However, we observed

a signi®cantly larger number of neurons ®ring more strongly for

FIG. 13. Predictability of responses to expansion and contraction stimuli from information about a cell's PD and its RF location. Optic ¯ow responses to (Aand C) expansion and (B and D) contraction of the two cells (upper and lower row) whose PD and RF are shown in Fig. 12 (same conventions as in Fig. 4).(A and B) The preference for an expansion stimulus of the ®rst cell can be predicted by a linear summation approach from the cell's PD and RF location. (Cand D) The preference of the second cell for an expansion stimulus cannot be predicted by such a linear summation technique. For details see main text.



expansion than for contraction stimuli (c2-test, P = 0.005). This is in

accordance with results from Schaafsma & Duysens (1996) and

comparable to results from area MST (Lappe et al., 1996).

Responses to frontoparallel motion and expansion andcontraction stimuli

Studies of the motion-sensitive area MST discussed whether `real

optic ¯ow neurons' should respond purely to classical optic ¯ow

stimuli (expansion, contraction, rotation, shear) or not (see, e.g., Saito

et al., 1986; Duffy & Wurtz, 1991a; Lagae et al., 1994). It is now

generally accepted that responsiveness to one or more of these

classical optic ¯ow stimuli and to frontoparallel motion should be

considered optic ¯ow sensitivity (Graziano et al., 1994; Duffy &

Wurtz, 1995; Lappe et al., 1996; Duffy & Wurtz, 1997; Duffy, 1998;

Upadhyay et al., 2000). Thus, in our study we did not apply the

former scheme of single, double and triple component responses as

originally introduced by Duffy & Wurtz (1991a; b). Instead, we

considered frontoparallel motion as well as expansion and contraction

stimuli to be single movement directions within the continuous

movement space as indicated, e.g. in Duffy (1998) (see their ®g. 3).

However, the question remains whether sensitivity to individual

optic ¯ow pattern such as, e.g., expansion and contraction stimuli, can

be predicted from information about other response parameters. Our

data provide no evidence for such a general predictability. For a

subpopulation of cells we determined the responses to large-®eld

stimuli, and in addition we mapped the neuron's receptive ®eld with a

small stimulus moving into the neuron's preferred direction. The

mapping of the receptive ®elds, together with information about the

preferred direction to frontoparallel stimulus motion, allowed valid

predictions for expansion vs. contraction selectivity in less than half of

the cases (46%). For the remaining 54% of the cases, the response to

expansion and contraction stimuli could not be deduced from

information about the other two response parameters (frontoparallel

PD, RF location and size). This means that large-®eld stimuli generate

local interactions, within and probably also beyond the classical RF,

which cannot be mapped with small single stimuli. Instead, it is the

visual stimulus as such and its local composition in the spatial and

temporal domain that causes a cell to respond or not. Our observation

is in good agreement with results from the study of Schaafsma &

Duysens (1996) who have shown that receptive ®elds of neurons with

responses to expansion (contraction) ¯ow stimuli were not compatible

with the `direction mosaic theory' implying that local preferences for

visual stimulus motion would match the local ¯ow pattern of

expansion (contraction) stimuli. Rather, directional preferences are

constant across the visual receptive ®eld, a result very similar to area

MST and employed implicitly in our RF mapping paradigm.

Possible functional differences between areas MST and VIP

Our results as well as those from a previous study on optic ¯ow

responses in area VIP (Schaafsma & Duysens, 1996) raise the

question why two nearby areas within the dorsal stream of the

primate visual system, i.e. areas MST and VIP, are both specialized

for the processing of optic ¯ow stimuli: both area MST and area VIP

receive strong projections from area MT. As shown above and in

accordance with Schaafsma & Duysens (1996), neurons in the two

areas have similar RF properties. In addition, a recent theoretical

study (Hilgetag et al., 1996) was able to show the same hierarchical

level of cortical processing of visual information for areas MSTd and

VIP. On the other hand, profound differences make the two areas

distinct. The most prominent difference between area VIP and area

MST is the responsiveness to tactile stimuli found in area VIP

(Duhamel et al., 1991; Colby et al., 1993; Duhamel et al., 1998).

These tactile responses are particularly noteworthy because they are

often directionally selective, with comparable PDs for visual and

tactile stimulus modalities. In a parallel study, we were able to show

that many neurons in area VIP are even trimodal, i.e. they respond to

tactile, visual and vestibular stimulation (Bremmer et al., 1997a). In

these cases, on-directions for all three sensory modalities were

codirectional. This polymodality of sensory responses led us to ask

about the coordinate system in which sensory information is encoded

in area VIP. Vestibular information is per se encoded in head-centred

coordinates. Neurons with tactile RFs on the face also encode in

head-centred coordinates. The question thus arose, whether visual

information could be encoded explicitly in head-centred coordinates

as well. In order to test this hypothesis, we let the monkey ®xate on

different locations on a screen, while visual RFs were mapped

quantitatively. Indeed, some RF locations remained stable on the

screen while the monkeys changed gaze direction, i.e. some VIP

neurons encode visual spatial information in head-centred coordinates

(Duhamel et al., 1997b). Such a head-centred encoding does not exist

in area MST (Bremmer et al., 1997c).

A functional hypothesis proposes a role of area VIP in the encoding

and visual guidance of movement in near extrapersonal space (Colby

et al., 1993; Bremmer et al., 1997a). For example, for a monkey

heading for fruits, visual objects (leaves of a tree) in the animal's

surrounding can become tactile stimuli touching its skin and thereby

generating tactile ¯ow codirectional with approaching visual stimuli

(leaves nearby). Such a functional consideration is supported by

recent physiological data (Bremmer & Kubischik, 1999) but also

anatomical ®ndings showing distinct projections between parietal and

premotor cortex, and especially connections between area VIP and a

region of premotor cortex controlling head and neck movements

(Rizzolatti et al., 1998). Therefore, a speci®c function of area VIP

could be to guide movements, based on polymodal sensory input, in

order to head for objects of interest in near extrapersonal space. Such

a function would be different from area MST (Roy et al., 1992),

where near and far extrapersonal space are equally represented. Thus,

two different optic ¯ow areas could be involved in encoding

predominantly one or the other part of extrapersonal space.

Acknowledgements

This work was supported by grants from the European Union (HCM:CHRXCT930267) and the Human Frontier Science Program (RG71/96B). Wethank Drs B. Krekelberg and M. Lappe for careful reading of an earlier versionof the manuscript and for valuable comments.

Abbreviations

cont, contraction; exp, expansion; LIP, lateral intraparietal area; MIP, medialintraparietal area; MST, medial superior temporal area; MT, medial temporalarea; PD, preferred direction; RF, receptive ®eld; RRFS, relative receptive®eld size; SOF, singularity of the optic ¯ow; VIP, ventral intraparietal area.

References

Ben Hamed, S., Duhamel, J.R., Bremmer, F. & Graf, W. (2001)Representation of the visual ®eld in the lateral intraparietal area ofmacaque monkeys: a quantitative receptive ®eld analysis. Exp. Brain Res.,140, 127±144.

Ben Hamed, S., Duhamel, J.R., Bremmer, F. & Graf, W. (2002) Visualreceptive ®eld modulation in the lateral intraparietal area during attentive®xation and free gaze. Cereb. Cortex, 12, 234±245.

Blatt, G.J., Andersen, R.A. & Stoner, G.R. (1990) Visual receptive ®eldorganization and cortico-cortical connections of the lateral intra parietalarea (area LIP) in the macaque. J. Comp. Neurol., 299, 421±445.



Boussaoud, D. & Bremmer, F. (1999) Gaze effects in the cerebral cortex:reference frames for space coding and action. Exp. Brain Res., 128, 170±180.

Bremmer, F., Thiele, A. & Hoffmann, K.-P. (1997c) Encoding of spatialinformation in monkey visual cortical areas MT and MST. Soc. Neurosci.Abstr., 23, 459.

Bremmer, F., Duhamel, J.-R., Ben Hamed, S. & Graf, W. (1995) Supramodalencoding of movement space in the ventral intraparietal area of macaquemonkeys. Soc. Neurosci. Abstr., 21, 282.

Bremmer, F., Duhamel, J.-R., Ben Hamed, S. & Graf, W. (1997a). Therepresentation of movement in near extra-personal space in the macaqueventral intraparietal area (VIP). In Thier, P. & Karnath, H.-O. (eds), Parietallobe contributions to orientation in 3D space. Springer Verlag, Heidelberg,pp 619±630.

Bremmer, F., Ilg, U.J., Thiele, A., Distler, C. & Hoffmann, K.P. (1997b) Eyeposition effects in monkey cortex. I. Visual and pursuit-related activity inextrastriate areas MT and MST. J. Neurophysiol., 77, 944±961.

Bremmer, F. & Kubischik, M., (1999) Representation of near extrapersonalspace in the macaque ventral intraparietal area (VIP). Soc. Neurosci. Abstr.,25, 1164.

Bremmer, F., Kubischik, M., Pekel, M., Lappe, M. & Hoffmann, K.P. (1999)Linear vestibular self-motion signals in monkey medial superior temporalarea. Ann. NY Acad. Sci., 871, 272±281.

Bremmer, F., Duhamel, J.R., Ben Hamed, S. & Graf, W. (2000) Stages of self-motion processing in primate posterior parietal cortex. Int. Rev. Neurobiol.,44, 173±198.

Colby, C.L., Duhamel, J.-R. & Goldberg, M.E. (1993) Ventral intraparietalarea of the macaque: anatomical location and visual response properties. J.Neurophysiol., 69, 902±914.

Duffy, C.J. (1998) MST neurons respond to optic ¯ow and translationalmovement. J. Neurophysiol., 80, 1816±1827.

Duffy, C.J. (2000) Optic ¯ow analysis for self-movement perception. Int. Rev.Neurobiol., 44, 199±218.

Duffy, C.J. & Wurtz, R.H. (1991a) Sensitivity of MST neurons to optic ¯owstimuli. I: a continuum of response selectivity to large-®eld stimuli. J.Neurophysiol., 65, 1329±1345.

Duffy, C.J. & Wurtz, R.H. (1991b) Sensitivity of MST neurons to optic ¯owstimuli. II: mechanisms of response selectivity revealed by small-®eldstimuli. J. Neurophysiol., 65, 1346±1359.

Duffy, C.J. & Wurtz, R.H. (1995) Response of monkey MST neurons to optic¯ow stimuli with shifted centers of motion. J. Neurosci., 15, 5192±5208.

Duffy, C.J. & Wurtz, R.H. (1997) Planar directional contributions to optic ¯owresponses in MST neurons. J. Neurophysiol., 77, 782±796.

Duhamel, J.-R., Colby, C.L. & Goldberg, M.E. (1991) Congruentrepresentations of visual and somatosensory space in single neurons ofmonkey ventral intra-parietal cortex (area VIP). In Paillard, J. (ed.), Brainand Space. Oxford University Press, Oxford, pp. 223±236.

Duhamel, J.-R., Bremmer, F., Ben Hamed, S. & Graf, W. (1997a) Encoding ofheading direction during simulated self-motion in the macaque ventralintraparietal area (VIP). Soc. Neurosci. Abstr., 23, 1545.

Duhamel, J.R., Bremmer, F., BenHamed, S. & Graf, W. (1997b) Spatialinvariance of visual receptive ®elds in parietal cortex neurons. Nature, 389,845±848.

Duhamel, J.-R., Colby, C.L. & Goldberg, M.E. (1998) Ventral intraparietalarea of the macaque: congruent visual and somatic response properties. J.Neurophysiol., 79, 126±136.

Gibson, J.J. (1950). The Perception of the Visual World. Houghton Mif¯in,Boston.

Graf, W., Bremmer, F., Ben Hamed, S. & Duhamel, J.-R. (1996) Visual±vestibular interaction in the ventral intraparietal area of macaque monkeys.Soc. Neurosci. Abstr., 22, 1692.

Graziano, M.S., Andersen, R.A. & Snowden, R.J. (1994) Tuning of MSTneurons to spiral motions. J. Neurosci., 14, 54±67.

Hilgetag, C.C., O'Neill, M.A. & Young, M.P. (1996) Indeterminateorganization of the visual system. Science, 271, 776±777.

Hoffmann, K.-P. & Distler, C. (1989) Quantitative analysis of visualrecreptive ®elds of neurons in nucleus of the optic tract and dorsalterminal nucleus of the accessory tract in macaque monkey. J.Neurophysiol., 62, 416±428.

Judge, S.J., Richmond, B.J. & Chu, F.C. (1980) Implantation of magneticsearch coils for mearurement of eye position: an improved method. VisionRes., 20, 535±538.

Klam, F., Duhamel, J.-R., Ben Hamed, S. & Graf, W. (1998) Vestibularresponse dynamics in macaque posterior parietal cortex neurons. Soc.Neurosci. Abstr., 24, 601.1.

Lagae, L., Maes, H., Raiguel, S., Xiao, D.K. & Orban, G.A. (1994) Responsesof macaque STS neurons to optic ¯ow components: a comparison of areasMT and MST. J. Neurophysiol., 71, 1597±1626.

Lappe, M., Bremmer, F., Pekel, M., Thiele, A. & Hoffmann, K.P. (1996) Optic¯ow processing in monkey STS: a theoretical and experimental approach. J.Neurosci., 16, 6265±6285.

Maunsell, J.H.R. & Van Essen, D.C. (1983) The connections of the middletemporal visual area (MT) and their relationship to a cortical hierarchy inthe macaque monkey. J. Neurosci., 3, 2563±2580.

Perrone, J.A. & Stone, L.S. (1994) A model of self-motion estimation withinprimate extrastriate visual cortex. Vision Res., 34, 2917±2938.

Rizzolatti, G., Luppino, G. & Matelli, M. (1998) The organization of thecortical motor system: new concepts. Electroencephalogr. Clin.Neurophysiol., 106, 283±296.

Roy, J.-P., Komatsu, H. & Wurtz, R.H. (1992) Disparity sensitivity of neuronsin monkey extrastriate area MST. J. Neurosci., 12, 2478±2492.

Saito, H., Yukie, M., Tanaka, K., Hikosaka, K., Fukada, Y. & Iwai, E. (1986)Integration of direction signals of image motion in the superior temporalsulcus of the macaque monkey. J. Neurosci., 6, 145±157.

Schaafsma, S.J. & Duysens, J. (1996) Neurons in the ventral intraparietal areaof awake macaque monkey closely resemble neurons in the dorsal part ofthe medial superior temporal area in their responses to optic ¯ow patterns. J.Neurophysiol., 76, 4056±4068.

Schaafsma, S.J., Duysens, J. & Gielen, C.C. (1997) Responses in ventralintraparietal area of awake macaque monkey to optic ¯ow patternscorresponding to rotation of planes in depth can be explained bytranslation and expansion effects. Vis. Neurosci., 14, 633±646.

Schlack, A., Sterbing, S., Hartung, K., Hoffmann, K.-P. & Bremmer, F. (2000)Auditory responsiveness in the macaque ventral intraparietal area (VIP).Soc. Neurosc. Abstr., 26, 1064.

Schmued, L.C. (1990) A rapid sensitive histochemical stain for myelin infrozen brain sections. J. Histochem. Cytochem., 38, 717±720.

Schoppmann, A. & Hoffmann, K.-P. (1976) Continuous mapping of directionselectivity in the cat's visual cortex. Neurosci. Lett., 2, 177±181.

Tanaka, K., Fukada, Y. & Saito, H.A. (1989) Underlying mechanisms of theresponse speci®city of expansion/contraction and rotation cells in the dorsalpart of the medial superior temporal area of the macaque monkey. J.Neurophysiol., 62, 642±656.

Tanaka, K., Hikosaka, K., Saito, H., Yukie, M., Fukada, Y. & Iwai, E. (1986)Analysis of local and wide-®eld movements in the superior temporal visualareas of the macaque monkey. J. Neurosci., 6, 134±144.

Tanaka, K. & Saito, H. (1989) Analysis of motion of the visual ®eld bydirection, expansion/contraction, and rotation cells clustered in the dorsalpart of the medial superior temporal area of the macaque monkey. J.Neurophysiol., 62, 626±641.

Upadhyay, U.D., Page, W.K. & Duffy, C.J. (2000) MST responses to pursuitacross optic ¯ow with motion parallax. J. Neurophysiol., 84, 818±826.



Heading encoding in the macaque ventral intraparietal area (VIP)

Documents

Transcript of Heading encoding in the macaque ventral intraparietal area (VIP)