Journal of Neurophysiology

Visual and Saccade-Related Activity in Macaque Posterior Cingulate Cortex

Heather L. Dean, Justin C. Crowley, Michael L. Platt

Abstract

Previous neurophysiological studies have reported that neurons in posterior cingulate cortex (PCC) respond after eye movements, and that these responses may vary with ambient illumination. In monkeys, PCC neurons also respond after the illumination of large visual patterns but not after the illumination of small visual targets on either reflexive saccade tasks or peripheral attention tasks. These observations suggest that neuronal activity in PCC is modulated by behavioral context, which varies with the timing and spatial distribution of visual and oculomotor events. To test this hypothesis, we measured the spatial and temporal response properties of single PCC neurons in monkeys performing saccades in which target location and movement timing varied unpredictably. Specifically, an unsignaled delay between target onset and movement onset permitted us to temporally dissociate changes in PCC activity associated with either event. Response fields constructed from these data demonstrated that many PCC neurons were activated after the illumination of small contralateral visual targets, as well as after the onset of contraversive saccades guided by those targets. In addition, the PCC population maintained selectivity for small contralateral targets during delays of up to 600 ms. Overall, PCC activation was highly variable trial to trial and selective for a broad range of directions and amplitudes. Planar functions described response fields nearly as well as broadly tuned 2-dimensional Gaussian functions. Additionally, the overall responsiveness of PCC neurons decreased during delays when both a fixation stimulus and a saccade target were visible, suggesting a modulation by divided attention. Finally, the strength of the neuronal response after target onset was correlated with saccade accuracy on delayed-saccade trials. Thus PCC neurons may signal salient visual and oculomotor events, consistent with a role in visual orienting and attention.

INTRODUCTION

Neurological studies have implicated posterior cingulate cortex (PCC) in visuospatial processing. Damage to PCC is associated with spatial disorientation and navigational impairments (Cammalleri et al. 1996; Katayama et al. 1999; Maguire 2001; Takahashi et al. 1997). Supporting this neurological evidence, recent functional imaging studies in human subjects have demonstrated PCC activation after illumination of a visual stimulus (Yamasaki et al. 2002), after a shift in visual attention (Hopfinger et al. 2000, 2001; Kim et al. 1999; Mesulam et al. 2001; Small et al. 2003), during the performance of visually guided saccade and pursuit eye-movement tasks (Berman et al. 1999), and during visuospatial navigation (Flitman et al. 1997; Ghaem et al. 1997; Gron et al. 2000; Pine et al. 2002).

Neurophysiological studies conducted in animals, however, have indicated that PCC neurons respond to visual stimulation under some conditions but not others. PCC neurons respond after stimulus onset in anesthetized cats presented with black and white bars of various sizes (Kalia and Whitteridge 1973), awake cats given small spots of light as saccade targets (Musil and Olson 1993; Olson and Musil 1992), and awake monkeys viewing large textured patterns (Olson et al. 1993). In a study in which monkeys were required to rapidly shift gaze from a central fixation stimulus to one of 4 (or in some cases 16) peripheral visual targets, most PCC neurons were activated shortly after saccade onset (Olson et al. 1996). However, PCC neurons failed to respond after stimulus onset when monkeys reported the dimming of a small peripheral visual cue by releasing a bar (Olson et al. 1993).

Taken together, the results of these studies suggest that activity in PCC is modulated by the timing and spatial distribution of visual and oculomotor events. One simple way to test this hypothesis would be to systematically manipulate these contextual factors in the same experiment. In the following study, we measured the spatial and temporal response properties of single PCC neurons in monkeys performing saccades guided by targets drawn from a pool of 100 potential target locations, both with and without a random delay intervening between target onset and movement onset. This delay permitted us to analyze changes in PCC neuronal activity after target onset that were temporally segregated from those associated with movement onset.

Under these conditions, PCC neurons responded in a lateralized fashion immediately after the onset of a broad range of small visual targets, as well as after the onset of gaze shifts to those targets. Moreover, the PCC population maintained broad spatial selectivity for small contralateral visual targets during delays of up to 600 ms. The magnitude of activation after target onset predicted the accuracy of the impending saccade. Additionally, overall responsiveness in the population decreased during the delay, independent of saccade direction or amplitude. Together, these observations suggest that the responses of PCC neurons depend on the behavioral context of visual and oculomotor events, consistent with a role in visual orienting and attention.

METHODS

Animal subjects

Two adult male rhesus macaques (Macaca mulatta) were used as subjects in these experiments. All animal procedures were developed in association with Duke University Medical Center veterinarians and were approved by the Duke University Institutional Animal Care and Use Committee. These procedures were designed and conducted in compliance with the Public Health Service's Guide for the Care and Use of Animals.

Surgical procedures

A head-restraint prosthesis and scleral search coil (Fuchs and Robinson 1966; Judge et al. 1980) were implanted in an initial aseptic surgical procedure performed under isoflurane anesthesia. First, the dorsal rostrum of the skull was exposed and six 2.5-mm holes were drilled through the skull with standard orthopedic surgical instruments. These holes were then tapped for 3.5-mm fine-thread orthopedic cortical bone screws. Sterile orthopedic bone cement (Biomet; Palacos) was used to bond a stainless steel head post (Crist Instruments) lowered to just above the skull surface to 6 titanium screws (Zimmer) inserted into the tapped holes. The Teflon-insulated scleral search coil (Cooner Wire AS634) was implanted beneath the conjunctiva, passing just rostral to the insertions of the extraocular muscles (Judge et al. 1980). The wire exited the conjunctiva temporally, exited the orbit subdermally, was embedded in the bone cement that formed the restraint prosthesis, and terminated in a gold and plastic electrical connector (Winchester Electronics/Litton). After surgery, animals received analgesics for a minimum of 3 days. Antibiotic prophylaxis was initiated intraoperatively and continued for 7–10 days. Animals were given a 4- to 6-wk recovery period after surgery.

A second aseptic surgical procedure was performed once animals could reliably execute all the behavioral tasks used in the study. A stainless steel recording chamber (Crist Instruments) was positioned stereotaxically perpendicular to the horizontal plane over a 15-mm craniotomy and bonded to 4–6 additional orthopedic bone screws and the original implant with orthopedic bone cement. The recording chamber was centered stereotaxically at position 0,0, the intersection of the midsagittal and interaural planes (cf. Olson et al. 1996). Postoperatively, animals received analgesics for a minimum of 3 days and antibiotics for 7–10 days. The recording chamber was kept clean with daily antibiotic washes and sealed with replaceable sterile Cilux caps. Single-cell recording experiments began after a 1-wk postoperative period.

Behavioral techniques

Access to water was controlled during training and testing, and animals were habituated to head restraint and trained to perform oculomotor tasks for a fruit-juice reward using a custom-built software interface (Ryklin Software). Visual stimuli consisted of light-emitting diodes (LEDs; LEDtronics), which could be illuminated to appear red, green, or yellow to normal human observers. The LEDs were fixed on a tangent screen placed 144.78 cm (57 in.) from the eyes of the animal, forming a grid of points, separated by 1°, spanning 49° horizontally and 41° vertically. These LEDs could be illuminated within 1 ms and extinguished within 7 ms by the computer system controlling the experiments.

Immediate-saccade (Fig. 1A) and delayed-saccade trials (Fig. 1B) were used to assess the firing patterns and spatial tuning of physiologically identified neurons in PCC. A tone presented along with the illumination of the central yellow LED, which subjects were required to fixate within 1,000 ms, signaled the beginning of the trial. Two hundred to 800 ms after gaze was aligned within 2° of the fixation stimulus, a single eccentric yellow LED was illuminated. In immediate-saccade trials, the illumination of the target coincided with the extinction of the fixation stimulus, cuing the subject to shift gaze to the eccentric target (monkey Ni: ±2°; monkey Br: ±3°) within 350 ms to receive reinforcement. During delayed-saccade trials, the subject was required to maintain fixation a further 200–600 ms after the illumination of the target until the fixation stimulus was extinguished, which served as the cue to shift gaze to the target (monkey Ni: ±2°; monkey Br: ±3°). In both trial types, fixation of the target for 500 ms was followed by the delivery of liquid reinforcement. A brief noise burst preceding juice delivery served as a secondary reinforcer on all correct trials.

FIG. 1.

Trial types. A: immediate-saccade trials: subjects fixated (±2°) a centrally located yellow light-emitting diode (LED). After a delay, the fixation LED was extinguished and a target LED illuminated immediately. Subjects then shifted gaze to the target LED within 350 ms and maintained gaze at the target LED (±2–3°) for 500 ms to receive a juice reward. B: delayed-saccade trials: after target onset, subjects were required to maintain fixation until the central light was extinguished (200–600 ms), shift gaze to the target (2–3°), and fixate the target for 500 ms to receive a juice reward.

Microelectrode recording techniques

The stainless steel recording chamber was opened under aseptic conditions and rinsed repeatedly with sterile saline before each experimental recording session. A plastic grid (Crist) was then inserted into the chamber over the dura. A 23-gauge hypodermic tube was used to puncture the intact dura through the grid, and a tungsten steel 10- to 14-MΩ electrode (Frederick Haer) was positioned and guided into the brain through the tube. An X-Y micropositioner (Crist Instruments) and hydraulic microdrive (Kopf) were then mounted to the chamber and attached to the electrode. Both hypodermic tube and electrode were cleaned with isopropyl alcohol before use. Power-line noise and the signals from magnetic fields were excluded by filtering (passband ≅ 500–2,000 Hz) the amplified electrophysiological signals.

An electrode was lowered, under physiological guidance, until neurons with visual and/or saccade-associated activity were encountered. Electrodes first passed through tissue containing neurons with somatosensory-related activity evoked by stimulation of the hindlimb or tail. Once an electrode was in the approximate vicinity of the PCC, the monkey performed immediate-saccade and/or delayed-saccade trials in a darkened room while the electrode was advanced in 10- to 20-μm increments until the waveform of a single neuron was isolated. Individual action potentials were identified in hardware by time and amplitude criteria and isolated from extraneous signals and noise (BAK Electronics). Times of spike occurrence were recorded by computer with the use of a 1-μs internal clock.

Once a cell was isolated, 100–300 delayed-saccade and/or immediate-saccade trials were presented. Trial types were interleaved when both were presented. The fixation stimulus was located at the center of the stimulus panel, whereas the location of the eccentric target varied randomly from trial to trial within a 36 × 36° grid of 100 LEDs spaced at 4° intervals, drawn from the 41 × 49° overall grid of 2009 LEDs spaced at 1° intervals. Horizontal and vertical eye position was sampled at 500 Hz (Riverbend Instruments) and recorded by computer.

Analysis

QUANTIFICATION OF SACCADE METRICS AND NEURONAL ACTIVITY.

The onset and offset times of task-required eye movements for each trial, the horizontal and vertical amplitudes of those movements, peak velocity, movement amplitude, and residual saccade error were computed off-line for analysis. The number of action potentials was measured on each trial during ten 200-ms intervals, each synchronized to a trial event: 1) immediately after the start of fixation on the central target; 2) after the illumination of the eccentric target; 3) preceding the onset of the required eye movement; 4) immediately after onset of the movement; 5) 200 ms after the onset of the required eye movement; 6) preceding the onset of the noise burst before juice reward; 7) immediately after the noise burst; 8) immediately preceding juice reward delivery; 9) immediately after the onset of juice reward; and 10) after the offset of juice reward. For each cell, a database was constructed from these measurements. Response-field plots for each neuron were later constructed from these databases.

To quantify delay period activity in delayed-saccade trials, the number of action potentials was measured in an additional 4 overlapping 200-ms intervals during the delay between target onset and the cue to move. The intervals were centered 200, 300, 400, and 500 ms after target onset. Because delays ranged from 200 to 600 ms, trials with delay periods terminating before the end of an interval were not included.

Because the horizontal and vertical amplitude of the required movement deviated slightly from the spatial location of the target on each trial, responses were initially plotted in both movement coordinates (firing rate as a function of horizontal and vertical saccade amplitude) and target coordinates (the horizontal and vertical position of the target). These response-field plots were used to assess the spatial tuning of each neuron during each of the measured intervals as a function of movement amplitude or target position. Because neither coordinate system provided a better planar fit (see Planar fits below) or Gaussian fit (see Gaussian models below) to the spatial distribution of neuronal responses, data are presented only in movement coordinates. For graphical display, response-field plots of firing rate data were averaged in 4 × 4° bins using an arbitrary color scale.

STATISTICAL METHODS.

To examine the laterality of neuronal responses, firing rates in each 200-ms epoch were computed separately for those trials with targets on the side contralateral to the recording site and those with targets on the ipsilateral side in every studied cell. For each epoch, the average firing rate for ipsiversive movements was subtracted from the average firing rate for contraversive movements (hereafter referred to as contra bias) and then averaged across the population. Contra bias was compared with a zero value with an SE value equal to that during the initial fixation epoch using Student's t-test. A criterion of P < 0.05 was required for statistical significance. ANOVAs were used to determine significant differences across epochs. Finally, multiple regression analysis was used to examine relationships between behavioral variables, such as saccade accuracy and latency, as well as between neuronal data and behavioral performance. Multiple regression analysis permitted us to examine the relationships between particular variables, independent of the effects of any other potentially correlated factors, such as subject animal (see results for details).

GAUSSIAN MODELS.

Response fields were quantified by fitting a 2-dimensional Gaussian model to firing rate plotted in either movement coordinates or target coordinates for each cell in each epoch (cf. Platt and Glimcher 1998). A custom program (Matlab) used a Nelder–Meade simplex iterative fit to minimize the squared Cartesian distance between the Gaussian model and the firing rate using 6 free parameters. These parameters were the horizontal and vertical position of the center, the horizontal and vertical SDs (sigmas), baseline (or tonic) firing rate (F0), and the amplitude of the Gaussian (Fmax) Math(1) where z is the firing rate in the measured epoch during a single trial. In movement coordinates, x is the horizontal movement amplitude on a trial and y is the vertical movement amplitude on a trial. In target coordinates, x and y represent the horizontal and vertical position of the target, respectively.

The fitting routine was run for 20,000 iterations on 10 sets of random initial seed parameters. These initial parameters were: the median horizontal and vertical amplitudes from the input data for the center; 10 for the horizontal and vertical components of sigma; half the maximum firing rate in the epoch for the amplitude (Fmax); and the median firing rate in the epoch for the baseline (F0). The center of the Gaussian was constrained to lie within 42° of the plot origin, twice the sampled range of movements. Sigmas were limited to a range of 2–88°, the peak firing rate to a range of 0 to twice the maximum recorded firing rate in that epoch, and the base to a range of 0 to the maximum firing rate recorded in that epoch. Average sigma was calculated as the mean of the horizontal and vertical components of sigma. The proportion of total variance accounted for (VAF) by the Gaussian model was computed for each fit (cf. Platt and Glimcher 1998) as Math(2)

PLANAR FITS.

Planar functions were also used to quantify the response fields constructed for each of the 10 epochs plotted in both movement coordinates and target coordinates. A 2-dimensional planar model was fit to response-field data using a Nelder–Meade simplex iterative fit that minimized the squared Cartesian distance between the model and the raw data using 3 free parameters. These parameters consisted of the horizontal and vertical slope and the intercept Math(3) Again, z represents the firing rate in the measured epoch during a single trial, and x and y represent either the horizontal and vertical amplitude of movement or the horizontal and vertical position of the target. The initial seed parameters were set to zero for all 3 variables, and all values were searched for within a range of −1,000 to 1,000. The fitting routine was run for 20,000 iterations. The proportion of total variance accounted for by the planar model was computed as above. Slopes for individual studied cells were averaged across the population in each epoch. The average slope in each epoch was compared with a zero value with an SE value equal to that during the initial fixation epoch using Student's t-test with a criterion of P < 0.05 for significance.

ULTRASOUND.

B-mode digital ultrasound imaging (Sonosite 180) was used to visualize neuroanatomical landmarks beneath the recording chamber in both subjects (Glimcher et al. 2001; Tokuno et al. 2000), as well as electrode paths. After a recording session, the chamber was flushed and filled with saline. An endocavitary ultrasound probe (center frequency 7.5 MHz) was then lowered into contact with sterile saline in the chamber. Ultrasound power and frequency were adjusted to permit visualization of the brain up to depths of 5–7 cm. The probe was translated or rotated to image within the plane containing the electrode penetration. Images were stored digitally and later downloaded to a computer.

RESULTS

Localization of recording sites

We recorded from neurons lying within grid penetrations 1–3 mm lateral and 0–3 mm rostral to the intersection of theinteraural and midsagittal planes, at depths 8 to 12 mm from the cortical surface (Fig. 2). The location of the recording chamber relative to neuroanatomical landmarks, as well as electrode paths, was visualized using B-mode digital ultrasound in both monkeys. In the midsagittal images shown in Fig. 2A, several major landmarks can be identified in white against the black background in both subjects (for more details see Glimcher et al. 2001; Tokuno et al. 2000). The red boxes in each figure indicate the ventral projection of the recording chamber in each subject. Red arrows indicate the intersection of the marginal ramus and the horizontal limb of the cingulate sulcus. Diagrams in the Fig. 2B illustrate major landmarks, including the corpus callosum (cc), cingulate sulcus (cs), and marginal ramus (mr) of the cingulate sulcus, traced from the ultrasound images on schematic midsagittal sections. Ultrasound images indicated that the cingulate sulcus lay about 8–10 mm below the cortical surface, in good agreement with travel indicated on the micropositioner for neurons with task-related activity. Figure 2C shows examples of hyperechoic tissue disruption induced by the passage of a recording electrode, indicated by the blue arrows, in each subject (Tokuno et al. 2000). These tracks correspond well with the location of the grid hole used to place the electrode on that day (monkey Br: 1 mm anterior, 2 mm lateral; monkey Ni: 1 mm anterior, 3 mm lateral) and clearly target tissue at or anterior to the intersection of the horizontal limb of the cs and the mr. Based on the ultrasound images, the stereotaxic placement of the recording chamber, and the depths at which task-related neurons were encountered, recordings appear to have been concentrated in area 31 in the ventral bank of the cingulate sulcus, as well as area 23c in the cingulate gyrus, coextensive with the location of neurons attributed to PCC in a prior study in monkeys (Olson et al. 1996).

FIG. 2.

Localization of recording sites. A: midsagittal ultrasound images in 2 subjects. B: landmarks traced from the ultrasounds on schematic midsagittal macaque brain sections. Red boxes in each figure indicate the area immediately below the recording chamber in each subject. Red arrows indicate the intersection of the marginal ramus and the horizontal limb of the cingulate sulcus. Shaded blue boxes indicate the approximate location of recorded neurons. C: midsagittal images showing tracks left by guide tube and electrode, indicated by the blue arrows. cs, cingulate sulcus; cc, corpus callosum; mr, marginal ramus of the cs. Note difference in scale (A, C).

Behavior

Saccades made by one of 2 monkeys were significantly more accurate on delayed-saccade trials than on immediate-saccade trials (residual saccade error: monkey Br, delayed saccade: 1.49 ± 0.01°; immediate saccade: 1.52 ± 0.01°, t = −2.34, df = 9,966, P < 0.02; monkey Ni, delayed saccade: 1.20 ± 0.03°, immediate saccade: 1.17 ± 0.02°, t = 1.09, df = 1,916, P > 0.27), suggesting that the subject used information about the location of the target provided before the cue to move on delayed-saccade trials. Moreover, on delayed-saccade trials, saccade latencies were inversely proportional to the length of the delay period between target onset and the cue to move for both monkeys (Fig. 3; rdelay = −0.2300, P ≪ 0.001; rsubject = −0.1092, P ≪ 0.001; df = 11,467), perhaps reflecting the additional time available to attend to the target and/or plan a gaze shift. In addition, the slope of the line relating peak saccade velocity to saccade amplitude (the main sequence) was higher for immediate-saccade trials than for delayed-saccade trials (ramplitude = 0.8352, P < 0.01; rtrial type = 0.1355, P < 0.01; df = 18,111), suggesting that improved saccade accuracy on delayed-saccade trials was associated with lower peak velocities, reminiscent of a speed-accuracy trade-off. Consistent with this idea, residual saccade error was inversely correlated with saccade reaction time on both immediate-saccade trials (rreaction time = −0.0834, P ≪ 0.001; rsubject = −0.0244, P > 0.05; df = 6,641) and delayed-saccade trials (rreaction time = −0.2133, P ≪ 0.001; rsubject = 0.1230, P ≪ 0.001; df = 11,467).

FIG. 3.

Plot of reaction time vs. length of the delay period between target onset and movement onset on delayed-saccade trials, averaged across both subjects (r = −0.230002, n = 11467, P ≪ 0.00001).

Single-neuron data

We studied the spatial and temporal profile of activity of 122 task-related neurons in PCC. Only neurons for which activity was measured for ≥50 trials were included. Data from ≥50 immediate-saccade trials were recorded in a total of 76 neurons (60 from monkey Br, 16 from monkey Ni), whereas data from 50 or more delayed-saccade trials were recorded in 111 neurons (58 from monkey Br, 53 from monkey Ni). On average, subjects completed 87 (±3) immediate-saccade trials and 103 (±5) delayed-saccade trials for each neuron studied with each trial type.

Some neurons in our population showed a change in activity immediately after movement onset but not immediately after target onset, as reported previously (Olson et al. 1996). Figure 4 plots data for an example neuron exhibiting a similar temporal profile on both immediate-saccade (Fig. 4, A and B) and delayed-saccade trials (Fig. 4, C and D). In both, activation peaked after saccade onset. The maximal differences in neuronal activity for contraversive (red) and ipsiversive (blue) movements occurred roughly 200–600 ms after movement onset in both trial types, and neuronal activity was greater after contraversive (rightward) than ipsiversive (leftward) movements. To illustrate the spatial distribution of these responses, we computed firing rate in a 200-ms epoch beginning 500 ms after movement onset (gray box in Fig. 4, B and D) and plotted this activity as a function of the horizontal and vertical amplitude of the movement in both immediate-saccade and delayed-saccade trials (Fig. 5, A and C, respectively). Neuronal activity in this epoch was greatest for contraversive movements.

FIG. 4.

Peristimulus time histogram (PSTH) for a neuron with exclusively postmovement activity in both immediate-saccade trials (A, B) and delayed-saccade trials (C, D). Mean neuronal activity (±SE) is plotted in 16 consecutive 100-ms bins. Trials in which the target appeared in the hemifield contralateral to the recording site are shown in red, and trials in which the target appeared in the ipsilateral hemifield are shown in blue. Plots are aligned on either target onset (A and C) or movement onset (B and D). Red and blue arrows indicate the average onset time of contraversive and ipsiversive movements (A and C) or the average target onset time (B and D). Significant differences between responses to contraversive and ipsiversive movements by t-test at P < 0.05 (*) and P < 0.01 (**) level are indicated.

FIG. 5.

Response-field plots and surfaces derived from 2-dimensional Gaussian models and planar models fit to the response-field data for neuron 1. Firing rate is plotted in color for a 200-ms epoch (Fig. 4, gray box) beginning 500 ms after movement onset for movements ending within 4° bins across the central 44° of the visual field.

To quantify response fields, data were fit with 2-dimensional Gaussian and planar models (see methods). Gaussian and planar surfaces derived from the models fit to the response fields shown in Fig. 5, A and C are plotted for both immediate-saccade (Fig. 5B) and delayed-saccade (Fig. 5D) trials. For immediate-saccade trials, the Gaussian functions fit to the response field shown in Fig. 5A accounted for slightly more of the spatial variance in firing rates than did planar functions (59.9 vs. 57.4%). For delayed-saccade trials, Gaussian functions again accounted for more of the spatial variance in firing rates than did planar functions (71.3 vs. 59.4%).

Other neurons in our population displayed different temporal patterns of activation. Specifically, many neurons exhibited a change in activity immediately after target onset as well as after saccade onset on delayed-saccade trials, but only after saccade onset on immediate-saccade trials. Figure 6 plots data for an example neuron in which peak differences in activity after contraversive and ipsiversive saccades occurred about 200–400 ms after saccade onset (gray box in Fig. 6, B and D). Response-field plots of neuronal activity measured in this epoch were constructed to illustrate the spatial distribution of neuronal activity (Fig. 7). In both trial types, this neuron was more strongly activated for movements to the contralateral lower left quadrant than to other areas of the visual field. Spatial tuning was similar for the epoch immediately after target onset in both trial types (not shown). Gaussian and planar fits to these response fields for immediate-saccade (Fig. 7B) and delayed-saccade (Fig. 7D) trials confirmed activation for contraversive movements. The Gaussian model accounted for the spatial variance in firing rate better than the planar model in this epoch (immediate-saccade trials, Gaussian 44.5%, plane 41.7%; delayed-saccade trials, Gaussian 45.5%, plane 41.9%).

FIG. 6.

PSTH for a neuron that responded after both target onset and movement onset in immediate-saccade trials (A, B) and delayed-saccade trials (C, D). Conventions as in Fig. 4. Significant differences between the responses to contraversive and ipsiversive movements by t-test at P < 0.05 (*) and P < 0.01 (**) level are indicated.

FIG. 7.

Response-field plots and surfaces derived from 2-dimensional Gaussian models and planar models fit to the response-field data for neuron 2. Conventions as in Fig. 5.

Population data

TEMPORAL PROFILE AND LATERALIZATION OF ACTIVITY.

Many of the neurons in our population appeared to respond consistently after the onset of contralateral targets as well as after contraversive movements on delayed-saccade trials, even though the visual event preceded the saccade initiation cue by up to 600 ms. To quantify the lateralization and temporal progression of neuronal responses in the PCC population, differences in average firing rate for contraversive and ipsiversive movements were calculated for each cell in each epoch (contra bias) and then averaged across the population for immediate-saccade (Fig. 8A) and delayed-saccade (Fig. 8C) trials. Average differences in firing rate did not vary significantly from zero during fixation, but increased after target onset and after movement onset for both trial types, indicating a contralateral bias in neuronal responses on these trials. Intriguingly, on delayed-saccade trials, contra bias persisted during the first few hundred milliseconds of the delay period before diminishing just before movement onset.

FIG. 8.

Average differences in firing rate for trials with contralateral and ipsilateral targets (contra bias) plotted as a function of time for several 200-ms epochs aligned on trial events for immediate-saccade (A) and delayed-saccade (C) trials; 200–600 ms intervened between target onset and the cue to move in delayed-saccade trials. Mean (± SE) slopes of planes fitted to response fields for neurons in immediate-saccade (B) and delayed-saccade (D) trials during the same 200-ms epochs. Positive slopes indicate a greater neuronal response on trials in which the target appeared in the hemifield contralateral to the recording site. Significant differences from zero by t-test at P < 0.05 (*) and P < 0.01 (**) level are indicated.

This contralateral bias was also evident in the horizontal slopes of the fitted planes, which varied significantly as a function of time for both immediate-saccade (Fig. 8B; ANOVA, F = 2.6256, df = 9, P < 0.001) and delayed-saccade trials (Fig. 8D; ANOVA, F = 2.6415, df = 13, P < 0.01). The average horizontal slope of the fitted planes tilted contralaterally after target onset, as well as after saccade onset, on both immediate-saccade and delayed-saccade trials. On delayed-saccade trials, however, the planar slopes showed persistent contralateral tilt throughout the unpredictable delay period, but returned to zero just before movement onset. Average vertical slope did not vary significantly over time in either immediate-saccade trials (ANOVA, F = 1.384, df = 9, P > 0.19) or delayed-saccade trials (ANOVA, F = 0.9338, df = 13, P > 0.51).

The temporal progression of lateralized neuronal responses was similar in both trial types, but with important differences. In immediate-saccade trials, the change in firing rate after target presentation overlapped in time with movement-related responses. On delayed-saccade trials, however, a delay of at least 200 ms and as much as 600 ms intervened between target presentation and fixation offset, effectively separating modulation in neuronal activity related to target presentation and movement onset. On these trials, there was an increase in activity in the PCC population after the illumination of contralateral targets that was temporally separate from the increase after the onset of contraversive movements. Moreover, spatial selectivity persisted through the delay between target onset and movement onset, but fell to zero just before movement onset. After contralateral movement onset, the population again responded strongly and this spatially selective response persisted for several hundred milliseconds before diminishing around the time of reward delivery.

BROAD SPATIAL TUNING.

The spatial selectivity of single PCC neurons appeared to be quite broad. We examined the breadth of spatial selectivity across the population in 2 ways. First, we compared the variance accounted for by planar models and 2-dimensional Gaussian models fit to response fields measured in both immediate-saccade trials (Fig. 9A) and delayed-saccade trials (Fig. 9B) during the 200-ms epochs immediately after target presentation and after movement. Although the Gaussian model had 3 more free parameters than the planar model, Gaussian fits (black bars) did not account for significantly more of the spatial variance in firing rate than planar fits (gray bars) in immediate-saccade trials (posttarget: t = 1.817, df = 150, P > 0.07; postmovement: t = 1.442, df = 150, P > 0.15) or delayed-saccade trials (posttarget: t = 1.690, df = 220, P > 0.09; postmovement: t = 1.426, df = 220, P > 0.15). Moreover, the fitted Gaussian functions were quite broad, indicated by the large average sigma in all epochs for both immediate-saccade (Fig. 9B) and delayed-saccade (Fig. 9D) trials. Thus PCC response fields were broadly tuned spatially, being nearly equally well described by tilted planes and very broad 2-dimensional Gaussian functions.

FIG. 9.

Average variance accounted for by fitted Gaussians (black bars) and fitted planes (gray bars) for the 200-ms epochs immediately after visual target onset and movement onset for immediate-saccade (A) and delayed-saccade (C) trials. Mean (± SE) sigmas of Gaussians fitted to response fields for neurons in immediate-saccade (B) and delayed-saccade (D) trials during 200-ms epochs.

OVERALL FIRING RATE CHANGES.

To determine whether spatially selective neuronal responses in PCC could be described solely by changes in the tilt of roughly planar response fields or instead were also accompanied by a spatially uniform increase or decrease in neuronal activity, we examined the intercept of the fitted planes as a function of time on immediate-saccade trials (Fig. 10A) and delayed-saccade trials (Fig. 10B). Whereas the slope of the fitted planes represents spatial selectivity in neuronal responses, the intercept is an estimate of average firing rate across all movements. On delayed-saccade trials, the intercept showed a decreasing trend after the coillumination of the fixation and target LEDs (ANOVA, F = 1.5187, df = 13, P > 0.10). On these trials, the intercept rebounded after offset of the fixation LED, peaking around the time of the reward. On immediate-saccade trials, in which only one LED was illuminated at a time, the intercept did not vary (ANOVA, F = 0.537, df = 9, P > 0.85).

FIG. 10.

Overall neuronal activity as represented by the baseline of the fitted Gaussian (circles) and the intercept of the fitted plane (triangles) for immediate-saccade (A) and delayed-saccade (B) trials. Significant differences from sigma during fixation by t-test at P < 0.05 (*) and P < 0.01 (**) level are indicated.

We also examined the time course of the base of the fitted Gaussians as an index of untuned baseline neuronal activity in the PCC population. The base did not vary significantly over time on immediate-saccade trials (ANOVA, F = 0.5273, df = 9, P > 0.86), but did so on delayed-saccade trials (ANOVA, F = 2.799, df = 13, P < 0.001), mainly because of a decrease from fixation at the end of the delay period (t = 3.758, df = 220, P < 0.001). Like the intercept of the plane, the base decreased during the delay period when both the fixation light and target LED were illuminated. These analyses suggest that the addition of a second visual stimulus is associated with a decrease in overall PCC activity, perhaps reflecting divided attention or saccade target probability.

RELATIONSHIP TO BEHAVIOR.

On delayed-saccade trials, the target was illuminated before the cue to initiate movement, permitting subjects to attend to and/or plan a gaze shift to the target in advance of movement initiation (see Fig. 3). If PCC neurons carry information related to these processes, then the strength of activation after the onset of the target on delayed-saccade trials might predict saccade accuracy or reaction time. Because residual saccade error increased with increasing eccentricity, as expected (recentricity = 0.1515, P ≪ 0.001; rsubject = 0.1439, P ≪ 0.001; df = 11,467), and reaction time decreased with eccentricity (recentricity = −0.0382, P < 0.001; rsubject = −0.1056, P < 0.001; df = 11,467), we included eccentricity in the regression model for saccade error and latency as a function of firing rate. Moreover, we confined our regression analyses to trials on which the target was illuminated in the hemifield contralateral to the recording site because most neurons in our study responded with an increase in activity on these trials.

Analyzed in this fashion, there was a significant relationship between neuronal activity after target onset and the accuracy of saccades. Residual saccade error decreased with increasing firing rate, even after the effects of eccentricity and subject were removed statistically (rfiring rate = −0.0315, P < 0.05; reccentricity = 0.1408, P ≪ 0.001; rsubject = 0.162258, P ≪ 0.001; df = 5,694). Additionally, there was a tendency, though not significant, for saccade latency to decrease with increasing firing rate (rfiring rate = −0.0117, P = 0.37; reccentricity = −0.0371, P < 0.05; rsubject = −0.0817, P ≪ 0.001; df = 5,694). Thus for a given target displacement, stronger PCC activation after target onset for a given neuron was associated with more accurate and slightly faster gaze shifts on delayed-saccade trials.

DISCUSSION

Summary

We investigated the spatial and temporal response properties of neurons in PCC to determine whether their activity is modulated by the timing and spatial distribution of visual and oculomotor events. To do this, we studied the activity of single PCC neurons while monkeys performed immediate-saccade and delayed-saccade trials guided by up to 100 potential target locations. As previously reported (Olson et al. 1996), PCC neurons responded with a contralateral bias after the onset of visually guided saccades. Many of these neurons also responded immediately after visual target presentation on delayed-saccade trials. Moreover, the population discriminated contralateral from ipsilateral visual targets throughout much of the delay between target onset and movement onset on delayed-saccade trials. On these trials, neuronal responses after target onset were temporally distinct from neuronal responses after movement, and the magnitude of these responses was correlated with saccade accuracy. Finally, overall neuronal responsiveness decreased during the delay when both the fixation and target LEDs were illuminated.

Relationship to prior neurophysiological studies

In a prior study of neuronal activity in PCC in behaving primates (Olson et al. 1996), monkeys performed immediate-saccade trials guided by 4, or sometimes 16, possible target locations. Neurons in that study typically showed an increase in activity after gaze shifts to one or more target locations. The directional tuning of neuronal activity after saccades was broad and generally contraversively oriented. These results led Olson and colleagues to conclude that an important function of PCC is to monitor eye movements and eye position, possibly serving to update spatial representations of visual objects between saccades. A recent study showing hemodynamic responses in human PCC related to eye movements supports this conclusion (Berman et al. 1999).

In an earlier report, Olson and colleagues (1993) also reported that, whereas PCC neurons did not respond to small spots of light used as peripheral cues on a covert attention task, these neurons were activated in monkeys after the presentation of a large (16° diameter) checkerboard stimulus. The authors also noted that postsaccadic activation of PCC neurons during immediate-saccade trials was strongest when they were performed in an illuminated room, suggesting that these neurons may be sensitive to visual stimulation or possibly task context.

Our results both corroborate and extend these findings. Consistent with the report by Olson and colleagues (1996), PCC neurons in our study increased their activity after a broad range of contraversive saccades on immediate-saccade trials. The broad spatial tuning of PCC neurons estimated by Olson and colleagues (1996) using between 4 and 16 targets was confirmed in our study using up to 100 visual targets, and quantified by the statistical equivalence of planar and 2-dimensional Gaussian fits to neuronal response fields.

We also made several novel observations. First, PCC neurons were activated after the onset of a broad range of small contralateral targets, as well as after the onset of saccades guided by those targets, when a delay intervened between these events. Second, spatial selectivity for contralateral targets was maintained throughout much of the delay. Third, overall responsiveness decreased during the delay when both the fixation and saccade targets were illuminated, irrespective of movement direction and amplitude. Fourth, stronger neuronal activity after target onset predicted more accurate saccades on delayed-saccade trials.

Some of these results are consistent with prior neurophysiological studies of PCC in cats. Kalia and Whitteridge (1973) first reported responses from PCC neurons in anesthetized cats to the presentation of white or black bars of various sizes. Olson and colleagues also studied the activity of single PCC neurons in alert cats and found neurons that responded to the onset of room lights as well as after visually guided saccades in an immediate-saccade task (Olson and Musil 1992). Although very few PCC neurons responded after the illumination of small saccade targets in an immediate-saccade task in a second study in cats by the same authors (Musil and Olson 1993), many neurons did respond to a visual stimulus flashed at low frequencies at fixation (<5 Hz).

Our results differ from the one prior study in monkeys in that PCC neurons responded after the onset of small contralateral visual targets on delayed-saccade trials, and maintained selectivity for these targets during delays before saccade onset. It seems unlikely that such differences are the result of recording from a different population of neurons from those studied by Olson and colleagues (1996), given that the pattern of activation reported here in immediate-saccade trials is consistent with their results using the same task (Olson et al. 1996). Additionally, ultrasound images of neuroanatomical landmarks beneath the recording chamber and electrode tracks place recordings in the same region as neurons recorded in that study. These observations suggest that the responses of PCC neurons to small spots of light reported here may have reflected their salience. We speculate that monkeys either covertly attended to these targets or planned overt gaze shifts to them during the unpredictable delay period between target onset and fixation offset on delayed-saccade trials, and that these processes were reflected by neuronal activity in PCC. Furthermore, we suggest that these covert processes were manifested in the improved accuracy of saccades on these trials and the inverse relationship between delay length and latency. This hypothesis is supported by the observation that stronger neuronal responses in PCC after target onset were associated with more accurate saccades on delayed-saccade trials and a tendency toward faster reaction times.

Relationship to activity in other eye-movement–related areas

Many cortical areas involved in oculomotor planning and control have direct connections with PCC. Oculomotor information in PCC could arise from its interconnections with the parietal lobe (Baleydier and Mauguiere 1980; Vogt and Pandya 1987), prefrontal cortex (PFC) (Barbas and Mesulam 1985; Vogt and Pandya 1987), the frontal eye fields (FEF) (Vogt and Pandya 1987), and the supplementary eye fields (SEF) (Huerta and Kaas 1990).

Like PCC, many areas with saccade-related responses also show activity linked to the onset of a visual stimulus, including the superior colliculus (SC) (Goldberg and Wurtz 1972), SEF (Schlag and Schlag-Rey 1987), FEF (Goldberg and Bushnell 1981), PFC (Boch and Goldberg 1989), and the lateral intraparietal area (LIP) (Colby et al. 1995). This visual activity is often modulated by task requirements. For example, FEF (Goldberg and Bushnell 1981), SC (Wurtz and Goldberg 1972), PFC (Boch and Goldberg 1989), and LIP (Colby et al. 1995) activity is enhanced when a visual stimulus is the target of a saccade.

We show here that target-related activity in many PCC cells is not transient but rather remains spatially selective through delays between target onset and saccade onset. Activity in many SC (Pare and Wurtz 2001), FEF (Bruce and Goldberg 1985; Sommer and Wurtz 2001), LIP (Ferraina et al. 2002), and PFC (Funahashi et al. 1990; Tsujimoto and Sawaguchi 2004) neurons also increases after target onset and persists during delays on delayed-saccade trials.

Although the target and delay activity of PCC neurons is similar to that in other areas, neuronal responses in this area before and after the saccade are somewhat unique. In many areas with eye-movement–related activity, including the SC (Wurtz and Goldberg 1972, 1971), LIP (Andersen and Gnadt 1989; Barash et al. 1991a), and SEF (Schlag and Schlag-Rey 1987), firing rate increases immediately before movement. In contrast, we report a decrease in the overall responsiveness of PCC neurons immediately preceding movement on delayed-saccade trials. A similar pause in firing rate preceding eye movement has been found in a small subset of movement-related cells in FEF (Sato and Schall 2001). The firing rate in many areas involved in oculomotor planning and initiation, including SC (Wurtz and Goldberg 1972, 1971), FEF (Bruce and Goldberg 1985; Segraves and Goldberg 1987), and LIP (Barash et al. 1991a), peaks around or soon after saccade onset, but responses generally fall off quickly after movement offset. In contrast, the results of the current study and others (Olson et al. 1993, 1996) demonstrate that PCC neurons respond most strongly after saccade onset, and this activity often persists for several hundred milliseconds. Thus although PCC responses share some similarities with those in other eye-movement–related areas, particularly after target onset and during premovement delays, the decrease in overall responsiveness before the saccade and the peak response after movement distinguish PCC from these other eye-movement–related areas.

Relationship to parietal area 7a

The response properties of neurons in PCC are remarkably similar to those in parietal area 7a, and these areas are reciprocally connected (Vogt and Pandya 1987). Much like neurons in PCC, cells in area 7a typically respond after movement onset on visually guided saccade trials (Barash et al. 1991a), as well as after the onset of peripheral visual stimuli when monkeys fixate a central stimulus (Andersen et al. 1987). Also, neurons in area 7a have been noted to be relatively inactive just before saccade onset (Barash et al. 1991a). Such decreases in activity in other brain areas have been argued to represent a “resetting” of the spatial coordinates of visual representations or attention before overt gaze shifts (Duhamel et al. 1992; Sato and Schall 2001). The activity of many neurons in area 7a is also sensitive to eye position (Andersen and Mountcastle 1983; Andersen et al. 1985, 1990) as well as visual salience (Constantinidis and Steinmetz 2001). Based on these response properties, it has been suggested that area 7a may be involved in covert attention (Steinmetz and Constantinidis 1995; Steinmetz et al. 1994) and sensorimotor transformations (Andersen et al. 1990).

The broad spatial tuning of neurons in PCC is also similar to the response-field properties of neurons in area 7a. Response fields in LIP (Barash et al. 1991b; Blatt et al. 1990), FEF (Bruce and Goldberg 1985), SC (Sparks et al. 1976), and PFC (Funahashi et al. 1990) are generally smaller and contralateral. However, response fields in area 7a, like those in PCC, tend to be large and, unlike PCC, bilateral (Blatt et al. 1990). Taken together, these observations suggest the hypothesis that the response properties of neurons in area PCC and 7a are functionally related.

Relationship to neuroimaging studies in humans

Recent neuroimaging studies suggest that visual activity in PCC may be related to shifting attention to a salient visual target. Several studies have found activation of PCC when attention is directed to a hemifield before the presentation of a visual target (Hopfinger et al. 2000, 2001; Kim et al. 1999; Small et al. 2003) or after the presentation of targets for attention (Yamasaki et al. 2002). Indeed, one recent study found a correlation between task-related signal changes in PCC and reaction time during a cued spatial attention task, similar to the weak relationship between neuronal responses and saccade reaction time reported here, implying a more direct role for PCC in allocating spatial attention (Mesulam et al. 2001). Our data also demonstrate a sustained selectivity in PCC for contralateral targets during the delay period between target onset and movement onset, perhaps reflecting attention to the target while the monkey plans but withholds his saccade. The de-crease in average and baseline activity during the coillumination of the fixation and target LEDs in our study is consistent with divided attention or saccade target probability (Basso and Wurtz 1998).

Relationship to motivational processing

It was previously suggested that PCC may be involved in associating motivational significance with environmental events (Freeman et al. 1996). In our study, overall responsiveness tended to increase around the time of juice reward delivery, suggesting that PCC neurons may respond in a spatially nonselective fashion to rewards. Consistent with this idea, a recent study from our laboratory reported that PCC neurons are activated after both saccade onset and reward delivery, and that the activity of these neurons is systematically modulated by the size and predictability of reward (McCoy et al. 2003). In primates, PCC may thus serve to highlight the motivational salience of visual targets for orienting attention.

In conclusion, the spatial and temporal response properties of single PCC neurons were measured in monkeys performing saccades guided by targets drawn from a pool of 100 potential target locations, both with and without a random delay intervening between target onset and movement onset. Response fields constructed from these data demonstrated that many PCC neurons were activated after the illumination of small contralateral visual targets, as well as after the onset of contraversive saccades guided by those targets. Moreover, the PCC population maintained selectivity for contralateral targets during delays between target onset and movement. Overall, PCC neurons were selective for a broad range of contralateral directions and amplitudes, and planar functions described response fields nearly as well as broadly tuned 2-dimensional Gaussian functions. Saccade accuracy was correlated with neuronal firing rate after target onset. The presence of more than one stimulus during the delay was associated with a decrease in overall responsiveness. Our results suggest that PCC neurons may signal the salience of visual stimuli in broad regions of space, consistent with a role in orienting and attention.

GRANTS

This work was supported by National Eye Institute Grant EY-013496, the McDonnell-Pew Program in Cognitive Neuroscience, the Whitehall Foundation, and the EJLB Foundation.

Acknowledgments

We thank H. Honore, S. Roberts, and J. Pantoja for expert technical assistance; N. Cant, E. Jarvis, and M. Nicolelis for use of their lab equipment; and M. Bendiksby, R. Deaner, C. Olson, J. Roitman, and A. McCoy for helpful comments on the manuscript.

Footnotes

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

REFERENCES

View Abstract