HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) All Versions of this Article: 97/1/387    most recent 00558.2006v1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Gardner, E. P. Articles by Ro, J. Y. Search for Related Content PubMed PubMed Citation Articles by Gardner, E. P. Articles by Ro, J. Y.

Neurophysiology of Prehension. I. Posterior Parietal Cortex and Object-Oriented Hand Behaviors

Department of Physiology and Neuroscience, New York University School of Medicine, New York, New York

Submitted 25 May 2006; accepted in final form 7 September 2006

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Hand manipulation neurons in areas 5 and 7b/anterior intraparietal area (AIP) of posterior parietal cortex were analyzed in three macaque monkeys during a trained prehension task. Digital video recordings of hand kinematics synchronized to neuronal spike trains were used to correlate firing rates of 128 neurons with hand actions as the animals grasped and lifted rectangular and round objects. We distinguished seven task stages: approach, contact, grasp, lift, hold, lower, and relax. Posterior parietal cortex (PPC) firing rates were highest during object acquisition; 88% of task-related area 5 neurons and 77% in AIP/7b fired maximally during stages 1, 2, or 3. Firing rates rose 200–500 ms before contact, peaked at contact, and declined after grasp was secured. 83% of area 5 neurons and 72% in AIP/7b showed significant increases in mean rates during approach as the fingers were preshaped for grasp. Somatosensory signals at contact provided feedback concerning the accuracy of reach and helped guide the hand to grasp sites. In error trials, tactile information was used to abort grasp, or to initiate corrective actions to achieve task goals. Firing rates declined as lift began. 41% of area 5 neurons and 38% in AIP/7b were inhibited during holding, and returned to baseline when grasp was relaxed. Anatomical connections suggest that area 5 provides somesthetic information to circuits linking AIP/7b to frontal motor areas involved in grasping. Area 5 may also participate in sensorimotor transformations coordinating reach and grasp behaviors and provide on-line feedback needed for goal-directed hand movements.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Prehension is an object-oriented behavior that underlies many skilled actions of the hand. It comprises four major components: reach, grasp, manipulation, and release that appear to be mediated by different areas of the cerebral cortex, and monitored by various sensory modalities. Visual information about an object's intrinsic properties—size, shape, and identity—and its spatial location in the workspace aids motor planning of grasping (reviewed in Fogassi and Luppino 2005; Jeannerod et al. 1995; Milner and Goodale 1995; Paulignan and Jeannerod 1996; Sakata et al. 1997). The initial view of the object strongly influences the kinematics of acquisition by determining the opening of the fingers during reach as the hand is preshaped for grasp, selection of contact points on the object to promote grasp stability and efficient manipulation, and the initial rate and magnitude of grip and load forces applied by the hand on contact. Hand preshaping to object size and shape during reach is characteristic of normal primate hand function (Chieffi and Gentilucci 1993; Jeannerod 1986, 1994; Jeannerod et al. 1995; Roy et al. 2000, 2002) and is disrupted by lesions of anterior zones of posterior parietal cortex (PPC) in humans and monkeys (Gallese et al. 1994; Goodale and Westwood 2004; Jeannerod et al. 1994, 1995; LaMotte and Acuña 1978; Milner and Goodale 1995; Pause and Freund 1989; Pause et al. 1989; Tunik et al. 2005) and their projection sites in the frontal lobe (Fogassi et al. 2001).

PPC is thought to be an important nexus in motor planning of prehension because its anatomical connections with both the dorsal stream of vision and the somatosensory areas of the anterior parietal lobe allow it to combine visual and postural information to develop a plan of action. Since the early studies of Mountcastle et al. (1975), it is well established that neurons in PPC play a significant role in reaching, pointing and grasping behaviors (reviewed in Andersen and Buneo 2002; Andersen et al. 1997; Battaglia-Mayer et al. 2003; Caminiti et al. 1998; Fogassi and Luppino 2005; Hyvärinen 1981; Jeannerod et al. 1995; Kalaska 1996; Kalaska et al. 1997; Wise et al. 1997). Collectively, these studies implicate PPC in sensorimotor transformations needed to direct the hand to objects of behavioral interest such as food and to acquire them for consumption.

Single-cell recordings in monkeys and functional imaging of human cerebral cortex indicate that reach and grasp are temporally synchronized but controlled by distinct networks of neurons in parietal cortex (Binkowski et al. 1998, 1999; Chieffi and Gentilucci 1993; Culham et al. 2003; Ehrsson et al. 2000; Frey et al. 2005; Grafton et al. 1996; Shikata et al. 2003; Tunik et al. 2005). Anatomical segregation of neurons tuned to reach and grasp behaviors was first reported by Mountcastle and co-workers (1975), who noted that hand-manipulation neurons in both area 5 and area 7 were more likely to be recorded on electrode tracks placed more laterally than those in which arm projection neurons were encountered. Subsequent investigations confirmed and extended these findings. Sakata and co-workers (Murata et al. 1996, 2000; Sakata et al. 1995, 1997, 1999; Taira et al. 1990) and Fogassi and Luppino (2005) demonstrated that neurons in the anterior intraparietal area (AIP) of the inferior parietal lobule responded to viewing objects as well as grasping them in trained tasks. Clear synergies occurred between observation and action in their task, and many of the cells responded preferentially to grasp and/or view of particular objects. They proposed that firing rates of AIP neurons might be used to select the appropriate grasp posture needed to acquire objects of specific sizes or shapes.

Hand manipulation neurons in area 5 were not studied after the original description by Mountcastle and co-workers (1975) until our laboratory adapted digital video to quantify hand behaviors during prehension (Debowy et al. 2001; Gardner et al. 1999, 2002; Ro et al. 1998, 2000). Using a grasp-and-lift task to compare firing patterns in primary somatosensory (S-I) cortex and PPC, we found that the onset of activity in PPC preceded that in S-I. This was a somewhat surprising finding to us, because the anatomical connectivity between S-I and PPC suggested that areas 5 and 7b occupied higher levels in the hierarchical organization of somatosensory areas of the cerebral cortex (reviewed in Felleman and Van Essen 1991). Furthermore, neurons in the rostral bank of the intraparietal sulcus (IPS) hand representation were shown to have more complex physiological responses to somatosensory stimuli than those of neurons in areas 3b, 1, and 2 (Darian-Smith et al. 1984; Duffy and Burchfield 1971; Iwamura and Tanaka 1978; Iwamura et al. 1993, 1995; Sakata et al. 1973).

In this report, we extended our original studies of prehension to additional animals to compare responses in the hand representation of area 5 to neurons in area AIP and the adjacent inferior parietal lobule (area 7b or PF/PFG). Each of the three monkeys studied used an individualized hand posture to grasp the objects, allowing us to determine whether the firing patterns observed in our initial studies could be generalized to the various muscle synergies used by each subject. The data presented in this report confirm and extend our earlier studies, indicating that neurons in both subregions of PPC serve a sensorimotor function during acquisition of objects by the hand. These neurons also receive somatosensory feedback that appears to confirm the expectations of reach and grasp actions, and enable corrective maneuvers of the hand if the desired action was unsuccessful.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Neurophysiological and behavioral data were obtained from three adult rhesus monkeys (Macaca mulatta, 2 male and 1 female, weight: 8–16 kg), trained to perform a prehension task. Experimental protocols used in this study were reviewed and approved by the New York University Medical Center Institutional Animal Care and Use Committee (IACUC) and are in accordance with the guiding principles for the care and use of experimental animals approved by the Councils of the American Physiological Society, the National Research Council, and the Society for Neuroscience.

Prehension task

The prehension task required the animals to grasp and lift objects using visual cues displayed on a computer monitor to select the appropriate one. The test objects were a set of four knobs mounted on a box placed at arm's length, 22–24 cm in front of the animal (Fig. 1). When testing the right arm, the knobs were arrayed left to right (1) in front of the monkey's left shoulder, (2) at the midline, (3) in front of the right shoulder, and (4) lateral to the right shoulder. The shape box was shifted to the left in recording sessions from the right hemisphere when testing the left arm. The knob shapes tested included rectangular blocks (20 x20 x 40 mm), large and small round knobs (30 or 15 mm diam), and a cylinder (40 x 15 mm diam). The total load was adjusted with weights inside the shape box and ranged were 108 g (small round), 137 g (rectangle), 140 g (cylinder), and 242 g (large round). The knobs were lifted using a whole-hand power grasp between fingers and palm. The animals could view the workspace, including all four knobs, and used visual guidance to position the hand on the objects. Some neurons were also tested with view of the shape box and hands blocked by an opaque plate inserted in the chair frame below the chin.

View larger version (44K): [in this window] [in a new window]   FIG. 1. Stages of the prehension task traced from sequential digital video images of a lateral reach trial. Labels describe the action performed in the top image of each set. This trial starts with the hand resting on knob 1 (left rectangle). Approach begins as the animal retracts his hand from the knob and lifts the wrist. The hand follows an arc-like path toward the right side of the workspace. Wrist rotation is coordinated with hand preshaping as the hand decelerates and descends near the large round knob. The fingertips contact the knob on its lateral side, and the hand moves downward to grasp it between the fingers and the palm. Once grasp is secured, the knob is rotated, and lift begins. Lowering the knob retraces the same path as lift. Grasp is subsequently relaxed (not shown), ending the trial. Methods for construction of the time series of hand kinematics are summarized in the text. Color matched video time code labels (minute:second:frame) mark the time elapsed between images; frame rate = 29.97 frames/s.

Trials were self-paced, without external time constraints on trial initiation or duration. Task performance mimicked natural grasping behaviors in that the animals were allowed almost complete freedom of execution so long as they fulfilled the basic goals of acquiring and lifting the designated object. The monkey could freely choose how to position its hand on the knob so long as the grasp posture secured it during lifting. Animals were not required to remove the hand from the workspace between trials and often left part of the hand touching the knob in anticipation of possible repeats of the same cue.

Digital video monitoring of hand kinematics

The monkey's hand movements were monitored by sets of digital video (DV) cameras at 29.97 frame/s and digitized in the camera itself (Canon XL-1 and Sony TRV900 Mini-DV camcorders) or with a digitizer board (Radius Video Vision Studio and Sony CCD-VX3 color Hi-8 camcorders). This system provided synchronized digitized records of neuronal spike trains that were correlated directly to matching video images using frame time codes (Debowy et al. 2002; Ro et al. 1998). The DV format provided high-definition image quality using consumer-grade, inexpensive DV camcorders that compressed 720 x 480 pixel video images to 3.1 MB with MPEG-2 sampling; spike trains were recorded and digitized at the same time on the camcorder's audio channels. Camcorders provided lateral, frontal, and/or overhead real-time views of the monkey and the workspace, and stored kinematic records of the animal's behavior on videotapes; simultaneously acquired neuronal responses, fed from the electrophysiological amplifiers, were recorded on the audio track.

Digital editing software (Final Cut Pro version 3 or Adobe Premiere version 5.1) was used to download clips of the experiment via the Firewire ports of Macintosh G4 or IMac computers and stored as QuickTime movie files. Hand behaviors were viewed in real time, at high speed, or in frame-by-frame mode. Forward and backward bracketing of sequential frames was particularly useful for visualizing how the hand posture changed over time and for compiling event logs of the start times of the task stages on each trial. These event time codes were stored in spreadsheets and were used subsequently by the software tools as markers for alignment of neural responses in rasters and peristimulus time histograms (PSTHs) and for bracketing task stages in statistical analyses of firing rates.

To further delinate the trajectory of hand movements during the task, we exported sets of sequential video frames as TIFF files that were placed in separate layers of Adobe Illustrator files. The pen tool was used to trace outlines of the monkey's arm, hand, and fingers in each frame, as well as the shape box and knobs, to construct a time series of kinematic drawings (Reitzen et al. 2004). Successive frames were aligned in separate layers and overlapped as shown in Fig. 1. The topmost drawings were made transparent, allowing the underlying layers to be visualized. In this way, the trajectory used by the hand to acquire and manipulate objects, and the dynamics of movement could be viewed directly in single images.

The kinematic drawings also provided an objective standard for parsing hand movements into seven distinct stages plus an intertrial interval (stage 0). Actions subsumed in each stage are illustrated in the hand tracings of Fig. 1. 1) Approach: the reach interval, which began as the animal projected the hand toward an object (red), and ended when the hand contacted it (yellow). The hand was preshaped during approach, assuming a posture that anticipated object acquisition. 2) Contact: the hand positioning interval that spanned the period betweeninitial touch (yellow) and full enclosure of the knob between the fingers and palm as it was grasped (orange). 3) Grasp: static enclosure of the knob in the hand prior to lift. Although the knob was sometimes rotated during the grasp stage, there was no further tangential motion of the hand over its surface. 4) Lift: upward displacement of the knob from rest (light blue) to the top position (magenta). 5) Hold: maintained elevation of the knob at the upper stop. 6) Lower: downward replacement of the knob through relaxation of grasp (cyan). 7) Relax: maintained hand contact on the knob in a relaxed posture.

The relax stage was succeeded by regrasp of the test object or by hand withdrawal from the knob. Release of a knob was followed by lateral reach to a new knob, initiating another trial, or removal of the hand from the workspace.

Stages 1–3 were required for object acquisition, stages 4 and 5 for manipulation, and stages 6–7 for release of the object. The time codes of these events were logged from the single frame views, and used as event markers for spike data analysis.

Surgical and recording techniques

Extracellular single-unit recordings were made in the left hemisphere of B2195 and H17094, and in both the left and right hemispheres of the third animal (N18588) as they performed the prehension task. Using techniques for chronic single-unit recordings developed by Gardner and Costanzo (1980) and Warren et al. (1986), a stainless steel chamber was permanently implanted over the postcentral gyrus hand area in an aseptic surgical procedure under general anesthesia (1.5–3% isoflurane mixed with 2–3 l/min of O₂). The dura was left intact to prevent infection and contain brain swelling. The recording chamber provided access to a 25-mm-diam region of cortex centered 2–4 mm posterior, and 18–20 mm lateral to the bregma; the rostral end was situated 2–5 mm anterior to the central sulcus, and the caudal end was located over the inferior parietal lobule. The chamber was sealed with a transparent Lucite cap, except during recording sessions when the cap was replaced with a sterile Silastic membrane held in a stainless steel ring. A pair of stainless steel screws (size 6–32) cemented to the occipital skull limited head movements during recording sessions to small vertical displacements to maximize the stability of spike train recordings.

Buprenex (buprenorphine hydrochloride, 0.01 mg/kg bid) was administered for a 4-day period after the surgery to alleviate postoperative pain. Solu-Medrol (methylprednisolone sodium succinate; 5 mg/kg im) was given immediately after surgery, and on the following day, to reduce brain swelling. Intraoperative antibiotics [Baytril (enrofloxacin) solution; 1 mg/kg] were supplemented with once-daily doses for 6–7 days postoperatively. The interior of the chamber was rinsed with 35–50 ml sterile saline before and after each recording session, and the wound margins were washed with surgical sponges and hydrogen peroxide. Topical antibiotics (gentamycin or Baytril) were applied as necessary to the implant site.

Extracellular recordings of spike trains in the left hemisphere were made with epoxylite-insulated tungsten microelectrodes (FHC, Model UEWLFELE2N1X, impedance = 2 M) advanced through the intact dura and into the brain by a remotely operated miniature stepping hydraulic motor (David Kopf Instruments, Model No. 607W). Microelectrode recordings in the right hemisphere of N18588 used a computer-controlled multiple electrode positioning system (Alpha Omega EPS-MT) that allowed simultaneous recordings from four independently mobile tungsten microelectrodes. Recording depth was calibrated from the microdrive reading; the depth at which the electrode exited the cortex at the end of the session was subtracted from that of the recording site to yield its approximate intracortical location.

Calibrated positioning guides placed within the chamber lumen specified the actual site of microelectrode insertion. In single-electrode studies, different combinations of two guides allowed us to establish a recording site's radial distance from the center of the chamber and angular displacement from the midline with 0.25 mm precision. The position of the multielectrode guide tube was indicated on a vernier in anterior-posterior and medial-lateral coordinates relative to the chamber center, accurate to 0.1 mm. The position of each penetration site was marked on photographs of the brain made during surgery, creating a functional micromap.

Spike trains were amplified and filtered (band-pass 100 Hz to 3 kHz, Grass P511 amplifiers or Cyberamps, Axon Instruments), displayed on oscilloscopes and/or computer monitors, and digitized at 16-bit resolution, 48 kHz, or 12-bit resolution, 32 kHz, by the DV camcorders. The same spike data were captured on all three cameras, allowing precise synchronization of their audio and video tracks. The digitized spike trains were downloaded to the lab computers together with the video clips of the animal's behavior and stored as both QuickTime audio signals and in Audio Interchange file format (AIFF) for quantitative analyses of firing patterns. The raw spike trains were displayed by the editing software as a strip chart in a separate window for the audio waveform, allowing us to correlate hand movements shown in the video window with the corresponding portion of the spike train. As video and spike trains were recorded and digitized simultaneously, both datasets spanned the same time interval. Hence, knowledge of the time code of each video frame in the clip provided a precise way to locate the matching firing patterns. Similarly, measurements of the timing of spikes with respect to the onset of the audio data sample placed each spike in a precisely designated video frame.

Firing patterns of cortical neurons were analyzed directly from the AIFF files using interactive clustering to distinguish neuronal action potentials from noise and to separate the spike waveforms of up to four different neurons recorded from each electrode into individual traces (Ro et al. 1998; Sherwood et al. 2006). A list of consecutive spike time stamps was obtained for each neuron to construct continuous displays of firing rates and perform other analyses. Neurons had to be recorded for 5 min to capture a sufficient number of task trials for statistical analyses.

Quantitative analyses of neuronal responses

Burst analysis graphs (Fig. 2) providing a continuous record of neural and behavioral events within a video clip were used to screen neural responses to the task. Burst analysis provided an objective mechanism for correlating periods of high neuronal firing with behavioral activity as it relied on the responses of the neuron as an alignment metric rather than subjective standardization of the animal's actions. Spike trains were represented as rasters and continuous binned firing rates together with markers of actions performed by the monkey and/or experimenter during the clip. Reverse correlation of periods of high firing (green "burst" trace) with the matching video images of the monkey's behavior at the burst start, peak, and end times were used to highlight the behaviors to which a neuron was most responsive. We chose 100 ms as the minimum burst duration because it spanned three complete video frames in NTSC format and 2.5 frames in PAL (30 and 25 fps US and European video standards). The burst threshold was set one SD above the mean rate per 100-ms bin compiled during the entire 2- to 3-min clip. This protocol allowed us to determine whether we could predict what the animal was doing by simply examining continuous spike train data. By repeating the process of frame captures for the largest bursts, we examined whether there was a reliable relationship between neuronal activity and the kinematics of prehension. The linkage between bursts and task kinematics suggested the most relevant task stage(s) on which to trigger PSTHs and rasters. Neurons without a clear response to the trained task or to spontaneous grasp behaviors were not subjected to further analyses.

View larger version (23K): [in this window] [in a new window]   FIG. 2. Burst-analysis graphs of continuous neural and behavioral activity. This example displays the spike train, continuous firing rate binned in 100-ms intervals (blue graph), and task actions performed during an 8-s period excerpted from a 2.5-min video clip. Yellow task stage trace: each stepped yellow pyramid marks a single trial. Upward deflections denote the start of stages 1–4; downward deflections mark the onset of stages 5–8. Orange knob trace: knob location on the shape box and duration of hand contact are depicted as downward pulses that span the contact through lower stages. The pulse amplitude is proportional to the knob distance from the left edge of the box. White burst threshold trace: firing rate set 1 SD above the mean rate during the entire 2.5-min analysis period. Green burst trace: upward deflections mark periods when the firing rate exceeds the burst threshold. The burst amplitude indicates the mean rate during the interval of high activity; in this example, the burst trace is displaced upward by 80 spikes/s for clarity. The neuron responded most vigorously during stages 1–3 on each trial.

Further quantification of task-related activity was obtained from measurements of mean firing rates per stage on each trial. Responses on all trials were averaged to compile firing rate profiles for each neuron during the pretrial interval and the individual task stages. Neuronal activity during successive actions was more accurately depicted by average firing rate graphs because the PSTH profile was somewhat dependent on the event selected to align spike trains and variable task stage duration across trials. Neurons were grouped by the stage(s) that evoked maximum firing and subdivided into classes tuned to single actions, two successive actions, or broadly tuned classes by statistical comparison of mean rates during sequential task stages.

The individual trial data were used for statistical analyses and to further classify responses. A repeated-measures ANOVA model (StatView, SAS Institute) analyzed whether each neuron demonstrated significant modulation of firing rates across the seven task stages and the pretrial interval (F-test, P < 0.05). Firing rates between stage markers on each trial were within-subject variables and the task conditions (knob grasped, approach style, hand used) between-subject variables. Nearly all of the task-related neurons yielded P < 0.001 on F-tests. In addition, task-related neurons were required to show significantly increased or decreased firing rates during at least one task stage compared with the pretrial rate in paired means comparisons (P < 0.05).

Histological localization of recording sites

Both physiological and neuroanatomical techniques were used to locate recording sites. As the monkeys were studied for periods of 2 yr, it was not possible to recover the precise electrode tracks for all but the last few in any brain. Instead we used the putative entry points of the microelectrodes into the cortex to localize the recording sites. These methods allowed us to reconstruct the antero-posterior (A-P) and medio-lateral (M-L) coordinates of electrode tracks; recording site depth along the track was approximate. Small errors in track localization were inevitable, especially at borders between areas. However, physiological properties of neurons recorded along the electrode tracks were consistent with the stated designations.

In selected sessions, we used electrodes coated with DiI or DiI-5 to mark recording sites of particular interest following protocols of DiCarlo et al. (1996). In addition, 16–20 days before death, injections of fluorescent dyes (fast blue, diamidino yellow, fluororuby) and dextran-conjugated tracers (dextran alexa, dextran biotin) were made in two of the three animals at sites that had yielded particularly interesting data on earlier penetrations to help localize those tracks in the cortex. Injection sites were also marked with microelectrodes dipped in DiO and inserted manually through the micropositioner. This left a blue trace that was visible during cryosectioning of the brain, helping to localize dye injections to particular histological sections.

The animals were killed by an overdose of intravenous barbiturate anesthetic (Nembutal, 120 mg/kg) following the current guidelines established by the American Veterinary Medical Association. The brain was prepared for histology by intracardiac perfusion with saline, followed by 4 l of 10% buffered formalin. The A-P and M-L boundaries of the recording chamber were marked using stainless steel hypodermic tubing (24 gauge) dipped in India ink or electrodes coated in DiO inserted through the micropositioner. The brain was photographed, and the reference marks used to align the map of recording sites on the cortical surface. Frozen histological sections were cut in the coronal or horizontal planes and stained with cresyl violet, or prepared for fluorescence microscopy.

Recording sites were reconstructed from serial sections. Histologically identified tracer injection sites were used to align the entry sites into the cortex and to extrapolate the locations of other recording tracks. The postcentral gyrus hand area was divided into three cytoarchitectural zones based on criteria set forth by Pons et al. (1985) and Lewis et al. (Lewis and Van Essen 2000a,b; Lewis et al. 1999). Area 3b-1, the most anterior zone of S-I cortex, included penetrations within a band 2 mm caudal to the central sulcus. Posterior S-I comprised the next 3–4 mm on the exposed gyrus, and was denoted as area 2. PPC comprised cortical areas surrounding the IPS and was divided into superior and inferior parietal lobules. The superior parietal lobule (SPL) included the rostral bank between area 2 and fundus of the IPS, corresponding to Brodmann's area 5. Neurons recorded in the exposed bank of the SPL were labeled area 5d following the terminology of Lewis and Van Essen (2000a), whereas those in the rostral bank of the IPS were labeled area 5v. Similarly, IPL neurons recorded in the caudal bank of the IPS near its anterior end were designated as in area AIP (Murata et al. 2000), whereas those in the adjacent lateral convexity of the IPL were designated as in area 7b (Cavada and Goldman-Rakic 1989a; Lewis and Van Essen 2000a); the latter region includes areas PF and PFG of Pandya and Selzer (1982).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Cortical recordings were made with the intent to maximize the number of neurons the firing patterns of which could be linked reliably to the actions of the hand during the prehension task. Tracks (294) were made in the cortices of the three animals as indicated in Fig. 3. The widest sampling of cortical areas was made in monkey B2195; recording sites in monkeys H17094 [GenBank] and N18588 were concentrated in posterior parietal cortex around the lateral IPS and in posterior S-I. This report describes neuronal firing patterns of 85 neurons in area 5d/5v and 43 cells recorded in area AIP/7b during the task. As responses to the four test objects were similar in time course, we pooled data from trials of the round and rectangular knobs for these analyses.

View larger version (13K): [in this window] [in a new window]   FIG. 3. Cortical recording sites in the 3 animals studied. Electrode entry points are shown as dots on each brain diagram; they were reconstructed from serial histological sections and dye markers placed at the chamber perimeter.

Figure 4 illustrates the kinematics of task performance captured during two successive trials for monkey H17094. [GenBank] In these examples, the trials began as the animal rested its hand on the chair frame and viewed the computer screen. The onset of reach coincided with or followed a saccade to the cued object. Approach was direct and rapid. Reaches had arc-like trajectories that spanned five to six video frames (167–200 ms) with peak velocity at the midpoint of travel. The fingers were preshaped for efficient grasp, and the hand simultaneously rotated downward so that the fingers contacted the side of the knob at the end of the reach (yellow). The animal grasped the knob in two frames (67 ms) by sliding digits 2–5 along its lateral face and pressing the base against the interdigital palm pads with the thumb on the top surface or parallel to the other fingers (orange). After rotation of the knob to a comfortable position, lift began. The knob was held above the box, as the animal consumed the juice reward and then was lowered and the grip relaxed. Note the common gestures used on the two trials.

View larger version (59K): [in this window] [in a new window]   FIG. 4. Kinematics of forward approach to the rectangular knob. Two trials with the same start and end position are illustrated (left and right). Note the similarity of reach kinematics and hand placement on the knob. Frames corresponding to the burst start, peak and end are labeled S, P, and E, respectively; maximum firing in these examples occurred as the hand was preshaped during approach.

View larger version (66K): [in this window] [in a new window]   FIG. 5. Burst analysis of neural responses to the actions shown in Fig. 4. These 2 neurons were recorded simultaneously on 1 electrode in area 5v in the left hemisphere. Task trials and stage onsets are indicated by stepped yellow pyramids; the knob location on the shape box and the duration of hand contact are marked by the orange trace. Other traces illustrate spontaneous actions of the 2 hands (red and blue) and gaze fixation on the target object (cyan, downward deflection). Both neurons reached peak firing rates during the early task stages as the hand approached and contacted the knobs (A–C); regrasps evoked weaker responses (D). The neurons also responded to an aborted reach to the large round knob that ended as he grasped of the chair frame instead (E). Video images to the right were captured at the peaks of bursts A, B, and C.

Regrasp of the same knob without a reach and preshape stage failed to produce the same high firing rates from either neuron. During burst D, the animal relaxed the grasp but did not break contact with the rectangle knob. Instead, as diagrammed in Fig. 6, he slid digits 4 and 5 under the knob and pushed it upward without fully grasping it. Neural responses were weaker than during bursts A–C, and only one of the two cells fired above threshold. Burst E occurred as the hand was withdrawn from the rectangle knob and projected laterally in the direction of the large round knob. However, the approach was spontaneously aborted, and the hand returned to the chair frame, which was approached and grasped in the same manner as the rectangular knobs. The neural response during burst E paralleled the kinematic sequence in the task, starting during hand withdrawal, peaking midway through reach, and ending prior to contacting the chair. Similar acquisition evoked activity was observed when the animal grasped other objects in the workspace such as food morsels (Babu et al. 2000).

View larger version (53K): [in this window] [in a new window]   FIG. 6. Actions performed during bursts D and E of Fig. 5. During burst D, the knob was not fully grasped but lifted instead with the digit tips. Because the hand was left in place on the knob at the end of trial A, there was no reach, and the accompanying neural response was weak. During burst E, the initial actions performed were identical to the start of a lateral approach between knobs, but the animal reached instead to the chair frame where the hand rested until the start of burst B. Neural responses during burst E were similar to task responses during reach.

View larger version (46K): [in this window] [in a new window]   FIG. 7. Rasters and peristimulus time histograms (PSTHs) aligned to contact with the knob for the neurons shown in Fig. 5. Colored lines on rasters and markers above the PSTH indicate the onset times of task stages relative to contact. PSTHs pool all trials of the four knobs. Only forward and lateral approach trials are shown in the raster. From top to bottom, trials 1, 7, 10, and 17 test knob 1 (left rectangle); trials 2, 3, 6, 11, 16, 18, and 19 test knob 2 (small round); trials 4, 5, 8, 13, 15, and 20 test knob 3 (right rectangle); trials 2, 9, 12, 14, and 21 test knob 4 (large round). All 4 knobs evoked similar responses: firing increased at the start of reach, peaked at contact, and declined during lift.

View larger version (51K): [in this window] [in a new window]   FIG. 8. Multiple-electrode responses recorded simultaneously from the right hemisphere in monkey N18588 as he used the left hand. Same format as Fig. 7. Neurons recorded on the same electrode (A and B) had very similar responses; spike trains recorded on different electrodes (C and D) had varied responses to identical hand actions. Knob 1 (left rectangle): trials 1–3, 8–11. Knob 2 (large round): trials 6, 7, 15, 16. Knob 4 (right rectangle): trials 4, 5, 12–14, 17–21. Knob 3 was not cued in the 1st 21 trials.

View larger version (35K): [in this window] [in a new window]   FIG. 9. Average firing rates per task stage (±SE) for the major response classes in area 5; responses were categorized by the stage(s) in which peak firing occurred. Stage 0 indicates mean firing rate during the pretrial interval. A–D: broadly tuned (type BT). Unit H17094-114-1.2 (A), vague response to passive wrist flexion. Unit H17094-70-3.2 (B), flexion of MCP joints and tactile receptive field on dorsum of proximal digit 3. Units N18588-313.2–3.1 (C) and N18588-313.4–2.1 (D), passive receptive fields were not determined. E and F: approach-tuned (type 1). Unit H17094-110-2.2 (E), receptive field on hairy and glabrous skin of digits, hand, and wrist. Unit N18588-94-7.3 (F), vague tactile receptive field on dorsal hairy skin of forearm, ulnar palm, and digit 5. G and H: approach-contact (type 1.5). Unit H17094-10-4.2 (G), receptive field on the dorsal hairy skin of the hand and digits 2–5. J–L: contact-tuned (type 2). Units H17094-131-3.1 (J) and 131–3.2 (K), tactile receptive fields located on the glabrous and hairy skin of the hand and digits. Unit N18588-313.1–2.1 (L), passive receptive field was not determined. M: contact-grasp (type 2.5). Unit N18588-313.2–2.1 (M), passive receptive field was not determined. Neurons recorded in area 5v are shown in E and F and J–K; the other cells were recorded in area 5d. Average rate graphs for the neurons in Figs. 7 and 8 are shown in J–K and D, H, L, and M, respectively.

View larger version (99K): [in this window] [in a new window]   FIG. 10. Burst analysis illustrating 16 s of continuous recordings from an anterior intraparietal area (AIP) neuron that responded strongly to task-related grasping actions. Video images captured at the start, peak, and end of bursts A, B, and G are shown below the neural records. The burst start coincided with the onset of reach regardless of the point of origin; responses peaked as the hand was preshaped over the knob, and ended when the knob was grasped. Task-evoked responses were bracketed by silent epochs when the animal engaged in spontaneous behaviors that failed to excite the neuron; red markers are keyed to examples shown in Fig. 11. The PSTH and average firing rate graph are shown in Fig. 12B.

View larger version (124K): [in this window] [in a new window]   FIG. 11. Images captured during epochs when the neuron in Fig. 10 fired at low rates or was silent (red A–C); other actions shown were recorded later in this clip. The cell was unresponsive to resting the glabrous hand on a flat surface (A) or to flexed hand postures unrelated to object acquisition (B, C, D, and F), or to striking the base of the shape box with the fist (E).

View larger version (47K): [in this window] [in a new window]   FIG. 12. PSTHs (left) and average firing rate graphs (right) from neurons recorded in area AIP of the 3 animals. Broadly tuned (BT) responses were more common than in area 5. A: unit H17094-157-2 recorded in the left hemisphere; receptive field not determined. B: unit B2195-121-2.2, receptive field located on glabrous skin of digits 2 and 3. Same neuron as in Figs. 11 and 12. C: unit N18588-315.3–4.1 recorded on electrode 3 of a multiple electrode array centered in area AIP of the right hemisphere; receptive field not determined. D: unit N18588-315.2–4.1 recorded on electrode 2 simultaneously with the neuron shown in C.

View larger version (95K): [in this window] [in a new window]   FIG. 13. Closeup images of the grasp postures used by monkeys H17094 (A and B) and N18588 (C and D) during trials of the small and large round knobs. Three sequential trials are shown for each knob. Neurons A–C were recorded in area 5, Neuron D in area AIP. Average firing rate graph for Neuron B shown in Fig. 9E; rasters and PSTHs for Neuron C are shown in Fig. 8.

Further quantification of task-related activity was obtained from measurements of average firing rates per stage on each trial. Sample mean firing rate graphs of hand manipulation neurons in area 5v/5d are shown in Fig. 9; the data illustrated are typical of 86% of SPL neurons and include responses of the neurons in Figs. 7 and 8 (J–K and D, H, L, and M, respectively). Epochs of high firing bridged the period from approach through lift but varied in relative intensity among individual neurons. All of the cells illustrated showed a significant rise in firing rate during approach, before the hand touched the objects (stage 1, P < 0.001). This activity was maintained or rose in intensity on tactile stimulation of the hand at contact; 9 of the 12 cells displayed the most intense firing at contact (stage 2). Elevated firing persisted during grasp and lift stages in most of these cells, but dropped precipitously during hold (stage 5) when hand movement ceased.

The intensity and specificity of firing during successive task actions was used to classify neuronal responses. Neurons were grouped by the stage(s) that evoked maximum firing and subdivided into classes tuned to single actions, two successive actions, or broadly tuned classes by statistical comparison of mean rates during sequential task stages. The distribution of response classes in the population studied is listed in Table 1. The most common type was broadly tuned, comprising 38% of the population. These neurons showed strong excitation during the task compared with baseline, but little distinction in firing rates between three or more successive actions in the preferred stages (Type BT, Fig. 9, A–D). 48% of neurons showed tuned activity focused on stages 1 and/or 2. The most commonly observed class was called contact-tuned (Type 2, Fig. 9, J–L) because its members fired at significantly higher rates during stage 2 than in the preceding approach stage or the following grasp stage (P < 0.05). PSTHs of contact-tuned neurons peaked midway through stage 2 (Fig. 7, 8A). Similarly, approach-tuned neurons (Type 1, Fig. 9, E and F) fired at higher rates during approach than in any other task stage; their PSTHs peaked before contact. Other area 5 neurons fired intensely during two successive stages, and were classified as approach-contact (Type 1.5, Fig. 9, G and H) or contact-grasp (Type 2.5, Fig. 9M). Mean firing rates did not differ significantly during these actions (P > 0.05), and their PSTHs peaked at the moment of contact (Fig. 8B) or grasp (Fig. 8C). Note that firing rates of the tuned neurons and dual-action cells were often significantly higher than baseline during other stages (P < 0.05) but failed to match the rates evoked during the preferred actions.

Responses to prehension in the adjacent hand representation of the IPL were similar to those recorded in area 5 of the SPL. As previously noted by Sakata and co-workers using a different grasp task, firing patterns of areas AIP and 7b neurons appeared to reflect the preparation and execution of grasping behaviors. We found that reaching, touching and grasping evoked stronger neuronal responses in IPL neurons than lifting, holding, and lowering the knobs. Each component stage of the task contributed to the evoked activity, as can be seen most clearly in burst analysis graphs.

Figure 10 illustrates 16 s of continuous recordings from an area AIP neuron in monkey B2195 as she alternated between task performance and other spontaneous hand actions. When not engaged in the task, the neuron was silent (red A, B, C). It fired strong bursts at rates >100 spikes/s on initiation of a task trial, particularly when the animal reached toward a knob from outside the immediate workspace of the shape box (blue A, G, H). The burst amplitude was high regardless of whether the hand started to reach from below (blue A) or above the knobs (blue G, H). Firing rates were independent of the shape of the target object. The bursts peaked as the hand was preshaped to grasp both round (blue A, G) and rectangular knobs (blue H) and ended when the hand touched and grasped them. Responses were stronger when the hand was properly aimed to the knob and grasp was completed (blue A, G) than when the hand fell between knobs and failed to secure one of them (blue H). In the latter case, the firing rate dropped abruptly, and the animal made a corrective lateral reach to the intended object accompanied by a second burst of impulses (blue I).

Lateral reaches between knobs evoked weaker bursts that began as grasp of one knob was relaxed, peaked during hand preshaping, and ended as the next object was contacted (blue B, I). Regrasp trials in which the animal did not relax the grasp evoked modest responses (blue C) as they lacked an approach or preshaping component, but trials in which the knob was released from grasp and the hand preshaped prior to regrasp on the next trial were quite strong (blue D, E).

The neuron was much less active when the animal engaged in behaviors other than object acquisition. Figure 11 shows actions that occurred during periods when the neuron was silent or fired sporadically. These included resting the hand on the base of the shape box (A), striking the base plate with the fist (E), or lifting the hand toward the face for inspection or grooming (B, C, D, and F). The neuron was sensitive to specific flexed hand postures only if they were used in the context of object acquisition. Indeed, even when the hand posture resembled that observed during preshaping (Fig. 11, B and C) or static grasp (D–F), the neuron remained silent if the goal of the action was something other than grasping objects. In this manner, the neuron signaled the coincidence of specific tactile and proprioceptive inputs with particular intentions.

Sensitivity to multiple task stages was also observed in the other two animals studied. Typical PSTHs and average firing rate graphs of area AIP neurons are shown in Fig. 12; similar responses were recorded in area 7b. AIP spike trains tended to last longer than those in area 5, and the task stages were less clearly distinguished. Task related activity typically began during stage 1, persisted through stage 3, when the object was fully secured in the grasp, and into stage 4 as the knob was lifted. Consequently, broadly-tuned neurons (Type BT) were the most common type observed in the population, comprising 51% of IPL neurons analyzed (Table 1). The broad sensitivity of these neurons to multiple task stages suggested that they participated in both the planning and execution of acquisition behaviors. Although the percentage of contact-tuned responses was similar to that observed in area 5, the other tuned and dual-action classes were less densely represented. In addition, 11% of IPL neurons were classified as grasp-inhibited (Type GI), because their activity was suppressed below baseline during task performance.

PPC responses are linked to motor schemas for grasping not hand postures

Although each animal developed personalized strategies and postures for grasping these objects, the hand kinematics of each one remained consistent both within a session and over the period of study. Our digital video methods allowed us to assess the detailed kinematics of prehension on a trial-by-trial basis. Images in Fig. 13, A and B, show the grasp postures used by monkey H17094 [GenBank] during three sequential trials recorded on tracks 10 and 110 spaced 6 mo apart. He grasped all of the knobs with similar postures, placing the hand on their lateral aspect with the radial surface upward. The round knobs were clasped tightly between the digits and palm along their lateral sides in a whole hand power grasp, but greater flexion occurred in trials testing the small round knob. The postures were nearly identical to those shown in Figs. 1 and 5 from track 131.

Monkey N18588 used a different hand posture to grasp the knobs, scooping them upward from below with the hand supinated or holding their shaft while pushing them upward (Fig. 13, C and D). He was somewhat clumsier than the other two animals, particularly when using his left hand, and showed greater variety in hand postures from trial to trial. Some of the intertrial variability may be related to the lack of visual guidance of his actions, as he rarely looked at the knobs before acquisition but instead focused on the computer monitor for cues or stared at other irrelevant targets.

Monkey B2195 used yet another grasp strategy. She aimed her hand toward the top surface of the knobs, made contact on the glabrous surface of the proximal phalanges, rather than on the digit tips, and used the heel of the palm to push the knobs upward (Fig. 10). As previously documented (Ro et al. 1998; Gardner et al. 1999), she used this overhand grasp posture throughout the 1-yr period of study.

Despite the different grasp styles used by these animals, the stage timing was roughly the same for all three monkeys (Table 2). Their task behavior remained within a narrow range from session to session; trial duration (approach through relax) ranged from 1.02 to 1.92 s in the three animals. The least variability in performance time occurred during acquisition stages that were crucial for reward. Static grasp was the shortest stage, spanning one to four frames as the animals transitioned rapidly from grasp to lift in this highly practiced behavior. Later stages were more variable in duration, from trial to trial and between individual monkeys because the principal action was consumption of the juice reward. Grasp usually was not released until licking ended.

Population responses to prehension in PPC

The consistent kinematic behaviors of each animal allowed us to compare neuronal behaviors between animals and across cortical areas over the period of study, and thereby measure population responses. Average firing rate graphs were used to quantify the distribution of preferred actions in the population of 128 hand manipulation neurons studied. Contact and hand positioning on the knob during stage 2 was the most strongly represented action in both the SPL and IPL populations (Fig. 14). 44% of area 5 neurons (37/85), and 28% of area AIP/7b neurons (12/43), fired maximally during stage 2. Another 28% of area 5 neurons (24/85), and 23% of AIP/7b cells (10/43), fired at highest rates during approach before the hand touched the knobs. In this manner, firing patterns correlated to the initial acquisition of objects were most salient in area 5, where 72% (61/85) of neurons fired at their highest rates during stages 1 and 2, and in AIP/7b, where 51% fired maximally. Static grasp was less effective than touching, particularly in area 5, where 16% of cells fired at peak rates in stage 3; more than half of these neurons were broadly tuned, meaning that their firing rates during contact and/or lift stages were not significantly lower. Similarly, while a higher percentage of AIP/7b neurons had peak activity during grasp (26%), nearly all of these cells were broadly tuned. Only 7% of hand manipulation neurons fired maximally during lift, but with only one exception, all were classified as broadly tuned. Holding was the least effective action; only 1 of 128 neurons responded maximally in stage 5.

View larger version (24K): [in this window] [in a new window]   FIG. 14. Stacked bar graphs of the number and classes of PPC neurons showing peak firing during each task stage. BT, contact-tuned (2), approach-contact (1.5), and approach-tuned (1) were the most common types. Neurons in both SPL and IPL were most likely to fire at peak rates in stage 2 (contact). Approach was the second most common period of peak activity in area 5. Stages 1 and 3 were equally common in area AIP/7b, reflecting the large number of BT neurons recorded there.

View larger version (19K): [in this window] [in a new window]   FIG. 15. Population normalized mean firing rates (±SE) averaged across the entire set of posterior parietal cortex (PPC) neurons tested. Normalized firing rates were computed for each neuron by dividing the mean rate per stage by the maximum value, and multiplying the resultant by 100. Like the peak rates, the mean normalized rate was highest in stage 2 in both areas. Mean firing rates were significantly higher than baseline (stage 0) during stages 1–4; significant inhibition occurred in stages 5–7 in area 5, and only in stage 6 in area AIP/7b.

View larger version (20K): [in this window] [in a new window]   FIG. 16. Bar graphs showing the percentage of neurons exhibiting significant excitation (gray bars) and inhibition (black bars) during the 7 task stages (P < 0.05). Excitation was strongest during stages 1 and 2 and decreased sharply following lift (stage 4); it persisted only in area AIP/7b. during later task stages. Inhibition began in stage 3 (static grasp), and peaked in stage 6 (lower) when the animal prepared to relax the grip.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
This report is the first study to directly quantify the firing patterns of hand manipulation neurons in both area 5 and AIP/7b of the same animals during prehension. Our digital video methodology allowed us to correlate neuronal firing patterns in the hand representation of PPC with skilled behaviors performed during a trained grasp-and-lift task. Hand actions in the task were divided into well-defined stages that encompassed reach, grasp, manipulation, and release of the test objects. We found marked similarities in responses of neurons recorded in the SPL and IPL during these behaviors. In both areas, object acquisition evoked the strongest responses. Eight-eight percent of area 5 neurons and 77% of AIP/7b neurons fired at their highest rates during at least one of the initial three stages of the task. In this period, the hand was brought to the object, touched, and grasped it. Although each of the animals we studied had a favorite method for grasping and holding the objects, their firing patterns followed a similar time course.

Neural activity in PPC began at or before the onset of reach. Eighty-three percent of task-related neurons in area 5 and 72% in AIP/7b were activated at the start of approach, showing significant increases in firing over baseline; similar proportions of neurons maintained elevated firing rates as the hand contacted the object and grasped it. Our kinematic analyses of the video images showed that as the hand was projected toward the object, it was rotated to a suitable orientation for efficient grasp, and the fingers opened to encompass the object in a smooth and rapid manner. Neural firing rates rose steadily during this period of hand preshaping, and typically peaked at contact. As discussed in the following text, we propose that activity during the approach stage reflects integration of visual information about the object, somatosensory information from the hand, and motor commands from frontal motor areas specifying the type of movement necessary to achieve the goal of grasping and manipulating that object.

Tactile contact with the object provided the strongest signal in the population. These high firing rates appeared to have both a sensory and a motor function. At the moment of contact, view of the object during reach was combined with the feel of the object, as the hand slid over the surface to grasp it. Object features such as surface curvature, edges, and texture were detected by mechanoreceptors in the hand and transmitted centrally to S-I cortex, and eventually to neurons in the hand representation of area 5, many of which had tactile receptive fields on the hand (see legend Fig. 9). This information provided feedback to the animal concerning the accuracy of the reach and helped guide the fingers to the preferred location(s) for grasping and subsequent manipulation. In cases where the animal missed the target, or conta