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: 95/1/144    most recent 00492.2005v1 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 Google Scholar Google Scholar Articles by Mason, C. R. Articles by Ebner, T. J. Search for Related Content PubMed PubMed Citation Articles by Mason, C. R. Articles by Ebner, T. J.

Purkinje Cells Signal Hand Shape and Grasp Force During Reach-to-Grasp in the Monkey

Department of Neuroscience, University of Minnesota, Minneapolis, Minnesota

Submitted 11 May 2005; accepted in final form 6 September 2005

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
The cerebellar cortex and nuclei play important roles in the learning, planning, and execution of reach-to-grasp and prehensile movements. However, few studies have investigated the signals carried by cerebellar neurons during reach-to-grasp, particularly signals relating to target object properties, hand shape, and grasp force. In this study, the simple spike discharge of 77 Purkinje cells was recorded as two rhesus monkeys reached and grasped 16 objects. The objects varied systematically in volume, shape, and orientation and each was grasped at five different force levels. Linear multiple regression analyses showed the simple spike discharge was significantly modulated in relation to objects and force levels. Object related modulation occurred preferentially during reach or early in the grasp and was linearly related to grasp aperture. The simple spike discharge was positively correlated with grasp force during both the reach and the grasp. There was no significant interaction between object and grasp force modulation, supporting previous kinematic findings that grasp kinematics and force are signaled independently. Singular value decomposition (SVD) was used to quantify the temporal patterns in the simple spike discharge. Most cells had a predominant discharge pattern that remained relatively constant across object grasp dimensions and force levels. A single predominant simple spike discharge pattern that spans reach and grasp and accounts for most of the variation (>60%) is consistent with the concept that the cerebellum is involved with synergies underlying prehension. Therefore Purkinje cells are involved with the signaling of prehension, providing independent signals for hand shaping and grasp force.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Lesion, imaging, and electrophysiological evidence show that the cerebellum is involved in the signaling of hand kinematics and grasp force during prehension (i.e., the act of gripping/hand use). Cerebellar patients exhibit a spectrum of kinematic impairments in the learning, planning, and execution of prehensile movements (Lang and Bastian 1999; Rand et al. 2000; Timmann et al. 1999, 2001; Zackowski et al. 2002). The deficits in reach-to-grasp include a breakdown in the coordination between transport and grasp components (Rand et al. 2000; Zackowski et al. 2002). Prehensile movements involving multijoint coordination, such as throwing and catching, exhibit impairments in timing (Timmann et al. 1999), variability in positioning (Timmann et al. 2001), and adaptation (Lang and Bastian 1999).

Various cerebellar structures have been implicated in the kinematics of reach and grasp. Reach-related activity, including the encoding of movement direction and amplitude, has been found in the cerebellar cortex and in both interpositus and dentate nuclei (Fortier et al. 1989; Fu et al. 1997; van Kan et al. 1993a). Single unit recordings have also implicated the interpositus nucleus in hand control during direct and indirect visual guidance of a manipulandum (Gibson et al. 1996; van Kan et al. 1994). The greatest modulations in interpositus neurons occurred when monkeys performed free reach movements to retrieve a raisin. Hand-related discharge has also been observed in the cerebellar cortical and dentate neurons in relation to grasp (Smith et al. 1993). Imaging studies have implicated the vermis, dentate nucleus, and intermediate and lateral cerebellar cortex during hand-shaping without a target (Colebatch et al. 1991; Takasawa et al. 2003). Regions of the nucleus interpositus anterior and posterior contribute differentially to reach-to-grasp, with the former involved in hand function and the latter with reach (Mason et al. 1998). Therefore cerebellar involvement in the kinematics of reaching and grasping has been validated across a range of methodologies.

The kinematic parameters, related specifically to reach-to-grasp, signaled in Purkinje cell simple spike discharge have not been well defined. Earlier studies concentrating on the discharge of nuclear neurons found strong involvement during coordinated hand and arm movements, especially for unconstrained reach-to-grasp movements (Goodkin and Thach 2003; van Kan et al. 1993b). Little is known, however, about grasp-related signals carried by cerebellar cortical neurons. There are no studies examining what aspects of hand-shaping are signaled by Purkinje cells during reach-to-grasp. The first aim of this study was to examine this question by having animals reach to and grasp an array of objects that varied in shape, size, and orientation.

The cerebellum is also involved in prehensile force control. Cerebellar patients have pronounced deficits in manual force control (Mai et al. 1988; Muller and Dichgans 1994a; Serrien and Wiesendanger 2000; Timmann et al. 2001). While the generation of peak force seems unaffected by chronic cerebellar lesions (Mai et al. 1988; Muller and Dichgans 1994a), tasks requiring the maintenance of isometric grip force (Mai et al. 1988) or changes in force production over time are chronically impaired (Serrien and Wiesendanger 1999; Timmann et al. 1999, 2001). Cerebellar patients also show impaired generation and adaptation of anticipatory force control (Lang and Bastian 1999; Serrien and Wiesendanger 1999). A patient's inability to control predictive grip forces often results in compensatory strategies involving unnecessarily elevated and uneconomical grip force levels (Fellows et al. 2001; Muller and Dichgans 1994b; Nowak et al. 2002; Serrien and Wiesendanger 1999).

Specific cerebellar structures and regions have been associated with different aspects of prehensile force control. Imaging (Kinoshita et al. 2000) and electrophysiological studies (Espinoza and Smith 1990) have shown positive correlations between the cell discharge occurring in lateral regions of the cerebellum with load-force requirements. Activation related to the coordination of grip-force/load-force coupling occurs in the ipsilateral anterior and superior cerebellum as well as the contralateral biventer lobule (Kawato et al. 2003). Preparatory increases in grip-force to predictable load-force perturbations are signaled in the discharge of paravermal and hemispheric cerebellar neurons (Dugas and Smith 1992) and in the dorsal anterior interpositus near, but not in, the dentate nucleus (Monzee and Smith 2004). The cerebellum's involvement may primarily be in active force discrimination, presumably related to volitional movement and feedback mechanisms, rather than passive force discrimination (e.g., static sustained pressure) at the fingertips (Bodegård et al. 2003).

The evidence for cerebellar involvement in force control of prehension leads to the prediction that cerebellar neurons are modulated by grasp force. Although cerebellar discharge modulates in relation to force control during pinch and lift tasks (Espinoza and Smith 1990; Smith and Bourbonnais 1981), no study of Purkinje cell firing and grasping has systematically varied whole hand grasp force across multiple levels and hand configurations. Understanding the relationship between hand-shaping and force production is needed because each plays a pivotal role in prehensile movements. Therefore this study examined the discharge of Purkinje cells during reach-to-grasp with systematic variation of grasp force as well as hand-shape.

The cerebellum may play a major role the coordination of complex movements (Bastian et al. 1996; Thach et al. 1992). In both humans and monkeys, kinematic synergies have been offered as a means of simplifying the control and coordination of hand movements (Mason et al. 2001; Santello et al. 2002; Soechting and Flanders 1997). Using dimensionality reduction techniques to examine the hand kinematics, only a few dominant hand postures were needed to explain most of the variability in hand-shaping (Mason et al. 2001, 2004; Santello and Soechting 1997; Santello et al. 2002). Furthermore, the amplitude of the dominant principle component remained relatively constant (i.e., a default posture) while the amplitude of secondary component scaled linearly with the object size (Santello and Soechting 1997). If hand-shaping is controlled by scaling a small number of hand postures, one would predict that the neuronal representation might exhibit analogous features such as a few dominant discharge patterns across task conditions. In addition to hand-shaping, data reduction techniques applied to multidigit forces and moments produce similar evidence of force synergies during grasp (Zatsiorsky et al. 2003). We hypothesize that the signaling of such synergies may be present in the discharge of cerebellar cortical neurons. Therefore this study applied one of these techniques (i.e., singular value decomposition) to examine whether the discharge of Purkinje cells was consistent with the synergy concept.

In this study, rhesus monkeys were trained to reach to and grasp a set of objects with explicit grasp force requirements. A detailed kinematic analysis of the wrist, hand, and finger movements during this task was published recently (Mason et al. 2004). Hand shaping began with the initiation of reach and continued throughout the transport phase. Furthermore, the hand shaping throughout the reach matched object properties (e.g., size) but showed no significant relation to the grasp force levels, leading to the additional hypothesis that the signaling of hand-shaping and grasp force are relatively independent. The results show how hand shape and grasp force are signaled in the simple spike discharge of Purkinje cells.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
The experimental protocol was approved and monitored by the University of Minnesota Institutional Animal Care and Use Committee and conformed to the "Guiding Principles in the Care and Use of Animals" of the American Physiological Society.

Two rhesus monkeys (1 female "G" at 5.2 kg, 1 male "L" at 6.8 kg) were trained to reach and grasp objects with an overhand grasp using specified force levels (Ojakangas and Ebner 1992). The animals sat in a primate chair with their heads fixed forward and facing a computer monitor (Fig. 1A). The animals initiated a trial by placing their hand on a start-pad located by their side. A "go" cue signaled the animals to reach (15 cm) and grasp the target object. The go cue also indicated the required grasp force level to be generated and maintained for that trial. As a result, the monkeys had a priori knowledge of the required force level before initiating the reach. Once the grasp was initiated, a red slider bar provided visual feedback of the grasp force being generated. If the monkeys successfully initiated and maintained the specified force level for 1.4 s, they received a juice reward. At the completion of 25 successful trials, the object was changed. The monkeys were not able to see their hands or the objects. However, before initiating the first trial of each block, the animals were allowed to identify the target object through touch. Therefore the animals were aware of the target object to be used before the initiation of a new block of trials.

View larger version (41K): [in this window] [in a new window]   FIG. 1. Reach-to-grasp paradigm and object classification. A: reach-to-grasp task time-line (top) and the corresponding computer monitor display sequence (bottom). The monkey began with its hand on the start-pad while viewing a blank monitor (left). After a variable delay, a large red rectangle on the right side of the monitor provided a "go" cue to initiate the reach-to-grasp (middle). Simultaneously, blue bars signaled the level of grasp force required at the target object. Once the grasp was initiated and force was applied to the object, the central red slider provided feedback on the exerted grasp force (right). The task epochs were 1) an initial start-pad hold segment varying from 1 to 1.5 s, 2) a reach segment beginning on movement onset and ending with grasp initiation (i.e., force onset on target object), and 3) a grasp segment beginning with grasp initiation followed by an object hold period (1.4 s). B: objects were grouped into 4 object classes (cubes, rectangular prisms, poly-sided prisms, and cylinders) and were presented with the x-y plane parallel to the frontal plane of the monkey. Grasp dimensions, measured along the z-axis (e.g., as shown by the white line segment on object 6), were 1, 2, 2.8, 3, 3.3, 4, or 4.5 cm.

The animals were required to exert five different levels of anterior–posterior (AP) grasp force (0.2, 0.4, 0.6, 0.8, and 1.0 N) with a tolerance range of ±0.1 N. The narrow range of the force windows, in addition to the rigorous behavioral training, required the monkeys to grasp and not pull on the object to control grip force. The five repetitions of the five forces were presented pseudo-randomly for each object with the requirement that the monkey successfully complete a trial before a new force window was introduced. A Force Sensing Resistor (FSR; Interlink Electronics, Camarillo, CA), 1.27 cm in diameter placed on the face of each object opposite the monkey, was used to detect forces applied to the object. The 5 x 5-cm start-pad was located by the monkey's side and had a 3.8 x 3.8-cm FSR on the superior surface to detect hand contact. Behavioral data collected included the force generated on the target object as well as reach and grasp timing information. Although both animals were highly trained and proficient at the task for objects and force levels, monkey L did not perform consistently on the smallest cube (object 1), and data on this object were not included for monkey L.

Surgical preparation and recording procedures

After the monkeys achieved an acceptable level of competency in task performance, they underwent an aseptic surgical procedure in preparation for cerebellar single cell recordings. The monkeys were anesthetized with a ketamine/xylazine (7/0.6 mg/kg, im) combination, supplemented as needed. An 18-mm circular chamber was placed over the parietal cortex targeting the intermediate zone of lobules V and VI. Hand- and arm-related Purkinje cell activity has been documented in these areas (Bauswein et al. 1983; Dugas and Smith 1992; Fortier et al. 1989; Fu et al. 1997). Furthermore, the intermediate zone projects to the interposed nuclei and these nuclei are involved with reach-to-grasp (Chapman et al. 1986; Mason et al. 1998; van Kan et al. 1994). The chambers were placed ipsilateral to the working hand. The stereotaxic coordinates of the center of the chambers for monkey G were 6.0 mm (both left and right sides) posterior to interaural zero and 10 (left side) or 8.0 mm (right side) lateral to the midline. The chamber placement for monkey L was on the right side using similar coordinates.

Parylene-coated tungsten microelectrodes were inserted vertically using a microdrive with an X-Y micropositioner mounted directly to the chamber. During recording sessions, the point of penetration, the depth of penetration, and transitional cell activities throughout the electrode penetration were monitored to determine the location of cerebellar Purkinje cell activity. Purkinje cell activity was identified by the presence of both simple and complex spike activity (Fu et al. 1997; Ojakangas and Ebner 1992). The cells were tested for the presence of peripheral receptive fields. Proprioceptive responses were identified by passively moving limb segments and fingers about each joint in isolation and noting changes in cell activity during the imposed movements. In addition, cutaneous receptive fields were also tested by stroking, lightly probing, and/or applying air puffs to the hand and arm.

Histological analysis

Small electrolytic lesions were made to further identify the recording areas after the recordings were completed. For each chamber placement, five lesions were made: one at the origin of the chamber and four at the extremes of each axis (Fu et al. 1997; Ojakangas and Ebner 1992). Each animal was initially anesthetized with ketamine (20 mg/kg, im) and xylazine (0.4 mg/kg, im), followed by a lethal dose of pentobarbital sodium (150 mg/kg, ip). Intracardiac perfusion with saline containing heparin was followed by perfusion with Zamboni's fixative (150 ml aqueous picric acid, 250 ml 8% paraformaldehyde, 2.81 g NaH₂PO₄ · H₂O, and 28.70 g Na₂HPO₄ · 7 H₂O, filled to remaining volume with distilled H₂O to 1.0 liter, titrate to pH 7.3). After removal of the brain, the cerebellar hemispheres were cut sagittally in 50-µm frozen sections on a microtome and stained with thionin. Recording tracks and electrolytic lesions were identified in the cut sections to localize cell recording sites.

Grasp force and movement kinematics

A critical issue is determining whether changes in the cell firing were related to changes in the reach as opposed to the grasp. An earlier analysis of the reach-to-grasp behavior for the same animals performing this task showed that reach varied only minimally with the different objects and grasp force levels (Mason et al. 2004). In this study, additional analysis of the magnitude and timing of the maximum grasp aperture and peak arm velocity were included to further quantify reach and grasp kinematics as a function of the objects and forces. The details of the data collection from the force sensors as well as the video-based tracking system used for the collection of the kinematic data were recently published (Mason et al. 2004). Monkey L kinematic recording sessions did not include the smallest cube (object 1). The kinematic recordings were performed in different experimental sessions than the Purkinje cell recordings for both monkeys.

Simple spike data analysis

The action potentials of the simple and complex spikes of isolated Purkinje cells were discriminated using a time-amplitude discriminator. The signals were digitized and stored at a resolution of 1 kHz during each recording session. The digitized signal was binned in 20-ms increments during off-line data processing.

All analyses of the Purkinje cell discharge were restricted to the simple spike firing and were based on the discharge during single trials. The complex spike activity was not analyzed in this study. The first step involved identifying in each individual trial the task related epochs. Using the start-pad force sensor, reach onset was determined as the time after which a peak force deceleration (lift-off) was detected, and the sensor reading dropped below 0.3 N. The reach segment ended, and the initiation of grasp began when the force sensor on the target was first activated. The average reach time durations were 542 ± 278 and 490 ± 106 ms for monkeys G and L, respectively. The analysis of the grasp segment was limited to the initial 1,400 ms of grasp initiation/object hold.

The simple spike firing of each trial was "normalized" by subtracting the averaged baseline firing rate for that trial from each time bin. Normalized firing provided for a measure of the change in firing with the task parameters independent of the background discharge. Baseline recordings began with the onset of each trial and ended 300 ms before movement onset to account for the subjects' reaction times. Intertrial intervals varied in duration across trials; therefore the periods over which baseline firing rates were collected also varied from trial to trial. The average baseline durations were 1,020 and 1,300 ms and firing rates were 32 and 23 spikes/s for monkeys G and L, respectively. The normalized firing rate was used for the regression and singular value decomposition (SVD) analyses described below.

Paired t-test relative to baseline

The aim of the first analysis was to determine whether a Purkinje cell simple spike discharge was significantly modulated during the task. A paired t-test was used to determine if the average simple spike firing in the reach and grasp segments was significantly different from the baseline firing rate of individual trials. The t-values ( = 0.05) were computed for each cell. The t-test also provided information about the number of Purkinje cells with increasing (positive values) versus decreasing (negative values) simple spike activity relative to baseline firing.

ANOVA of normalized firing rates during epochs

The second analysis determined whether the Purkinje cell simple spike discharge was significantly modulated in relation to the two task variables, objects and grasp force. Therefore ANOVA of the normalized firing rates of individual trials were computed for each cell, with object and force level as treatment factors and = 0.05. This was done for both the reach and grasp segments, which are generally accepted as the two components of prehension (Jeannerod 1981). The epoch-based analysis constitutes the first step in determining how Purkinje cell simple spike discharge was modulated in relation to the task variables and during what epochs.

The objects grasped resulted in significant differences in hand-shaping parameters, such as hand orientation, shape, and finger spreading across objects (Mason et al. 2004). One parameter of hand shaping, grasp aperture (e.g., the distance between the thumb and middle finger), was found to vary linearly with the object grasp dimension. Similarly, the object grasp dimension is a convenient method of reducing the various object properties to a single continuous variable. An ANOVA of the normalized firing rates as a function of the object grasp dimension was also performed in which the 16 objects were grouped by a grasp dimension of 1, 2, 2.8, 3, 3.3, 4, or 4.5 cm (Fig. 1B).

If an experimental variable (i.e., object, grasp dimension, or grasp force) was associated with a significant F-value, a post hoc analysis (Bonferoni t-test with adjusted P value) was performed on all levels of the variable. For significant object-related cell modulation, the post hoc analyses resulted in object groupings in which objects falling within a grouping did not have significantly different firing rates from each other but could have firing rates that were significantly different from objects outside their group. Consequently, certain objects (usually extreme values) could be included within several object groupings. The reporting of object groupings was necessitated because the object numbers do not represent quantitative ordered values of a variable. In contrast, grasp force (N) and grasp aperture (mm) did represent continuous variables, and a linear regression analysis was also performed as a post hoc analysis on these parameters. While the epoch-based ANOVA provided evidence of significant differences in discharge across the independent variables, the post hoc analyses provided insight into the nature of this relationship (e.g., whether linear). Performing one type of analysis in the absence of the other may lead to inappropriate conclusions about the relationships examined; for example, a relationship could be significant but not necessarily linear (Howell 1987). The linear regression was considered significant if the change in the regression slope was greater than the overall variance (noise) and the coefficient of determination (R²) explained 50% of the total variance.

Multiple regression analysis

The epoch-based analysis described above showed that the simple spike discharge was significantly modulated by objects and grasp force. However, the durations of the epochs were relatively long, warranting more precise timing information on the relation of the cell firing to the task parameters. To determine the times at which the simple spike discharge was correlated with various objects and grasp force levels, a temporal multiple regression analysis was used. The temporal regression evaluated the firing in each time bin in relation to the objects and grasp force (Eq. 1). A similar approach has been used during a reaching task to examine the timing of direction and distance correlations occurring during reaching (Fu et al. 1995, 1997; Johnson and Ebner 2000). This method is equivalent to performing an ANOVA on the averaged firing rate for each 20-ms time bin rather than the entire reach or grasp segment as done with the epoch analysis. The firing, F, in each 20-ms bin (t_i) in an individual trial was regressed to the variables

(1)

where n is the number of objects completed, FL is the grasp force level the monkeys needed to maintain during the task, and OB represents the individual objects. Both force levels and individual objects were coded as dummy variables, variables that do not necessitate quantitative ordered values for the analysis. Total and partial R² values were calculated using a multiple regression analysis for a factorial experiment with two factors (Edwards 1979) and tested for significance at = 0.05. A time bin was considered significant only if the model or partial R² remained significant for at least five consecutive time bins (i.e., 100 ms). The multiple regression analysis was performed for each individual cell. In addition, the percentage of cells with significant regression results was determined for each time bin to capture the overall trend in significant R² across the individual cells.

Singular value decomposition

Examination of the simple spike firing suggested that the temporal pattern of each Purkinje cell discharge was relatively stereotypic, with the amplitude of firing increasing or decreasing with objects and/or grasp force. To verify and quantify this impression, an SVD analysis (Hendler and Shrager 1994) was used to detect patterns of Purkinje simple spike discharge during reach-to-grasp across objects and force levels. SVD is a more generalized method of a principle component analysis (PCA). PCA is equivalent to performing an SVD on the covariance matrix of the data set, whereas SVD is performed on the actual data. Our main reason for using SVD was that the methodology provides a temporal weighting of the eigenvalues that allows a determination of how a firing pattern changes over time. Also, the SVD was used in a previous analysis of the hand kinematics (Mason et al. 2004).

The calculation of the SVD was based on matrix X, where each column of data represented the mean Purkinje cell simple spike firing rate over time (20-ms bins from –800 to 1,200 ms) averaged across all repetitions of each object-by-force treatment combination (i.e., 80 columns = 16 objects x 5 force levels). Trial data were first filtered using a 9-Hz low-pass Butterworth ninth-order filter before averaging, because emphasis was placed on identifying and statistically evaluating lower frequency response patterns (e.g., <9 Hz). Matrix X was deconvolved into three matrixes using X = UV^T. The diagonal matrix contained the resulting eigenvalues, in greatest-to-least rank order. The relative variance explained by each eigenvector was obtained by squaring each eigenvalue and dividing by the sum of squares of all eigenvalues. The columns of matrix U contained the temporal weightings for each eigenvalue. Matrix V contained score values and the superscript T denotes the transpose. The reduced eigenvalues (REVs) were computed for each eigenvalue to determine if the factor was random (i.e., noise) or a significant pattern component (Elbergali et al. 1999). SVD calculations were also performed across grasp force levels and object grasp dimensions as a function of time. If stereotypic patterns are represented in the discharge of Purkinje cells, one would expect a dominant discharge pattern(s) remaining largely invariant across experimental conditions with secondary patterns regulating the signal across object and force conditions, as observed psychophysically (Santello and Soechting 1997).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Kinematics of reach and hand-shaping

Relating the modulation in the simple spike discharge to the objects, hand shape, or grasp forces requires that the reach remain relatively constant. An examination of the speed and time parameters of the reach confirmed invariant reach kinematics. For monkey L, for example, reach speed was similar across grasp force levels and objects (Fig. 2, A and B, respectively). Specifically, the time of peak speed was constant across both grasp force [F(4,1248) = 0.68, P > 0.05] and object grasped [F(14,1248) = 1.17, P > 0.05; Fig. 2A and B, respectively, gray bars]. In addition, peak speed was constant across force levels [F(4,1248) = 0.10, P > 0.05; Fig. 2A, white bars]. Significant differences in peak speed were noted across the objects grasped [F(14,1248) = 11.71, P < 0.05; Fig. 2B]. Similar results were noted for monkey G, with no significant differences in the time of peak speed [F(4,121) = 1, P > 0.05 and F(10,121) = 1.55, P > 0.05 for force and object variables, respectively]. Also, peak speed did not vary with force [F(4,121) = 0.75, P > 0.05] but varied with object [F(10,121) = 5.78, P < 0.05]. Although significant differences in the magnitude of peak speed were noted across objects, three points need to be stressed. First, the magnitude of the change in peak speed as a function of objects was quite small, with maximum variation of only 2.7 cm/s (5.5% of the actual peak). Second, the variation in peak speed across objects accounted for only 15% of the total variance in peak speed within a recording session (i.e., R² = 0.15). Third, variations in peak speed with objects were not consistent across all sessions. Additional measures of reach kinematics, including reach trajectory, also support the finding that reach was similar across the experimental conditions (Mason et al. 2004).

View larger version (42K): [in this window] [in a new window]   FIG. 2. Reach kinematics remained invariant across grasp force levels and objects. A: plots of reach speed (i.e., forearm displacement) as a function of time showed similar traces across grasp force levels. Bar plots and SD of the mean time to peak speed (gray bars) and magnitudes (white bars) revealed no differences across grasp force levels. B: similar plots of reach speeds across objects. Although results indicated a significant difference of the magnitude of the peak speed across objects [F(14,1248) = 11.71, P < 0.05], the bar plot shows that effect size was small. Data in A and B are from individual trials for 1 experimental session for monkey L.

View larger version (38K): [in this window] [in a new window]   FIG. 3. Hand-shaping kinematics remained invariant across grasp force levels but varied with objects grasped. A: on top are 4 snapshots of whole hand shaping during the task for 3 representative objects: objects 1 (blue), 5 (green), and 12 (red). Yellow spheres represent marker positions used to track the primate's hand movement during the task. Each hand configuration is the average of all trials for the middle grasp force level (0.6 N) and for each object represented. Bottom plots of grasp apertures as a function of time show variations in aperture with objects (middle: trace colors match object colors) but not grasp force. In both sets of plots, each trace represents an individual trial for objects 1, 5, and 12 and the low, medium, and high grasp force levels. B: bar plots and SD of mean times (gray bars) and magnitudes (white bars) of peak grasp aperture revealed no differences across grasp force levels. C: similar plots revealed peak grasp apertures varied with objects grasped. Plots in A for monkey G and bar graphs in B and C for monkey L are each based on individual trials from 1 experimental session.

Data base and simple spike firing patterns

Recording sessions in which 8 or more of the 16 objects were successfully completed for all five force levels with at least five repetitions per treatment combination (e.g., a minimum of 200 trials per cell) were used in this study. In addition, for a cell to be included in the study, its simple spike activity during the reach and/or grasp segment had to be significantly different from the baseline activity (paired t-test). A total of 40 Purkinje cells were analyzed for monkey G and 37 for monkey L, for a total of 77 cells. The majority of the Purkinje cell receptive fields identified were proprioceptive (78%), with fewer cells having cutaneous or cutaneous/proprioceptive fields (22%). Cells had receptive fields generally relating to the hand (88%), forearm (11%), and/or shoulder areas (1%). The histology results indicated that cell recordings for both monkeys were centered on the intermediate zone of lobules IV–V of the anterior lobe (Fig. 4).

View larger version (50K): [in this window] [in a new window]   FIG. 4. Histology results. A: left and right chamber placements for monkey G indicating the penetration sites of the 40 cells used in the study. B: right chamber placement for monkey L with penetration sites of the 37 cells.

An example of the simple spike firing pattern for an individual cell, G104, across representative objects from each object class and across all grasp force levels is shown in Fig. 5. Individual trial firing rates are represented as color plots to the right of each histogram. Several aspects of the discharge modulations were noteworthy. First, the overall pattern of modulation was similar across objects and force levels with a significant increase in the discharge rate during both the reach and grasp segments compared with the baseline (paired t-test: t = 14.46, P < 0.05, mean 67 spikes/s and t = 9.87, P < 0.05, mean 28 spikes/s, respectively). Cell modulation increased around reach onset (550 ms before force onset), obtaining a peak firing rate just before the grasp initiation (time 0). The firing rate decreased as the grasp was initiated and reached a steady, although slightly decreasing, level of modulation during the object hold period.

View larger version (76K): [in this window] [in a new window]   FIG. 5. Purkinje cell (G104) simple spike firing histograms and grasp force profiles for representative objects from the cube (object 4), rectangular prism (object 9), poly-sided prism (object 12), and cylinder (object 16) class sets. Histograms in this and subsequent figures represent the averaged firing rate across all 5 repetitions for the object at 5 force levels (left column). Histograms are ordered with the lowest force (0.2 N) on the top and highest force (1.0 N) on the bottom. Grasp force profiles are located above histograms and are the average across all repetitions of each force level. Histograms and force profiles were aligned on grasp initiation (time 0). Negative time represents the start-pad hold and reach segment. Positive time represents the grasp segment (e.g., grasp initiation and target hold). Each time line in this and other figures have the same time markers as well as a vertical dashed line representing the average time at which movement onset occurred. SD of movement onset is indicated by a horizontal bar at the base of the dashed line. Color plots show individual trial firing rates ordered with the repetitions of the lowest force (0.2 N) on the top and repetitions of the highest force (1.0 N) at the bottom.

Cell modulation varied with object

The simple spike discharge for the vast majority of Purkinje cells had significant differences in firing across objects during the reach (36/40 or 90% for G and 32/37 or 86% for L) and grasp segments (34/40 or 85% for G and 30/37 or 81% for L). The histograms for cell L139 show such modulation differences across objects 4 and 5 with the least change in normalized firing rate (2 ± 7 and 1 ± 6 spikes/s during grasp, respectively) and objects 8 and 1 with the largest change (17 ± 10 and 21 ± 16 spikes/s during reach, respectively; Fig. 6A). For this cell, the normalized firing rate during both reach and grasp segments were significantly different across objects [F(14,315) = 7.11, P < 0.05 and F(14,315) = 4.19, P < 0.05, respectively]. The two groupings of objects for the reach segment, having the highest (objects 1 and 8 in descending order) and the lowest (objects 4, 2, 16, 6, 7, 5, 15, 12, and 14 in ascending order) normalized firing rates, were identified (Fig. 6B, left). Two similar groupings were also identified during the grasp segment (Fig. 6B, right). Although firing differed significantly across objects during both reach and grasp segments, the changes in firing were greater during the reach segment (e.g., significant object groupings had less overlap). The greater differences in firing during reach was consistent for all cells with significant object related modulation during reach and grasp segments. This is shown (Fig. 6C) by a second example (cell G125) for which the normalized firing rate was significantly different across objects during the reach [F(9,209) = 7.04, P < 0.05] but not during grasp [F(9,209) = 1.80, P > 0.05]. The objects grouped with the highest mean firing included 8, 2, 9, 13, and 3 (in descending order) and those with the lowest mean firing included 14, 5, 11, 12, 4, and 3 (in descending order).

View larger version (46K): [in this window] [in a new window]   FIG. 6. Simple spike modulation with object. A: histograms for objects with the 2 lowest means (objects 4 and 5—grasp) and 2 highest means (objects 8 and 1) show object-related modulation for cell L139. B: normalized firing rates and SE across objects show significant object-related modulation during reach and grasp [F(14,315) = 7.11, P < 0.05 and F(14,315) = 4.19, P < 0.05, respectively]. C: for cell G125, normalized firing rates and SE across objects during reach and grasp segments show significant object-related modulation during reach [F(9,209) = 7.04, P < 0.05] but not grasp [F(9,209) = 1.80, P > 0.05]. Highest and lowest object groupings generated by a post hoc multiple comparison t-test for the reach segment are indicated by * and x, respectively, in this and other figures showing groupings. Objects falling within a grouping did not have significantly different firing rates from each other but could have firing rates that were significantly different from objects outside their grouping.

Cell modulation with object grasp dimension

The next question was whether differences in simple spike discharge modulation as a function of the objects reflected hand shaping properties such as grasp aperture. Grasp aperture is a key measure of hand shape that also reflects the dimension of the object to be grasped (Jeannerod 1991; Jeannerod et al. 1995; Mason et al. 2004; Muir and Lemon 1983; Roy et al. 2002). Examples of two Purkinje cells with significant grasp dimension related modulations are shown in Fig. 7. The simple spike discharge of cell G28 (Fig. 7, A and B) showed significant differences in firing with grasp dimension for both reach and grasp segments [F(6,425) = 12.06, P < 0.05 and F(6,425) = 10.23, P < 0.05, respectively]. The discharge decreased linearly with grasp dimension during both reach (R² = 0.62, P < 0.05) and grasp segments (R² = 0.64, P < 0.05). In contrast, the simple spike discharge of cell L152 (Fig. 7, C and D) was characterized by a linearly increasing relationship between firing and grasp dimension (R² = 0.78, P < 0.05). During the grasp, the cell firing rate also varied significantly with the grasp dimension [F(6,433) = 2.1, P < 0.05] but the relation was not linear (R² = 0.002, P > 0.05).

View larger version (29K): [in this window] [in a new window]   FIG. 7. Simple spike modulation with object grasp dimension. A and B: simple spike firing of a cell (G28) during both reach (A) and grasp (B) decreased linearly with object grasp dimension. C and D: for cell L152, the simple spike firing increased linearly with object grasp dimension during reach (C) but not during grasp (D). Each data point is the mean normalized firing rate and SE. Dashed lines are the resultant regression line having equation parameters and R2 values indicated within each plot of this and consecutive figures.

Cell modulation with grasp force

The task design explicitly manipulated the magnitude of grasp force during the task. The simple spike discharge was found to be significantly modulated with grasp force in 26% (20/77) of the Purkinje cells during reach (15/40 for G and 5/37 for L) and in 51% (39/77) during the grasp (22/40 for G and 17/37 for L). Cell G117, for example, showed an increase in simple spike discharge with grasp force as shown for selected objects (3, 6, 8, and 12) in Fig. 8A. Plots of the normalized firing rates across force levels show a monotonic and linearly increasing relationship (Fig. 8B). The modulation of simple spike firing with force was significant for both reach [F(4,168) = 9.43, P < 0.05] and grasp [F(4,168) = 10.80, P < 0.05] and the relationship was linear and positive (R² = 0.96, P < 0.05 and R² = 0.80, P < 0.05 for reach and grasp, respectively). In a second example (Fig. 8C), cell G54 showed force related modulation confined to the grasp [F(4,357) = 2.46, P < 0.05] and not the reach [F(4,357) = 0.49, P > 0.05]. During the grasp, similarly defined relations in simple spike firing were also noted (i.e., increasing monotonically and linearly with force, R² = 0.97, P < 0.05). Simple spike firing increased with increasing force levels (e.g., positive slope) in all but two cells (2/39 cells during grasp).

View larger version (45K): [in this window] [in a new window]   FIG. 8. Simple spike modulation with grasp force. A: consistent increases in peak firing rates (G117) with grasp force are shown from top (0.2 N) to bottom (1.0 N) in the histograms of 4 representative objects (3, 6, 8, and 12). B: normalized firing rates and SE across force levels during the reach and grasp segments were positively correlated with grasp force during both reach (R2 = 0.96) and grasp (R2 = 0.80) segments. C: normalized firing rate (G54) across force levels was significant and linear during grasp but not reach. Normalized firing rates across force levels for this and consecutive figures are averaged across all objects.

Timing of the object and force modulations

The epoch-based analyses show that Purkinje cell simple spike discharge modulated in relation to objects, grasp dimension, and grasp force. However, the firing histograms indicated that the cell firing changed within epochs as well as during the transitions between epochs (Figs. 5, 6, and 8). Furthermore, briefer periods of cell modulation were certainly masked by averaging the firing rate over the entire grasp segment (1,400 ms). Therefore a linear multiple regression model (Eq. 1) relating cell firing to object and force variables over time was used to more precisely define the timing of the signals.

An example of the temporal regression analysis for cell L145 showing the model and partial R² profiles and associated significance levels is shown in Fig. 9, A–E. The model R² first peaked at the time of the go cue (300 ms before movement onset), at movement onset, and just before grasp initiation. The model R² then decreased to an intermediate level during the initiation of the grasp followed by an additional reduction (380 ms) during the object hold period (Fig. 9A). The model fit was significant and sustained over an extended period including movement onset, reach, and grasp initiation. The model R² was lowest and failed to reach significance during the object hold period. The partial R² profiles for object (Fig. 9B) and force (Fig. 9C) showed that the modulation was related exclusively to the objects. The force partial R² was low and not significant, also evident in the color plots of normalized firing across force levels (Fig. 9E). The temporal multiple regression results are consistent with the epoch analysis for this cell, with significant differences across objects [F(11,252) = 12.97, P < 0.05 and F(11,252) = 3.85, P < 0.05 for reach and grasp, respectively] but not force levels [F(4,252) = 1.46, P > 0.05 and F(4,252) = 0.30, P > 0.05 for reach and grasp, respectively]. The nonsignificant object-related modulation during the hold period was consistent with the color plot and object partial R² at that time.

View larger version (38K): [in this window] [in a new window]   FIG. 9. Temporal multiple regression analysis showing object- and force-related modulation. A–C: for cell L145, the model and partial R2 profiles for object (B) and force (C) parameters indicated that the model R2 was driven by the object, not force, parameter. Above each bar plot, a raster display indicates times at which the discharge was significantly correlated (minimum of 5 consecutive bins) with the model parameter(s). D and E: color plots of the firing rates over time, averaged across objects (D) and force levels (E), show that the peak R2 value at approximately –100 ms was caused by large differences in the firing rates across objects and not force levels. F–J: for cell G117, the model and partial R2 profiles for object (G) and force (H) parameters indicated that the model R2 was driven by both object and force parameters. Color plots of the firing rates over time also show that differences in the firing rates across both objects (I) and forces (J) contributed to R2 peak values.

Across the cell population, the number of cells with significant model and partial R² at each time bin confirms that object-related modulation began early in the task and was maximal during the reach (Fig. 10). In contrast, force modulation occurred in a small number of cells. It is important to note that this analysis is based on the percentage of cells that had significant levels of modulation rather than the depth of modulation. For each monkey, the number of cells with significant model R² (Fig. 10A) and significant object partial R² (Fig. 10B) increased with reach initiation and peaked throughout the reach and grasp initiation. The number of cells with significant modulation declined throughout the grasp segment. In contrast to this, the force partial R² was significant for approximately a quarter of the cells and, for monkey G, the number of cells with significant modulation was greatest during the reach and grasp initiation (Fig. 10C). Similar results were observed for monkey L during the grasp, but fewer cells reached significance during reach. The percentage of cells with significant force-related modulation was lower for the temporal regressions compared with the epoch analysis. The difference is probably caused by the improvement in signal-to-noise as a result of averaging over an epoch and a more stringent criterion for significance (e.g., dependent on 5 consecutive bins reaching significance) for the temporal regression.

View larger version (15K): [in this window] [in a new window]   FIG. 10. Temporal profile showing the percentage of cells with significant model and partial R2. A–C: percentage of cells with significant model R2 (A) for monkeys G and L (gray circles and black diamonds, respectively) and corresponding percentages of partial R2 for object (B) and force (C) parameters. Horizontal lines represent the range (±SD) over which reach onset occurred for both monkeys.

The temporal profile of the simple spike firing for most Purkinje cells was similar across objects and force levels (Figs. 5, 6, and 8), varying primarily in amplitude. The similarity in the temporal firing pattern suggests that the cell discharge may also be signaling a more global parameter of the task. We hypothesize that only a few (i.e., 1 or 2) significant firing profiles will be identified across the experimental conditions for a Purkinje cell. To test this hypothesis, an SVD analysis was completed for each cell's discharge across all object grasp dimensions and force levels to reduce data complexity and identify salient patterns of activity in Purkinje cell discharge. The first step was to compute the eigenvectors and identify significant eigenvalues. The eigenvalues were proportional to the variability in the original firing pattern that can be explained by the corresponding eigenvector.

For cell G117, the first eigenvector (E1) explained 94% of the variance, whereas the second and third eigenvectors (E2–E3) reflected only 3 and 0.7%, respectively. The REVs and associated F and P values showed that the eigenvalues for E1 and E2 were significant factors [REV = 0.66, F(1,39) = 18.65, P < 0.05 and REV = 0.13, F(1,38) = 4.48, P < 0.05, respectively], whereas E3 was not [REV = 0.06, F(1,37) = 2.32, P > 0.05]. The number of significant factors reflects the complexity of the system. Of the total cell population, 64/77 cells had signal-to-noise ratios that were large enough to detect at least one significant factor (F values with P < 0.05). On average, the significant E1s accounted for 62 ± 18% of the variance. Approximately one-half of those cells, 29/64, had significant E2s accounting for 12 ± 7% of the variance. Therefore the discharge pattern of most cells consisted of a single dominant temporal profile of which 45% were also modulated by a second pattern.

The temporal weighting patterns for cell G117 showed the overall pattern of the simple spike firing was accommodated by a dominant first component (E1), which explained the vast majority (>94%) of the signal. Variations from this dominant pattern covered by a secondary component (E2; Fig. 11A, solid and dashed lines, respectively). Specifically, the temporal weighting profile for E1 was near zero while the monkey's hand was on the start-pad (–800 ms) and increased during reach to attain a maximum value toward the end of the reach segment (approximately –100 ms). Just before force onset (0 ms), the weighting for E1 decreased rapidly on grasp initiation. During the subsequent object hold period, the weighting gradually decreased and reached a relatively stable (although slightly decreasing) weighting (400 ms to end). The weighting for E2 reached peak (absolute) values just before and after the peak weighting in the dominant pattern, thereby accommodating deviations from this ideal firing pattern about the time of peak firing. The temporal weighting profiles for E1 and E2 are reflected in the simple spike firing for G117 (Figs. 8A and 9, I and J).

View larger version (40K): [in this window] [in a new window]   FIG. 11. Singular value decomposition of the normalized firing rate. A: temporal weighting profiles for the 1st 2 eigenvectors (E1–E2) for cell G117. A default firing pattern described by E1 (solid line) peaked toward the end of the reach and before grasp initiation (time 0). Variations from E1 are described by E2 (dashed line), indicating that most of the variation occurred just before and after the peak firing described by E1. B: E1 weightings (black circles) remained relatively constant across grasp force levels (top) and object grasp dimensions (bottom), whereas E2 weightings (white circles) show a gross scaling effect. C and D: plots of E1 (top) and E2 (bottom) weightings across grasp force levels (C) and object grasp dimension (D) for monkey G cells with significant E1 and E2 eigenvalues. E and F: similar plots for monkey L. All E1 and E2 weightings were adjusted to reflect positive initial weighting values so that comparisons could be made across the cell population.

As described above for grasp force, the SVD analysis was performed as a function of the grasp dimension. For example, E1 remained relatively constant, whereas E2 tended to decrease with the object grasp dimension (Fig. 10B, bottom). Across the cell population, E1 weightings remained similar across object grasp dimensions, with E2 showing a gross linear scaling across conditions (Fig. 11, D and F). The resulting SVD findings across object grasp dimensions showed greater variation, particularly for the monotonic trends in E2, compared with force-related effects. This finding may reflect that grasp force was explicitly controlled during the experiment, whereas the object grasp dimension was one of many object-related properties (i.e., shape, volume, orientation) defined by each of the 16 objects. Furthermore, positive-to-negative E2 weightings across conditions should not be interpreted to imply decreasing firing rates across these same conditions. The weightings simply describe a scaling effect that could represent an increase or decrease in firing across experimental conditions.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Unique to the study was the evaluation of Purkinje cell discharge during whole hand reach-to-grasp to a large set of objects, varying in size and shape, and across a wide range of grasp force levels. The major findings are that simple spike discharge is significantly modulated by objects, grasp dimension, and grasp force. These observations are consistent with the proposal that the cerebellum plays a major role in the signaling of prehensile movements (Dugas and Smith 1992; Gibson et al. 1985a,b; Grafton et al. 1996; Imamizu et al. 2003; Mason et al. 1998; Smith and Bourbonnais 1981) encoding both reach-to-grasp kinematics (Gibson et al. 1985b, 1996; van Kan et al. 1994) and grasp force (Kawato et al. 2003; Monzee and Smith 2004; Smith and Bourbonnais 1981). The SVD findings of a single dominant discharge pattern explaining most of the variance in discharge modulation across all objects and force levels is consistent with the kinematic observations of a dominant hand posture across the same conditions (Mason et al. 2004; Napier 1956; Santello and Soechting 1997; Santello et al. 2002). Finally, object- and force-related modulations were independent in support of the proposal that reach-to-grasp kinematics and grasp force are signaled separately (Mason et al. 2004).

Reach versus grasp related modulation

In designing the study, the assumption was that variations in the geometric properties of objects would elicit variations in hand-shaping kinematics during reach-to-grasp but have little or no effect on the transport of the hand itself. The analyses in this and our previous study (Mason et al. 2004) show reach kinematics remained largely invariant across objects grasped and grasp force levels. We infer from this invariance that the changes in simple spike firing are not related to the reach but instead to the grasp. It is possible that cell discharge is related to aspects of the reach other than those specifically measured such as patterns of muscular activity, including co-contractions. However, it is unlikely that these highly trained animals varied their patterns of arm EMG activity on each reach while maintaining the same trajectory, timing, and speed parameters. Furthermore, interpositus neurons are highly modulated during reaching to grasp, and this discharge is independent of variations in reach parameters such as trajectory, amplitude, and direction of reach (Gibson et al. 1996; van Kan et al. 1994). Finally, the greater depth of modulation noted during reach compared with grasp is also consistent with Purkinje cell signaling of hand shaping during reach. A similar conclusion was reached for the modulation of cerebellar nuclear neurons during the reach in preparation for grasp (Gibson et al. 1996; van Kan et al. 1993b, 1994). Therefore simple spike modulation during the task is most likely related to changes in hand-shaping or grasp force.

Modulation in relation to object and grasp dimension

Several of the key findings relate to the modulation of the simple spike discharge during reach and grasp as a function of the target objects. Conversely, the shaping of the hand and grasp aperture were highly dependent on the objects. The simple spike discharge of the majority of Purkinje cells, 97% during reach and 91% during grasp, was significantly modulated by the objects. Although the simple spike modulation with objects persisted and retained object preferences during the grasp, the depth of modulation was most marked during the reach. Similarly, the temporal regressions showed peak R² during reach rather than grasp, and these peaks were attributed to object related modulation (i.e., the object partial R²). For most cells, significant modulation with grasp dimension occurred during both reach (75%) and grasp (62%). As with objects, the modulation with grasp dimension was most marked during the reach.

The observation that the simple spike discharge was significantly and linearly modulated in relation to grasp dimension for most cells (75%) provides the strongest evidence yet that the firing is signaling hand shape, both during reach and grasp. A large proportion of Purkinje neurons was linearly related to grasp aperture during both reach (65%) and grasp (46%). Grasp aperture is linearly related to grasp dimension (Mason et al. 2004). Therefore Purkinje cell activity provides an explicit representation of the grasp aperture and provides further support of the signaling of hand shape. A similar conclusion was reached for the discharge of magnocellular red nucleus neurons during a reach to grasp task (van Kan and McCurdy 2002). Comparing the discharge using a precision grasp (thumb and forefinger) with that of a whole hand grasp, the firing was most likely to relate to the extension of the metacarpophalangeal (MCP) joints. Flexion and extension of the MCP joints contributes significantly to the overall hand shaping observed in this task (Mason et al. 2004). We interpret the behavioral and neurophysiological observations to imply that Purkinje cell simple spike firing is unlikely to signal objects specifically. Instead, the modulation is more likely to reflect hand shapes or finger positions associated with preshaping and/or shaping used to grasp the objects.

Distinguishing between feedforward (i.e., commands) and feedback signals (i.e., afferent input) is inherently difficult in behavioral electrophysiology experiments and was not the focus of this study. Most cells had receptive fields with proprioceptive and/or cutaneous properties. Therefore it is possible that t