Prehension kinematics in humans and macaques

Abstract

Nonhuman primates, especially rhesus macaques, have been a dominant model to study sensorimotor control of the upper limbs. Indeed, human and macaques have similar hands and homologous neural circuits to mediate manual behavior. However, few studies have systematically and quantitatively compared the manual behaviors of the two species. Such comparison is critical for assessing the validity of using the macaque sensorimotor system as a model of its human counterpart. In this study, we systematically compared the prehensile behaviors of humans and rhesus macaques using an identical experimental setup. We found human and macaque prehension kinematics to be generally similar with a few subtle differences. Although the structure of the preshaping hand postures is similar in humans and macaques, human postures are more object-specific and human joints are less intercorrelated. Conversely, monkeys demonstrate more stereotypical preshaping behaviors that are common across all objects and more variability in their postures across repeated presentations of the same object. Despite these subtle differences in manual behavior between humans and monkeys, our results bolster the use of the macaque model to understand the neural mechanisms of manual dexterity in humans.

NEW & NOTEWORTHY Macaques have been a dominant animal model to study the neural mechanisms of human dexterity because they exhibit complex manual behavior. We show that the kinematics of prehension, a critical dexterous behavior, are largely similar in humans and macaques. However, human preshaping postures are more object-specific and the movement of human digits is less correlated with each other. The thumb, index, and wrist are major drivers of these interspecies differences.

INTRODUCTION

The human hand is a remarkably precise and versatile manipulative organ, enabling interactions with objects that range from the mundane—like grasping an object—to the extraordinary—epitomized by virtuosic pianism. Manual dexterity is made possible by the complex anatomy of the hand, which comprises 27 bones articulated by 52 muscle actuators, and by expansive neural circuits involved in motor control and sensing (1).

Our understanding of the neural mechanisms that mediate manual dexterity has relied heavily on experiments with nonhuman primates, and in particular, macaques, whose hands and sensorimotor systems resemble their human counterparts (1–3). Indeed, the anatomy of human and macaque hands is largely similar and homologs to the neural structures that support the sensorimotor control of the hand in humans have been identified in macaques.

Qualitatively, manual behavior is also similar in humans and monkeys. Like humans, monkeys preshape their hands before grasping in an object-specific way and are capable of some degree of finger individuation (4, 5). In both humans and monkeys, hand preshaping is so object-specific that the identity of an object can be accurately inferred from the (precontact) hand posture (6–9). However, to our knowledge, human and monkey manual behavior has never been compared quantitatively using approaches that have been used to characterize manual behavior of either species separately. Careful comparison of human and monkey manual behavior will provide a further assessment of the validity of the macaque sensorimotor system as a model of human manual behavior.

With this in mind, we systematically compared the prehensile behavior of humans and rhesus macaques using an identical experimental paradigm. In brief, humans and monkeys grasped a set of objects handed to them by a robot while we monitored the time-varying joint angles of their hands using a computer vision algorithm. This paradigm was designed to eliminate any differences in prehension due to reaching, thereby isolating the grasping component. We then compared the hand kinematics of humans and monkeys using a variety of quantitative approaches. We found that, although hand kinematics are similar in humans and monkeys during prehension, these differ in subtle but systematic ways.

METHODS

Behavioral Task

All procedures were approved by the Institutional Review Board of the University of Chicago. Two rhesus macaques were trained to perform a grasping task while seated, with their head fixed (Fig. 1). Animals were required to keep their arm stationary on armrests between trials, as monitored by a photosensor embedded in the armrest. At the beginning of a trial, the experimenter attached an object to a robotic arm (MELFA RV-1A, Mitsubishi Electric, Tokyo, Japan) using a magnet. The arm then presented an object to the monkey at a constant speed always stopping in the same position. Although the object was approaching, the animal preshaped its hand and grasped it without removing the arm from the photosensor. After 1–3 s, the robotic arm retracted, and the animal had to hold the object with enough force to overpower the magnet.

Figure 1.

Experimental workflow and grasped objects. A: overview of the grasped objects. B: positions of the tracked markers on macaque forearm. C: example trial for human and macaque subjects.

Twenty-four different shapes were presented, some of which were used at different orientations, totaling 27 different objects. Each object was presented 8–11 times in a single experimental session. If the animal failed to hold onto the object or lifted its arm, the trial was aborted. Monkeys completed 8 sessions (monkey 1: 6 sessions, monkey 2: 2 sessions), for an average 10 trials per object, respectively. For details on the experiment, see Ref, 10.

Two right-handed adult human subjects (male and female, 23 and 28 yr old) performed the same grasping task as the rhesus macaques during a single experimental session. The humans were instructed to grasp the objects tightly without moving their arms off the armrest, but they were not required to hold onto the objects as they retracted.

The two human subjects also performed a standard reach-to-grasp task with a set of 25 everyday objects (yielding 30 different grasping conditions because some objects were presented at multiple orientations). For more details, see Ref. 9. We found that most analyses yielded similar conclusions with the two data sets (Supplemental Figs. S1, S2, and S3; all Supplemental material is available at http://dx.doi.org/10.6084/m9.figshare.17113892).

Recording and Processing of Kinematic Data

We used an infrared motion tracking system to record macaque hand movements (MX T-Series, Vicon, Los Angeles, CA). Thirty-one reflective markers were placed on the bony landmarks of the animal’s arm and hand and tracked in three dimensions by 14 cameras at a sampling rate of 100 Hz. Each marker was labeled to a corresponding anatomical location using the Vicon Nexus software (Nexus, Vicon, Los Angeles, CA). For details on the data acquisition, see Ref. 10.

We used a machine-learning method to track the kinematics of the human subjects (11). The movements were recorded by eight video cameras (Blackfly S USB3 BFS-U3-16S2C-CS, FLIR Integrated Imaging Solutions, Inc.) at a spatial resolution of 1,440 × 1,080 and a sampling rate of 50 Hz. A machine vision neural network was trained on 80 images per camera per subject to locate 31 manually labeled bony landmarks distributed over arm and hand (12). The network labeled all frames, producing two-dimensional (2-D) trajectories of the landmarks in the camera plane. We triangulated the 2-D maker positions using the relative locations of the cameras, to obtain the time-varying three-dimensional (3-D) positions of the markers.

For both humans and nonhuman primates, we used OpenSim Inverse Kinematics tool (13) to obtain the time-varying joint angles of the hand from the 3D marker trajectories based on a scaled musculoskeletal model of primate hand and forearm (14, 15). We tracked the time-varying angles of 23 degrees of freedom around 16 joints: flexion-extension, rotation, and radioulnar deviation of the wrist; flexion-extension and abduction-adduction of the carpometacarpal (CMC), metacarpophalangeal (MCP), and interphalangeal (IP) joint of the thumb; and flexion-extension and abduction-adduction of the MCP and flexion-extension of the proximal interphalangeal (PIP) and distal interphalangeal (DIP) of each of the four fingers. The angles were smoothed with a 50-ms moving average. We only analyzed kinematics from the start of movement (indicated by the wrist movement from rest) to 100 ms before contact. Joint angles in that period reflect visually guided volitional movements and are not affected by contact with the objects.

Subspace Similarity

To investigate the similarity in kinematics, we used cross-projection similarity (16). In brief, we calculated the amount of variance in the kinematics of one subject that could be explained by the first N principal components (PCs) of a different subject. This value was then normalized to the variance explained by those PCs in the original subject. If the kinematics subspaces were identical for the two subjects, the cross-projection similarity would be 1. Note that this metric is not commutative: the projection of kinematics of one subject onto the PCs of the other is not necessarily equal to the projection of the kinematics of the second onto the PCs of the first. For that reason, we computed it both ways for each pair of subjects and used their average as the cross-projection similarity.

We first computed the average cross-projection similarity between different human subjects (SH), different monkey subjects (SM), and pairs of humans and monkeys (SHM) at a fixed dimensionality (9 dimensions, which explains ∼90% of the variance). We calculated the average of SH and SM, which represents the mean within-species similarity, and calculated the normalized interspecies difference D = (SH + SM)/2 – SHM. D is the normalized subspace difference between species. We then calculated the change in D if we remove all joints of a particular digit (e.g., thumb) or a joint group across all digits (e.g., MCP joints). We interpreted a decrease in D upon removal of a digit or joint group as evidence that this digit or joint group contributes to the dissimilarity between monkeys and humans.

We also computed an alternative measure of subspace similarity based on subspace geometry, namely, the principal angle (17). The principal angle ranges from 0° to 90°, where 0° denotes identical subspaces and 90° denotes orthogonal subspaces. As an index of dissimilarity, we report 90° minus the mean principal angle as a function of the dimensionality of each subspace.

Classification

To quantify the degree to which the hand volitionally conforms to each object, we attempted to classify the grasped objects based on the hand posture 100 ms before contact. To this end, we first calculated PCs from that posture for each subject. Then, we used linear discriminant analysis (LDA) on kinematics projected on all the PCs or on subsets of PCs (9), using leave-one-out cross-validation. For each object, one trial per object was randomly left out of the training data, and the classifier was tested on those trials. The procedure was repeated with each trial left out. To quantify how much each joint group or digit preshapes to the object, we examined classification accuracy using only one digit or joint group.

Linear Dependence of Fingers and Joint Groups

To quantify the degree to which individual joints move independently of other joints, we first regressed the trajectory of each joint on that of every other joint. We then calculated the linear dependence index by averaging all R² values of each joint in a digit or joint group (18). Values near 0 denote nearly complete independence and values near 1 denote full linear dependence.

Variance Decomposition

Although principal component analysis (PCA) identifies axes of maximum variance, these axes are not ascribed any functional significance. To fill this gap, we used a modified version of PCA, demixed PCA (dPCA), which finds axes of variance that relate to task conditions (19, 20). Here, we looked for variance in the kinematics that related to object identity and trial progression, the component of the kinematic trajectory that is common to all objects. All remaining variance in the kinematics was attributed to noise. We used the MATLAB package “dpca” by Kobak et al. (20) to perform the dPCA analysis.

RESULTS

We examined the hand movements (23 joint angles) of two humans and two rhesus macaques as they grasped objects of varying shapes and sizes (Fig. 1), focusing on the preshaping epoch before contact with the object is established. We found, as has been previously shown, that joint kinematics were similar for repeated grasps of the same objects and differed across objects (Fig. 2). We then set out to quantitatively compare the prehensile behavior of humans and monkeys.

Figure 2.

Example kinematics traces. Time-varying joint angle traces of digits 2 through 4 in humans and monkeys. Each row corresponds to a different object and each curve to a different trial. The joint angles shown are MCP flexions of the fingers. MCP, metacarpophalangeal.

Basic Features of Prehension Kinematics

First, we examined the parameters of movement for monkeys and humans, including the range of motion, speed, and acceleration of each joint, as well as the duration of grasping movements (Fig. 3). We found that the ranges of motion were similar for monkey and human joints, but the speeds and accelerations differed in joint-specific ways. For example, the distal interphalangeal joints (DIP) tended to accelerate and move faster in humans than monkeys (Fig. 3) (Mann–Whitney U test, speed: n = 12,135 joint angles × trials, P < 1e-06; acceleration: n = 12,135 joint angles × trials, P < 1e-06). The reverse was true for the proximal finger joints (MCP flexion/extension), which tended to accelerate more rapidly and reach higher peak speeds in monkeys (P < 1e-06). Overall, monkeys performed the task more rapidly than did humans (Mann–Whitney U test, n = 2,427, P < 1e-06). Viewed through this lens, humans tend to move their distal joints more than monkeys do, who prefer to complete the movement using proximal hand joints. This difference may be attributed to the more common use of the hook grip in monkeys than in humans, which involves the distal joints remaining in a hook configuration during the hand closing.

Figure 3.

Basic kinematic statistics of grasping for monkeys and humans. A: average range of motion of each joint for monkeys and humans. Joint group is coded by color and digit by marker shape. We used filled shapes for MCP joints’ abduction-adduction and open shapes for MCP flexion-extension. B: maximum speed of each joint. C: maximum acceleration of each joint. On subplots A, B, and C, the values were averaged across trials and subjects. D: cumulative distribution of grasp durations for monkeys (red) and humans (blue). MCP, metacarpophalangeal.

The Structure of Prehension Kinematics

Next, we examined and compared the structure of manual behavior in the two species. To this end, we performed a principal component analysis (PCA) on the prehension kinematics of monkeys and humans. We found that human kinematics tended to be more complex as evidenced by a shallower cumulative variance plot; more principal components (PCs) were required to account for the same proportion of variance in human than monkey kinematics (Fig. 4A). However, this separation was relatively modest. Furthermore, the first two PCs were remarkably similar for humans and monkeys (accounting for 63% and 78% of the variance, respectively). The first involved changes in hand aperture, and the second mainly involved movements of the wrist (Fig. 4B).

Figure 4.

Principal component analysis of kinematics. A: cumulative variance explained as a function of the number of PCs. PCs are arranged in descending order of eigenvalues. B: the hand movements produced by the first and second PCs in monkeys and humans. PC, principal component.

Dependence of Joints

The kinematic subspaces reflect the linear dependence of different hand joints. Indeed, each PC denotes a specific pattern of correlated joint angles. To compare patterns of interjoint correlations across humans and monkeys, we analyzed the linear dependence of each digit and joint group on the other joints, measured as the mean fit (R²) of a regression of the time-varying angle of one joint on that of all the other joints. We found that monkeys yielded consistently higher linear dependence across all digits than did humans (Fig. 5A) (Wilcoxon signed-rank test, n = 40 joints, P = 8.9e-05). In other words, monkey digits were significantly less individuated during grasp than were their human counterparts, consistent with the results from the PCA. In both monkeys and humans, the thumb was the most individuated digit, yielding a lower linear dependence index than the other digits (Fig. 5A). Similarly, the wrist, MCP, PIP, and DIP joints of monkeys were more linearly dependent on the other joint types than were their human counterparts (Wilcoxon signed-rank test, n = 40 joints, P = 5.23e-05) (Fig. 5B).

Figure 5.

Dependence index of digits and joint groups. A: linear dependence of each digit for monkeys vs. humans. B: linear dependence index of each joint group for monkeys vs. humans. Blue dashed line is the identity line; gray dashed line is the regression line.

Note, further, that the interdependence of joints at the level of joint groups is somewhat idiosyncratic, whereas the interdependence of joints grouped by digits is rather consistent (Supplemental Fig. S4).

Kinematic Subspaces of Monkeys and Humans

Next, we aimed to quantify the similarity of the kinematic subspaces in humans and monkeys separately from the mean and scale of the data. To this end, we calculated how much variance in the hand kinematics of one subject could be accounted for using the PCs of another human or monkey, a measure known as cross-projection similarity. We found that, on average, the first 10 PCs of one species could account for ∼80% of the (explainable) variance in the kinematics of the same species and 75% of the variance of the other species (Fig. 6A), suggesting a high degree of similarity in the structure of human and monkey prehensile kinematics. Nonetheless, the cross-species similarity was significantly lower than its within-species counterpart (human-human vs. human-monkey: Wilcoxon signed-rank test, n = 46 principal components, P = 6.92e-04, monkey-monkey vs. human-monkey: Wilcoxon signed-rank test, n = 46, P = 3.09e-05) (Fig. 6A).

Figure 6.

Subspace similarity between humans and monkeys. A: cross-projection similarity (measured in proportion variance explained) as a function of the PCs added. Error bars denote SE. B: interspecies difference with individual digits removed. The difference is measured as the distance between the human-monkey similarity to the average of the human-human similarity and the monkey-monkey similarity. The bar marked as “None” represents the difference value calculated without removal of any digits. C: interspecies difference with different joint groups removed. PC, principal component.

Given the relatively subtle difference in human and monkey kinematics, we set out to examine which digits or joint groups contributed most to this difference. To this end, we computed an “interspecies difference” measure by subtracting the mean across-species similarity from the mean within-species similarity. We then applied this metric after removing a digit (Fig. 6B) or a joint group (wrist, MCP, PIP, and DIP) (Fig. 6C). To the extent that removing a digit/type led to a decrease in interspecies difference, this digit/joint group contributes to the observed difference between humans and monkeys, and vice versa. We found that the thumb, index finger, and wrist joints contributed strongly to the subspace difference, while digit 3 ∼ 4 and PIP did not (Fig. 6, B and C). We repeated these analyses with a different subspace similarity measure, principal angle, and obtained similar results (Supplemental Fig. S5).

Complexity of Prehensile Behavior

One manifestation of dexterity is that manual behavior is tailored to the task. The object specificity of precontact hand postures during prehension is an example of such task specificity. With this in mind, we compared the object dependence of human and monkey preshaping postures and examined how this dependence was distributed over the joints. To these ends, we used linear discriminant analysis to classify the grasped object based on the hand posture 100 ms before contact, when the hand is preshaped to the object but before volitionally adopted hand posture is distorted by object contact. We found that, for both monkeys and humans, objects could be classified with high accuracy based on hand posture. Both human subjects exhibited hand preshaping postures that were highly object-specific (Fig. 7A). However, the degree of object specificity was very different for the two monkeys: one yielded classification performance commensurate with that of humans, and the other exhibited preshaping hand kinematics that was far less object-specific, yielding lower classification performance.

Figure 7.

Object specificity of hand preshaping. A: object classification performance based on postures measured 100 ms before object contact, with an increasing number of high-variance PCs removed. B: classification accuracy based on the kinematics of each digit individually for monkeys vs. humans. C: classification accuracy based on the kinematics of each joint group individually for monkeys vs humans. PC, principal component.

We then assessed how this object-specificity was distributed over the space of hand postures in two ways. First, we measured classification performance as we gradually removed high-variance PCs. Preservation of classification after removal of PCs suggests that prehensile behavior occupies a high-dimensional space (9). In both humans and monkeys, we found that object information was distributed over a large number of PCs (Fig. 7A) (cf. Ref. 9). Even after removal of all but 1 PC, classification performance was above chance (∼7% > 3.7%). Second, we examined the degree to which different joint groups adopted object-specific postures. To this end, we calculated the accuracy of object classification based on the postures of subset of joints, grouped joints by digit (Fig. 7B), or type (Fig. 7C). Again, humans and monkeys yielded similar results, with subtle differences. For example, the index finger was less informative in monkeys than in humans (Mann–Whitney U test, n = 300 repetitions, P = 1.65e-04). Contrary to expectation, the thumb was more informative in monkeys than in humans (Mann–Whitney U test, n = 300 repetitions, P = 4.18e-09). The object specificity of the joint groups, however, was consistent across humans and monkeys, with the MCP being the most informative and DIP the least (Fig. 7C). Note, however, that these indices of informativeness also depend on the objects, as evidenced by the fact that they differ in human subjects across different object sets (Supplemental Fig. S3). Thus, prehensile behavior in both humans and monkeys occupies a high-dimensional kinematic space and is distributed over joint groups in similar ways.

As mentioned earlier, the hand postures of one monkey were far more object-specific than those of the other. In contrast, the two humans exhibited highly similar prehensile behaviors (Fig. 7A). In light of this, we repeated the above analyses on the kinematics of individual subjects to assess the degree to which our conclusions were skewed by the observed differences across monkeys. We found that basic prehensile features and linear dependence indices were consistent across monkeys (Supplemental Figs. S4, S6, and S7), despite these significant differences in the object-specificity of hand preshaping (Supplemental Figs. S8 and S9).

Decomposition of Hand Kinematics Variance

Finally, we sought to understand the determinants of human and monkey prehensile kinematics. To this end, we used demixed PCA (dPCA) (19, 20) to decompose the kinematics of each subject into components that were stereotypical across objects (time-dependent), components that were depended on the object to be grasped (object-dependent), and components that varied across repeated presentations of the same object (noise). We found that monkey kinematics comprised more time-dependent and noise variance and less object-dependent variance than did their human counterparts (Fig. 8). In other words, monkey prehensile postures were less object-specific, more stereotyped, and noisier than their human counterparts. Furthermore, the decreased object-specificity observed in the classification analysis for monkey 2 (Fig. 7A) is driven by greater noise in this monkey’s prehension kinematics than in those of the other monkey or in humans (Supplemental Fig. S9). In other words, that monkey exhibited a tendency to grasp the same objects in different ways.

Figure 8.

dPCA analysis and variance decomposition. A: human variance decomposition. Bar plots show the variance explained by each dPC arranged in descending order of eigenvalues. For each bar, the contribution by each type of variance is shown by color. Pie chart shows the overall proportion of each type of variance. B: monkey variance decomposition. dPCA, demixed principal component analysis.

DISCUSSION

In summary, we find that, although the basic statistics of prehensile kinematics, effective ranges of motion, speeds, and accelerations, differ somewhat between humans and monkeys, the structure of manual behavior is similar for the two species. Both humans and monkeys exhibit high-dimensional preshaping kinematics, though monkey kinematics are more stereotyped across objects and noisier than their human counterparts.

Kinematic Similarities and Differences

Among nonhuman primates, macaques arguably possess the most human hands except for the (nonhuman) apes (2). Macaque hands are endowed with a semi-opposable thumb, share much of the musculature with humans, and can individually move their fingers (21–23). Human and macaque hands differ most strikingly in the respective morphologies of their thumbs. First, the macaque thumb is much shorter than the human one compared with the index and middle fingers (24, 25), putatively to prevent it from interfering with arboreal locomotion. Second, the pronounced saddle shape of the trapezium bone supporting thumb carpometacarpal joint in humans produces considerable rotation of the thumb as it abducts, presenting its distal pad to that of the fingers, allowing greater opposability (26). Third, more muscles articulate the human than the macaque thumb (21). All of these differences contribute to the increased manipulative ability of humans, exemplified by tool-making and strong precision grip (27–31). We found that the thumb was the most independent digit in both humans and monkeys and the thumb was a significant contributor to the dissimilarity between human and monkey prehensile kinematics.

The index finger also significantly contributed to the difference in kinematics, possibly due to its higher individuation in humans, enabled by the anatomy of the extensor indicis muscle. In humans, this muscle only extends the index finger, whereas, in macaques, it also extends the middle finger (21). Interestingly, the kinematics of the index finger is the second least object-specific in macaques, whereas the kinematics of the thumb is the least object-specific in humans. The greater stereotypy in the thumb and index may reflect the fact that these digits contribute the most to grasp. As macaques commonly use hook grasps, the index finger provides stability; as humans favor the power grip, the thumb plays the equivalent stabilizing role.

Grasping is commonly characterized by measuring hand aperture, commonly defined as the distance between the tips of index and thumb (32). Macaques change their aperture with higher velocity and acceleration than humans during reach and grasp tasks (33). We found that this difference in aperture dynamics is produced by the difference in the movement of the proximal finger joints (Fig. 3, A, B, and C), an effective strategy as proximal joint deflections result in larger end point deflections.

Often overlooked, the wrist is the most complex joint of the hand, connecting up to 17 bones in primates and transferring the load between hand and arm for locomotion and manipulation (26). Functionally, the carpal bones of nonhuman primates can lock together for support during terrestrial locomotion or allow for a high range of motion during other behaviors. In humans, however, the wrist is not used for locomotion, which has led to the fusion of some wrist bones, a decrease in bone volume, and changes to its microanatomy, making the human wrist somewhat simpler anatomically, although better suited for manipulation and tool making (34–37). One might thus expect large differences in wrist kinematics. Indeed, we found human wrist movements to be more independent than their macaque counterparts (Fig. 5B), and differences in wrist kinematics were a major contributor to interspecies differences in prehensile behavior (Fig. 6C).

Macaque and human hands differ in not only their morphology but also their size. In this study, we used identical objects for experiments with humans and macaques and found their prehension to be similar. However, some of the identified differences may have been an artifact of differences in the size of the objects relative to the hand. To test this possibility, we compared the kinematics of humans grasping large objects to those of monkeys grasping small ones and found no systematic differences (Supplemental Figs. S10, S11, and S12). Thus, relative object size did not shape our findings.

Control of Volitional Movements

In primates, large swaths of cortex are dedicated to hand control, much larger than in other species (3, 38–42). The motor cortex of many primates, mostly Old World monkeys and apes (including humans), sends projections directly onto spinal motoneurons and thus has more access to the muscles than does the motor cortex of other mammals (43–46). That prehensile kinematics occupy a high-dimensional space in both humans and macaques likely reflects these expansive and specialized neural circuits that mediate manual dexterity in humans and macaques (1). Indeed, the representation of hand posture in primary motor cortex is higher dimensional than the already high-dimensional hand postures themselves (9, 47, 48).

Macaques and humans also possess highly developed cortical fields in posterior parietal cortex (PPC) that convert visual information about the object into a motor plan (39, 49–51). Specifically, the anterior interparietal area is crucial for the visuomotor transformation from shape to grip type (52–56). These mechanisms of visuomotor transformation are likely the substrate that underlies the ability of humans and macaques to preshape their hands to objects to be grasped and, more generally, to exhibit manual behaviors that are appropriate for specific interactions. In both humans and macaques, a large proportion of the variance in hand movements was object-specific and, though this proportion was higher in humans than monkeys, differences were dwarfed by similarities. Rats and mice have a comparatively small PPC and do not exhibit such sophisticated visuomotor transformations (57).

Conclusions

We show that prehensile behaviors in humans and monkeys are very similar when examined through a detailed, quantitative lens. This similarity likely reflects the similarity in end effectors and associated neural circuits. We expect that such detailed behavioral comparisons across multiple species may help disentangle links between neural mechanisms and behavior.

SUPPLEMENTAL DATA

Supplemental Figures S1–S12: https://doi.org/10.6084/m9.figshare.17113892.

GRANTS

This work was supported by National Institute of Neurological Disorders and Stroke Grant NS122333.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Y.Y., A.R.S., and S.J.B. conceived and designed research; Y.Y. performed experiments; Y.Y. analyzed data; Y.Y. and A.R.S. interpreted results of experiments; Y.Y. prepared figures; Y.Y. and S.J.B. drafted manuscript; Y.Y., A.R.S., and S.J.B. edited and revised manuscript; Y.Y., A.R.S., and S.J.B. approved final version of manuscript.