Deficient internal models for planning hand–object interactions in apraxia

Abstract

Motor imagery (MI) has been associated with planning stages of motor production, and in particular, with internal models that predict the sensory consequences of motor commands and specify the motor commands required to achieve a given outcome. In this study we investigated several predictions derived from the hypothesis that ideomotor apraxia (IM), a deficit in pantomime and imitation of skilled actions, may be attributable in part to deficits in internal models for planning object-related actions, in the face of relatively intact on-line, feedback-driven control of action. This hypothesis predicts that in IM, motor imagery should be (a) strongly correlated with other motor tasks not providing strong visual, tactile, and proprioceptive feedback from objects, i.e., object-related pantomime and imitation; (b) poorly correlated with performance tasks providing strong environmental feedback about the locations of effectors and targets, i.e., actual interaction with objects; and (c) particularly deficient in conditions that are computationally difficult for the motor planning system. Eight left fronto-parietal stroke patients with IM, five stroke patients without IM, and six healthy matched controls imagined grasping dowels and widgets presented at varying orientations, and actually grasped the same objects. The experimental predictions were confirmed. In addition, patients with IM and motor imagery deficits were significantly more likely than the non-apraxic group to have lesions in the intraparietal sulcus, a region previously implicated in imagery for hand–object interactions. The findings suggest a principled explanation for the deficits of IM patients in object-related gesture pantomime, imitation, and learning of new object-related gestures.

Introduction

Several models of motor performance distinguish a mode of action concerned with planning, learning, and motor prediction, and another specialized for motor execution and control (see Keele, 1981). One influential account, for example, distinguishes semantic representations necessary for motor learning and planning from pragmatic representations subserving the control and execution of action (Jeannerod, Arbib, Rizzolatti, & Sakata, 1995). The planning mode has been proposed to generate movement parameters by way of internal representations of action. These internal models are modified based on practice and experience, and are requisite for motor learning and for the generation of skilled actions (Rosenbaum, Engelbrecht, Bushe, & Loukopoulos, 1993; Wolpert, 1997; Wolpert, Ghahramani, & Flanagan, 2001).

The execution mode, in contrast, emphasizes on-line control that is sensitive to current environmental conditions, and is equally relevant for skilled and novel actions. In motor execution stages of action, sensory and proprioceptive information about the locations of effectors and targets is fed back and compared to internal models of action. The disparity between actual and predicted behavior can be used to correct motor performance and to improve learning and accuracy (Wolpert et al., 2001).

It has been proposed that the relative contribution of planning processes decreases and the relative contributions of execution (feedback-driven) processes increases as actions unfold temporally (Glover, in press; Wolpert & Ghahramani, 2000). Thus, for example, as the hand approaches a target object, visual and proprioceptive feedback loops operate to adjust and correct movement error. For instance, the time-course of the early acceleration phase of a reaching movement is assumed to reflect planning, whereas the deceleration phase is typically assumed to reflect the on-line integration of sensory feedback (e.g., Arbib, 1981; Jeannerod, 1981, Keele, 1981).

It has also been suggested that motor imagery (MI) is a form of movement planning in which action is not executed (Blakemore & Sirigu, 2003; Glover, in press; Jeannerod, 2001, Jeannerod, 2003; Johnson, 2000a, Prinz, 1994). The posterior parietal cortex (PPC) appears to be of particular importance in planning intended movements. For instance, when non-imagery processes are properly controlled, imagined reaching activates a putative homologue of the macaque parieto-frontal reach circuit (Johnson, Grafton, Hinrichs, Gazzaniga, & Heinze, 2002). Motor imagery of finger movements activates posterior portions of left inferior and superior parietal lobe more than does motor execution (Gerardin et al., 2000). Additionally, whereas healthy subjects exhibit a strong correlation between the imagined time to generate finger or pointing movements and actual movement execution, patients with posterior parietal lesions are significantly less accurate (Sirigu et al., 1996).

Damage to the left parietal lobe also frequently results in ideomotor apraxia (IM). Patients with IM are deficient in production of complex transitive (object-related) movements whether upon imitation or in response to command or sight of objects. IM patients also have diminished knowledge of skilled object manipulation when assessed with multiple choice tasks, suggesting the deficit is not limited to movement execution stages of performance (Buxbaum, Sirigu, Schwartz, & Klatzky, 2003). In contrast to their deficient memory for skilled object-related movements, IM patients perform relatively normally in on-line reaching and grasping tasks when visual information about the limb and target is available (Haaland, Harrington, & Knight, 1999).

We have suggested previously that IM may reflect a specific deficit in generating and maintaining internal representations of object-related actions (Buxbaum, 2001). There are several lines of evidence supporting this hypothesis. First, neuroimaging studies, neurophysiological studies with non-human primates, and behavioral studies with healthy human subjects suggest that the posterior parietal region damaged in many apraxics is an important locus of internal representations of object-related hand movements (Decety et al., 1994; Rushworth, Krams, & Passingham, 2001). Inferior parietal and intraparietal sulcus (IPS) regions are involved in imagined grasping movements (Grafton, Arbib, Fadiga, & Rizzolatti, 1996), preparation of hand shape for object grasping (Sakata, Taira, Murata, & Mine, 1995), prospective judgments about object manipulation (Johnson, Grafton, Hinrichs, Gazzaniga, & Heinze, 2002) and planning of object-related pantomimes (Johnson-Frey, Newman-Norland, & Grafton, 2004). Second, in contrast to their difficulties with knowledge and production of transitive gestures, patients with IM are frequently less impaired in the performance of non-object-related (i.e., intransitive), symbolic gestures such as signaling “stop” or waving goodbye (Foundas et al., 1999, Haaland and Flaherty, 1984). Third, despite their deficits in pantomiming, imitating, and responding to multiple choice questions about object-related gestures, IM patients’ performance frequently improves to some degree when they are permitted to contact and use objects directly (Poizner, Mack, Verfaellie, Rothi, & Heilman, 1990). Pantomime, imitation, and declarative (e.g., multiple choice) questions about action plausibly emphasize the contribution of internal models of movement, whereas actual contact with objects, and indeed, even the approach to objects, affords additional sensory and proprioceptive feedback about the locations of the effector and object that can be used to shape the response via feedback. Fourth, IM patients are able to generate appropriate prehensile hand actions to objects based on shape and size information currently available to the visual system, despite deficits in retrieving the learned hand posture (HP) associated with the same objects (Buxbaum, Sirigu, Schwartz, & Klatzky, 2003).

Because motor imagery involves action planning in the absence of overt execution, it is an ideal context for evaluating the integrity of internal action representations in apraxics. The hypothesis that IM reflects deficient generation of internal models of object use predicts that apraxics should perform poorly on tasks associated with planning object-related movements; thus, it predicts deficient motor imagery of interactions with objects. Consistent with this possibility, it has been argued that gesture representations are stored in the form of motor images, and that motor imagery deficits may underlie some aspects of apraxia (Roy et al., 1993). Second, to the degree that feedback-driven motor execution stages of movement are relatively intact, IM patients should perform better when permitted to actually execute these same movements with objects. Third, to the degree that IM patients are particularly deficient in generating internal models of the hand posture component of skilled object-related movements, motor imagery should be particularly deficient on tasks putatively burdensome for the computation of hand posture. Thus, IM patients should perform substantially more poorly when imagined hand posture must be computed de novo on each trial than when a single hand posture can be used repetitively. Fourth, if IM, as assessed by praxis tasks, reflects an underlying deficit in internal models associated with planning object-related movements, there should be a strong association of particular scores on tests of object-related gesture imitation and pantomime (i.e., praxis tasks not having strong feedback from objects in the environment) and scores on measures of object-related motor imagery.

To assess these predictions, we employed variants of tasks used by Johnson with healthy subjects (Johnson, 2000a) and neurologic patients (Johnson, Corballis, & Gazzaniga, 2001; Johnson, Sprehn, & Saykin, 2002; Johnson-Frey, 2004). In these tasks, subjects are required to make prospective judgments about how they would grasp objects, in the absence of overt hand movements. Although no mention of imaging or simulating actions is made, several previous studies by our group suggest that the tasks are likely to involve motor imagery.

In the first study reported by Johnson (2000a), healthy subjects were asked to judge how they would grasp dowels presented in several orientations were they permitted to actually act, and in a second condition, to actually grasp the dowels in a power grip (i.e., as one would clench a hammer). Subjects’ preference for prospective underhand or overhand grips varied systematically with dowel orientation. Moreover, prospective and actual grasping positions were strongly correlated, and both respected the biomechanical constraints of the arm and hand. That is, in both prospective and actual grasping tasks, subjects preferred grips that would cause no discomfort to execute. This suggests that prospective grasping accesses an analog internal representation, which has been characterized as the hallmark of motor imagery (Kosslyn, Thompson, Kim, & Alpert, 1995). In a second related study, Johnson (Johnson, 2000b) demonstrated that a majority of patients with acute hemiplegia exhibited strong correspondence between the grip preferences of the hemiplegic and intact limb, and that both were influenced by biomechanical constraints. However, there were exceptions. Poor correlations between limbs were observed in two patients, one with parietal and the other with frontal lesions. The possible role of apraxia was not assessed. Subsequent work demonstrated that chronic upper limb immobility does not compromise the ability of stroke patients without frontal or parietal damage to accurately simulate movements on these tasks (Johnson, Sprehn, & Saykin, 2002).

In the present study, apraxic and non-apraxic subjects were asked to make prospective judgments concerning how they would grasp dowels and pinch “widgets” presented in different orientations (motor imagery condition), and to actually perform these same actions (motor control (MC) condition). In addition to blocks consisting of all dowels and widgets, subjects also performed motor imagery and motor control tasks in mixed block conditions requiring the selection of imagined hand position on a trial-by-trial basis. We reasoned that because both the orientation and shape of the stimulus objects varied, these mixed block conditions would require subjects to plan both hand orientation and hand configuration. This should place greater demands on internal representations of hand movements than the blocked conditions, where there was no need to plan hand configuration across trials. To evaluate the accuracy of these judgments, response preference was compared with those postures chosen by subjects when actually reaching to grasp these same stimuli in the same orientations. Finally, participants performed a short battery of praxis, reaching, and language measures to assess the association between these and their Congruence score from the experimental task.