TOWARD A THEORY OF APRACTIC SYNDROMES* J. A. Scott Kelso+ and Betty Tuller+ Abstract. Theory development on human motor behavior has, for the most part, occurred independently of data on pathological movement disorders. This paper is an initial attempt to interface findings from studies of apraxia and those of normal motor behavior in order to formulate a common theoretical framework. Such an integration should further the understanding of the nature of skill acquisition and provide insights into the organization of motor systems. Three theoretical models of movement control are discussed with reference to apractic syndromes. The most commonly accepted view--the hierar chy--involves, for example, linear transitivity and unidirectionali ty of information flow, properties that render it an inadequate explanation of functional plasticity in the central nervous system. The heterarchy, incorporating reciprocity of function and circular transitivity, is a more likely candidate but cannot regulate the degrees of'freedom of the system. Our favored candidate is the coalition model, which embodies heterarchical principles, but in addition, offers a solution to the problems of degrees of freedom and context for motor systems. Evidence is reviewed from apraxia of speech and limbs in terms of a coalitional style of control, and an experimental approach, consonant with coalitional organization, is developed. We promote the claim that an understanding of apractic behavior--and perhaps motor systems in general--will benefit when clinicians and experimenters embrace a theory of context and con straints rather than a theory of commands such as those in vogue. INTRODUCTION It is an interesting but perhaps distressing feature of science that two different areas of study, each bearing a strong potential relation to the other, can function independently, each in its own oblivion. Except for one or two isolated cases· (e.g., Grimm & Nashner, 1978; Roy, 1978), such a situation appears to exist between those who would seek to understand the motor functions of the central nervous system via investigations of clinical disorders and those who seek to understand the underlying behavioral processes * A preliminary version of this paper was given as an invited presentation to the International Neuropsychology Symposium, Oxford, England, 1978. To appear in Brain and Language. +Also University of Connecticut, Storrs. Acknowledgment: This work was supported by NINCDS grant NS-13617, BSRG grant RR-05596 and NIH grant AM-25814. We wish to thank K. S. Harris, K. Poeck and D. Shankweiler for comments on an earlier version of this paper. [HASKINS LABORATORIES: Status Report on Speech Research SR-61 (1980)] 175 involved in the acquisition of skill and the control of movement in normal human populations. The motor behavior area has undergone a good deal of theoretical development in the last decade (t:;.g , Adams, 1971; Pew, 1974; Schmidt, 1975; Turvey, 1977), but y enough, with a total for of movement. The latter , have ramifications comprehending motor and the organization of motor s (e.g , Geschwind, 1975) The so-called syndromes seem relevant in this , for they arise as a result of cerebral insults that interfere with the generation and elaboration of voluntary movement. An analysis of apraxia therefore affords a unique opportunity for us to derive important insights into the neural control and coordination of learned movements. Perhaps a major reason for the absence of a viable interface between the research findings of normal and pathological movement behavior is the absence of a theoretical framework that a style of organization appropriate to both Neither closed-loop nor motor program models--currently popular in the motor behavior area (e.g, Kelso & Stelmach, 1976 for review)--address the problem of controlling a complex system containing multiple degrees of freedom Thus they are unlikely candidates for an adequate s of apractic behavior. In this paper, we will promote a reconceptualization of apraxia based on a perspective on coordinative movement developed primarily by Soviet theorists (e.g., Bernstein, 1967; Gelfand, Gurfinkel, Fomin, &Tsetlin, 1971), and under both theoretical (e.g., Greene, 1972; Turvey, 1977) and empirical (Kelso, Southard, & Goodman, 1979a, 1979b; Nashner, 1976; Nashner & Grimm, 1978) elaboration in this country. In brief, the style of control that we shall propose is coalitional; namely, one in which apraxia is viewed, not simply as a breakdown in function of the nervous system itself (cf. Brown, 1972; Hecaen, 1968), but rather as a decomposition of the syngergistic relationships that hold between the organism and its environment. Current views of apraxia (and indeed motor behavior in general) are based on hierarchical notions of central nervous system organization. This perspective, we shall argue, offers only a limited conceptual framework for analyZing apractic syndromes, being a partial systems approach that fails to account for some important and intriguing phenomena in apraxia. The first part of this paper will be directed towards sUbstantiating this claim and elaborating a coalitional style of control that for us is the minimal organization possessing functional integrity (Turvey, Shaw, &Mace, 1978). This leads us, in the second part, to a reinterpretation of the nature of apraxia and consequent proposals for experimentation that may yield significant dividends in the quest for understanding how actions are coordinated and controlled. Hierarchical Styles of Motor Organization Let us first consider some commonly accepted organizational concepts of the motor system that we feel warrant careful scrutiny and reexamination. This is an often-ignored, but important preliminary step since our conceptual ization of pathology is likely to correspond closely to our views of central nervous system (CNS) organization. The notion that the motor system is hierarchically organized has a long history that stems from Hughlings 176 Jackson's (1976 ) initial observations and insights; it forms the basis for much of our thinking about the nature of skill and the control of movement. There have been numerous recent expressions of the hierarchical viewpoint (e.g., Bruner, 1973; Connolly, 1977), but the basic idea is explicitly defined in Tinbergen's (1950) description of the nervous system as "...Higher centres controlling a number of centres at a next lower level, each of these in turn controlling a number of lower." Hierarchical assumptions form the basis for investigations in modern neurophysiology and neurology as well as providing an interpretative backdrop. For example, in answer to the question of where the assembly of "action programs" takes place in the CNS, Brooks (1974) suggests, on anatomical grounds, that we look to the limbic system for the drive to move--and thence to frontal and parietal cortex for formation of the needed association. From there the output would be channeled via the cerebellum and basal ganglia through the ventrolateral nucleus of thalamus to the motor cortex, and then finally to the spinal cord and the muscles. Similarly, in clinical neurology circles, Geschwind (1975) seeks an anatomical (and hierarchical) framework for the analysis of movement organization in response to verbal commands. Since Wernicke's area is responsible for spoken language comprehension, when the subject receives an instruction to execute a right handed movement, "...this order is probably transmitted from Wernicke's area through the lower parietal lobe to the left premotor region. The premotor region, in turn, probably controls the precentral motor cortex, which gives rise to the pyramidal tract--a major pathway for motor control--which sends fibers to the spinal cord, where it activates the nerve cells controlling the muscles" (p. 189). These examples clearly illustrate a style of organization that obeys hierarch ical principles (or more accurately, linear chaining). Thus the role each element plays in the anatomical chain is fixed and the ordering of dominance relations is immutable. More recently, Roy (1978) has related the presumed hierarchical structure of the nervous system to the functional disorders observed in apraxia. At the top of the hierarchy, damage to frontal and parietal-occipital areas of cortex results in the disordered planning and sequencing of motor acts characteristic of ideational and ideomotor apractics (Liepmann, 1920). At the next level, lesions in pre-motor cortex lead to an inability to execute the movement sequence properly (premotor apraxia) even though, according to Roy (1978), the patient is still able to plan the motor activity since the frontal areas and the pathways to premotor cortex are preserved. At the lowest level of the hierarchy comes limb-kinetic apraxia--disruption of individual movements with in a sequence--which is thought to occur following trauma to the primary areas of the motor cortex. In all three of the above examples and in the last particularly, we see an effort to relate the functional organization of the nervous system to its neuroanatomical substrate. But the argument from anatomy does not require presupposing a style of organization that is based on hierarchical principles. Although any characterization of movement must be consonant with anatomical fact, the hierarchical description provides an inadequate representation of both anatomical organization and the functional deficits observed in apractic disturbances. We can buttress this claim on several grounds. Consider first the logic behind attempts to relate apractic behavior to site of brain damage. 177 Its premises are as follows. (1) Certain lesions in the central nervous system are associated with certain functional deficitse (2) The effect(s) of the lesion depends on its locus in a neuroanatomi cally structured hierarchy, i.e@, the higher or lower the lesion, the of lesser the of behavioral deficit (3) Therefore, the organization of the system is also hier- The latter bances reinforce the need to Yt organi into models of motor skill" (Roy, 1978) on a curious tautoiogy. Although we are not that structure and function are unrelated, we submit that the conclusion of a functional hierarchy in apractic and indeed in normal behavior is predicated upon the a 9 9 assumption that the motor system is hierarchically organized. But the notion that the movement control system is hierarchi organized subject to more serious flaws. While hierarchical organization offers economy of control as its principal quality, it pays for this dearly. Thus in the conventional view, information in the eNS undergoes continuous transformation into progressively less abstract levels of descrip tion as it flows through the system from the 'hierarch' to the peripheral musculature.1 In this manner, the system may use a single degree of freedom on the central level to regulate many degrees of freedom at the periphery. Whilst an efficient means of reducing the degrees of freedom in a system, hierarchical principles of CNS function obviously dictate unidirectionality of information flow (Davis, 1976). Assuming, for example, that two structures, ~ and X, are arranged in a hierarchy at two different levels, the higher system ~ will always command X; the opposite cannot occur. Furthermore, if ~ and X are at the same level in the hierarchy, there can be no cross-talk between them (Turvey, Shaw, &Mace, 1978). In other words, in a hierarchy there is no provision for internal communication within central nervous system networks, yet the evidence for unidirectional pathways is shaky at best (e.g., Brown, 1972; Kelso, 1979; Roland, 1979), while the evidence for such 'feedforward' or 'internal feedback' throughout the CNS is virtually unassailable (see Evarts, 1971; Kelso &Stelmach, 1976, for reviews). A further, and related problem with hierarchical control is that prestige rests with the highest level in the system, i.e., the so-called executive plan or program.2 Although this is a tempting approach to conceptualizing apractic disorders (which are, after all, 'higher-order' in nature), it has severe limitations if taken seriouSly.3 Hierarchies, by definition, dictate that once the superordinate state is lost (say by brain lesions), all subordinate portions governed by it will be left uncontrolled regardless of whether the role of the superordinate state is viewed as excitatory or inhibitory. The absence of excitation or the release from inhibition would render the system unmanageable. In fact, disruption at any level in a hierarchically organized system affects the functioning of all those below it. This shortcoming arises because the relation, 'is governed by' or 'controls,' follows the principle of linear transitivity in hierarchical networks. 178 But there are data on apractic patients that illustrate the inadequacy of such a view. For example, the abnormally long stretches of jargon speech associated with one form of Wernicke's aphasia have been explained as due to a relatively isolated, and therefore free-running, Broca's area (Geschwind, 1969). But this notion implies a totally random stringing together of elements that is not only an inaccurate description of the grammatical patterns of fluent aphasics, but does not explain why the initial phrase of sentences is often produced correctly (Buckingham & Kertesz, 1974; see also Buckingham, 1979). Similarly, damage to frontal cortex--which typically is assigned superordinate status in the planning of actions (e.g Luria, 1973; Of Milner, 1964)--should invariably lead to gross disruption of motor activity, if the hierarchical model is valid . Indeed one could think of few better tests of hierarchical organization than to damage the structure thought to contain the 'hierarch'4; motor behavior should disintegrate But this is not the case: So-called 'habitual skills' (Luria, 1966; Roy, 1978), while rarely produced in response to clinicians' requests, are effectively carried out in the proper context. This finding demands an adequate explanation, but for the time being let us emphasize how damaging it is for hierarchical conceptualiza tions of motor skill in general, and for apraxia in particular. To preserve the theory, the hierarchist must provide some rationalization for why damage to the structure involved in planning all acts (in this case the frontal or parietal-occipital area) has different effects on some acts than on others. At a minimum, the theorist must provide some basis for distinguishing 'planned' and 'voluntary' acts from those that are 'habitual' or 'automatic' (cf. DeRenzi, Pieczuro, & Vignolo, 1966). More important, we must understand why certain acts may be performed 'given the proper context' (e.g., Roy, 1978; see also Luria, 1966, 1973). Such a rationale does not exist, nor, from our perspective, can it exist within a hierarchical framework. The language of hierarchical systems is one of command, not context· If the motor system is hierarchically organized, then it is, we would argue, independent. The broader message of this analysis, then, is that hierarchies provide no means for explaining the functional plasticity in biological systems. Apractic deficits, for example, as pointed out by Geschwind (1975), are often difficult to detect because of this very factor. Alternative pathways not normally used can be brought into operation following insults to primary brain mechanisms. Conventional (hierarchical) accounts of apraxia assume plastici ty. But assumed plasticity is incompatible with hierarchical organization because the ability to recruit reserve or back-up structures violates the linear transitivity principle, instead favoring circular transitivity (Turvey et al., 1978). Circular transitivity is a characteristic of heterarchical organization (McCulloch, 1945), a topic to which we briefly turn. Heterarchical Styles of Organization Heterarchies embody many features that contrast directly with hierarchies: Rather than ascribing control to a hierarch, heterarchies are characterized by reciprocity of function. In neurology circles, the concept of reciprocity is similar "to what Luria (1966) termed "functional pluripoten tialism.. Thus, in a heterarchical style of organization, a system--by virtue tv of the reciprocal interconnections amongst its elements--is allowed to assume a variety of roles or functions depending on task demands. Conversely, a particular function may be manifested in a variety of structures; there is no 179 zation of function in a consequence of such distributed function These features of and distributed function--enable a ication of function and the presence of extensive interconnections thus reduce the vulnerabil of the in~ults, and preserve behavioral If ization sour characterization of [is and the limitations are necessary before making a functional correlation motor disturbance." The above statement nicely the tion and shifts the away from the common we believ preoccupation with linking function directly to structures is not to say that a heterarchical of zation is an framework for understand apractic deficits. well-motivated ection of the so-called (Luria, 1973 irectional of information and compartmentalization of function--all of which, strictIy speaking, defy the emergence of functional plasticity, heterarchies bring with them a new set of problems. Perhaps the major one is that the heterarchy--by virtue of its free dominance capabil and the fact that locus of control is free to reside anywhere in the system--is too flexible. Thus, while a hierarchy is an effective solution to regulating the degrees of freedom of a system (admittedly with serious consequences) a heterarchy has the problem of managing a potentially infinite number of degrees of freedom The degrees of freedom problem is not trivial (cf. Bernstein, 1967), belonging as it does to the class of "non-deterministic polynomial-time complete" problems (Lewis & Papdimetrios, 1978) More simply, the time e necessary to regulate a set of independent variables increases as an exponen tial function of the number of variables to be regulated. Thus, for any living system, the cost (in time) of controlling a number of degrees of freedom would outweigh the benefits of heterarchical organization. If coordi nated movement is to follow heterarchical principles, the number of degrees of freedom to be controlled individually must be reduced. The question arises as to how this may be accomplished. One possibility is that a reduction of the degrees of freedom occurs when a set of variables are linked to form self regulating autonomous subsystems (cf. Greene, 1972) Wherever the locus of e control at any given moment, regulation of the entire subsystem entails only one degree of freedom. Moreover it makes no ,sense for variables to be 9 randomly linked to form biologically and behaviorally irrelevant subsystems Rather there must be a principled basis for constraining variables into appropriate functional units. A good candidate from which such constraints may arise, and one that is motivated by descriptions of apraxia, is the 180 situational context within which an act is performed. Given this hypothesis, the traditional dichotomy between "habitual" or "automatic" acts and "planned" or "voluntary" acts becomes less tenable. For us then, understanding how a heterarchical system operates entails understanding how a system may be contextually constrained. We feel that the issue of context has been skirted too long in explanations of apractic behavior. What follows is an attempt to conceptualize the notion of 'context of constraint' as it applies to the functioning of the nervous system (or, more appropriately, animals and humans), and to promote an experimental approach that is consonant with it. What does it mean for ~ system to be contextually constrained? "The meaning of a particular action cannot be explained by a narrow concentration upon the physical movement in isolation. The meaning is given by the context of the action, or complex of actions, of which it can be observed to form a part. Precisely the same physical movements may have quite different meanings, i.e., it may be different actions in different contexts" (Best, 1978). A popular way of conceptualizing 'context of constraint' is in terms of the activation of a motor image or plan (which itself may be either hierarchical or heterarchical) in the rather restricted sense of a stimulus activating a response. Geschwind (1975) , for example, considers a verbal command as an inadequate stimulus for a patient with destruction of the anterior four-fifths of the corpus callosum, in that the experimenter's verbal command cannot reach the patient's right hemisphere. Hence, the verbal stimulus cannot initiate correct responses by the patient's left limbs. In contrast, an object placed in the patient's left hand constitutes a visual stimulus to the right hemisphere, which can evoke the correct movement response. In both cases the movement is considered functionally equivalent regardless of whether it occurs as the response to a verbal or a visual stimulus. By considering verbal commands and situational context as stimuli for functionally equivalent movements, motor apraxias, so it seems, can be understood as a breakdown of the stimulus-response relations that normally hold. Notice that in this view the relationship between the object (the stimulus) and the actor (the response) is not truly interactive but rather is unidirectional, being characterized by an immutable, hierarchical dominance relation. In order to be effective, the stimulus must activate the response via pathways that are responsible for the interpretation of verbal or visual information or by the motor system itself. We wish to support an alternative theoretical perspective--based on work by Bransford, McCarrell, Franks, and Nitsch (1977)--in which a movement (or any event) is not defined independently of the context in which it occurs. In this view, in response to questions like "Who wants to go with me?", "How many oranges do you have?" and "How high can you reach?" the gesture of an t outstretched hand with all five fingers extended has very different meanings. In short, the hand gesture is not functionally equivalent in different contexts. 181 In our view, the significance of a movement, and its functional role, are integral to the process of linking free variables into coordinated sub s Just as the situational context provides boundary conditions or constraints on the possible meaning of the movement, so also do the possible s of the movement provide boundary conditions on the movement's ic form. We will examine this notion in more detail later. Our view of organization, in which actions cannot be considered dently of their context is captured by what may be termed a coalition (see 9 Turvey, Shaw, & Mace, 1978, for detailed discussion) We have gone to some e length in attempting to establish that heterarchical and hierarchical notions consider, in effect, only part of the total system that the movement. In contrast, a coalitional framework stresses the mutual ityor fit between the individual and the environment. Whilst a coalitional of control embodies the advantageous characteristies of heterarchies--namely, free dominance, reciprocity, and distributed function--it possesses the addi tional control advantage of effectively reducing the of freedom of the system.5 Thus, unlike heterarchies where environmental variables are tially indifferent to the organism and vice versa (hence magnifying the degrees of freedom problem) in coalitions the environment is just as j thoroughly organized as the organism and is specific to it (Gibson, 1977; Turvey et al 1978), Thus as Turvey et ale (1978) point out member Of j j of the synergy is properly constrained without the other nor may the total j system be defined without their closure. From our , then, reduc tion of the degrees of freedom is accomplished by the contextual framework that operates as a constraint on possible movements.· Accordingly, the interaction between the individual and the context or environment must be an adaptive one whose fit is functionally defined by the particular behavioral goal. As a consequence, the significance of this interaction must be an important variable in the coalitional system. If we are correct in claiming that a coalition represents the minimal organization that possesses functional integrity, then apractic deficits may be more properly viewed as a breakdown in the synergistic relationship between the individual and the environment as defined by the behavioral goal. To summarize, we view the role of informational support or context in a system as providing boundary constraints on the specifics of an action. Our definition of context is very broad and may be applied to both coarse-grain and fine-grain analyses of the nature of control. Accordingly, the significance of a movement, as well as the specifics of the movement, are a function of the coordinative relationship between any particular movement and a set of contextual boundary constraints. Verbal commands, imitation, and even object use, in our view, leave too many degrees of freedom unconstrained. Context, defined globally and locally, provides boundary conditions that specify exactly how the degrees of freedom of meaning and movement must be constrained. Let us illustrate how this may be the case. Various forms of apraxia (e.g., ideational, ideomotor, constructional) may be characterized broadly as disorders in which the meanings of objects and events are disrupted. We have seen that the meaning of an act in the absence of an appropriate contextual framework is quite different from the of 182 an act embedded in a particular context, even though the kinematic details may be superficially the same and analyzed such by the clinician/experimenter The kinematic sequence exhibited by an apractic patient to hammer a nail in a clinical setting is not functionally equivalent to the (possibly identical) kinematic sequence that occur when actually hanging a The former is extrinsically specified and applies to only a part of the system, namely the patient The latter is specified as a function of the G interactions within the total system; the significance, or meaning is an intrinsic feature of the whole act. EVIDENCE IN SUPPORT OF COALITIONAL CONTROL: TOWARD AN RE- ANALYSIS ~APRAXIA Thus far we have argued that given appropriate context the of an act is uniquely specified. We would like to examine some experimental evidence supporting the notion that contextual constraints serve to specify precisely the parameters of the motor system. This, we believe, forces a novel, but principled approach to the perimental of apraxiao insights into this problem are provided by i, Gurfinkel Pal'tsev's (1967) demonstration that during the reaction time movement, the muscles of the trunk and lower limbs undergo a highly and specific series of changes Note that these muscles to 0 those actually involved in the volitional act, but characteristically change before any actual limb movement. Nevertheless, the postural that occur depend on the requirements of the intended limb movement such that those changes specific to raising a leg cannot be identical to those specific to raising an arm. The requirements of the intended movement the necessary postural adjustments thereby reducing the number of t decisions required. In other words, the boundary constraints applied to the postural organization minimize what Bernstein (1967) called "the of freedom problem." In the Be1en'kii et ale experiment, the complex of postural adjustments is uniquely specified by the nature of the upcoming movemento We also see from the Belen'kii et al. study that the between postural adjustments and the specific limb movement is not one of immutable, unidirectional dominance. The requirements of the limb movement do not simply impose boundary constr~ints on the postural mechanisms. While the latter are indeed specific to a movement, e.g., lifting an arm, they in turn the occurrence of a number of other possible activities. e.g. lifting a As t Fowler (1977) points out, an individual in this state of "feedforward" is constrained to produce one of a limited class of acts. The postural context provides boundary conditions on what movements are possible while fies of the intended movement constrain the postural organization. Hence, the rela tionship between postural adjustments and limb movement must be viewed as at least reciprocal. It remains to appreciate an additional variable in the relati between postural adjustments and limb movements before necessarily desc the style of control as coalitional. We have seen how fics of the intended movement bias the postural system. Nevertheless, the particular movement comprises only part of the contextual framework for postural adjust ments: The nature of the support surface must also constrain such modifica tions. Those changes appropriate for lifting the arm when the individual is 183 on solid terrain are inappropriate for 1 the arm when in \oJater. the coalitional style of control, then, the coordinative constraints that exist between the individual, the activ , and the environment The so-called 'lower-level' adjustments occur before volun- tary movement warrant, in our mind, much more detailed examination Conceivably, some apractic di from brain that the complex of supraspinal influences on in brainstem spinal organization Kots (1977), among examined the changes in spinal organization before and during voluntary movement the excitability of spinal motoneuronal pools. A monosynaptic Hoffman-reflex is elicited by direct electrical stimulation of an afferent nerve. The strength of the reflex provides the information from which one can infer the state of excitability, or gain in the motoneuronal pool. On the basis of empirical evidence, Kots has divided the complex of spinal (and presumably brainstem) changes into three basic processes: pretuning, tuning, triggering. Pretuning occurs before the signal to move and extends throughout the latent period of the movement It involves a "background" increase in the reflex 0 excitability of all motoneuron pools and is generally associated with postural adjustments in anticipation of a movement. The pretuning process is associat ed with processes, in pretuning is absent from the agonist motoneuron pool during the latent period of an elicited reflex Changes in the spinal apparatus fic to the future movement are described by the processes of tuning and triggering. Approximately 50 to 60 msec before the onset of electromyographic activity in the agonist of the impending movement, there is a smooth and progressive tuning increase in the reflex excitability of the motoneuron pool of the agonist. During the last to 30 msec of this interval, the "fast" motoneurons of the agonist show a sharp increase in reflex excitability simultaneous with a depression of the inhibitory interneuronal system acting on the motoneuronal pool of the future agonist (triggering). Are impairments of these processes evident in apractic disorders? In patients with cortically localized pyramidal lesions, the background pretuning change in the motoneuron pool of a paralyzed muscle is absent during any attempt to move the paralyzed limb. However, a small background increase in reflex excitability of the same motoneuronal pool occurs during the latent period of voluntary movement of the healthy limb.' These observations suggest separate supraspinal mechanisms for changes specific to the limb to be moved and those changes having more global consequences. An observation by Geschwind (1975) may be relevant here. He observed that axial movements involving bilateral actions of the eyes, neck, or trunk are often executed correctly by apractic patients in response to a verbal command. In contrast, movements of individual limbs, or of the lips, tongue or larynx cannot be produced. He attributes this to the availability of a non-pyramidal motor'system which, while capable of elegantly executing axial movements, can only roughly perform discrete movements of individual limbs. In the current perspective, we do not wish to dichotomize "pyramidal" and "non-pyramidal" motor command systems, but rather prefer to consider the possibility of selective impairment of cerebral influences on spinal and 184