motor skill acquisition - freee.guigon.free.fr/rsc/article/newell91.pdf · 216 newell power law is...

Annu. Rev. Psychol. 1991.42:213-37Copyright © 1991 by Annual Reviews Inc. All rights reserved

MOTOR SKILL ACQUISITION

K. M. Newell

Department of Kinesiology, University of Illinois at Urbana-Champaign, Urbana,Illinois 61801

KEY WORDS: motor learning, coordination, motor control, information, representation andaction

CONTENTS

INTRODUCTION ..................................................................................... 213A GENERAL LAW OF MOTOR LEARNING ................................................. 215

Motor Learning as a Power Law .............................................................. 216The Power Law--A Special Case of Motor Learning? .................................... 217

WHAT IS LEARNED WITH PRACTICE ....................................................... 219One-To-One Recall-Recognition Processes .................................................. 219Schema Representation of Action .............................................................. 221Problems with Prescriptive Accounts of Motor Learning ................................. 223Motor Skill Acquisition as a Search Strategy ............................................... 224

INFORMATION AND MOTOR SKILL ACQUISITION ..................................... 225Information as a Prescription .................................................................. 226Information as Feedback ........................................................................ 227Information to Channel the Search ............................................................ 231

CONCLUDING REMARKS ........................................................................ 233

INTRODUCTION

Motor skills are usually distinguished from perceptual skills, cognitive skills,communicative skills, and other skill categories; but clearly these traditionaldistinctions have been made as a matter of heuristic convenience. As aconsequence, skill categories reflect primarily differences in scholarly empha-

sis rather than mutually exclusive avenues of scholarly enquiry. The term

213

0066 -43 08/91/0201-0213 $02.00

www.annualreviews.org/aronlineAnnual Reviews

http://www.annualreviews.org/aronline

214 NEWELL

motor skills usually refers to those skills in which both the movement and theoutcome of action are emphasized.

In this chapter I review certain of the major theoretical issues that haveguided the study of motor skill acquisition during the previous 20 years or so.Noble (1968) provided the last related review in this series. An encompassingoverview of a century of motor skill acquisition research is that of Adams(1987).

Traditionally, the study of motor skill acquisition is viewed as distinct fromthe study of the related subdomains of motor control and motor development.Motor learning originated as a branch of experimental psychology and waslabeled accordingly to distinguish it from what used to be called verballearning. The term motor control originated in physiology and was taken torepresent the neurophysiology of the motor system. A behavioral focus withinthe study of motor control was initiated with the influential edited book ofStelmach (1976), which examined the processes that support the control movement. Motor control has since become the predominant theoreticalinterest of researchers with a behavioral interest in motor skills. Furthermore,with physiology increasingly using macrolevel behavioral experimental strat-egies and psychology increasingly using more microlevel experimental strat-egies, it is becoming difficult to draw a line between the physiology andpsychology of motor control---an issue Woodworth (1899) thought was irrele-vant to the study of movement. In contrast, motor development has alwaysbeen viewed as distinct from motor learning and motor control because itexamined children’s motor skills and placed special importance on the studyof phylogenetic movement patterns.

In this review I adhere to the traditional domain distinctions and focus onadult motor skill learning. This decision is made at some conceptual cost,however, because it is becoming increasingly clear that the three subdomainsof study--motor learning, motor control, and motor development--holdconsiderable common theoretical ground (see Wade & Whiting 1986). Thelinkage among these heretofore distinct areas of movement research has beenstimulated by contemporary theoretical developments regarding perceptionand action.

The motor skill acquisition domain also falls on the boundaries of in-structional theory, especially with respect to the role that a change agent (suchas a teacher, instructor, or coach) may play in facilitating the acquisition ofskill. This area of study is sometimes called training, particularly in engineer-ing psychology or human factors research. The contemporary study of motorskill acquisition has tended to deemphasize the role of instructional concepts,implicitly focusing instead on self-generated motor performance enhancementwithin a variety of contextual and task constraints.

The basic actions of posture, locomotion, and manipulation (and their



MOTOR SKILL ACQUISITION 215

variations) allow the learner to engage in a variety of motor skills defined by wide range of task constraints (e.g. those found in athletic, musical, in-dustrial, military, self-help, and vocational contexts). Differences in taskconstraints have promoted the establishment of separate study domains inboth theory and practice. Certain activities, such as those found in sport, tendto emphasize the natural whole-body actions of posture and locomotion,whereas other activities, such as those found in military man-machine con-texts, tend to emphasize the manipulative activities of the hands, sometimeswith the direct minimization (or even elimination) of the role of movement the task. Models and theories of motor skill acquisition have tended, as aconsequence, to be task and hence context specific. One theme of this reviewis that a general understanding of task constraints is necessary for the develop-ment of broad theoretical perspectives on the acquisition of motor skills.

Like many domains of study in psychology, motor skill acquisition hasbeen influenced, albeit somewhat indirectly, by the information processingframework of the 1950s and 1960s and the subsequent development of generalcognitive perspectives on action. Accordingly, I review the major contribu-tions of these theoretical perspectives to some key issues in motor skillacquisition. These issues include: the applicability of a general law to motorskill acquisition; the question of what is learned as reflected in the acquisitionof representations of action; and the role of information in motor skilllearning. A more recent theoretical influence in motor skill learning researchhas been the ecological theory of perception and action. Once more relevantto the study of motor control, this theory now also applies to issues in motorlearning.

Currently, there is no prevailing theoretical view of motor skill acquisition;indeed, there has not been one since Hull’s theory fell from favor during the1950s. This chapter, therefore reflects the theoretical eclecticism of the last 20years, but it also indicates significant issues that a general theory of motorskill acquisition must address.

A GENERAL LAW OF MOTOR LEARNING

Investigators of motor skill acquisition have continued to eschew all-incompassing theories of learning. However, the principle that skill learningis continuous has remained an important (although often implicit) generalproposition. The established continuous functions created from the learningcurves for perceptual-motor skills have been reexamined and compared withthe learning functions for a range of cognitive skills. As a consequence of thissynthesis, A. Newell & Rosenbloom(1981) have proposed that the power lawfunction is a general law of learning. A contrasting perspective holds that the



216 NEWELL

power law is simply a special case of the general nonlinear integro-differentialform of learning (Shaw & Alley 1985).

Motor Learning as a Power Law

It has been known since at least the study of Snoddy (1926) that performancetime, when considered as the task criterion in perceptual-motor skills, tends todecrease with practice as a function of a power law. This finding has beenreplicated in a number of motor performance tasks. The most well-knownexample of the power law function for performance time is Crossman’s(1959) between-subject demonstration of the reduction in the operator’scycle-time in making 10,000,000 cigars over a 7-year period.

A. Newell & Rosenbloom (1981) have discerned the power law functionfor practice effects in skills beyond those usually classified as perceptual-motor. Indeed, they showed that the general power law fits the practice datamuch better than exponential functions for a range of cognitive tasks in whichperformance time is the critical dependent variable. They explained the powerlaw description of leaming across tasks with a chunking model of informationprocessing in skill learning, based in part upon Miller’s (1956) classic accountof inlbrmation capacity limitations.

The generality of the power law for practice over a range of performancetasks leads naturally to the proposal that it is a universal law of learning notlimited to a particular behavioral subdomain; but several limitations of thepower law interpretation must be addressed. First, the power law function istypically demonstrated in tasks where time is the dependent variable. There islittle evidence of a power law for other motor performance variables. Second,a power law cannot accommodate practice effects where the task dependentvariable is on an ordinal scale, such as producing a given set of relativemotions. In this situatibn, the qualitative properties of the coordination modemay change from trial to trial, leading to discontinuous changes over practicetime in performance measures. Third, even if performance changes on onevariable (even the task criterion variable) as a power law, parallel qualitativechanges may occur on other dimensions of performance.

These potential limitations of power law interpretations for both motor skillacquisition and learning in general have not been examined directly owing to(a) the narrow range of task constraints currently used to examine motorperformance; (b) the fact that multiple dependent variables are rarely mea-sured in motor learning studies; and (c) the fact that long-term practice studiesare now rarely conducted, even in the motor skills domain. In spite of thesereservations, the power law function for learning has gained a considerablefoothold as the most robust and best-known feature of motor learning (Logan1988; Salmoni 1989). Indeed, Logan (1988) has remarked that any theory motor skill acquisition that does not accommodate the power law function forlearning can be rejected immediately.




The Power Law--A Special Case of Motor Learning?

A different approach is to view the power law for practice effects in per-ceptual-motor skills as a special case of learning. That is, in spite of theapparent generality of the power law for practice effects across tasks, studieshave still only examined tasks in a narrow segment of the potential state spaceof learning. In this view, learning is discontinuous or nonlinear, and theconstraints of the experimental tasks chosen for study have allowed only thelinear portions of that state space to be reflected in the emergent motorperformance. Thus, n~rrow task selection may have served to confirm the apriori theory of a continuous state space for learning that has been ex-emplified, for example, in information processing models of performance. Inthe same way, narrow theorizing may have confirmed the a priori selec, tion ofhighly constrained perceptual-motor and cognitive laboratory tasks for ex-amination. Theory construction and task selection in motor learning havebeen mutually supporting but limiting endeavors (Newell 1989; Newell et al1989).

The traditional associative theories of learning (Guthrie, Hull, Thorndike),upon which early motor learning theories were predominantly based, causallyspecified cumulative continuous changes in behavior. The strength of therepresentation of movement, such as an S-R bond, was seen to change in acontinuous fashion. The more recent motor learning theories of Adams (1971)and Schmidt (1975) have implicitly reiterated this position by accepting thegradual build-up, over practice, of the strength of their respective memoryconstructs for movement control. Thus the prevailing theoretical bias,together with the accompanying limitation of research tools that emphasizelinear analysis techniques (Greeno 1974), has helped sustain continuity as thenull hypothesis of the laws of learning.

Shaw & Alley (1985) have outlined, from an ecological perspective perception and action, a discontinuous or nonlinear approach to consideringthe laws of learning. They treat perception and action as dual entities in themathematical sense that the values assumed by the function for perceptionconstrain the course of values assumed by the function for action, and viceversa. Learning is the lawful operation that increases the coordination be-tween the perception and action functions. This proposition treats learning asa functional rather than a function because the learning-to-learn source ofvariance that coordinates the perception and action dual is a function offunctions. Shaw & Alley draw on the field of hereditary mechanics to providea physical analogy to model the dissipative nature of the learning functionaland propose that it will prove to be nonlinear. Within this broader nonlinearview of the laws of learning, the continuous changes in motor performanceare considered a special case.

The continuity-discontinuity controversy surrounding the laws of learningcan only be examined empirically through consideration of a broader set of



218 NEWELL

task constraints than have been manipulated in traditional and contemporarymotor learning studies. The tests to date of the laws of learning have beenbiased toward confirming the continuity hypothesis as a consequence of thefew tasks selected (usually single-degree-of-freedom tasks, such as linearpositioning) and the limited amount of practice conducted over a singlesession (often no more than i00 trials). These task and practice limitationstypically require the subject to learn only the scaling of movement amplitude,movement time, or force output in an already established coordination mode(Newell 1985). Rarely studied are the qualitative shifts in the movementdynamics that occur early in the exploratory phase of practice, or the quali-tative change.s that occur late in practice as a function of the flexible andadaptive qualities of the skilled performer.

Examinations of Shaw & Alley’s (1985) proposal of nonlinear integro-differential learning functions for motor skill acquisition require an enrichedperceptual-motor environment rather than an impoverished one. An enrichedperceptual environment, following Gibson (1966, 1979), has several in-formational invariants available that could be utilized to organize a stablecoordination mode. In the same way, an enriched action environment, follow-ing Kugler & Turvey (1987), has several coordination modes that may sufficeto provide stable solutions to the task constraints. The mapping of theperceptual informational invariants to the movement kinetic invariants reflectsthe learning or search through the perceptual-motor workspace.

Shaw & Alley’s (1985) proposal emphasizes motor skill acquisition as thelearning of the laws that map the dynamics of perception and action, ratherthan the memory intensive rule-based procedures that reflect what have beenloosely called the laws of learning. No direct tests of the Shaw & Alleyhypothesis have been conducted, although qualitative changes in the move-ment dynamics as a function of practice have recently been demonstrated inlearning to write (Newell & van Emmerick 1989), throw darts (McDonald al 1989), juggle (Beek 1989), and ride a ski-simulator (van Emmerick et 1989). The examination of tasks with a richer perceptual-motor environmentthan provided by the traditional single-degree-of-freedom tasks (such as linearpositioning) is likely to open the door to a more ecologically relevant descrip-tion.and explanation of motor skill acquisition.

As a contrast to Logan’s (1988) statement about the primacy of the powerlaw, I suggest that a theory that cannot explain the qualitative discontinuouschanges in movement dynamics with practice will lack generality. The ex-perimental strategy of examining the discontinuities of practice effects willalso afford an understanding of their continuities. Current evidence suggeststhat the reverse theoretical and empirical strategy, which has dominated themotor skill acquisition field for the last 100 years, has failed to offer generalaccounts of the laws of learning.




WHAT IS LEARNED WITH PRACTICE

A fundamental assumption of traditonal and most contemporary theories ofmotor skill acquisition is that learning is a consequence of the acquisition ofmore appropriate representations of action. That is, the improved perfor-mance over time is due to the acquisition of prescriptions for action thatspecify the movement dynamics in relation to the task demands. In general,theoretical propositions about the representation of action have shifted overthe last two decades from a one-to-one to a one-to-many relation betweenwhat is represented and the movement sequence that is produced.

One-To-One Recall-Recognition Processes

A major influence on the motor skill acquisition domain was Adam’s (1971)closed-loop theory of motor learning, which proposed a two-state representa-tional scheme for the learning of self-paced positioning movements. Onememory state was the memory trace, whose role was to select and initiate themovement. The strength of the memory trace was seen as a function ofstimulus-response contiguity; it grew as a function of practice trials. Thesecond memory state was the perceptual trace which was the image of thecorrectness of the desired movement; it was based upon prior experience ofthe sensory consequences (both exteroceptive and proprioceptive) of action.Each movement created a set of associated sensory consequences, and thedistribution of these sensory consequences from practice trials formed amodal perceptual trace or memory reference for matching the task demands athand. In single-degree-of-freedom positioning movements the memory tracehad a limited role (what Adams called a modest motor program), and the skillin the task in essence resulted from the formation of an appropriate perceptualtrace to evaluate the on-going sensory consequences of movement.

Adam’s proposal of two independent memory states for movement initia-tion and movement evaluation was based primarily on two considerations.First was the logical argument that error hulling closed-loop procedures couldnot occur if a single mechanism both initiated and evaluated the movement.This was because the sensory consequences of the on-going movement wouldalways be compared against a representation of itself. Second, evidence in theverbal learning domain that recall and recognition could be independentlymanipulated by certain learning variables supported the proposition that theseprocesses were based on independent memory states. The theory thus pro-posed dual one-to-one memory processes for the recall and recognition ofmovement.

Adam’s theory stimulated a number of empirical studies of the two-statememory proposal for motor learning in positioning tasks (Adams & Goetz



220 NEWELL

1973; Adams et al 1972). Other investigators branched out from the confinesof slow positioning movements and examined the recall and recognitionprocesses that support rapid short-duration movements, because such move-ments allowed a cleaner operational distinction between movement initiationand evaluation processes (Newell 1974; Schmidt & White 1972). Theseexperiments demonstrated that the development of movement recognitionprocesses over practice trials with knowledge of results (KR) paralleled thosethat had traditionally been found for movement recall processes.

A direct test of the two-state recall and recognition memory proposal formovement was conducted by Newell & Chew (1974). After sufficient prac-tice with KR to develop the movement recall and recognition processes,withdrawal of visual and auditory feedback of the movement producedan immediate decrement in movement recognition but not in recall. Thisdifferential effect of the feedback variables on recall and recognition pro-cesses was taken as evidence for a two-state memory system for movementcontrol.

The concept of independent recall and recognition processes for movementcontrol was preserved in Schmidt’s (1975) schema theory of motor learning(see below). The direct empirical examination of the dual-state memoryconcept for motor learning did not, however, receive subsequent empiricalattention. The general influence of the Adams’s (1971) theory of motor skilllearning faded rapidly with the arrival of schema theory, in part becauseAdams’s theoretical contributions were largely encapsulated within the sche-ma theory. Furthermore, theoretical emphasis in motor control switched fromclosed-loop to open-loop processes (Keele & Summers 1976; Schmidt 1976)where the role of recognition in both motor control and motor learning wasdownplayed.

The beginnings of the ecological approach to perception and action wereintroduced via a challenge both to the two-state concept and to representation-al accounts of action in general. Turvey (1974) proposed that the problemsfacing perception theorists and action theorists were very similar. Con-sequently, similar principles were needed for their solution that would prob-ably prove indigenous to neither. Fowler & Turvey (1978) subsequentlyoutlined a new approach to motor skill acquisition that was based on theemerging ecological perspective on movement coordination and control (Tur-vey 1977; Turvey et al 1978). A key principle of this perspective was thatcoordination is a relation defined over the organism and the environment, andthat control is the exclusive prerogative of neither. This idea provided a majorchallenge to prescriptive theories that viewed motor skill learning as theaccumulation of more appropriate representations of action, even if theserepresentations were of the one-to-many variety, such as proposed by schematheories.




Schema Representation of ActionSchmidt’s (1975) theory of motor learning retained the independent recall andrecognition mechanisms for movement advocated by Adams (1971) but gavethem a generalized (one-to-many) memory construct through the concept the schema. The schema is seen as a rule that represents the relations betweenvariables rather than the absolute instantiations of the variables themselves.The schema is an old concept in psychology and neurology (e.g. Bartlett1932; Head 1920) but its links to movement were indirect until Schmidt(1975) merged the intuitions of Pew (1974) on schema and motor learningwith the two-state closed-loop theory of Adams (1971).

Two primary theoretical problems motivated the proposal for generalizedmemory states in motor learning (Pew 1974; Schmidt 1975). First was thestorage problem; how many representations of motor programs and closed-loop references could the CNS store? Although this concern was used insupport of the schema concept, there was no evidence on this issue. Secondwas the novelty problem; how could a given motor program in a one-to-onememory framework generate new movement configurations? The schemaconcept finessed these two theoretical problems. Generalizable schema rulesreduced the representation demands on the memory stystem and provided thenecessary principles to accommodate new aspects of movement dynamics.

In Schmidt’s (1975) theory the relative contributions of the recall andrecognition schemata to movement output varied with the task constraints, butthe recall schema clearly played a stronger role in determining movementoutput than in Adams’s (1971) theory. The strengths of the recall and recogni-tion schemata were postulated to be built up over practice trials and feedback.The recall schema was based upon the past experience of the relationsbetween the actual outcome and the response specifications. The recognitionschema was based on the relations between the initial conditions for action,the movement-produced sensory consequences, and the actual outcomes(KR). The generalized rule allowed the production of so-called new move-ments within the class of movements for which the schema was established. Amajor limitation to examining the schema concept in terms of the acquisition,transfer, and retention of motor skill was that no principles were establishedabout what represented a class of movements.

Schmidt (1975) proposed that the schemata rules became more representa-tive of the movement class if a range of movement conditions were ex-perienced in practice. This principle gave rise to the hypothesis that retentionand transfer would be facilitated by variable practice in acquisition, becauseunder these conditions the schema rule was both more impervious to decayand more generative in execution. The concept of variability was not limitedhere to the natural variability that a subject produces in movement dynamicsover repeated attempts to attain the same criterion but, in addition, included



222 NEWELL

the variability that accrues from structured practice under a range of taskconditions.

The evidence for the benefits of variable practice has largely come fromsingle-degree-of-freedom positioning or timing tasks in which variations inthe amplitude or movement-time task constraints were made in both training(usually KR acquisition trials) and transfer (often no-KR trials). For example,Newell & Shapiro (1976) showed that variable practice in a timing taskfacilitated transfer to a criterion movement time that was outside the range ofthe initial movement-time practice conditions. McCracken & Stelmach (1977)also found transfer benefits in a timing task from variable practice over arange of amplitudes in producing the criterion movement time. Other sim-ilarly designed studies have reported either weak trends for the benefits ofvariable practice in transfer (e.g. Wrisberg & Ragsdale 1979) or no effects all (e.g. Zelaznik 1977).

The early evidence for the benefits of variable practice on the transfer ofmotor skill was not strong (Newell 1981). Some schema studies also tended confound the manipulation of variability of practice with differences in thesimilarity between acquisition and transfer task criteria. This is a significantproblem because similarity of task stimulus-response conditions is the corner-stone of most traditional accounts of motor skill transfer (Holding 1976).Shapiro & Schmidt (1982) have suggested that the evidence for the benefits variable practice is stronger in young children than in adults. Certainly anumber of motor learning studies in children have demonstrated the facilita-tive effects of variable practice on subsequent transfer performance (e.g.Kelso & Norman 1978; Moxley 1979), but no study has conducted a directtest of the interaction of age and variable practice.

It appears that the structure of the variable practice schedule is veryimportant in determining the resultant benefits in transfer (Lee et al 1985; Lee1988). Random practice at a range of task criteria affords better transfer thanblocked practice over the same range of practice conditions. Lee (1988) hasgiven a transfer-appropriate processing interpretation to these and relatedtransfer findings. He posits that learning is optimal when the processingactivities promoted by the practice conditions are similar to the processingactivities required by the transfer test. This view shifts transfer away from thesimilarity of task stimulus-response properties to similarity of informationprocessing activities--an interpretation concordant with recent attentionaltheories (not reviewed here) of the acquisition of skill (Schneider & Fisk1983; Schneider & Shiffren 1977; Shiffren & Schneider 1977).

The process of schema formation is poorly understood. Schmidt (1975) andKeele & Summers (1976) assumed that the schema rule was based on theinvariant relations of the movement dynamics, which were independent of thespecific muscle groups involved. These invariant characteristics of a move-



ment sequence may not, however, require representation. Indeed, the demon-stration of movement invariances in a range of tasks may be a reflection ofemergent properties of some other level oLorganization of the motor system.

Motor learning needs to be reconsidered within a broader framework ofhow the learner solves the motor problem. This has been a general propositionof cognitive approaches to skill acquisition, but the topic has not beenexamined systematically from this point of view. The change in performancewith practice needs to be examined on an individual-subject basis overindividual practice trials. The analysis-of-variance approach to examiningperformance on blocks of trials does not sufficiently inform about the changein behavior. There have been few studies of the between-trial performance indiscrete movement tasks.

Another schema view of action was outlined by Rumelhart & Ortony(1977), an extension of their schema theory for comprehension and knowl-edge representation. By virtue of its attempt to accommodate the transferbetween subactions or classes of movement, Rumelhart & Ortony’s theoryprovided the basis of a view of knowledge and action broader than that offeredby Schmidt (1975). On the other hand, its strong roots in the comprehensiondomain led to its isolation from the motor skill acquisition community, andthe theory failed to attract empirical or theoretical attention. A similar fatebefell other schema-like cognitive theories either directly or indirectly con-cerned with motor skill acquisition (Anderson 1976, 1982; Arbib 1980;Norman & Rumelhart 1975; Rumelhart & Norman 1982). It remains to beseen in what way the connectionist approach to cognition will influence thestudy of human motor skill acquisition.

Problems with Prescriptive Accounts of Motor Learning

The theoretical perspectives on motor learning reviewed above are all pre-scriptive in the sense of having representational schemes at some level ofanalysis that prescribe the movement sequence in relation to the task con-straints. Learning is viewed as the acquisition of prescriptions for action thatwill more appropriately satisfy the realization of the task goal. The fun-damental differences among the preceding accounts of motor skill acquisitionlie in the nature of the representations they posit, which are seen as reflectionsof what is learned with practice.

Prescriptive approaches to motor skill learning have been challenged on anumber of different grounds over the last two decades. The primary challengecame from the enaerging ecological approach to perception and action (Fowler& Turvey 1978; Kugler et al 1980; Kugler & Turvey 1987; Turvey & Kugler1984). Central concems have been the appropriateness of rule-based accountsof action and the logical difficulties of mapping symbols and dynamics in aprincipled fashion (Carello et al 1984). Advocates of the ecological



224 NEWELL

position seek the solution in the mapping of perception and action withminimal resort to intelligent operations. Where representation is invoked, it isto be fashioned from law-based dynamic accounts of perception and actionrather than from a discrete nonholonomic (non-integrable) symbol systemlogically separate from the dynamics. This central concern with prescriptivetheories of motor skill learning gives rise to a number of subsidiary problems.

Schema theories of motor learning cannot account for the acquisition ofnew coordination modes or movement forms. As Kugler et al (1980) noted, is logically difficult for rule-based schema theory to account for the in-stantaneous production of quadruped-like locomotion by centipedes after allbut two pairs of their legs are amputated. Expressed another way, where doesthe centipede’s instantaneous representation of a quadrupedal gait originate?In principle, the schema rule can only accommodate the new scaling of anestablished coordination mode, and even here, the logic for generalizability ofthe changing movement d.ynamics is questionable.

Prescriptive theories of motor learning also have trouble handling logicallythe compensations to perturbations of an ongoing movement sequence. Howdoes a template-reference-of-correctness concept account for the in-stantaneous, anatomically distant-in-time-and-space, and functionally specif-ic compensations evident, for example, in sudden and novel perturbations tothe jaw in ongoing speech production (Kelso et al 1984)? Expressed anotherway, where does the closed-loop representation of correctness arise thatenables compensation during novel and unexpected perturbation?

The challenge of the ecological position to extant prescriptive accounts ofmotor skill learning has undoubtedly weakened the influence of information-processing and cognitive accounts of motor skill acquisition in the 1980s. Onthe other hand, Adams’s (1971) theory was limited by design to a very narrowset of task constraints, and this narrowness was fundamentally the cause of itsdemise. The potential generality to motor skill acquisition of the Schmidt(1975) schema theory was considerably broader but it only stimulated empir-ical activity on the variability of practice issue. Thus, the Adams and Schmidttheories were already a waning influence in the motor skill acquisition domainby the time the challenge to the prescriptive views arrived.

Motor Skill Acquisition as a Search Strategy

Bernstein’s (1967) insights into coordination strongly influenced the develop-ment of the ecological approach to action. Bernstein (1967:127) viewed "thecoordination of movement as the process of mastering redundant degrees offreedom of the moving organ, in other words its conversion to a controllablesystem." The process of practice was characterized as the search for theoptimal motor solutions to the problem at hand. It is important to note thatpracticing was seen as repeating the solving of the motor problem rather thanrepeating a particular solution to thee problem.




Consistent with the Bernstein proposals, Fowler & Turvey (1978) in-terpreted skill acquisition as the search for the optimization of the coordina-tion and control function of several variables. Search strategies reflect the waythe perceptual-motor workspace is explored to "solve" the motor problem(Newell et al 1989b). The perceptual-motor workspace is the interface be-tween the relatively high-energy movement-kinetic field and the relativelylow-energy information-kinematic flow field-~an interface that arises fromthe complementary influences of the perception-action cycle (Kugler & Tur-vey 1987; Shaw & Alley 1985). Learning is the coordination of the perceptualenvironment with the action environment in a way consistent with the taskconstraints.

In this perspective, Gibson’s (1966, 1979) insights about the informationalproperties that organize the perceptual environment are extended to thecomplementary action environment. The dynamic organization of the per-ceptual-motor workspace can be examined through defining the layout of thegradient and singular properties of the perceptual-motor fieldlike spaces thatsupport the macrolevel coordination pattern. The search through this work-space can be analyzed via established search and optimization proceduresfrom biology (Gelfand & Tsetlin 1962) and physical systems.

The significance of this orientation for motor skill acquisition is that itpromises to provide a principled way to accommodate the adaptive nature ofthe dynamics of movement control without resort to the computation-intensive procedures advanced by the prescriptive accounts of skill learningreviewed above. The theoretical and experimental challenge becomes one ofidentifying critical perceptual and kinetic variables that are being exploited tochannel the search for the appropriate mapping of information and movementkinetics in the perceptual-motor workspace. The promise is that there arerelatively global macrolevel variables of few degrees of freedom that organizethe many microlevel degrees of freedom harnessed in support of action.

A central hypothesis of the search strategy approach (Newell et al 1989b) motor skill acquisition is that the learning, retention, and transfer of skill indifferent tasks is dependent on the similarity among the correspondingsearches through the equilibrium regions of the perceptual-motor workspace,and relatively independent of the specific effector and manipulanda utilized.The role of search strategies in motor skill acquisition, both in tasks where theperceptual-motor workspace can be specified a priori (Krinskii & Shik 1964)and in the more natural tasks where the workspace can only be modeled posthoc, is currently being examined.

INFORMATION AND MOTOR SKILL ACQUISITION

Information facilitates the changes in motor performance that reflect learning.Different types of information are used in motor skill acquisition, and there



226 NEWELL

have been a number of theoretical interpretations of the role of information.This section is organized around three themes: information as a prescription;information as feedback; and information to channel the search through theperceptual-motor workspace.

Information as a Prescription

The prescriptive accounts of motor skill learning hold that informationstrengthens the development of the respective memory constructs for action.The development of task-relevant prescriptions for action can be strengthenedthrough the presentation of prior-to-movement information that specifies theto-be-produced outcome and movement dynamics. Instructions or demonstra-tions convey this prescriptive information (Newell 1981; Newell et al 1985a).Recent empirical work has focused on the role of demonstrations in motorskill: learning.

The .general evidence in support of the facilitating effect of demonstrationsin motor skill acquisition is not strong. This is in part because experimentshave tended to use tasks already familiar to the learner. In effect, theinformation conveyed via the demonstration is often redundant to the learner’stask-relevant knowledge.

Bandura’s (1977, 1986) social learning theory has stimulated a systematicset of empirical studies on the role of demonstrations in motor skill learning.The theory offers a generative schema-like account of motor learning in whichthe spatial and temporal elements of the movement are symbolically codedthrough perceptual cues. The coding of the perceptual cues allows the de-velopment of the reference against which movement may be successivelymodified by appropriate feedback, and used for covert rehearsal techniques.Thus, in this view feedback information is only useful in learning when theappropriate movement reference has been developed.

Carroll & Bandura (1982, 1985, 1987) have examined aspects of thesetheoretical ideas in a series of experiments that required subjects to learnseparate arm and hand postures with specific movements between postures asa function of different modeling and visual control conditions. The resultshave shown that -visual feedback is not useful in the early trials of learning thepostural sequences, but it facilitates learning subsequently in the practicesequence; delayed visual monitoring of the just-produced movement does notaffect the acquisition process; and the stronger the movement representation(as determined by independent procedures) the more accurate are the subse-quent recognition and reproduction of the action patterns. The use of a taskthat requires the production of a novel movement coordination sequence wasinstrumental in revealing these systematic effects of observational learning.However, early in practice, learners may not always be able to produce thenew coordination mode demonstrated by the model (Martens et al 1976).



MOTOR SKILL ACQUISITION ’ 227

Movement related demonstrations can be made to the learner via sensorysystems other than the visual system. Newell (1976a) and Zelaznik et (1978) have shown that auditory demonstrations of the sound associated withrapid movements can effectively convey information about the task move-ment dynamics. Indeed, in these auditory demonstration protocols, subjectsreduced their timing error over a series of practice trials in the absence of KR.

Demonstrations can also provide information about procedural aspects ofthe task demands that are not directly related to the movement dynamics. In arecent study of different types of augmented information in learning a video-game task, Newell et al (1989a) found that the demonstration of gameprocedural information was more effective in improving game performanceover a 10-hr training period than information and specific practice on theisolated components of the movement dynamics. This finding reveals anotherway task properties mediate the nature of the appropriate information tosupport motor skill acquisition.

A major challenge for the motor skill acquisition domain is to understandthe nature of the information conveyed in a demonstration. Bandura’s (1977,1986) social learning model fails to address this issue and, in effect, holds thatall aspects of the movement dynamics are in some way coded in a-memoryconstruct. Schmidt’s (1975) schema theory proposed that the relative motioninvariances are stored in memory, but no perceptual recognition tests of thishypothesis have been conducted. A perceptual orientation to demonstrationsthat attempts to understand what information for action is conveyed by amodel offers a new approach to this problem (Newell 1985; Scully & Newell1986).

Information as Feedback

Information feedback is available to the learner both during the ongoingmovement sequence (concurrent feedback) and on completion of the move-ment sequence (terminal feedback). In either case, information can be pro-vided about the outcome of the movement (KR) and/or some aspect of theongoing movement dynamics (sometimes called knowledge of performance).Information feedback can also be naturally available through the inherentproperties of the task constraints, or it can be supplemented via augmentedinformation (i.e. information not normally available from engagement in thetask). The distinction between natural and augmented information feedback isnot absolute and is inevitably task dependent. The working assumption of thefeedback literature has been that the natural and augmented types of informa-tion feedback operate on the same principles, but no direct tests of thisintuition have been conducted.

CONCURRENT INFORMATION FEEDBACK There is a long tradition of study-ing the influence of concurrent information feedback on motor skill learning



228 NEWELL

(Annett 1969; Armstrong 1970). Many studies have demonstrated the facilita-tive effect of concurrent information feedback on motor performance. Thecontribution of information from the different sensory systems, whether it isinherent in the task or augmented, varies with the particular task constraints.The positive influence of augmented concurrent information feedback onmotor performance is often negated once the supplementary informationfeedback is removed.

The Adams (1971) and Schmidt (1975) theories of motor skill acquisitionincorporated concurrent information feedback into closed-loop accounts ofmotor learning. In both of these theoretical formulations the ongoing sensoryconsequences to movement were used in a closed-loop error detection andcorrection framework. The more information available from sensory channelsthe stronger and more representative was the development of the respectiverecognition memory state, which led to a better performance during both KRacquisition trials and when KR was withdrawn.

Adams et al (1972) proposed that the development of the perceptual tracefor movement recognition allowed the movement to be executed accuratelythrough concurrent information feedback without external informational sup-port in the form of KR. The perceptual trace provided the evaluation of themovement internally--a process that Adams (1971) called subjectivereinforcement. Adams et al (1972) provided evidence for this proposition showing that positioning movements were more accurate, and that the learnerwas more able to evaluate the correctness of the movement, when all thesensory channels were available.

The question of what information is available from different sensory sys-tems during movement was never addressed in the closed-loop accounts ofmovement control. This limitation, together with the increased emphasisgiven to centralist accounts of motor control during the 1970s and 1980s,contributed to the decline in the study of augmented concurrent informationfeedback. Concurrent information feedback can clearly have a potent impacton performance according to the task constraints and the nature of theinformation provided.

TERMINAL INFORMATION FEEDBACK The presentation of information oncompletion of the movement sequence has traditionally proved to have a verystrong influence on motor skill acquisition (Adams 1971; Bilodeau 1966;Newell 1976b). KR of the outcome of the action has continued to be studiedduring the past 20 years, but there has also been a new emphasis on informa-tion about the dynamics of the just-completed movement.

Adams’s (1971) closed-loop theory of motor learning gave a strong role KR as information in strengthening the two-state memory process. Both therecall and recognition states were postulated to be strengthened over KR




practice trials. Schmidt’s (1975) schema theory proposed similar learningprinciples in regard to the necessity and contiguity of KR in motor learning.The informational interpretation of KR was examined by studies that provideddirect tests of the processing of KR during motor learning.

Rogers (1974) examined the idea that different precision levels of should have differential effects on learning according to the minimal amountof time allowed the learner for processing the information during the post-KRinterval. The findings from a micrometer positioning task showed that in-creased precision of KR up to some point facilitated learning, beyond whichdecrements in performance occurred. The most beneficial level of KR preci-sion could be changed by varying the duration of the time for informationprocessing during the post-KR interval. Thus higher levels of KR precisioncould be used effectively with more time for information processing. Similarmanipulations have also shown that older children can more effectively utilizemore precise levels of KR precision (Newell & Kennedy 1978).

Single-degree-of-freedom positioning or timing tasks do not require muchtime to process the relevant KR provided. For example, Barclay & Newell(1980) showed that children of 10 years of age only used about 1.5 sec in self-paced post-KR interval of a timing task study. This finding demonstratesthat the manipulations of the post-KR interval have generally been too long toinduce information processing effects (Boucher 1974; Magill 1973).

The information processing hypothesis has also been tested by imposingsecondary task activity during either the KR-delay interval or the post-KRinterval. The basic rationale for this manipulation was that a competingsecondary task restricts the capacity remaining for processing the KR.Boucher (1974) showed that reading 4- and 5-syllable words during thepost-KR interval produced a detrimental effect in learning a positioning task.In contrast, Magill (1973) failed to find interference effects as a consequenceof inserting counting backwards by 3s during the post-KR interval of anangular positioning task. Marteniuk (1986) showed that the influence information processing activity during the KR-delay interval depended on therelative difficulty of the task and the secondary activity. Thus, the findings insupport of the information processing idea that the subject actively operateson the KR information are suggestive rather than decisive, and are stronglyinfluenced by task properties.

The general interpretation of KR studies has been challenged by Salmoni etal (1984), who argue that learning effects for KR can only be inferred if theperformance difference from practice with KR is sustained during a subse-quent no-KR test phase, such as on a retention test. They have picked up onan earlier finding by Lavery (1962), who showed that while absolute frequen-cy rather than relative frequency of KR presentation was the variable thatdetermined performance level when KR was available, the reverse effect was



230 NEWELL

apparent when performance was subsequently examined over a series ofno-KR trials. Using a number of experimental protocols, Schmidt and col-leagues (Schmidt et al 1989; Wulf & Schmidt 1989) have provided evidencethat the presentation of KR on every trial may not be the most effective KRschedule if performance is to be subsequently evaluated under no-KR con-ditions.

This proposal for the benefits of relative versus absolute KR effect probablyonly holds once the learner has produced a performance that is in the ballparkof the task goal. In other words, the intermittent schedule is more appropriateto the maintenance of performance than to the acquisition of new performancestates. Furthermore, performance under no-KR conditions is only one possi-ble scenario for transfer and retention tests and therefore should not be takenas the single measure of motor learning. The findings of Schmidt and col-leagues clearly suggest some modification to the traditional interpretation ofthe frequency effects of KR. However, they do not require the formation ofnew laws of KR as proposed by Salmoni et al (1984).

KR is very effective in single-degree-of-freedom tasks or tasks where thescaling of a given coordination pattern is all that is required to satisfy the taskconstraints. The usefulness of KR to a learner in acquiring whole bodyactions, or in tasks where the learner needs to establish a stable coordinationmode for the task at hand, has been increasingly questioned over the last 20years (Fowler & Turvey 1978; Gentile 1972; Newell & Waiter 1981). In thissituation the learner requires knowledge of performance, or information aboutthe dynamics of the just-produced movement, in addition to KR of theoutcome of the action.

Fowler & Turvey (1978) suggested that the information required in thefeedback must contain as many degrees of constraint as there are degrees offreedom in the action to be coordinated. This proposal attempts to explainwhy the single degree of constraint provided by KR is sufficient in single-degree-of-freedom positioning or timing tasks. Newell & McGinnis (1985)suggested a framework by which to determine what information is required bya learner in a given task situation. This framework requires an understandingof the sources of constraint upon action, particularly the role of task con-straints (Newell 1986). A number of experimental demonstrations of how taskconstraints determine the nature of information feedback required by thelearner have been provided. It has also been shown that in many taskconditions, the use of kinematic and kinetic information feedback facilitatesmotor learning and performance beyond those reached by means of thepresentation of KR alone (Newell & Carlton 1987; Newell et al 1985b;Newell et al 1983, 1987).

The experimental study of kinematic and kinetic information feedbackhas been limited to one and two-degree-of-freedom task constraints. The




generalization of this framework to whole body multiple-degree-of-freedomtasks requires an increased understanding of the nature of what is beingregulated in the coordination mode. Furthermore, it needs to be understoodthat both KR and knowledge of the just-produced dynamics are informationfeedback and only tell the learner what has happened in regard to themovement dynamics and the outcome. These forms of information feedbackdo not directly inform the learner about what and how the action should bechanged on the next trial. This limitation upon information feedback isparticularly evident when a change in the qualitative properties of thecoordination mode is required. In this situation a different form of in-formational support for motor learning is required and is discussed in the nextsection.

Information to Channel the Search

Information in the ecological approach to perception and action is interpretedas the means via which the learner channels the mapping of information andmovement dynamics in the perceptual-motor workspace in a way consistentwith the task demands (Kugler & Turvey 1987). After Gibson (1966, 1979), is assumed that the invariant properties of the environment act as informationto guide the exploratory activity of the learner. These informational propertiesare qualitative in nature and attune the learner to the layout of the perceptual-motor workspace. One important aspect of perceptual learning is the contin-ued differentiation by the learner of the properties of the perceptual-motorworkspace. The natural learner-generated search of the perceptual-motorworkspace can be supplemented with various forms of augmented informa-tion, as described previously, to facilitate the search strategy.

The idea of a natural search through the workspace by the learner isconsistent with the traditional concept of discovery learning. The evidencesuggests, however, that self-discovery does not always enable the learner tolocate a task-appropriate mapping of information and dynamics in the per-ceptual-motor workspace. Furthermore, even on the occasions where self-discovery affords attainment of the task goal, the process of learning or thesearch behavior can be very inefficient.

Information can be used, therefore, to channel the search through theperceptual-motor workspace to locate a task-relevant solution to the coordina-tion function. This theoretical framework involves a three-component con-sideration of augmented information and skill learning (Newell 1990). Thefirst component is understanding the nature of the perceptual-motor work-space in terms of the attractor equilibrium and gradient regions. The secondcomponent is understanding the natural search strategies used by learners toexplore the space. The third component is the application of augmented



232 NEWELL

information to facilitate the search. These three components are in-terdependent. This orientation provides a new look at the strengths andweaknesses of traditional prescriptive and feedback accounts of augmentedinformation techniques.

Demonstrations provide some information about the nature of the desiredequilibrium set of the perceptual-motor workspace, but they do not inform thelearner how to navigate through the space to arrive at the task-relevantsolution. In static analogy consider the problems of a traveler being given amap marked only with her current location and final destination. Furthermore,demonstrations do not accommodate the individual nature of the layout of theperceptual-motor workspace.

Information feedback, such as knowledge of results, only informs thelearner of performance error in relation to the task criterion. This can be veryeffective when the perceptual-motor workspace supporting that activity islinear--a condition well approximated in the traditional laboratory tasks thatproduce the power law for learning. However, feedback cannot provide directinformation about how one might search the nonlinear portions of the work-space to produce the qualitative changes necessary to realize new coordinationmodes.

It has been suggested that a new class of augmented information is requiredto promote systematic qualitative changes in the coordination mode (Newellet al 1985a; Newell, in press). This information category was labeled transi-tion information. In effect, this information acts as another source of con-straint on action in anticipation of producing a qualitative change in thecoordination mode. This kind of information should prove valuable in theearly stages of acquisition where the learner is attempting to find a new stableequilibrium region in the perceptual-motor workspace.

Instructors of physical activities often provide this information throughinstructions. The beginning golfer, for example, may be told "Keep yourelbow in." The instructor does not intend the learner to keep the elbow in tothis degree after she has attained the desired coordination configuration.Rather, this informational constraint acts as a control parameter to change theconfiguration of the coordination mode. Thus the nature of the informationrequired by the learner seems to depend on the stage of learning. Thisinteraction has not been examined.

Perceptual-motor skills in context have rich sources of information avail-able, but the traditional operational strategy in motor skill acquisition ex-periments has been to strip away from the learner the support of this informa-tion and construct impoverished environments in which skill learning is totake place. This experimental strategy has burdened the learner by providingthe constraints to learning. The result has been an emphasis on cognitiveoperations. The emphasis of the ecological approach to information and motor




learning is intimately tied to understanding the natural dynamics of theperceptual-motor workspace.

CONCLUDING REMARKS

The information processing and cognitive frameworks have had considerableimpact on concepts of motor skill acquisition domain over the last twodecades. In many respects, however, this influence has been indirect. Theinformation processing approach has been concerned primarily with perfor-mance, not learning. It has emphasized the processes that support perform-ance. Only the attentional accounts of skill learning (not reviewed here) havetreated issues of direct relevance to the changes that occur with practice.

The information processing approach has also given little or no emphasis tothe question of what information is processed in motor skill acquisition. Thefocus has been on the how of information processing, with no direct examina-tion of the informational support required from the learner. The schema viewof motor learning promised a new way to examine the role of information inskill learning but failed to stimulate any empirical activity on this importanttopic. The revival of interest in learning from a cognitive perspective duringthe 1980s has largely been oriented to so-called cognitive tasks.

The strong influence of task constraints on motor skill learning is a keypoint to emerge from the foregoing synthesis. It is usually accepted that skillis specific. This conclusion may arise from the fact that even small changes inthe experimental task constraints can lead to large changes in the performanceof the learner. The ecological approach to perception and action has offeredthe beginnings of a new way to consider task analysis: in terms of theperceptual-motor workspace.

The traditional approaches to motor skill acquisition have failed to capturemany of the dynamic qualities of the stages of motor skill acquisition exhib-ited by novice and expert performers. Skill is a reflection of a dynamicexploratory activity, not the stereotypic reproduction of a static representationof action. Current views of skill learning, such as those embedded in a powerlaw view, have failed to capture the richness of the essence of skill and thefullness of the constraints that shape it. The effort to understand the ecologi-cally relevant aspects of task constraints, in relation to the dynamic interfaceof information and movement, opens the door to a more general theory ofmotor skill acquisition.

ACKNOWLEDGMENTS

The preparation of this paper was supported in part by NIH grant HD 21212. Iwould like to thank Richard van Emmerik and Vernon McDonald for helpfulcomments on an earlier version of the paper.



234 NEWELL

Literature Cited

Adams, J. A. 1971. A closed-loop theory ofrnotor learning. J. Mot. Behav. 3:111-50

Adams, J. A. 1987. Historical review andappraisal of research on the learning, reten-tion, and transfer of human motor skills.Psychol. Bull. 101:41-74

Adams, J. A., Goetz, E. T. 1973. Feedbackand practice as variables in error detectionand correction. J. Mot. Behav. 5:217-24

Adams, J. A., Goetz, E. T., Marshall, P. H.1972. Response feedback and motor learn-ing. J. Exp. Psychol. 92:391-97

Anderson, J. R. 1976. Language, Memoryand Thought. Hillsdale, NJ: Edbaum.546 pp.

Anderson, J. R. 1982. Acquisition of cogni-tive skill. Psychol. Rev. 89:369-406

Annett, J. 1969. Feedback and Human Be-havior. Baltimore: Penguin. 196 pp.

Arbib, M. A. 1980. Interacting schemas formotor control. In Tutorials in Motor Be-havior, ed. G. E. Stelmach, J. Requin, pp.71-82. Amsterdam: North-Holland. 680 pp.

Armstrong, T. R. 1970. Feedback aud per-ceptual-motor learning: a review of in-formation feedback and manual guidancetraining techniques. Tech. Rep. No. 25,Hum. Perform. Cent., Univ. Michigan.66 pp.

Bandura, A. 1977. Social Learning Theory.Englewood Cliffs, NJ: Prentice Hall. 247PP.

Bandura, A. 1986. Social Foundations ofThought and Action: A Social CognitiveTheory. Englewood Cliffs, NJ: PrenticeHall. 617 pp.

Barclay, C. R., Newell, K. M. 1980. Chil-dren’s processing of information in motorskill learning. J. Exp. ChiM. Psychol.30:98-108

Bartlett, F. C. 1932. Remembering: A Study inExperimental and Social Psychology. Lon-don: Cambridge Univ. Press. 317 pp.

Beek, P. J. 1989. Juggling Dynamics. Am-sterdam: Free Univ. Press. 222 pp.

Bernstein, N. A. 1967. The Coordination andRegulation of Movement. London: Per-gamon Press. 196 pp.

Bilodeau, I. McD. 1966. Information feed-back. In Acquisition of Skill, ed. E. A.Bilodeau, pp. 255-96. New York: Academ-ic. 539 pp.

Boucher, J.-L. P. 1974. Higher processesin motor learning. J. Mot. Behav. 6;131-37

Carello, C., Turvey, M. T., Kugler, P. N.,Shaw, R. E. 1984. Inadequacies of the com-puter metaphor. In Handbook of CognitiveNeuroscience, ed. M. S. Gazzaniga, pp.229-48. New York: Plenum. 416 pp.

Carroll, W. R., Bandura, A. 1982. The role ofvisual monitoring in observational learn-ing of action patterns: making the unobserv-able observable. J. Mot. Behav. 14:153--67

Carroll, W. R., Bandura, A. 1985. Role oftiming of visual monitoring and motor re-hearsal in observational learning of actionpatterns. J. Mot. Behav. 17:269-81

Carroll, W. R., Bandura, A. 1987. Translat-ing cognition into action: the role of visualguidance in observational learning. J. Mot.Behav. 19:385-98

Crossman, E. R. F. W. 1959. A theory of theacquisition of speed-skill. Ergonomics2:153-66

Fowler, C. A., Turvey, M. T. 1978. Skillacquisition: an event approach with specialreference to searching for the optimum of afunction of several variables. In InformationProcessing in Motor Control and Learning,ed. G. E., Stelmach, pp. 1-40. New York:Academic. 315 pp.

Gelfand, I. M., Tsetlin, M. L. 1962. Somemethods of control for complex systems.Russ. Math. Surv. 17:95-116

Gentile, A. M. 1972. A working model ofskill acquisition with application to teach-ing. Quest 17:3-23

Gibson, J. J. 1966. The Senses Considered asPerceptual Systems. London: George Allen& Unwin. 335 pp.

Gibson, J. J. 1979. The Ecological Approachto Visual Perception. Boston: HoughtonMifflin. 332 pp.

Greeno, J. G. 1974. Representation of learn-ing as a.discrete transition in a finite statespace. In Contemporary Developments inMathematical Psychology. Vol. 1: Learn-ing, Memory, and Thinking, ed, D. H.Krantz, R. C. Atkinson, R. D. Luce, P.Suppes, pp. 1-43. San Francisco: Freeman.299 pp.

Head, H. 1920. Studies in Neurology, Vols. I& II. New York: Macmillan. 862 pp.

Holding, D. H. 1976. An approximate transfersurface. J. Mot. Behav. 8:1-9

Keele, S. W., Summers, J. J. 1976. The struc-ture of motor programs. In Motor Control:Issues and Trends, ed. G. E. Stelmach, pp.109-42. New York: Academic. 232 pp.

Kelso, J. A. S., Norman, P. E. 1978. Motorschema formation in children. Dev. Psy-chol. 3:529-43

Kelso, J. A. S., Tuller, B., Vatikiotis-Bateson, E., Fowler, C. A. 1984. Func-tionally specific articulatory cooperationfollowing jaw perturbations during speech:evidence for coordinative structures. J. Exp.Psychol.: Hum. Percept. Perform. 10:812-32




Krinskii, V. I., Shik, M. L. 1964. A simplemotor task. Biophysics 9:661-66

Kugler, P. N., Kelso, J. A. S., Turvey, M. T.1980. On the concept of coordinative struc-tures as dissipative structures. I. Theoreticallines of convergence. In Tutorials of MotorBehavior, ed. G. E. Stelmach, J. Requin,pp. 3-48. Amsterdam: North-Holland.68O pp.

Kugler, P. N., Turvey, M. T. 1987. Informa-tion, Natural Law, and the Self-Assembly ofRhythmic Movement. Hillsdale: Erlbanm.481 pp.

Lavery, J. J. 1962. Retention of simple motorskills as a function of type of knowledge ofresults. Can. J. Psychol. 16:300-11

Lee, T. D. 1988. Transfer-appropriate pro-cessing: a framework for conceptualizingpractice effects in motor learning. In Com-plex Movement Behavior: The ’Motor-Action’ Controversy, ed. O. G. Meijer, K.Roth, pp. 201-15. Amsterdam: North-Holland

Lee, T. D., Magill, R. A., Weeks, D. J. 1985.Influence of practice schedule on testingschema theory prediction in adults. J. Mot.Behav. 17:283-99

Logan, G. D. 1988. Toward an instancetheory of automatization. Psychol. Rev. 95:492-527

Magill, R. A. 1973. The post-KR interval:time and activity effects and the relationshipof motor short-term memory theory. J. Mot.Behav. 5:49-56

Marteniuk, R. G. 1986. Information process-ing in movement learning: capacity andstructural interference effects. J. Mot. Be-hav. 18:55-75

Martens, R., Burwitz, L., Zuckerman, J.1976. Modelling effects on motor perfor-mance. Res. Q. 95:277-91

McCracken, H. D., Stelmach, G. E. 1977. Atest of the schema theory of discrete motorlearning. J. Mot. Behav. 9:193-201

McDonald, P. V., van Emmerik, R. E. A.,Newell, K. M. 1989. The effects of practiceon limb kinematics in a throwing task. J.Mot. Behav. 21:245-64

Miller, G. A. 1956. The magic number sevenplus or minus two: some limits on ourcapacity for processing information. Psy-chol. Rev. 63:81-97

Moxley, S. E. 1979. Schema: the variabilityof practice hypothesis. J. Mot. Behav.11:65 70

Newell, A., Ros6nbloom, P. S. 1981.~Mech-anisms of skill acquisition and the law ofpractice. In Co~,nitive Skills and TheirAcquisitioh, ed. J. R. Anderson, pp. 1-55.Hillsdale: Erlbaum. 386 pp.

Newell, K. M. 1974. Knowledge of resultsand motor learning. J. Mot. Behav. 6:235-

Newell, K. M. 1976a. Motor learning without

knowledge of results through the develop-ment of a response recognition mechanism.J. Mot. Behav. 8:209-17

Newell, K. M. 1976b. Knowledge of resultsand motor learning. In Exercise and SportScience Reviews, ed. J. Keogh, R. S. Hut-ton, 4:195-228. Santa Barbara: JournalPublishing Affiliates. 394 pp.

Newell, K. M. 1981. Skill learning. In HumanSkills, ed. D. H. Holding, pp. 203-26. NewYork: Wiley. 303 pp.

Newell, K. M. 1985. Coordination, con-trol and skill. In Differing Perspectives inMotor Learning, Memory, and Control,ed. D. Goodman, R. B. Wilberg, I. M.Franks, pp. 295-317. Amsterdam: Elsevier.340 pp.

Newell, K, M. 1986. Constraints on the de-velopment of coordination. See Wade &Whiting 1986, pp. 341 60

Newell, K. M. 1989. On task and theoryspecificity. J. Mot. Behav. 21:92-96

Newell, K. M. 1990. Augmented informationand the acquisition of skill. In Motor Learn-ing and Training, ed. R. Daugs, K. Bliscke.Schormdorf: Hofmann. In press

Newell, K. M., Carlton, M. J. 1987. Aug-mented information and the acquisition ofisometric tasks. J. Motor Behav. 19:4--12

Newell, K. M., Carlton, M. J., Fisher, A. T.,Rutter, B. G. 1989. Whole-part trainingstrategies for learning the response dynam-ics of microprocessor driven simulators.Acta Psychol. 71:197-216

Newell, K. M., Chew, R. A. 1974. Recall andrecognition in motor learning. J. Mot. Be-hav. 6:245-53

Newell, K. M., Kennedy, J. A. 1978. Knowl-edge of results and children’s motor learn-ing. Dev. Psychol. 14:531-36

Newell, K. M., Kugler, P. N., van Emmerik,R. E. A., McDonald, P. V. 1989. Searchstrategies and the acquisition of coordina-tion. In Perspectives on the Coordination ofMovement, ed. S. A. Wallace, pp. 85-122.Amsterdam: Elsevier. 455 pp.

Newell, K. M., McGinnis, P. M. 1985.Kinematic information feedback for skilledperformance. Hum. Learn. 4:39-56

Newell, K. M., Morris, L. R., Scully, D. M.1985. Augmented information and theacquisition of skill in physical activity. InExercise and Sport Sciences Reviews, ed.R. L. Terjung, 13:235-61. New York: Mac-millan. 596 pp.

Newell, K. M., Quinn, J. T., Carlton, M. J.1987. Kinematic information feedback andtask constraints. Appl. Cogn. Psychol,1:273-83

Newell, K. M., Quinn, J. T., Sparrow, W.A., Walter, C. B. 1983. Kinematic in-formation feedback tbr learning a rapid armmovement, Hum. Mov. Sci. 2:255-70

Newell, K. M., Shapiro, D. C. 1976. Vail-


236 NEWELL

ability of practice and transfer of training:some evidence toward a schema view ofmotor learning. J. Mot. Behav. 8:233-43

Newell, K. M., Sparrow, W. A., Quinn, J. T.1985. Kinetic information feedback forlearning isometric tasks. J. Hum. Mov.Stud. 11:113-23

Newell, K. M., van Emmerik, R. E. A. 1989.The acquisition of coordination: preliminaryanalysis of learning to write. Hum. Mov.Sci. 8:17-32

Newell, K. M., van Emmerik, R. E. A.,McDonald, P. V. 1989. On simple move-ments and complex theories (and vice-versa). Behav. Brain Sci. 12:229-30

Newell, K. M., Walter, C. B. 1981. Kinemat-ic and kinetic parameters as informationfeedback in motor skill acquisition. J. Hum.Mov. Stud. 7:235-54

Noble, C. E. 1968. The learning of psy-chomotor skills. Annu. Rev. Psychol. 19:203-50

Norman, D. A., Rumelhart, D. E., and theLNR Research Group, eds. 1975. Explora-tions in Cognition. San Francisco: Freeman.430 pp.

Pew, R. W. 1974. Human perceptual-motor~ "performance. In Human Information Pro-

cessing: Tutorials in Performance andCognition, ed. B. H. Kantowitz, pp. 1-39.Hillsdale: Erlbaum. 365 pp.

Rogers, C. A. 1974. Feedback precision andpostfeedback interval duration. J. Exp. Psy-chol. 102:604-8

Rummelhart, D. E., Norman, D. A. 1982.Simulating a skilled typist: a study of skilledcognitive-motor performance. Cognit. Sci.1:1-36

Rummelhart, D. E., Ortony, A. 1977. Therepresentation of knowledge in memory. InSchooling and the Acquisition of Knowl-edge, ed. R. C. Anderson, R. J. Spiro, W.E. Montegue, pp. 99-135. Hillsdale:Erlbaum. 448 pp.

Salmoni, A. W. 1989. Motor skill learning. InHuman Skills, ed. D. H. Holding, pp. 197-227. New York: Wiley. 334 pp.

Salmoni, A. W., Schmidt, R. A., Walter, C.B. 1984. Knowledge of results and motorlearning: a review and critical reappraisal.Psychol. Bull. 95:355-86

Schmidt, R. A. 1975. A schema theory ofdiscrete motor skill learning. Psychol. Rev.82:225-60

Schmidt, R. A. 1976. Control processes inmotor skills. In Exercise and Sport Sciences_Reviews, ed. J. Keogh, R. S. Hutton,4:229-61. Santa Barbara: Journal Publish-ing Affiliates. 394 pp.

Schmidt, R. A., Shapiro, D. C., Young, D.E., Swinnen, S. 1989. Summary knowledgeof results for skill acquisition: support

for the guidance hypothesis. J. E:~p. Psy-chol.: Learn. Mem. Cogn. 15:352-59

Schmidt, R. A., White, J. L. 1972. Evidencefor an error detection mechanism in motorskills: a test of Adams’ closed-loop theory.J. Mot. Behav. 4:143-54

Schneider, W., Fisk, A. D. 1983. Attentiontheory and mechanisms for skilled perfor-mance. In Memory and Control of Action,ed. R. A. Magill, pp. 11944. Amsterdam:North-Holland. 395 pp.

Schneider, W., Shiffren, R. M. 1977. Con-trolled and automatic human informationprocessing. I. Detection, search, and atten-tion. Psychol. Rev. 84:1-66

Scully, D. M., Ncwell, K. M. 1986. Observa-tional learning and the acquisition of motorskills: toward a visual perception perspec-tive. J. Hum. Mov. Stud. 11:169-86

Shapiro, D. C., Schmidt, R. A. 1982. Theschema theory: recent evidence and de-velopmental implications. In The Develop-ment of Movement Control and Coordina-tion, ed. J. A. S. Kelso, J. E. Clark, pp.113-50. New York: Academic. 370 pp.

Shaw, R. E., Alley, T. R. 1985. How to drawlearning curves: their use and justifica-tion. In Issues in the Ecological Studyof Learning, ed. T. D. Johnson, A. T.Pietrewicz, pp. 275-304. Hillsdale: Erl-baum. 451 pp.

Shiffren, R. M., Schneider, W. 1977. Con-trolled and automatic human informationprocessing. II. Perceptual learning, auto-matic attending and a general theory. Psy-chol. Rev. 84:127-90

Snoddy,(3. S. 1926. Learning and stability. J.Appl. Psychol. 10:1-36

Stelmach, G. E., ed. 1976. Motor Control:Issues and Trends. New York: Academic.232 pp.

Turvey, M. T. 1974. A note on the relationbetween action and perception. In Psycholo-gy of Motor Behavior and Sport, ed. M. G.Wade, R. Martens, pp. 307-13. Cham-paign: Human Kinetics. 356 pp.

Turvey, M. T. 1977. Preliminaries to a theoryof action with reference to vision. In Per-ceiving, Acting and Knowing, ed. R. Shaw,J. Bransford, pp. 211-66. Hillsdale:Erlbaum. 492 pp.

Turvey, M. T., Kugler, P. N. 1984. An eco-logical approach to perception and action.In Human Motor Actions--Bernstein Revis-ited, ed. H. T. A. Whiting pp. 373-412.Amsterdam: North-Holland. 633 pp.

Turvey, M. T., Shaw, R. E., Mace, W. 1978.Issues in the theory of action: degrees offreedom, coordinative structures and coali-tions. In Attention and Performance, ed. J.Requin, 7:557-95. Hillsdale: Erlbaum. 730PP.

van Emmerik, R. E. A., Brinker, B. P. L. M.,




Vereijken, B., Whiting, H. T. A. 1989.Preferred tempo in the learning of a grosscyclical action. Q. J. Exp. Psychol.41A:251-62

Wade, M. G., Whiting, H. T. A., eds. 1986.Motor Development in Children: Aspects ofCoordination and Control. Boston: Marti-nus Nijhoff. 555 pp.

Woodworth, R. S. 1899. The accuracy ofvoluntary movements. Psychol. Rev.Monogr. Suppl. 3:1-114

Wrisberg, C. A., Ragsdale, M. R. 1979.Further tests of Schmidt’s schema theory:development of a schema rule for a coinci-dent timing task. J. Mot. Behav. 11:159-66

Wulf, G., Schmidt, R. A. 1989. The learningof generalized motor programs: reducing therelative frequency of knowledge of resultsenhances memory. J. Exp. Psychol.: Learn.Mem. Cognit. 15:748-57

Zelaznik, H. N. 1977. Transfer in rapid timingtasks: an examination of the role of variabil-ity in practice. In Psychology of Motor Be-havior and Sport, ed. D. M. Landers, R.W. Christina, pp. 3(~43. Champaign: Hu-man Kinetics. 286 pp.

Zelaznik, H. N., Shapiro, D. C., Newell, K.M. 1978. On the structure of motor recogni-tion memory. J. Mot. Behav. 10:313-23



motor skill acquisition - freee.guigon.free.fr/rsc/article/newell91.pdf · 216 newell power law is...

Documents