Handbook of neurological rehabilitation

level of movement coding occurs. For example, a person’s handwriting is equally distinctive whether produced in a miniature note-book or on a flip chart, even though very different combinations of muscles are used. Such phenomena suggest that the brain encodes a motor skill in an abstract form, which Schmidt (1982) termed a schema. However, it seems likely that no single learning process will adequately explain all aspects of a complex motor skill. A general theory of skill acquisition Anderson (1982, 1995) has proposed a theory of skill acquisition which can be applied to non-motor as well as to motor skills. In the cognitive stage the learner approaches the task as an explicit problem to be solved. There are often opportunities to break down the major goal into subgoals. For example someone with hemiplegia learning to stand from a sitting position will first learn to adjust the position of the trunk and feet. Similarly, in the cognitive stage of learning to drive the process of starting the engine and getting into first gear must be memorised as a series of discrete stages. Verbal instructions from a teacher can be crucial at the cognitive stage (Adams, 1971; Anderson, 1982, 1995). It is more efficient (and safer) to tell a new driver which pedal is the brake, rather than waiting for her to discover by trial and error. Similarly, it is helpful to give verbal descriptions in the early stages of recovering sitting balance or learning to use a new walking aid. Later, such facts may play no direct part in executing the skill of driving, standing, or walking. In the associative stage the learner is lumping discrete stages into “condition–action pairs” (one might call them reflexes) of the type: “for the goal of standing up, shift weight forward and tuck feet in” or “for the goal of starting the car, turn the ignition-key and depress the clutch”. The formation of these automatic “production rules” is termed proceduralisation. Procedures lump together subgoals and appear to “cut corners”. Finally, in the autonomous stage, a series of production rules are amalgamated into a programme which represents the skill (for example standing, or driving). Since the cognitive component at this stage is reduced, the task makes less demands on attention and other tasks can be carried out simultaneously. At this stage skills are accomplished not through the conscious use of learned information but through unconscious application of procedures. We have seen that motor skills may be encoded as abstract representations of movement termed “schemata” (see earlier). Anderson’s more general theory of skill-learning postulates the formation of “productions”, a term borrowed from artificial intelligence. Essentially this involves statements of the type “If X . . . then Y”. Rules of this type provide a general structure for skills as diverse as parsing a sentence or riding a bicycle. Factors affecting skill learning Verbal learning. In the cognitive stage of skill learning, all the factors which promote efficient verbal learning are relevant (see earlier). Active learning. Much skill learning is procedural (nonverbal), but this does not imply that optimal learning occurs automatically. Active participation enhances procedural learning (Vakil et al., 1998). Skill learning is enhanced by greater engagement in the task. Not all forms of engagement are helpful, however: Conscious (verbal) analysis of performance can be distracting and can impede learning. Practice. Practice schedules have been studied in many learning situations, including verbal learning (see Baddeley, 1997) and motor skill learning (see Kelso, 1982). The total time hypothesis states that the amount of learning achieved is a function of the total time devoted to learning. Another principle is that distributed practice—learning in multiple sessions—is preferable to fewer more prolonged practice sessions. Fatigue is one obvious reason for using distributed practice schedules; another is that during a prolonged practice session a particular pattern of response, and of error, may predominate and be perpetuated. A third advantage of distributed practice may be that intervals between practice sessions are used for covert or mental rehearsal. There is growing evidence particularly from research on the acquisition of sporting skill, that active mental rehearsal enhances subsequent performance (Garza & Feltz, 1998). Context. Learning is facilitated by varying the performance and the context. At a different practice session the context will be different, other types of performance will occur and the resultant learning be more general. The inherent variability of each performance of a motor skill—the changes in initial conditions—may help to explain why motor learning tends to be more robust than information learning: the encoding conditions are less specific. Feedback: Knowledge of performance and knowledge of results. One prerequisite for improving a motor skill is information about the outcome of each performance, termed knowledge of results (KR). However, knowing that you scored an ace does not in itself improve your tennis serve. KR must be combined in some way with a mechanism for retaining and reproducing good performance and rejecting erroneous trials. Such data will usually be derived firstly from a mental image of the intended movement (feed-forward) and secondly from a memory record of the performance, typically based on visual and proprioceptive feedback. For reasons given in the preceding section on motor learning, these data are likely to be coded in the form of an abstract schema. In motor skills there is a complex relationship between KR and KP (Brisson & Alain, 1997). The processing of KR is one aspect of motor skill learning which is open to cognitively based experiments since it involves acquiring information and competes with LEARNING AND SKILL ACQUISITION 139 other information-learning for working memory resources (Marteniuk, 1986). The issue of KR is relevant to all skills but the interaction between KR and other learning factors is complex (see Swinnen, 1990; Swinnen et al., 1990; Winstein & Schmidt, 1990). In non-motor skills KR and KP contribute similarly to learning. KP is derived from a comparison between intentions and a record of performances (for example chess moves which led to a current position). Positive and negative transfer. In normal learning, specific aspects of one task frequently cause negative effects on related tasks. For example, driving a car whose indicator switch is on the left of the steering column deleteriously affects subsequent driving of a car with the opposite arrangement. In a study of telegraphists, negative transfer was demonstrated when, after learning one letter-symbol code, the telegraphists learned another code. Subsequently, however, performance in the second task benefited from the previous exposure to the original task. The improvement in general aspects of the skill—for example speed of finger or eye movement—represented positive transfer (Siipola & Israel, 1933). The concept of positive and negative transfer applies to all skills. It is easier to demonstrate that transfer has occurred than to pinpoint precisely what has been transferred during skill learning. Among the candidates as vehicles of transfer are verbal information; other cognitive data such as motor schemata or abstract rules termed “productions”; and conditioned reflexes. Learning new skills in rehabilitation Motor skill learning. Learning or relearning a skill such as walking following a hemiplegic stroke is not based purely on memorising a sequence of basic movements such as stepping but is likely to involve the formation of an abstract “schema”. An efficient schema must take account of altered biomechanical conditions, for example failure of the hip to flex on the paretic side. Such factors may make pre-existing “coordinative structures” irrelevant or inoperative, and presumably new ones are formed. It therefore seems implausible that motor relearning in rehabilitation consists solely in the revival of memory traces laid down prior to the stroke. Motor learning takes place against a background of innate, “hard-wired” reactions which can sometimes be exploited but may often be deleterious—for example unselective flexion of the spastic arm. General aspects of skill acquisition. In rehabilitation, motor learning is just one aspect of the more general problem of skill acquisition, and a number of general learning principles apply both to motor and to non-motor skills. Many lessons can be drawn from the theoretical background of skill acquisition, even though evidence is fragmentary. First, as mentioned earlier, it must be recognised that verbal learning is fundamental to many skills. Preservation of “procedural memory” in experimental tests of brain-injured patients should not disguise this fact. Even in motor skill acquisition it may be necessary to devise ways of compensating for deficiencies in retaining information which is required especially in the early stages of learning. Therapeutic programmes may need to be prolonged specifically in order to allow basic verbal information required for skill acquisition to “sink in”; or alternatives such as written instructions may be needed. During rehabilitation the environment for learning tends to be standardised—the same parallel bars, the same bathroom—and the materials for learning are often equally routine: a limited set of words for speech therapy; the same clothes for dressing; etc. But abilities acquired in hospital are often inappropriate to natural conditions. People with severe brain injury tend to have limited ability to generalise from one learning situation to another. Insofar as operant processes contribute to learning, it is important, as we have seen, to vary the context of learning to minimise negative aspects of stimulus control. The same principle applies to the learning of motor skills. For exa