Levels Of Representation

Learning to modify actions based on sensory inputs and previous experiences is a complex computational problem. In fact, the pervasiveness of skill acquisition makes it difficult to define. For example, learning a novel, multicomponent series of movements, such as when we throw a baseball, is certainly considered motor skill learning. However, do we include other skills, such as learning to use a mouse or blink in anticipation of a puff of air? In all these cases, new relationships between particular actions and environmental events must be encoded. However, in some cases, the new behavior requires the precise timing and modulation of different muscles. In other cases, the movements are already well entrained and acquiring the new skill requires learning the appropriate goal that is associated with a particular set of environmental conditions. For example, in eye-blink conditioning, the movement is well established and learning primarily consists of determining the stimulus conditions in which this response should be emitted.

Another behavior, handwriting, serves as a useful example when considering the different processes involved in motor learning. Learning to write requires the complex coordination of multiple effectors to produce elaborate spatiotemporal patterns of output. Handwriting is usually produced by a set of effectors including the muscles of the fingers and wrist of the dominant hand. However, it can also be produced —and with considerable skill—with other, nonover-lapping sets of effectors. For example, writing can be performed relatively effortlessly with the muscles of the right arm and shoulder, such as when writing large words on a chalkboard.

These transfer effects have led to the hypothesis that skilled movements are constrained by abstract motor programs. An abstract motor program does not specify the commands sent to the muscles or even which muscles must be activated; this information must be provided by downstream systems that implement a motor program. Exactly what a motor program represents remains controversial. However, the absence of effector-specific commands makes the abstract motor program so powerful. Once established, it can be useful in a variety of contexts, depending on the goals and circumstances confronting the organism. Given the complexity of our environments, it is advantageous since learning acquired with one set of effectors can transfer to other sets, thus allowing behavior to be both skilled and flexible.

However, transfer of knowledge to different effectors does not always appear to be complete. When right-handed individuals are asked to write with their left hands, their movements are comparatively slow and awkward, and the resulting trajectories can be considerably distorted. Such difficulty indicates that many components of the learned behavior do not transfer well to this new set of effectors. Even so, many features of an individual's unique style are present in their production with the nondominant hand. That is, the shape of the letters appears to match, although with less precision, the same template used to generate those produced by the dominant hand.

The partial transfer of handwriting to novel effectors suggests that multiple representations can subserve a single learned behavior. Representations about the desired trajectory of the movement appear to be encoded in spatial coordinates that can be accessed by various effectors. However, such representations are not sufficient to rapidly produce fluent performance. In addition to requiring a representation of the desired goal of the movement, individuals must also learn how to accomplish the goal with a specific set of muscles.

This example illustrates the distinction between the two classes of knowledge required to perform motor acts, often referred to as "knowing what to do'' and "knowing how to do.'' Although it is tempting to associate the former with declarative knowledge and the latter with procedural knowledge, several lines of evidence suggest that "knowing what to do'' can be encoded without awareness. That is, in many cases, decisions about what to do appear to be facilitated by implicit knowledge.

Understanding how these multiple levels of representation are supported by the neural structures associated with motor learning is critical for a comprehensive theory of motor skill learning. In this article, we review some experimental tasks that have been used to identify the neural loci of motor learning and develop functional hypotheses regarding the computations performed by these structures.

Breaking Bulimia

Breaking Bulimia

We have all been there: turning to the refrigerator if feeling lonely or bored or indulging in seconds or thirds if strained. But if you suffer from bulimia, the from time to time urge to overeat is more like an obsession.

Get My Free Ebook


Post a comment