Music provides an especially interesting laboratory for the study of cognition. Because so much musical behavior is non-linguistic in nature, music tends to challenge dominant linguistic paradigms, which reduce all cognition to rational thought processes such as problem solving, deductive reasoning, and inference. Unfortunately, most research in music perception and cognition has focused on a very narrow band of human musical phenomena, namely the tonal concert music of pre-20th-century Western Europe, as filtered through contemporary European-derived performance practices. Hence we have an abundance of tonal-music-inspired models and representations for perceptual and cognitive phenomena, focusing almost entirely on pitch organization in the large-scale time domain. Some examples are theories of recursive formal hierarchies (Lerdahl & Jackendoff 1983) and of musical meaning from deferred melodic or harmonic expectations (Narmour 1990, Meyer 1956). Lerdahl and Jackendoff (1983) contend that in the way that information-processing stages are organized, musical cognition is fully analogous to linguistic cognition. Such models suppose that the cognition of music consists of the logical parsing of recursive tree structures to reveal greater and greater levels of hierarchical organization.

A side effect of the traditional linguistics-based approach has been the adoption of a major tenet of the Anglo-American "analytic" philosophical tradition, namely that rational language as a logical system has been taken to form the basis of virtually all thought and meaning (Prem 1996). Even in the work of Lakoff (1987, discussed in the following chapter), which purports to frame linguistic understanding and meaning in terms of the human body and its environment, the theory centers around language. Such work would have it that no non-linguistic modes of discourse exist, and that all meaning is attached to language. However, music provides a clear counterexample, with its basis in bodily activity and its strong emotional component.

To be sure, some musical features may have elements in common with language. Bregman (1990) and Handel (1990) suggest independently that similar functions operate in the perception of both speech and music at early levels of processing in order to organize the complex acoustic signal into extended patterns of events, presumably subject to some rules. At a higher level, the expressive manipulation of rhythmic timing in musical performance invites comparison with the use of timing in speech for semantic emphasis. Similarly, the regular grouping of pulses into an intelligible template occurs both in music and in metered verse. But such functions are of a different order from the abstract formulations of linguistic theories. Expressive timing and grouping have less to do with rules governing sentence structure than they do with either low-level auditory processing or the high-level shades of meaning possible in spoken language. Although aspects of musical behavior may have elements in common with linguistic behavior, a large part of musical understanding seems to operate overall quite separately from the realm of rational language and inferential reasoning.

Though often posited as musical universals, many of the results of Lerdahl and Jackendoff (1983) do not carry over effectively to the vast majority of non-Western musics, nor do they account fully for the perception and cognition of Western tonal or atonal music. The inapplicability of these linguistics-derived models to other musics is quite glaring in the cases of West African and African-American musics such as jazz, rumba, funk, and hip-hop. In these cases, certain salient musical features, notably the concept of groove, seem to have no analogue in rational language. Although groove is a highly subjective quality, music that grooves can sustain interest or attention for long stretches of time to an acculturated listener, even if "nothing is happening" on the musical surface. A prime example is James Brown s music [CD-2], [node:3750 teaser] which frequently has precious little melodic or harmonic material and is highly repetitive, but would never be described as static. The fact that groove carries enough weight to override other musical factors in certain kinds of musical experience suggests that the traditional linguistics-based viewpoint does not suffice in describing the entirety of music cognition.

A major reason for this mismatch between tonal-music grammars and most music of the world is not (as is commonly thought) differing levels of musical sophistication or complexity, but rather a major cultural disparity in approaches to rhythmic organization and musical form. I claim that an essential component of this disparity is the status of the body and physical movement in the act of making music. The role of the body in various musics of the world becomes clearer when one observes the function that music and dance assume in these cultures, the common cultural/linguistic metaphors associated with musical activity. All of these observations have led us to study the role of the body in cognition in general.

Cognitivism & traditional cognitive science. Until the late 1980s, generally in cognitive science, and particularly in artificial intelligence, the logic paradigm had prevailed. The cognitivist or objectivist point of view involved the assumption that thought amounted to the mechanical manipulation of abstract symbols, and that the mind was an abstract machine, manipulating symbols by algorithmic computation in the way a computer does. All meaning arose via correspondences between symbols (words, mental representations) and things in the external world. These symbols formed internal representations of external reality, independent of any limitations of the human body, the human perceptual system, and the human nervous system. The mind was seen as a mirror of nature, and human thought as abstract and disembodied. Human bodies and their environments were thus incidental to the nature of meaningful concepts or reason. The materials (or "hardware," in computer terminology) with which brains think and senses perceive were believed irrelevant to the abstract processes ("software") that these systems conduct. The brain was merely a specific instance of a computing engine, of which all such manifestations would be formally equivalent (von Neumann 1951). Hence, in principle, a brain could be replaced by a computer, and in particular computers can do anything that brains can do; machines that mechanically manipulate symbols that correspond to things in the world were believed capable of meaningful thought and reason. (Lakoff 1987: xii-xiii) The concept of a representation, an abstract data structure for storing and manipulating such symbolic information, is central to such cognitivist theories of mind. Typically, a cognitive representation has taken the form of a "message board" -- a disembodied abstraction whose shared contents provide an interface among various independent information-processing units.

These tacit assumptions spawned generations of experiments in artificial intelligence in which symbolic, language-based reasoning was taken to be the cornerstone of intelligent behavior. A classic example of this approach appears in the medical consultation program, MYCIN (Shortliffe 1976). As a so-called "expert system," the program was provided with task-specific data taken from real-life experts, thereby enabling it to achieve substantial competence in a restricted domain of analytical reasoning. However, when running such an expert-system program, the computer merely manipulates abstract symbols according to a body of explicitly formulated rules and guidelines. One could never claim that the program "understood" the ins and outs of human health issues, but merely that it contained a representation of such information and a model for manipulating this information based solely on its symbolic characteristics. Furthermore, its expected input was highly restricted in range, such that grammatical errors or ambiguous terminology could, for example, cause it to diagnose a rusty Chevy as having the measles (Lenat and Feigenbaum 1992: 197; see also Clancey 1997: 29-45). Such text-inspired approaches to knowledge have modeled intelligence as disembodied symbol manipulation.

The above description raises the hotly debated issue of the difference between "real" intelligence and "simulated" intelligence. Presumably, MYCIN could pass some restricted kind of Turing test, in which an impartial expert might judge the computer s input-output behavior to appear intelligent. However, I must stress that a microworld representation of intelligence is not human intelligence. For intelligent behavior encompasses not just symbolic manipulation and deductive reasoning, but also interaction with others, attunement to one s surroundings, and awareness of the relationship between oneself and one s world, not to mention intelligent action, creativity, physical coordination, emotion, and countless other behavioral manifestations. One finds that in traditional cognitive science, the emphasis on things "mental" has been at the expense of the physical, and the disproportionate attention paid to "the rational" has often occurred in opposition to the emotional and the intuitive. Yet everyday human intelligence includes all of these dimensions.

Embodied cognition. Thankfully, recent conceptual developments in cognitive science move towards the inclusion of such dimensions. In particular, cognitive scientists have begun to infer connections between the structure of mental processes and physical embodiment. The viewpoint known as embodied or situated cognition treats cognition as an activity that is structured by the body and its situatedness in its environment -- that is, as embodied action. In this view, cognition depends upon experiences based in having a body with sensorimotor capacities; these capacities are embedded in an encompassing biological, psychological, and cultural context. Sensory processes (perception) and motor processes (action), having evolved together, are seen therefore as fundamentally inseparable, mutually informative, and structured so as to ground our conceptual systems. (Varela et al. 1991: 173)

The embodiment hypothesis suggests an alternative basis for cognitive processes. Perception is understood as perceptually guided action. Our eyes move to frame the visual field and to focus on regions of that field; our heads move to facilitate binaural localization. Such behavior is facilitated through elaborate feedback mechanisms among sensory and motor apparatus. Hence, cognitive structures emerge from the recurrent sensorimotor patterns that enable the perceiver to guide his or her actions in the local situation. That is, the emergent, reinforced neural connections between the senses and the motor system form the basis for cognition. The mind s embodiment provides natural biases for inductive models and representations, and thus automatically grounds cognititive processes that might normally be considered disembodied. This view provides a sharp contrast from the standard information-processing viewpoint, in which cognition is seen as a problem of recovering details of the pre-given outer world. (Varela et al. 1991: 173)

In this light, the mind is no longer seen as passively reflective of the outside world, but rather as an active constructor of its own reality. In particular, cognition and bodily activity intertwine to a high degree. In this perspective, the fundamental building blocks of cognitive processes are control schemata for motor patterns that arise from perceptual interaction with the body s environment. The drives for the cognitive system arise from within the system itself, in the form of needs and goals. (Prem 1996)

Neuroscientific evidence corroborates this viewpoint. As was pointed out earlier, we can make more sense of our brains and bodies if we view the nervous system as a system for producing motor output. The cerebellum is connected almost directly to all areas of the brain -- sensory transmissions, reticular (arousal/attention) systems, hippocampus (episodic memories), limbic system (emotions, behavior). All areas of our brain seem geared to coping with their functions as they pertain to problems of motor control. (Hardcastle 1996: 7) Such evidence from neuroscience allows for postulating shared mechanisms for low-level control of embodied action and higher-level cognition; motor plans for limb movement could interact with goal-oriented abstract plans. The mind thereby becomes a distributed entity, an emergent characteristic of the whole sensory-central-motor neural system, existing in the elaborate network of interconnections that extend throughout the body.

Situated cognition. The above characterization of the embodied mind covers merely half of the picture. If we grant that cognition is structured at least to some degree by bodily experience, then we must understand the body to be immersed in an environment that shapes its experience. Hence the philosophy of embodiment also stresses temporal, physical, and sociocultural situatedness.

It has been shown that the framing of cognition in terms of the body and its environment provides not only limiting but also enabling constraints for cognition. Work in animal behavior has addressed the potential links between sensory and motor systems, as in the classic experiment by Held & Hein (1958). In this study, a group of kittens were raised in the dark and exposed to light only under controlled conditions. A first group of animals was allowed to move around normally, but each of them had to pull around a cart in which rode a member of the second group. The two groups thus shared nearly identical visual experience, but one group experienced the world actively and the other passively. Upon release after a few weeks, the first group of kittens behaved normally, but the second group behaved as if blind, bumping into objects and falling over edges. Hence objects in the world are apprehended not simply by visual extraction of features, but rather by the visual guidance of action. (Varela et al. 1991: 175) In a similar but more humane fashion, it has been observed that infants who can walk have qualitatively different reactions to certain stimuli, such as slopes and falloffs, than infants who can t (Thelen & Smith 1994: 217-220).

There is reason to believe that such programmability of the brain extends far beyond childhood. While it is commonly known that cognitive development proceeds rapidly along with brain growth and cortical myelinization during the crucial first years of life (Passingham 1982: 112ff), it is not often recognized that many networks of the brain retain a susceptibility to reprogramming throughout an individual s life (Laughlin et al. 1992: 41). The remarkable case studies of Ramachandran and Blakeslee (1998) and Sacks (1985) further attest to the adaptability and plasticity of the brain throughout adulthood. Hence we may discern a continuum of neural structures ranging from "hard-wired" evolutionary traits to highly flexible, environmentally adaptive features (Shore 1996:17). The existence of this continuum supports the embodied cognitive paradigm, which encompasses the body and its environment.

In addition to the universals of cognition based upon having a body and its sensorimotor systems, we study the particular social and cultural factors that contribute to the development of mind:

[T]here is reason to suspect that what we call cognition is in fact a complex social phenomenon... The point is not so much that arrangements of knowledge in the head correspond in a complicated way to the social world outside the head, but that they are socially organized in such a fashion as to be indivisible. 'Cognition' observed in everyday practice is distributed -- stretched over, not divided among -- mind, body, activity and culturally organized settings. (Lave 1988: 1)

We may rely upon various attributes of our physical, social, and cultural environment to support or augment our mental capacities. Lave (1988) studied the arithmetic of adults of various backgrounds in the grocery store. Lave s results showed that in making purchase decisions, these situated agents employed a flexible real-time arithmetic in order to select better prices per unit weight, continually taking into account the constraints imposed by the layout of the stores, the capacities of their home refrigerators, and the dietary requirements of their family members. Such shopping prowess -- skill at situated arithmetic -- was rarely reflected in subjects performance on grade-school math problems. Cognition as demonstrated in practice was found to be not at all in the same realm as cognition in an abstract, un-situated setting. Similarly, Clark (1997: 213-216) discusses the ambiguity of the boundary between "mind" and "world," along the lines of a distinction between "user" and "tool." When a bird drops a nut from a great height to crack it open, does the ground become a tool? (Clark 1997: 214) Rather, the bird is exploiting an aspect of its environment to extend its physical capabilities; the concept of a tool dissolves.

Along with the embodied/situated paradigm came the gradual recasting of the notion of a representation. In a classic paper entitled "Intelligence without Representation," Brooks (1991) reported research in situated robotics, in which artificial intelligence was "approached in an incremental manner, with strict reliance on interfacing to the real world through perception and action" such that "reliance on representation disappears." (Brooks 1991: 139) In Brooks s view, the intelligent system is decomposed into independent, parallel activity producers, which all interface directly to the environment (rather than to each other) such that their respective perceptions and actions can override one another in the system s resultant observed behavior. Brooks aimed to simulate entire embodied, situated intelligent systems. Abandoning the cause of simulating human intelligence, he considered simulated insect intelligence to be the most feasible first step. In doing so, he cast his work in an evolutionist s light; in his opinion, nature has conducted research on situated agents for billions of years, so he modeled his work after the gradual accrual of complexity in the evolution of species.

While rather extreme in its views, Brooks s work broke down the traditional notion of a representation. In particular, an important distinction is made in the embodied viewpoint between the point of view of the subject and that of an "objective" observer of the subject in its environment. (Prem 1996) The notion of a representation was seen to stem directly from this supposedly objective position of the observer, and hence its actual role in cognition for the situated agent was questioned. Later, Coelho (1995) suggested that we not throw out the baby with the bathwater. While it was granted that purely representational accounts were probably "too limited for modelling the recurrent processing that is the norm in the brain" (Coelho 1995: 311), the validity of the partial role of representations in the mind is proven by our capacity for memory formation and recall. Above all, the notion of representation has been retained as a valuable descriptive tool in the study of cognition, but not as a structural necessity in many cases of cognition itself. Hence caution must be used in relying on representations to define a cognitive model.

In sum, the theory of embodiment encompasses both neuropsychological and socioenvironmental views of cognition. Embodied cognition stresses physical, temporal, and functional situatedness, and enforces interaction between the agent s body and its environment. Such a holistic view prevents some inappropriate simplifications and unrealistic assumptions because it enforces dealing with unexpected contingencies, provides specificity, and incorporates energetic and resource considerations. (Mataric 1996) And, quite significantly, the embodied view of cognitive science allows for direct cultural interaction, which is undeniably crucial for both language and music.