Three dogmas of embodiment: Cognitive linguistics as a cognitive science
The three dogmas
Embodiment as an eliminative reductionism. When I was a graduate student in philosophy, we passed around to each other what we considered a philosophical “howler” of a paper. The paper in question sought to analyze love in terms of symbolic logic. Its end result was a formula that the author proclaimed, without irony, constituted the meaning of the “love” relationship.
To reduce human love to symbolic logic may now seem an obvious absurdity, but I fear we are now running similar risks in Cognitive Linguistics. For example, we hear certain extreme formulations of, for example, Lakoff’s Neural Theory of Language, in which it is has been proclaimed that “cognitive linguistics isn’t cognitive linguistics if it ignores relevant structure about the brain” (Dodge and Lakoff 2005). In the section entitled “Coda” in that article, Dodge and Lakoff argue that as “linguistic structure reflects brain structure,” the underlying brain structure causally “imposes” its image schemas, superpositions, and universal primitives onto language. The obvious implication that many more socio-culturally-minded cognitive linguists draw is that theories of language will be eliminated in favor of theories of how brain structure determines language. Just as the symbolic logicians sought to replace Shakespeare’s love sonnets with logical formulae, so do some cognitive linguists run the risk of advocating translating polysemy, syntax and grammatical construals into talk of neural models, neural circuits, axonal firings and parietal-hip-pocampal networks.
Cognitive linguists are right to question this dogma on both philosophical and substantive grounds. However, oftentimes the reaction to this dogma is a counterproductive iconclasm, in which no shrift is given to evidence of how the brain shapes language. It is equally unpardonable to commit the opposite sin and eliminate our neurophysiological embodiment in favor of our sociocultural embodiment, though I know of no one proclaiming that “cognitive linguistics isn’t cognitive linguistics if it ignores relevant facts about the sociocultural context.” The right antidote to such extreme formulations is to realize that even brain structures are not isolated, static entities, but are embedded within a nervous system and body that interacts with both physical and sociocultural space. As the American pragmatist philosopher Dewey famously asserted, “to see the organism in nature, the nervous system in the organism, the brain in the nervous system, the cortex in the brain is the answer to the problems which haunt philosophy” (Dewey 1925: 198); or, and more appropriately for the present conversation, “haunt cognitive linguistics.” In studying language and conceptual structure with the new tools of cognitive neuroscience, we do not necessarily have to engage in an eliminativist reductionism of language to the biophysical. From the perspective of American pragmatism, both the brain and the sociocultural situation have their roles in the interactions which shape language.
Embodiment as temporally static. The second false dogma dovetails neatly with the first. For example, if we aim to offer some insights to linguistics from the study of brain structures, it would help if we understood how brain structures develop and change across time. A brain is fundamentally an organic entity; it grows, matures, changes, suffers injury and recovers, adapts in response to a changing environment; it evolves over generations. It is altogether too facile to take findings about brain structure in right-hand dominant adults and generalize them inappropriately. If we fail to attend to the limitations of the methods and the scope of current brain research, we risk reifying “brain structure” into a static causal force that determines language, rather shaping and motivating it. Brain structure is dynamic; and while it is certainly not infinitely malleable there are important differences between the adult brain and the developing brain, between the adult brain and the aging brain, between the injured brain and the uninjured brain, between brains of left- and right-hand dominant individuals, and so on and so forth. This organic dynamicity has real impact on cognition and language.
When we see the brain and organism in context, we realize that embodiment is not static. We are fundamentally live creatures who are embodied-in-time; unlike Athena, we do not spring full-grown from the head of Zeus. As newborns we know no language and have only the barest rudiments of what will eventually become the preconceptual basis for adult language. If in Cognitive Linguistics we hold that semantic structure is largely shaped by preconceptual structures such as image schemas, then we must inquire as to how these image schemas are formed on a developmental timeline and whether they might differ cross-linguistically, however slight a difference that might be. For example, Bowerman and Choi (2003) have shown that while nine-month-old Korean- and English-speaking infants can make the same spatial discriminations, at eighteen months their acquisition of language has solidified their spatial categories enough so that they are no longer able to make the discriminations which their language does not. To me this suggests that the developmental course of image schemas is such that their neural instantiations in the brain ‘hardens’ both with age and exposure to the dominant language of the sociocultural environment. This hypothesis about the neural structure underlying image schemas is further warranted given the evident other differences between early and late language learning experiences in humans, and given what we know from animal research into how different spatial schemas cause different patterns of growth in juvenile versus adult brains.1 Similarly, languages change across time; every individual speaker is situated within a specific sociocultural and historical context that influences how their metaphors are shaped (Geeraerts and Grondelaers 1995). Thus, there are multiple temporal forces at work, from phylogenetic change to historical language change to learning and other ontogenetic changes.
Embodiment as consciousness (or as unconscious). Having skirted the shoals of an eliminativist Scylla, we must also avoid being engulfed by the Charybdis of consciousness. A standing temptation of academics is to assume that because we can describe and study some mental process, we are (or can be) ordinarily conscious of it. This is in fact not the case. For example, consider how we process sound waves into phonemes. We do not consciously attend to the peaks and valleys of the sound waves at different frequencies, but the fact that we do parse sound waves into phonemes implies that our brains do something very much like that in the auditory neural maps for pitch. However, conscious introspection is almost entirely useless when it comes to determining the sound shape of a phoneme at different frequencies; instead, scientists are forced to rely on sophisticated electronic measuring equipment. Such processes are certainly beneath the level of ordinary conscious awareness, and inaccessible to consciousness without the tools of science.
Nonetheless, Zlatev has recently argued that Lakoff and Johnson’s notion of an embodied “Cognitive Unconscious” is seriously mistaken and “conceptually incoherent” (Zlatev in press). He rejects the existence of unconscious mental processes entirely, arguing that every mental process must at least be in principle accessible to conscious introspection. Following Searle (1992), he asserts that if we cannot be aware of a mental process then it is neurophysiological and not mental. However, note that this argument is merely definitional– it simply equates introspective consciousness with all mental processes, a philosophical move which is both trivial and unproductive.
There is more merit to other parts of Zlatev’s argument that mental processes are in principle accessible to consciousness, such as those concerning learning and development. To return to the previous example, we are not ordinarily aware of the sound shape that distinguishes one phoneme from another. However, when as adults we grapple with hearing a new phoneme while learning a new language, we do become much more conscious of subtle differences in the shape of its sound. We might also become aware that it is easier to learn it if we also try to speak it and associate its sound with the motor movements which produce it. Eventually we will automate this mental process well enough that we can listen and converse in the new language without having to attend to it. Moreover, at some point, as infants we likely had to “consciously” attend to learning each and every phoneme we now have automated– though the infant’s “consciousness” indubitably has a very different quality to it than how we would typically construe as “consciousness” in the case of the adult cognition; i.e., as a discursive reflective introspection.2
While the philosophical arguments concerning “consciousness” are notoriously interminable, Zlatev is right that embodiment theory must explicate the relationships between at least some of its major theoretical constructs and, as I prefer to designate them, matters of awareness, attention, and feeling. The role of attention in language learning is one example, while another is that the image schemas that underlie conceptual metaphor mappings are construed by Johnson as affect-laden patterns of felt experience– that is, patterns of felt experience of which we can be consciously aware (Johnson 2005, 1987). If we break down the mysterious and ephemeral notion of “consciousness” into its constituent parts (as did the American Pragmatist William James (1900)), we can develop methods to assess when and how it is important to Cognitive Linguistics.
Given this discussion of learning to distinguish phonemes, it should now be clear how the dogmatic assertions about the importance of consciousness (or unconsciousness) to embodiment are intertwined with the dogma that embodiment is temporally static. In his critique, however, Zlatev also points out that banishing the discussion of the various forms of consciousness from Cognitive Linguistics is intertwined with the eliminativist dogma:
Third, there is a dangerous tendency to underestimate the role of consciousness in many– though not all– embodiment theories. There seems to be some sort of fear that in appealing to anything that is irreducible to either biology or behavior, one is bound to fall into the clutches of “Cartesian dualism”. The consequence is, however, that such “non-dualistic” approaches run the risk of one form or another of physico-biological reductionism, which as pointed out by Itkonen … is deeply misguided. (Zlatev in press: 2)
Obviously, these three dogmas are clearly interwoven. To counter them, we must cultivate a philosophical framework for embodiment that is methodologically pluralistic and dynamically attentive.
1. I have in mind several series of studies, but in particular the Knudsen and colleagues studies of juvenile and adult barn owls, some of whom who wore prismatic glasses that distort the visual field in the xy-plane (e.g. 23 degrees to the right) results in an bifurcated neural growth pattern in the cross-modal map of the ICX in their brains. (Knudsen 2002, 1998). To make the connection between this and language more explicit: Just as the juvenile owls who wore glasses in their youth become bi-‘visual’, so do the bilingual speakers who are exposed to more than one language from birth are more likely to become more adroit speakers of those languages than those who start off monolingual and acquire a second or third language later in life. It seems likely, then that the many different neural structures underlying second language acquisition also ‘grow differently,’ particularly when we are talking about distinguishing the subtle sorts of spatial relations differences that underlie using the spatial relations terms perfectly in a second language. For further discussion, see Johnson and Rohrer (in press) and Rohrer (2005).
2. Infants have what appears to be a near-innate ability to distinguish between the full range of human phonemes on a perceptual basis, but between months 10–12 lose that raw perceptual ability (Bohn 2000: 6–10). Instead, they automate and “zoom in on those contrasts which are functionally relevant (i.e. phonemically distinctive)” in their own peculiar linguistic environment (Kristiansen 2003: 85). Such infant evidence is not only an example of how the categorization of phonemes in the infant mind emerges from the perceptual categorization processes available in the infant brain, but also demonstrates that the infant’s mental categorization of phonemes is neither a result of nor is accessible to the processes of conscious introspection that Zlatev takes as essential for language. (I wish to thank Gitte Kristiansen and René Dirven for their help in elucidating this point.)
