Many philosophers, psychologists, physiologists, and others have written on knowing-how. Merleau-Ponty, Heidegger, and Nietzsche are but three of the more prominent philosophers who have discussed what it is to know how to do something. Gestalt psychologists have also written about the whole of a learned action being greater than the sum of its parts. For example, skiing is not merely moving your knee one way or the other, holding your body in a certain position, and so on. Rather, it is learing to “flow” with the act of skiing. In Pike’s terms, an action has more than a static existence; it also has dynamic existence (how it is performed) as well as “field” existence (how it relates to other actions within a given cultural system).
For any knowing-how, there exists a process to be mastered, mainly by doing, usually learned in parts. Riding a bike entails learning to mount, pedal, steer, turn, dismount, brake, and so on, without falling. We cannot learn such things by reading about them or being lectured about them. We need (painful) practice, and we benefit from examples (partially due to mirror neurons that communicate with our muscles so as to prepare them to act as we perceive another’s action).
Over time, our basal ganglia, synaptic connections, even our muscles—through specific patterns of use and excitation, even creating new fiber nuclei in some cases (Bruusgaard et al. 2010)—register these actions. We build up such memories by integrating individual apperceptions of parts of action into a whole. Then we have mastered bike riding. It has become an emic action rather than merely etic actions. The involvement of the muscles, nerves, and the body more generally, along with the seeming irrelevance of propositional knowledge, make bike riding a prototypical case of knowledge-how—the kind of knowledge one more effectively “demonstrates” than “tells.”
But are these ingredients completely lacking from what is referred to as knowledge-that? Consider a simple proposition such has “John knows that an apple is a fruit.” How can we determine whether John in fact knows the fruity nature of apples? We can only know it by what assertions he makes, how he responds to an apple in his environment, how he infers properties of the apple based on his knowledge and so on. And another question: Then how has John acquired this knowledge of apples? In just the same way, by doing and watching actions done. As John acts, he affects his synapses, his body (e.g., salivation, olfactory sense, hunger pangs), and his physiology and cognitive states more generally. In other words, to say that John knows x, we have to also know how John reacts to x, uses x in a sentence, draws inferences from x, and the like. We must know how John knows how to do these things. Yet if this is so, then the knowledge-how vs. knowledge-that distinction is at best weak, at worst nonexistent.
There is another sense in which the distinction between knowing-how and knowing-that is weak. Knowledge-how can be translated into knowledge-that. It is possible (and often done), for example, to develop a bike-riding algorithm that can enable a machine to try to ride a bike.2
When I began to learn to ride a bike, I remember focusing during each attempt on what I thought were the important steps—from my intentionally white-knuckled grip on the bar, to keeping up the minimum speed necessary to maintain forward motion without risking a lurch into the intersection, to avoiding objects by focusing ahead sufficiently to plan my steering. Conscious concentration on every subroutine of the larger act of “riding” tired me out. But as I internalized my knowledge by use of my body, before long, I “got it”—my brain’s system of equilibrioception adjusted to this new form of locomotion, and I came to “know how” to ride a bike. Riding a bike is a paradigm example of knowing-how. But is this knowing-how utterly distinct from knowing-that? To answer this question, let’s consider another problem that seems superficially like knowing-how: the knowledge of a linguistic sound system, a phonology. The phonological knowledge of native speakers is both knowing-how and knowing-that simultaneously. Before looking at phonology, however, we need to better understand “knowing-that.”
Knowing-that is generally assumed to be quite different from knowing-how. The former is propositional knowledge, while the latter encompasses skills. Thus knowing-how is taken by some philosophers and psychologists not to be knowledge at all but only habits, skills, or capacities developed over time through practice, or repetition and imitation—muscle memory, largely—whereas knowing-that is taken to be part of cognition proper. Knowing how to ride a bike, again, in this view, is not properly speaking cognition, but merely muscle routines. And yet if knowing that a proposition means x requires that we know how to use it in x-related inferences, as Brandom (1998) urges, then the how/that distinction is again weakened, if not eliminated.
Because there is a clear sense in which I have “internalized” or “emicized” the bike riding process, yet I am nevertheless unable to give a fully adequate explicit description of what I am doing when so engaged, it may not be possible to reduce knowing-how to knowing-that. There is some part of all knowledge that is not reducible to propositions. But you might reply “It has already been demonstrated that we can devise a set of algorithms for bike riding that, when fed into a robot, enables it to actually ride a bike. Wouldn’t that prove that bike riding is ultimately a knowing-that rather than a knowing-how, and thus that knowledge is not ineffable in principle?”
No. And for largely the same reason that Google Translate cannot be said to speak a language or Searle’s (1980b) “Chinese room” cannot be said to speak Chinese.3 Whatever algorithm the robot is following, I am not using these algorithms to ride my bike. How do I know? I know first because they are never conscious, and they are impossible to make conscious for the average bike rider. Second, no scan or other image of my brain is going to turn them up. They are literally not in my brain. But third and most important, the algorithms add nothing to our understanding of human bike riding. The way to understand that is to study the human body and learn what the muscles do, how the brain controls them in bike riding, and how the entire body adjusts to angles, wind, and so on, to maintain balance, speed, and directedness without overt commands. We do not linguistically represent all problems; we “resonate” with many of them (Gibson 1966, 1979). Our bodies adjust to the movement of the bike, the contour of the earth, the holes in the pavement, and so on. Perhaps the notion of resonance might even help us to understand why trying to make what we are doing explicit while riding the bike actually interferes with our riding and can make us fall. The algorithms describe the physics of bike riding, let us say, but not the contents of my brain or how I actually accomplish the physical task. Of course, I cannot escape the constraints of physics—speed, trajectory, energy, gravity, and so on. But if you ask me what I am doing, or if you were to look at an image (fMRI, CAT scan, etc.), no such algorithms turn up. They are not causally implicated in what I am doing. They may describe what I am doing abstractly, but when we externally describe the steps to accomplish a behavior, we are not describing the emicization of the behavior. There is of course no problem with that so long as we do not confuse the two. A robot riding a bike has not emicized bike riding. It has no conceptual choices and no sense of “this is the way we do it.”
To see this more clearly, consider the difference between descriptions of riding a bike vs. riding a bike “very well.” (Later I will contrast descriptions of grammar and language.) I can offer some account, however inadequate, of my bike riding, what I am doing, what I hope to do, what I think my legs, eyes, arms, hands, and so forth, are doing, but I cannot explain how to ride a bike perfectly, except to prohibit certain effects of poor bike riding: Don’t fall down!” “Go faster!” “Turn sharper!”
And yet the question still lingers—namely, “Why do we pretend that the dividing line between knowing-how and knowing-that is so clear?” I think that the answer is straightforward: it’s linguistic. For the past sixty years or so, cognitive scientists have adopted the idea that the brain represents knowledge in some way, most likely propositionally. Knowing-that knowledge is just that which can be attributed in propositional form to a co
gnizing being. Knowing-how, on the other hand, is behavior that cannot easily be summarized in this form and so is usually not considered to be knowledge. But what if this is just a distinction without a difference, simply a reflection of what we are able to describe rather than any meaningful distinction in the individual? Riding a bike and proving a theorem perhaps best represent the two poles of this purported knowledge distinction.
Setting aside this controversy, one way the two types of knowledge are the same is in cultural-activity slotting. Once we have learned something, whether knowing-how or knowing-that, the thing learned occupies a “slot” in our cultural matrix, a position in a grid of similar things known. This concept of matrix and slot are ideas from Pike’s (1967) analysis of language (and, by extension, culture) as “particle, wave, and field.”
Knowledge as Particle, Wave, and Field
A particle can be either or both a unit of language and behavior in Pike’s theory. An easy example is a distinctive feature of sound, such as [+voiced], an indication that at some point during the sound in question, the native speaker notices that the vocal folds are vibrating.4 This example will do double duty, reinforcing the idea that knowing-how and knowing-that are not easily distinguished, if at all, and corroborating the nature of a unit in Pike’s sense. Part of linguists’ understanding of the particle [+voiced] is its set of positive or active properties—“the vocal cords vibrate; the vocal cords are relaxed; the air pressure above the glottis is less than the air pressure below the glottis; the Bernoulli effect sets the vocal cords in rapid motion,” and so on. The average speaker will have a less technical awareness, knowing merely that /b/ is not /p/, /t/ is not /d/, and so on. Like average speakers, a crucial part of the scientist’s understanding of a particle, however, is what it is not. For example, if every sound in a language were [+voiced]—such as /b/, /g/, /a/, /i/, /m/.—this feature would not be distinctive, and it would not therefore be necessary to learn it as a particle of the language. To do so would be superfluous, adding nothing to our understanding of the language. On the other hand, if a language has both [+voiced] and [–voiced] sounds—the latter represented by sounds such as /p/, /t/, /k/, /s/, /h/—then we would need to learn, as part of learning this language, whether a sound is + or –voiced. Particles are thus understood positively and negatively, by what they are and what they are not. Properties of particles cannot be known in isolation from other particles of the language, but as part of a system.
Each particle in a language or a culture must also be studied as a “wave.” This is the dynamic perspective, the particle-in-use perspective. Thus when we say something in which /b/ or /p/ is found in different contexts of use—for example, ba, ab, and aba; pa, ap, and apa—the voicing and articulation of the consonant and the vowel will look different spectrographically according to their relative positions in words and phrases. Thus, in action, details are revealed about the particles that cannot be seen in isolation. In particular, the vocal cords will begin to vibrate differently in different contexts. This differential onset of voicing is referred to as voice-onset timing (VOT), and it has played a fundamental role in our understanding of the phonetic differences between sounds within and across languages for decades.
For example, in English VOT begins later than in Spanish. In Spanish a b is fully voiced, by and large, from start to finish. But in English a b has a later VOT. Aspiration (a puff of air following a sound) is often found in languages like English because this makes it easier—especially in word or syllable-initial positions—to hear the difference between, say, p and b. Thus English, but not Spanish, has aspirated voiceless stops (spreading the glottis more widely during phonation, indicated as: [–voiced, + spread glottis]). Again this aspiration serves largely as an aid to hearing the difference between [–voiced] and [+voiced] segments. We can understand these features of individual sounds of English and Spanish only by seeing them used in context, a dynamic perspective. It is never enough to study them strictly in isolation.
The wave perspective of bike riding is how bike-riding varies in different occasions of riding, whether that is how different cultures or individuals ride, or how turning corners may vary, or the relative speed of the rider. It is the knowledge of bike riding in action.
And yet even once we understand a unit as a particle and as a wave, we still have not grasped the meaning of any behavioral particle until we see how it fits into the language or culture as a whole. How is it slotted? Slotting of units entails a field perspective. The easiest way to illustrate a field perspective would be to show a cultural or linguistic matrix. In the case of individual sounds we have what structuralist linguists referred to as a “phoneme chart.” (table 1.1) Phoneme charts are iconic. Left to right, they represent the front of the mouth moving toward the back of the mouth. Reading vertically, we see differences in mode of articulation (voiced, voiceless, occlusive, fricative, etc.). Thus, in the chart, a p is contrasted in voicing with b and in place of articulation with t. All vowels are voiced, so that i, a, and u contrast in the rounding of the lips and their position in the mouth—front (i) to back (u), top to bottom (a).
Table 1.1. Pirahã Phonemes
Consider each position in each chart a “slot.” Though the phonemic squares are not cultural slots per se, the entire chart is itself (at least partially) a cultural selection—the sounds a language has “decided” to use at some point in time. Consider each position in a given context a slot as well. Thus there are two kinds of slot that every behavioral particle must fit into—a matrix slot (i.e., within the phoneme chart) and a context of usage slot (where it goes in a word). These are also known as paradigmatic and syntagmatic slots, respectively. By slotting the behavior or the linguistic unit in this way, we understand it more effectively both within a given culture or language and across cultures and languages. For example, Pirahã syllable structure makes use of these phonemes in allowing only syllables that potentially allow a single initial consonant, followed by one or two vowels (with the added condition that a syllable must have at least one consonant and vowel or two vowels), given in the following moraic-phonotactic constraint:
Pirahã phonotactics: IV(V)
Moraic constraint: Pirahã syllables must be greater than one mora, where voiced consonants have half a mora; voiceless consonants are one mora in length, and vowels are one and one half moras (all relative durations, no absolute timing implied).5
Likewise, English words have a syntagmatic arrangement along the lines of [[Prefix-[[Root] Stem]-Suffix] Word], as in [[en [[light]en]]].
How might bicycle riding or any other nonlinguistic behavior relate to these concepts of particle, field, and wave? As a particle, bicycle riding can be described as “locomotion via a two-wheeled, manually powered vehicle that one sits on, requiring balance.” But it is also understood as not being “motorcycle riding,” “riding a unicycle,” “riding a ‘girl’s bike’ vs. a ‘boy’s bike,’” “riding a delivery bike,” “tricycle riding,” “walking,” or “driving.” It is a behavioral particle when seen as a single action according to the categorization of a particular culture.
Likewise, bicycle riding must be slotted in a chart of “means of locomotion” and another of “forms of recreation” at a minimum. Only by examining behavioral units in relation to other behavioral units of the same type can we understand ourselves or others. Of course, the meta-issue is that fields are themselves cultural constructs—cultures are not merely knowledge, values, actions, and so on, but the arrangement of these things.
Thus our tacit knowledge includes knowledge structures. If our dark matter is largely composed of what we learn as individuals via participation in a given culture, is there any other kind of dark matter? For example, are we born with innate dark matter as well as those ideas and lessons we glean from our lives? What are the conceptions of tacit knowledge that have dominated thought and discussion in history? Before returning to my empirical proposals, it is necessary to undertake a brief historical survey of the notion
s of tacit knowledge that have influenced current discussions and debate, in order to contextualize and thus better understand the discussion of the remainder of the book. I am going to argue later that such facts lead us to conclude that some forms of knowing, such as phonology, are at once knowing-how and knowing-that, an ability that embodies dynamic and static cognition (D. Everett 1994).
THE PLATONIC TRADITION OF INNATE KNOWLEDGE
In what follows, we survey different concepts of tacit knowledge. This survey reveals that concepts vary so extensively as to call for another term, my dark matter of the mind: knowledge and values relating to our environment, ranging across the unspoken and the ineffable, the bodily and the mental, engaging the full individual.
Moving toward our brief history of dark matter, we need to mention at the outset an underlying assumption upon which much theorizing about human knowledge is based—namely, the belief that all humans share a basic psyche. Adolph Bastian, of whom we will learn more presently, developed the influential idea of the “psychic unity of mankind” that profoundly affected the thinking of people such as Carl Jung, Joseph Campbell, Franz Boas, and others. This “unity” is one of the first modern proposals of nativism—in this case, that there are concepts universal to all humans. In modern linguistics, this idea has strongly influenced the development of the natural semantic metalanguage theory of Anna Wierzbicka (1996) and her colleagues (see chap. 9) and finds resonance in Chomsky’s universal grammar, as well as compatible concepts of a variety of instincts proposed in recent literature (e.g., a moral instinct [Hauser 2006], religious instinct [Haidt 2013], art instinct [Dutton 2010], and language instinct [Pinker 1995]).
Dark Matter of the Mind Page 5