Dark Matter of the Mind

by Daniel L. Everett

  This chapter illustrates the work of the previous chapters by considering the concept of scientific progress in light of a theory of dark matter. To take a concrete example of a science, we focus on linguistics, because this discipline straddles the borders between the sciences, humanities, and social sciences. The basic idea to be explored is this: because counterexamples and exceptions are culturally determined in linguistics, as in all sciences, scientific progress is the output of cultural values. These values differ even within the same discipline (e.g., linguistics), however, and can lead to different notions of progress in science. To mitigate this problem, therefore, to return to linguistics research as our primary example, our inquiry should be informed by multiple theories, with a focus on languageS rather than Language. To generalize, this would mean a focus on the particular rather than the general in many cases. Such a focus (in spite of the contrast between this and many scientists’ view that generalizations are the goal of science) develops a robust empirical basis while helping to distinguish local theoretical culture from broader, transculturally agreed-upon desiderata of science—an issue that theories of language, in a way arguably more extreme than in other disciplines, struggle to tease apart.

  The reason that a discussion of science and dark matter is important here is to probe the significance and meaning of dark matter, culture, and psychology in the more comfortable, familiar territory of the reader, to understand that what we are contemplating here is not limited to cultures unlike our own, but affects every person, every endeavor of Homo sapiens, even the hallowed enterprise of science. This is not to say that science is merely a cultural illusion. This chapter has nothing to do with postmodernist epistemological relativity. But it does aim to show that science is not “pure rationality,” autonomous from its cultural matrix.

  In 1996, the late Peter Ladefoged—at the time of his death perhaps the world’s leading phonetician—and I published an article in the main journal of linguistics, Language, entitled “The Status of Phonetic Rarities” (Ladefoged and Everett 1996). In this paper we discussed unique sounds I had brought to the attention of linguists, which had been found in the Amazonian languages Pirahã, Wari', and Oro Win, but in no other language of the world. The sounds we discussed were [] and []. Both sounds are unique to Wari' and Oro Win on the one hand, and Pirahã on the other. [] is a voiceless alveolar-bilabial trill and [] is a voiced apico-labial double flap.

  To some, these sounds would be little more than outliers in the range of human sounds, merely curiosities. But to phoneticians, they are more than this. Phoneticians and phonologists have theories of how sounds come to be incorporated into human languages and how such sounds fit into phonetic or phonological theories more broadly. The majority of phonologists believe that all sounds in human languages are decomposable into “distinctive features.” Thus a sound like [t] is a [–voiced, + coronal, –continuant].10 What Ladefoged and I showed, however, was that no extant combination of distinctive features could describe these rare sounds. Thus we were faced with a choice: we could refer to these sounds as exceptions or counterexamples to modern phonological theory, or we could modify the otherwise universally accepted list of distinctive features in order to accommodate these sounds. The former possibility would in effect be claiming that we should throw out all distinctive features on the basis of two sounds. The latter would be to so widen the theory that it could accommodate any sounds, or so we argued, but of course a theory that can describe anything explains nothing. We argued then that these sounds were exceptions, not counterexamples, but exceptions that could not be incorporated into the theory in any interesting way. What they show us, we argued, is that theories cannot always account for everything. Theories leak.

  But this conclusion illustrates exactly the point of the black swan—namely, that counterexamples and exceptions are etically the same, but emically very different. For example, if we expand a universal phonology to account for all phonetic rarities, we weaken it because in so doing, it will predict things never found. At the same time, these rarities show us that our theory can never account for everything. We must resign ourselves to having exceptions at all times that in principle ought not to be incorporated into or analyzed by the theory. To see what I mean, let’s examine these phonetic rarities in Pirahã and Wari' in more detail.

  When I was first planning to visit the Pirahãs, I read what little information there was about them. I learned that among other things they have a sound [l], which apparently is not found in any other language in the world. This had first been written up by an American missionary, Arlo Heinrichs, and subsequently observed by Heinrichs’s successor, Steve Sheldon. When I arrived among the Pirahãs in 1977, therefore, I was eager to hear this sound. Since Sheldon and Heinrichs had left lists of words in which this sound appeared, I asked the Pirahãs for exactly those words that featured [l]. After observing the word myself, I realized that no one had ever published an article on the sound, so I wrote up a small paper and submitted to the Journal of the International Phonetics Association, JIPA, which published my four-page description of the sound, as a voiced apico-labial double flap. It did not occur to me that this sound could have theoretical significance; it was merely a discovery that I wanted to share. And there was little reaction to the sound until I began to work with Ladefoged.

  Likewise for the [tp] trill of Wari' and its closely related language Oro Win. The phoneme had been known for several decades by missionary-linguists who worked on these languages. Yet no one had realized that both [l] and [tp] were not predicted by anyone’s phonological or phonetic theory. This failure to predict the sounds means that no phonological theory had any “slot” or “matrix” in which these sounds could fit—they were anomalies. But were they counterexamples or exceptions? If a counterexample is an analysis of facts that a particular theory cannot explain, what would an exception be? Exactly the same: an exception is an analysis of facts that a particular theory cannot explain. The difference between a counterexample and an exception, of course, is that the former is taken to be a serious problem that a theory must address, while the latter is a hiccup that the theory might have to account for at some point, but for now can safely ignore.

  (The subtle distinction I am attempting to draw here does not include pseudo-exceptions. For example, if I claim that I have discovered an arithmetic system in some isolated community in which 2 + 2 = 5, I have in fact made a mistake. By definition any such claims is a misunderstanding and thus is neither a counterexample nor an exception. Many purported “counterexamples” are simply misunderstandings. For example, claims like “The word haggis is cultural, therefore grammar is influenced by culture” would be the linguistic equivalent of an arithmetic error. Another example of an actual counterexample or exception is found in the general statement and specific statement pair: “All birds fly. Ostriches do not fly.”)

  Whether we classify an anomaly as counterexample or exception depends on our dark matter—our personal history plus cultural values, roles, and knowledge structures. And the consequences of our classification are also determined by culture and dark matter. Thus, by social consensus, exceptions fall outside the scope of the statements of a theory or are explicitly acknowledged by the theory to be “problems” or “mysteries.” They are not immediate problems for the theory. Counterexamples, on the other hand, by social consensus render a statement false. They are immediately acknowledged as (at least potential) problems for any theory. Once again, counterexamples and exceptions are the same etically, though they are nearly polar opposites emically. Each is defined relative to a specific theoretical tradition, a specific set of values, knowledge structures, and roles—that is, a particular culture.

  One bias that operates in theories, the confirmation bias, is the cultural value that a theory is true and therefore that experiments are going to strengthen it, confirm it, but not falsify it. Anomalies appearing in experiments conducted by adherents of a particular theory are much more likely to be interpreted as
exceptions that might require some adjustments of the instruments, but nothing serious in terms of the foundational assumptions of the theory. On the other hand, when anomalies turn up in experiments by opponents of a theory, there will be a natural bias to interpret these as counterexamples that should lead to the abandonment of the theory. Other values that can come into play for the cultural/theoretical classification of an anomaly as a counterexample or an exception include “tolerance for cognitive dissonance,” a value of the theory that says “maintain that the theory is right and, at least temporarily, set aside problematic facts,” assuming that they will find a solution after the passage of a bit of time. Some theoreticians call this tolerance “Galilean science”—the willingness to set aside all problematic data because a theory seems right. Fair enough. But when, why, and for how long a theory seems right in the face of counterexamples is a cultural decision, not one that is based on facts alone. We have seen that the facts of a counterexample and an exception can be exactly the same. Part of the issue of course is that data, like their interpretations, are subject to emicization. We decide to see data with a meaning, ignoring the particular variations that some other theory might seize on as crucial. In linguistics, for example, if a theory (e.g., Chomskyan theory) says that all relevant grammatical facts stop at the boundary of the sentence, then related facts at the level of paragraphs, stories, and so on, are overlooked.

  The cultural and dark matter forces determining the interpretation of anomalies in the data that lead one to abandon a theory and another to maintain it themselves create new social situations that confound the intellect and the sense of morality that often is associated with the practice of a particular theory. William James (1907, 198) summed up some of the reactions to his own work, as evidence of these reactions to the larger field of intellectual endeavors: “I fully expect to see the pragmatist view of truth run through the classic stages of a theory’s career. First, you know, a new theory is attacked as absurd; then it is admitted to be true, but obvious and insignificant; finally it is seen to be so important that its adversaries claim that they themselves discovered it.”

  In recent years, due to my research and claims regarding the grammar of the Amazonian Pirahã—that this language lacks recursion—I have been called a charlatan and a dull wit who has misunderstood. It has been (somewhat inconsistently) further claimed that my results are predicted (Chomsky 2010, 2014); it has been claimed that an alternative notion of recursion, Merge, was what the authors had in mind is saying that recursion is the foundation of human languages; and so on. And my results have been claimed to be irrelevant.

  Yet in spite of such characterizations, the discussion of whether a language has recursion or not is vital to psychology and linguistics (Futrell et al., forthcoming). And if, as I claim, the manifestation of recursion or other properties of grammar are constrained by culture and dark matter, it is also vital to anthropology. One of the oldest and most important empirical programs in cognitive science and linguistics aims to characterize the range of possible human languages. Linguistic universals—if any exist—would point to deep properties of the cognitive mechanisms supporting language; at the same time, the search for possible universals and violations of universals creates rich data for linguistic theory.

  One aspect of this controversy is whether a purported universal (recursion, in this case) actually needs to be observed in every language. Some linguists think so, depending on the nature of the universal, while others claim that linguistic universality is a claim of abstract cognitive abilities and not about the formal inventory of any specific language. The question is not whether this or that language actually has recursion, for example, but whether the speakers of that language are capable in principle of speaking a recursive language. Thus is a language without recursion a counterexample, an exception, or irrelevant to current syntactic theory?

  To try to see how this is answered culturally, we must begin with two culturally distinct understandings of universals of language, Greenbergian or Chomskyan universals. One of the most common objections raised by critics of the idea that Pirahã falsifies the suggestion that recursion is universal is that the absence of recursion superficially (in what people actually say) does not mean that the language could not be derived from a recursive process mentally. And this is correct. (See chap. 5 for technical discussion.) Some then proceed from this banal observation to conclude that the claim that Pirahã lacks recursion is either deliberately or ignorantly failing to understand the difference between Greenbergian and Chomskyan universals. This is an old accusation, one that I and others have rebutted in numerous publications.

  The late Joseph Greenberg was a professor at Stanford University and was the first researcher to make serious proposals on linguistic universals: forms or implications between forms actually observed in all or most of the world’s studied languages. Thus Greenbergian universals refer to things that can actually be observed and thus easily falsified. Chomskyan universals are quite different. Chomsky’s concept of universals includes the notion of what he refers to as “formal universals.” Formal universals are grammatical principles, processes, or constraints common to all languages—that is, supposedly following from UG—at some level of abstraction from the observable data. Thus these refer to things that cannot be seen except by the appropriate theoretician. Unfortunately, this makes formal universals difficult to falsify because they can always be rescued by abstract, unseen principles or entities; for example, so-called empty categories (which frankly I find reminiscent of Kepler’s “epicycles”—invisible to all but the initiate).

  Take recursion. The Chomskyan claim would be that all languages are formed by a recursive process, even though the superficial manifestation of that process may not look recursive to the untrained eye. So long as we can say that a sentence is the output of Merge, limited in some way, then it was produced recursively, even though superficially non-recursive. The Greenbergian way, on the other hand, would be to say that either you see recursion or it is not there.

  Both positions are completely rational and sensible. But the Chomskyan view renders the specific claim that all languages are formed by Merge/recursion untestable. In Chomsky’s earlier writings, he claimed that if two grammars produce the same surface strings (weak generative capacity), we can still test them by examining the predictions of the structures they predict for the strings (strong generative capacity). Since most of my work on Pirahã recursion has been to show that the predictions Merge makes are all falsified (see chap. 5), I have dealt exclusively with strong generative capacity. Of course, clever lads and lasses can always add epicycles to the accounts to save Merge but, again, with two effects: (i) it loses all predictive power, and (ii) it provides a longer, hence less parsimonious, account of the same structures.

  The Chomsky-Greenberg split is only apparent in this case. Pirahã falsifies the Chomskyan formal universals predictions/account (sans epicycles, i.e., the bare claim of Hauser, Chomsky, and Fitch [2002]) and is irrelevant to a Greenbergian account, exactly the opposite of the normal dialogue occurring among my critics.

  This cultural divide between those whose value ranking is THEORY >> DATA vs. DATA >> THEORY surfaces again in theoreticians’ criticisms of N. Evans and Levinson (2009). Again, a feature could be abstractly present, even though it is superficially absent. This is correct. But the critics then make the mistake of moving from this banal observation to conclude that they/we are either deliberately or ignorantly failing to understand the difference between Greenbergian and Chomskyan universals.

  In the value system of the theory of dark matter here, there are the following values:

  1. Understanding particulars is vital and is the first step in developing an etically valid basis for emic science.

  2. There is no atheoretical research, so be informed.

  3. Using insights from multiple theories can mitigate the counterexample vs. exception quandary.

  4. Never be too sure.

  5. T
he same structure can be a counterexample in one language but a pseudo-exception in another, depending on the “field/matrix” view (or in some cases, the “dynamic” or “wave” view).

  Summary

  This chapter applied the concept of dark matter as the primary tool for interpreting all of our experiences, focusing on visual perception among various societies, with a special focus on Pirahãs’ perception of two-dimensional representations. The chapter demonstrated that Pirahãs’ cultural background makes it difficult for them to recognize photographs that are degraded in rather minor ways, contrasting strongly with the perception of other populations. This was argued to be a result of the different apperceptional background of Pirahãs compared to, say, North Americans. At the same time, the chapter discussed ways in which North Americans (me in particular) are unable to perceive things that the Pirahãs are in fact able to perceive quite easily.