Dark Matter of the Mind

by Daniel L. Everett

  How does one know that five-month-old infants or even one-month-old infants have not experienced and disliked selfishness or the withholding of something desired from them—such as their mother’s breast? Why appeal to a vague nativism until experience has been ruled out? The dark matter theory of mind maintains that the sources of our knowledge are apperceptions, reasoning in a general sense (though likely Bayesian), and culture. Until these three broad possible sources of knowledge have been ruled out, until the semantics of the moral judgment have been sorted out, until the evolutionary story has been properly worked out, claiming that infant glances in one direction or another are evidence of some sort of innate knowledge is unconvincing.

  The proponents of semantic universals, innate phonological knowledge, and moral instincts—like other nativist approaches to human epistemology—all fail to systematically evaluate alternatives to their nativist accounts. More important, their theories crash upon three massive rocks: Occam’s razor (they are unnecessary unless proven to be so); empirical results (they are wrong or untested); and evolutionary theory (they are just-so stories in the absence of serious evolutionary origin accounts).

  With regard to the first problem, Occam’s razor, or “do not multiply entities beyond necessity” (a summary of the views of William of Ockham, first presented with this name in 1852 by Sir William Hamilton), the problem is clear. Everyone agrees that we need learning apart from instinct—learning to find square roots, learning to like fine whiskey, learning to raise cattle, and so on. If learning can be extended to all human knowledge, that is, if every kind of knowledge could be learned (including grammatical structures, skills, and concepts) then we would obviously not need instincts. Instincts make sense only if we have clearly shown that an animal brings unlearnable knowledge or skills from their genome. If it cannot be clearly and uncontroversially demonstrated that learning of a particular skill or other form of knowledge is impossible, then proposing an instinct is proposing an unnecessary entity, a violation of Occam’s razor. And yet arguments for a “poverty of stimulus” or instincts, especially barring origin arguments, are nothing more than arguments from silence. They raise intriguing problems but settle nothing. Those who find a more compelling case for learning as opposed to knowledge-in-the-genes approaches will perhaps be forgiven for being underwhelmed by nativism for this reason alone.

  But this is not the only reason for rejecting nativism. Beyond Occam’s blade, proposals for nativist epistemologies are unconvincing empirically. There are never exceptionless cases, for example. If you are claiming that some knowledge is innate and yet we find evidence to the contrary, then this is an empirical problem. D. Everett (2012a) is chockablock with counterexamples to nativist claims about language. Of course, there are possible answers for any perceived empirical shortcoming raised. The question any such answer raises, though, is how many ancillary hypotheses are required. We have seen some of the empirical problems for nativist linguistics and morality above. The literature is full of more (e.g., Buller 2006; R. Richardson 2007; Gopnik 2010; V. Evans 2014; Blumberg 2006; Lieberman 2013). The works just mentioned all raise many more varied and detailed empirical problems for a wide range of nativist accounts.

  But even if Occam’s razor and empirical shortcomings were not problems for any nativist account of human knowledge, the evolutionary question, the question of origins, is fatal. I return to this directly, following a more detailed discussion of Chomsky’s universal grammar.

  No Universal Grammar

  I want to summarize our discussion to this point on universal grammar before moving on. It is interesting that in sixty-plus years of generative linguistics, not a single analysis ever proposed causally implicates UG. By this I do not mean that there has been no analysis where the linguist assumes that what he or she is proposing is universal because it is part of UG. Rather, I refer to an analysis of a phenomenon that could not be proposed without assuming that grammar is innate. There are many theories of what UG would look like, but no predictions and no analysis of any syntactic structure in which UG is crucial.13

  Before beginning a detailed criticism of universal grammar, however, we should note that there are several definitions of it. Here is one from Chomsky in an e-mail to me on April 8, 2006: “UG is the true theory of the genetic component that underlies acquisition and use of language.” This seems innocuous enough. In fact, it could easily be translated as meaning that the cerebral cortex itself is universal grammar. But more than this is meant, if one searches for other definitions. For example, in an interview with Wiktor Osiatyński, Chomsky offers a less biological and more epistemological, or Bastian, view:

  I think the most important work that is going on has to do with the search for very general and abstract features of what is sometimes called universal grammar: general properties of language that reflect a kind of biological necessity rather than logical necessity; that is, properties of language that are not logically necessary for such a system but which are essential invariant properties of human language and are known without learning. We know these properties but we don’t learn them. We simply use our knowledge of these properties as the basis for learning. (Chomsky 1984, 96)

  In fact, however, the concepts of “innate” and “instinct” are both imprecise and overused with little scientific or philosophical rigor. Philosopher Matteo Mameli, in a series of papers, has revealed the widespread imprecision in these terms, as have Buller (2006), R. Richardson (2007), and Blumberg (2006), among others, from a more empirical perspective. I want to examine their criticisms here and then explain why I think that instinct-based accounts of human behavior are inferior to the dark matter account I have been developing until this point.

  Problems with Instincts: Definitions and Origins

  Instinct is a difficult concept to pin down. Animals seem to be born able to do things that they had no opportunity to learn—baby turtles finding the sea; dogs pointing or herding; ducks imprinting on their mothers; human infants’ production of most speech sounds in the worlds’ languages early in their lives, followed by a great winnowing process as they home in on their mothers’ speech sounds; and so on.

  Often it is believed that animals have instincts in far greater numbers than humans. In Darwin’s day (Blumberg 2006, xiii), instincts were “God’s way of manifesting rationally intelligent behavior in otherwise irrational animals.” And as Blumberg goes on to say, “The term instinct is often merely a convenience for referring to complex, species-typical behaviors that seem to mysteriously emerge out of nowhere” (ibid, xiv). In any case, the strongest reason for retiring the terms instinct and innate from scientific thought is the devil of the details, which shows them to be unnecessary at best.

  But there is also a definitional problem—that is, how to express with clarity what the term innate actually means operationally. For example, here is a partial list of possible definitions of innate:

  A trait is innate if it is not acquired; a trait is innate if it is present at birth; a trait is innate if it reliably appears during a particular, well-defined stage of life; a trait is innate if it is genetically determined; a trait is innate if it is genetically influenced; a trait is innate if it is genetically encoded; a trait is innate if its development doesn’t involve the extraction of information from the environment; a trait is innate if it is not environmentally induced; a trait is innate if it is not possible to produce an alternative trait by means of environmental manipulations; a trait is innate if all environmental manipulations capable of producing an alternative trait are abnormal; a trait is innate if all environmental manipulations capable of producing an alternative trait are statistically abnormal; a trait is innate if all environmental manipulations capable of producing an alternative trait are evolutionarily abnormal; a trait is innate if it is highly heritable; a trait is innate if it is not learned; a trait is innate if (i) the trait is psychologically primitive and (ii) the trait results from normal development; a trait is innate if it is generatively ent
renched in the design of an adaptive feature; a trait is innate if it is environmentally canalized, in the sense that it is insensitive to some range of environmental variation; a trait is innate if it is species-typical; a trait is innate if it is pre-functional; etc. (Mameli 2008; see also Mameli and Bateson 2006)

  Mameli and Bateson (2006) show in painstaking detail that all of these definitions are inadequate.

  But let’s suppose that, in spite of this daunting problem, we can find a workable definition of instinct or innate. Even then, however, we would still not be in a position to use these terms. The reason is that we cannot attribute something to the human genotype convincingly without some evolutionary account of how it might have gotten there. And such an account would have to offer a scenario by which the trait could have been selected. To do this we would need information about the extent and character of variation in ancestral forms as well as differential survivorship and reproduction of those forms. To know how something was selected, however, we need to know something about the ecology under which the selection took place, such as an answer to questions like what were/are the ecological factors that explain the innate trait, either in the biological or social or other abiotic environment? Next, to use instinct or innate we would need to know how the traits could be passed on to subsequent generations. There should be a correlation between phenotypic traits of parents and offspring greater than chance. Then we would need to know about the population during the time of selection. Any evolutionary biologist also knows that we must have information concerning population pressures, gene flow, and the environment leading to the diffusion of the trait. We do not know the answers to these questions. We are in no position at present to know the answers. And we will never be able to know some of the answers. Therefore, there simply is very little utility to the terms instinct and innate.

  Alison Gopnik, a professor of psychology at the University of California at Berkeley whose research highlights the child’s learning abilities, as opposed to native instincts, offered a similar assessment of nativism:

  It’s commonplace, in both scientific and popular writing to talk about innate human traits, “hard-wired” behaviors or “genes for” everything from alcoholism to intelligence. Sometimes these traits are supposed to be general features of human cognition—sometimes they are supposed to be individual features of particular people. The nature/nurture distinction continues to dominate thinking about development. But its time for innateness to go. (Gopnik 2014)

  She offers an apt illustration:

  [Researchers] took two different but genetically identical strains of mice which normally develop different degrees of intelligence and cross-fostered them—the smart mice mothers raised the dumb mice pups. The result was that the dumb mice developed problem-solving abilities similar to those of the smart ones and this was even passed on to the next generation. So were the mice innately dumb or innately smart? The very question doesn’t make sense. (ibid.)

  And finally:

  The increasingly influential Bayesian models of human learning, models that have come to dominate recent accounts of human cognition, also challenge the idea of innateness in a different way. At least since Chomsky, there have been debates about whether we have innate knowledge. The Bayesian picture characterizes knowledge in terms of a set of potential hypotheses about the world. We initially believe that some hypotheses are less probable and others are more so. As we collect new evidence we can rationally update the probability of these hypotheses. We can discard what initially looked very likely and eventually accepting ideas that started out as longshots. (ibid.)

  To sum up, the evidence against evolutionary psychology and massive modularity includes not only our earlier points, but also the following:

  1. Evolutionary psychology is incompatible with cerebral and mental plasticity.

  2. EP confuses arguments between innatism and modularity—that if something is innate, it is an argument for modularity or highly specific epistemological cognition, as opposed to very general perceptual or emotional biases.

  3. EP fails to offer any serious account for profound culturally influenced variability in so-called module-directed behaviors (or even much field research to check cross-cultural variability), such as in non-Indo-European cultures with little contact with the outside world.

  4. There is no physical portion of the brain dedicated to any module—not even of the five senses—that cannot be used for something else instead.14

  5. Most of EP’s thesis of “massive modularity” is based on the reasoning used in arguments for a language instinct and linguistic module, as argued by Chomsky, Fodor, Pinker, and others. But as I show in D. Everett (2012a) and as we saw above, there is no compelling evidence for any such instinct or module. There is not a single component of language that cannot be explained by either (i) the nature of the communicative task; (ii) the effect of culture; or (iii) general biological constraints on humans (see also Dascal 2012; Enfield 2013a; D. Everett 2013a).

  People often raise the issue of “solution space” as an argument for UG and other innate knowledge. If UG is correct, then the task of the child faced with learning its native language is considerably easier. If a child has to search at random for generalizations, the story goes, without innate guidance, it might never learn its first language. Of course, though solution spaces are important, so is motivation. UG has nothing to say about motivation, assuming that all children learn their first language equally well, which is not at all clear (D. Everett 2012a). But the solution space could emerge from images (see V. Evans 2014 among many other works on this), from cultural meanings, from mathematical properties of language, from very basic, easily acquirable input for a Bayesian general learning mechanism common to many tasks humans face, and so on. Nativism by no means is the only perspective that can explain child knowledge acquisition. Goldsmith (2015) offers a detailed assessment of solution space needs and comes down on the side of what he calls “the new empiricism” rather than nativism—that solution spaces need not be hardwired.

  Having discussed some of the problems with proposals of innate knowledge in different domains, I want to discuss what would be required to argue for instincts from an evolutionary perspective. Many of the points discussed below are from R. Richardson (2007). They are essential to any theory of innate knowledge, for if there is no potential explanation of how traits could have evolved, proposals that such traits are innate are weakened.

  1. Selection: We need information about the extent and character of variation in ancestral forms, as well as differential survivorship and reproduction of those forms.

  2. Ecology: What are the ecological factors that explain the innate trait, either in the biological or social or other abiotic environment?

  3. Heritability: The mechanisms for genetic transmission must be understood clearly. There should be a correlation between phenotypic traits of parents and offspring greater than chance.

  4. Population structure: There must be information concerning population structure, gene flow, and the environment leading to the diffusion of the trait.

  5. Trait polarity: Which traits are primitive and which are derived? As is well known, adaptive features are themselves not necessarily adaptations (e.g., the usefulness of skull sutures for human parturition—these did not evolve for that, however useful).

  Other questions that require answers before accepting instincts or nativistic accounts of traits anywhere in nature include the following (also from R. Richardson 2007):

  6. How often and in what circumstances is a species-typical trait developmentally and post-developmentally environmentally canalized (insensitive to environment)?

  7. How often and in what circumstances is a species-typical trait generatively entrenched (resisting disturbance because it is closely connected to other traits)?

  8. How often and in what circumstances is a species-typical trait the result of stable developmental sequences?

  9. How often and in what circumstances ar
e developmental and post-developmental environmental canalization both present?

  10. How often and in what circumstances does natural selection result in developmental or post-developmental canalization?

  11. How often and in what circumstances does natural selection result in generative entrenchment?

  12. How often and in what circumstances does natural selection result in stable developmental sequences?

  13. How often and in what circumstances does natural selection result in adaptive plasticity?