(Originally appeared in Lognet 95/1)
Let me describe my point of view on cognition and linguistics theory, my view of the current scene in theoretical linguistics (including the “mentalese hypothesis” and, by contrast, an alternative hypothesis), and my reasoning and conclusions after reading Steven Pinker’s The Language Instinct (Harper-Perennial, 1995; W. Morrow and Co., 1994).
Then, let me recommend some actions.
I am a theoretician. Often, it is difficult for me to decide how seriously to take a book, like Steven Pinker’s The Language Instinct, that is aimed at the general public yet purports to deal with theoretical subjects. Pinker has much to say about issues that concern many members of the Loglan Institute. Since the book is a national best-seller and has gotten rave reviews in the rest of the English-speaking world, we ought to take note.
I have decided to take what Pinker says (about cognition and linguistic theory) seriously. I assume that he says what he means and means what he says (to the extent that it is possible to be sure of such things in English). However, unlike Pinker, I am not attempting to write for the general public. Clarity and conciseness are as important to me as they are to Pinker, but I will not avoid abstractions and other tools of the theoretician’s trade.
I will discuss one chapter, entitled “Mentalese,” from Professor Pinker’s book. To keep this essay short I will not summarize the chapter. Readers already familiar with the book or the chapter will readily recognize the issues under discussion here. Readers not familiar with the book or chapter should be able to interpolate from what I write to what is written there. A later reading of the book should be doubly rewarding.
In the Mentalese chapter of his book, Pinker presents a “mentalese hypothesis” (see below for a definition of ‘mentalese’). His authorship of that hypothesis is irrelevant to this discussion so I will refer to it as the mentalese hypothesis (M). In my master’s thesis, I presented a semiotic manifold hypothesis that provides an alternative explanation for some of the same cognitive phenomena. My authorship of that hypothesis is irrelevant to this discussion so I will refer to it as the “semiotic hypothesis” (S).
In the rest of this essay I will try to show that the flaw in M is its fundamental misrepresentation of the roles that language plays in thinking and in the development of thinking skills during each individual’s lifetime.
In the glossary of his book, Pinker defines ‘mentalese’ as follows: “The hypothetical language of thought, or representation of concepts and propositions in the brain, in which ideas, including the meanings of words and sentences, are couched.” M is essentially an affirmative answer to a rhetorical question Pinker asks:“... are our thoughts couched in some silent medium of the brain—a language of thought, or mentalese—and merely clothed in words whenever we need to communicate them to a listener?”1 But M also asserts that language is instinctual, “... a distinct piece of the biological makeup of our brains ...”2 Further, “Language is a complex, specialized skill, which develops in the child spontaneously, without conscious effort or formal instruction, is deployed without awareness of its underlying logic, is qualitatively the same in every individual, and is distinct from more general abilities to process information or behave intelligently.”3 Clearly, M asserts that mentalese is a genetically determined cognitive component of the human brain.
A description of M is commercially available but a description of S is not. In addition, S is addressed to audiences with very specialized technical knowledge and even a summary would go beyond what can be presented in this short essay. Consequently, I will present here only the pertinent results of S by contrasting them with M.
Both M and S assert a relation between the reach of our genetically endowed capacity for thought (genetically determined cognitive reach) and the reach of our syntactically endowed capacity for thought (syntactically determined cognitive reach). M asserts that syntactically determined cognitive reach is a proper subset ofgenetically determined cognitive reach as shown in Figure 1.
S asserts that they are intersecting sets as shown in Figure 2.
A direct consequence of the relation that M asserts between the genetically and syntactically determined components of cognitive reach is a denial ofthe strong version of the Sapir-Whorf hypothesis and a trivialization of the weak version. To understand how this is so, imagine a dynamic version of Figure 1 (like a cartoon or movie) that would represent what happens as language is acquired. Basically, by acquiring the grammar and lexicon of a natural language we expand what we can express in language to include much (although perhaps not all) of what we can understand in mentalese. Graphically, we can imagine the size of the proper subset expanding within the area of its superset as shown in Figure 3.
The dynamic version of S is very different. As in M, the syntactically determined cognitive reach of an individual begins as a growing proper subset of his or her genetically determined cognitive reach. Then, two things happen: syntactically determined cognitive reach extends beyond the boundary of genetically determined cognitive reach and a new component of cognitive reach is created like an expanding ring around the genetically determined component as shown in Figure 4. This new component shares some features with each of the other components. Like the genetically determined component, much of this component remains beyond the reach of language. S uses the technical term ‘transattentional’ to qualify this feature. For our purposes, the term ‘intuitive’ captures the salient characteristics. Like the syntactically determined component, this component is learned. So, in S,we can talk about the syntactically determined and the learned-intuitive extensions of cognitive reach. S asserts that both of these extensions are dependent on particular components of the syntax of the native languages of monoglots.
Apparently mentalese, as defined in M, plays the role in theoretical linguistics that the ether played in theoretical physics before the Michelson-Morley experiments. At that time, the quantification of the features of the ether (its density, its relative velocity with respect to the earth, etc.), was a subject of vigorous debate and the motivation for Michelson and Morley’s experimental design. However, the existence of the ether was little discussed among scientists because it seemed to them to be required as an explanatory principle for theoretical physics.
Mentalese, too, seems to be required as an explanatory principle for the brands of theoretical linguistics (like M) that Dr. Pinker prescribes. Like the concept of the ether in physics, mentalese is so broad that it conveniently fills in or covers holes in any linguistic theories that might use it. Pinker refers to mentalese as “some silent medium of the brain”4 and characterizes the internal symbolic representation used by a Turing Machine as “a kind of mentalese.”5 He does not claim that the Turing Machine kind of mentalese is the only possible kind. So, even if we argue successfully that the human brain cannot be modelled by a Turing Machine, mentalese will endure. Proponents of M will simply argue that the nature of human mentalese is not like the Turing Machine kind.
So, Pinker sees being scientific in linguistics as dependent on participation in the debate on the nature, not the existence, of mentalese.6
In physics, Einstein provided a theoretical framework in which the existence of the ether was not required. An analogy between theoretical physics at the turn of the last century and contemporary theoretical linguistics, raises the possibility of a theory, historically analogous to Einstein’s theory of special relativity, that could provide a theoretical framework in which the need for mentalese is eliminated.
The first step toward such an elimination might be a reduction in the scope of the term ‘mentalese’. The theoretical purpose of M is to explain how genetically determined cognition can support the acquisition and use of language. The role of mentalese after the acquisition of language should be left open. In particular, we should not insist that mentalese delimits our thinking. For this purpose, if mentalese exists at all, it need not be the medium of thought. Mentalese should only be defined as a medium for some thought because such a definition is sufficient for the stated purposes of M (or theories like M such as S). In fact, for these theoretical purposes, it is sufficient for the “some” to be bounded below by non-human cognition (primate cognition may provide the boundary) and from above by syntactic cognition.7 So, a theoretically sufficient definition of mentalese is: a medium for Homo Sapiens-specific, presyntactic cognition. This limited conception of a genetically determined component of human cognition is viable even for S-like hypotheses.
Some observations about the relation between M and Loglan and M and Universal Grammar help us to understand the implications of M.
Loglan for Turing: Pinker considers the observation that no spoken tongue could be the basis of a Turing Machine that makes logical deductions8 as definitive support for the contention that spoken language cannot be the basis of thought. He concludes that mentalese must be the foundation for all thought. However, that supporting observation is demonstrably false: Loglan can provide such a basis.
Pinker is evidently ignorant of Loglan, or at least fails to take it seriously. I come to this conclusion based on the following facts:
No entry exists for Loglan in the index.
No entry for James Cooke Brown (or any other Loglan Institute author) exists in the bibliography.
Pinker shows an appreciation of linguistics as a potential experimental science, and he discusses the Sapir-Whorf hypothesis at some length. However, he does not mention Loglan, a language which was designed to address in an experimental setting the issues raised by the Sapir-Whorf hypothesis.
Although this may simply be a consequence of his professionalism, that is to say, the narrowness of his scientific reading, it may also be an indication of a marketing problem for the Loglan Institute. How is it possible that a professor of Cognitive Neuroscience at MIT has never heard of Loglan? Perhaps the other alternative is even worse. Could it be that he has heard of Loglan but has dismissed it entirely?
Universal Grammar and Mentalese: In M, mentalese is even more fundamental to cognition than Chomsky’s Universal Grammar (UG). UG is hypothesized by Chomsky to be an “innate”, that is, a pre-experiential, template of the principal features of the particular grammars that human individuals actually learn, but certainly not a complete description of any of them. In short, UG is not an absolute. In fact, its assumed adaptive function is to set boundary conditions for the acquisition of a particular grammar in a very restricted time frame. Since we are never again able to learn a language with the same velocity, our grasp of UG evidently diminishes after about the age of seven. According to Pinker, however, without mentalese, the language-learning child would not have the cognitive skills required to set the switches on his inherited Universal Grammar to adapt it to the surrounding linguistic culture. In addition, no thought can occur, even after language acquisition, that is not representable in mentalese.
In this sense, Pinker’s thought transcends Chomsky’s yet depends on it. With Pinker we have moved from the Chomskyan hypothesis of the existence of a Universal Grammar template that aids children in their attempts to learn their natural languages, to an entire species-wide mentalese that allows us, and in weaker dialects all animals, to learn and behave intelligently.
Pinker scoffs at the Sapir-Whorf (S-W) hypothesis.9 This attitude is based on the argument that humans think without language in “mentalese”.10 For Pinker, mentalese is a species-wide, innate function that is unaffected by the process of learning a language or the process of using a language. All actual and possible thought is, or would have to be, conducted in mentalese. Clearly, if what can be thought is limited to the reach of our genetically determined cognitive component, then the syntax of the language or languages we speak can’t possibly limit what we think (the strong version of the Sapir-Whorf (S-W) hypothesis). At most it can limit what we say or write.
Similarly, the weak version of S-W can be reduced to the trivial observation that it is sometimes difficult to translate between languages. The important point is that mentalese is always available as the common foundation of human thought and communication both among and within all languages and cultures.
By denying both linguistic determinism (the strong version of the S-W hypothesis) and linguistic relativism (the weak version of theS-W hypothesis), Pinker adopts a position (reminiscent of Plato’s Doctrine of Forms) that takes mentalese as a pre-experiential given. In addition, since he claims that mentalese is not affected by later learning but, on the contrary, provides the foundation for such learning, mentalese, in M, is a complete mechanism, not merely a template to be used in further learning like Chomsky’s UG.
According to M, language is not a tool for thinking but merely a tool for translating my mentalese into yours. As a tool for understanding the world, language, in M, is like a pair of eyeglasses. At best, language can help us to “see” more clearly the abstractions that we can already “see” without language. In S, language is like an extensible toolkit: a pair of eyeglasses, a microscope, a telescope, and a lens-making tool. Language not only clarifies what we could understand by other means, but lets us “see” things that are beyond our genetic cognitive reach and even programs our nonsyntactic understanding to extend our intuition. This is the misrepresentation of M that I mentioned in the opening paragraphs.
An important consequence of M is that mentalese protects us against the limitations of thought that the S-W hypothesis claims our spoken languages impose on us. As we can see by the contrast with S, the limitations of thought are actually imposed by M. According to M, our genetically determined cognitive medium is not extensible. S, on the other hand, offers the prospect of unlimited extension of both our explicit and our implicit cognitive medium.
In 1928, Hilbert enumerated four prominent open problems in the theory of the foundation of mathematics. Within two years, Kurt Godel had resolved all of them. This essay has contrasted two hypotheses, M and S, about the relation between thought and language. The problem of which one can be falsified in a definitive way is an open problem in linguistics and cognitive science. I believe that Loglan is the instrumental medium that could contribute to the design of experiments that, like the Michelson-Morley experiments in theoretical physics, would test both M and S-W and determine if mentalese exists. If S could play a role in such an experiment, I would be delighted.
Experimental design is never easy. In Loglan 1 James Cooke Brown devotes a chapter to a discussion of the S-W hypothesis and how to test it with Loglan. I think it is time to revisit that experimental design and seek funding.
[All page references are to S. Pinker, The Language Instinct.]
Copyright © 1995 by The Loglan Institute. All rights reserved.