The Science of Language

by Noam Chomsky

  Chapter 9

  Page 56, On inclusiveness as a computational principle

  Inclusiveness is the principle that after selection (technically, “numeration”) of a set of lexical items (with their features) as ‘input’ to a linguistic computation, nothing new is introduced until the computation yields a complex sound–meaning pair, or the computation crashes. It can be thought of as an aspect of one version of modularity: that a computation cannot draw on outside information, systems, or principles to make a linguistic computation come to the ‘right’ outcome, or crash. It is also – as pointed out by Chomsky below – an aspect of simplicity, for it incorporates a notion of computational economy. The label “inclusiveness” was first introduced by Chomsky (1995b), but in various forms (such as ‘lexical projection’), it has long played a role in Chomsky's view of a linguistic derivation (the computation of a sentence or complex expression). For an introduction to the issue of complexity and simplicity in linguistics, see Appendix IX.

  Page 57, On making up new words

  “Making up new words” can amount to several things. One is the trivial and theoretically uninteresting matter of changing associations – associating the sound “arthritis” with the concept STOMACH DISEASE, for example (see “Internalist Explorations” in Chomsky [2000] for discussion). Another is making up new – new to an individual or to the community, or in history – sounds or concepts. New sounds are constrained by any parameters that might apply: “strid” is ok for English speakers (it is a possible English-appearing sound), but not for Arabic speakers. There may be fewer or more constraints on novel (in either way) concepts; no one knows, because we don’t know much about what concepts are, as noted. It is clear we introduce new ones: COMPUTER (in the sense understood these days by child and adult, perhaps something like “artifact (something made) that is used (TOOL) for . . .)”) was introduced at a time, and is readily understood or acquired by anyone. We can surely introduce others, such as ones for which we have no use now, but might later. And we should expect some kinds of internally and/or third factor determined constraints, for otherwise we would not come to develop the virtually universal conceptual resources (capacities to quickly mobilize concepts) we seem to have. In any case, at least at this stage of inquiry, the possibility of introducing novel words on an occasion does not figure in discussion of linguistic computation. That is, one assumes that a computation ‘starts’ with established lexical items or packages of phonological, semantic, and formal (including parameters) ‘information.’ Distributed morphology (depending on what that amounts to) introduces variants on the theme, but does not address novel introductions.

  Chapter 10

  Page 59, Chomsky's efforts to convince philosophers

  See Chomsky (1975, 1980/2005, 2000) for his efforts to convince philosophers. See also the discussions in Appendices III, V, and VI.

  Page 61, On connectionism and behaviorism

  Connectionism is in part a thesis concerning the architecture of brain wiring: that brains are interconnected neurons wherein the firing rates of individual neurons serve through their connections to other neurons to increase or decrease the firing rates of those other neurons. However, this is arguably not the central theme of those who call themselves connectionists. The central theme seems to amount to a learning thesis – a thesis about how ‘connections’ come to be established in the (assumed) architecture. The learning thesis is a variation on behaviorism and old-style associationism. Training procedures involving repetition and (for some accounts) ‘backpropagation’ (or some variant) lead to differences in ‘connection weights’ in neural pathways, changing the probability that a specific output will occur, given a specific input. When the network manages to produce the ‘right’ (according to the experimenter) output for a given input and does so sufficiently reliably under different kinds of perturbations, the network has learned how to respond to a specific stimulus. Chomsky seems to focus on the architectural thesis of connectionism in the discussion here. His views on the learning thesis amount – so far as I can tell – to his negative views of behaviorism.

  On the behaviorist/connectionist learning thesis: it is remarkable how much those who devote large amounts of their intellectual work to pushing that program as far as they can have been able to accomplish while working with quite simple ‘neural’ models (computer models simulating the supposed architecture and endowed with one of several possible statistical sampling algorithms). The work of Elman and his students and colleagues gives some idea of what has been accomplished with what he calls “simple recurrent networks” (SRNs). That said, the accomplishments give no reason, at least at the moment, to abandon the rationalist nativist research strategy in favor of connectionist or other empiricist ones with regard to language or other internal systems that appear to be innate. Specifically, there is no empirical reason, for there is no reason to think that the connectionist learning thesis offers anything that could count as an adequate view of the child's course of linguistic development. To say this is not to deny that the ‘choice’ procedure that a child's mind carries out in setting parameters might rely on what looks to be a form of statistical sampling like that noted above and remarked upon in what Chomsky says about the work of Charles Yang, mentioned before. But that is irrelevant, for the set of possible options seems to be fixed – and fixed by either biology (the genome) or third factor matters – we don’t know yet which (although there are independent reasons to place them in third factor considerations). That fact rules out pursuing a strategy that assumes from the start something like Joos's and other empiricists’ view that variation is unconstrained. See also Appendix VIII, and note Chomsky's (2009) remark about the possibility that there are infinite numbers of parameters. If there are any parameters at all, there are ‘choice’ points with fixed options.

  Chapter 11

  Page 66, Chomsky's linguistic rules vs. those of (many) philosophers

  There are many philosophers who endorse the idea that ‘the mind’ does indeed ‘follow rules,’ although without in any way taking ‘mind’ and ‘rule’ to be anything like ‘rule’ and ‘mind’ as Chomsky conceives them. Sellars and his followers, for example, think of the mind as a neural net that has been trained to follow the “rules of language,” where these are understood in epistemological terms. That is, the net is trained to follow what a community of language-users take to be the epistemically correct language-entrance (perceptual), language-exit (action-related) and language-internal inferential rules that a particular community endorses as “the rules of the language game.” For Chomsky, rules are principles embodied in the operations of the language faculty, the language faculty is a naturally occurring biophysical organ, and there is no sense in asking if the rules/principles it observes are the epistemically correct ones for dealing with the world. For further discussion, see Appendix VI. For the moment, the important point is that Sellars essentially takes the commonsense notion of people following rules (which for him is readily understood) and applies it to the idea of training people to follow rules (in the behaviorist manner, again readily understood) to (he believes) establish inferential “connections” in a neural net. He makes the same mistake that those who believed that action must be understood in terms of a contact mechanics did: he believes (without, of course, acknowledging to himself that he is doing so) that commonsense notions suffice for understanding “the mind.” Chomsky's view is that language and its rules must be understood in terms of the nature and operations of a biophysical system, and they can only be understood by appealing to the methods of the natural sciences – a view that he sometimes calls “methodological monism,” opposed to methodological dualism (adopting a different methodology – here drawn from common sense – for the study of mind). Behaviorists and their offspring make the same error.

  Sellars also held (in his “Philosophy and the Scientific Image of Man”) that the concepts that figure in our ‘ordinary’ languages are artifacts, inventions over millennia
of social beings – people – and that these concepts have the characters that they do because of the roles of the correlative terms, with the roles fixed by the rules of usage. The concept PERSON is what it is because of the way the term “person” is used in those languages. Concepts – ways of understanding the world – and prominently the concepts PERSON and AGENT as creatures of the “original” and “manifest” images that humans have invented – will inevitably conflict with the “scientific image” of the world that emerges in the developing sciences. So he would to an extent agree with Chomsky who, later in this discussion, endorses the idea that the commonsense concept PERSON is not one that can sustain scientific investigation – as too the commonsense concept LANGUAGE, and so on. But the agreement is superficial. Chomsky's commonsense concepts are innate, products of some kind of system or systems (perhaps including the language faculty) that are themselves biophysically based. And Chomsky sees no difficulty in holding that PERSON and LANGUAGE (as understood in folk psychology) will continue to have their uses in our commonsense thoughts and dealings with the world even though the sciences of mind do not have these concepts, and give them no roles to play. Metaphysics (and in particular Sellars’ eliminativist form of scientific realism) gets no grip.

  Chapter 12

  Page 71, ‘Meaning’ computation vs. ‘sound’: Chomsky's intellectual contribution

  One point Chomsky makes here is to emphasize that the operations involved in providing ‘information’ (primarily semantic features) to the semantic interface don’t need to rely on linearity. That is needed only when dealing with the phonetic interface, where linguistic signals have to be produced and perceived in a temporal, linear fashion. On the meaning side, processes can proceed “in parallel.”

  On Chomsky's intellectual contribution more generally, note that he begins by revitalizing and placing in the much broader context of how to go about making sense of language as a biological organ his 1959 devastating review of B. F. Skinner's Verbal Behavior. He earlier construed his contribution as not only the construction of a plausible theory of language within the rationalist tradition, but a criticism of empiricist dogma about how to construct theories of the mind. The criticism of behaviorism is generalized now to criticism of the naïve form of evolution one finds in at least some views of the role of selection and adaptation, backed up by a developed biolinguistic (or as indicated elsewhere, bio-physico-computational linguistic) account of language, its evolution in the species, and its development/growth in individuals. See also Appendix II.

  Chapter 13

  Page 72, Chomsky, simplicity, and Goodman

  It is useful to read this section and the next (on Chomsky's relationship to Nelson Goodman) together, for Chomsky got some of his formal techniques from Goodman's “constructionist” project (one that goes back to Carnap and his Aufbau) and his appreciation for and pursuit of the notion of simplicity from Goodman. There are important differences, of course, primarily having to do with Goodman's behaviorism and refusal to countenance the nativist foundations that his ‘solution’ to projection (an aspect of induction) and his account of simplicity clearly needed. These points are anticipated in the discussion immediately below.

  For one view of the connection between Chomsky's ‘internal’ simplicity measure and the Minimalist Program (and also Chomsky's debt to Goodman), see Tomalin (2003). Chomsky comments on these and related matters below. Tomalin draws from Chomsky's 1970s introduction to the published version of his The Logical Structure of Linguistic Theory (1955) – surely the best place for readers interested in Goodman's and others’ influences on Chomsky's early work to start. For a historical study focusing on the influence of formal sciences on the emergence of transformational grammar, see Tomalin (2006).

  Chapter 15

  Page 87, On creativity and its basis in a fixed nature; the puzzle of empiricism

  Evo-devo, canalization, etc. emphasize the point at the developmental level in ways that might have surprised – but also gratified – Descartes and his rationalist followers, all of whom, with the addition of a few of the Romantics (Wilhelm von Humboldt, A. W. Schlegel, Coleridge) seemed to have some kind of recognition that individual and cultural diversity and creativity required fixed natures. Chomsky offers an illuminating historically oriented account of the point and its implications in his Cartesian Linguistics, and I touch on these matters in my introductions to the second (2002) and third (2009) editions of this important work. Given how obvious the point is, the popularity of the empiricist view of a plastic mind is a puzzle. At times (see Chomsky 1996, for example) Chomsky suggests that the ‘solution’ is not argument, but Wittgensteinian therapy-for-philosophers (and similarly inclined psychologists and linguists). Wittgenstein held that philosophical problems are hopelessly confused conundrums created by the oddities of the ways we speak of ourselves, time, the world, etc: “ordinary usage” is fine for solving practical problems, but hopeless for theoretical purposes – a conclusion Chomsky endorses. The only way to get ‘solutions’ to these philosophical ‘problems’ is to stop insisting on trying to solve them. There are no solutions; the problems are Scheinstreiten (pseudoproblems).

  Page 89, A possible moral faculty

  Marc Hauser has published some of the work he and others have done on the topic (2006). Mikhail should be given the credit for pioneering much of this work in his Cornell PhD thesis in philosophy, completed with some direction from Chomsky. (See www.law.georgetown.edu/faculty/mikhail/.)

  Chapter 17

  Page 98, On Hume, the missing shade of blue, and the study of human nature

  Hume seems to have held that the colors we can array and order – given his assumption that every ‘simple’ idea in the mind must be derived from a simple ‘sense impression’ – must include only those that have actually been experienced. If someone has not experienced a specific shade of blue, then, he or she should not – strictly speaking – be able to form an idea of that blue when presented with a spectrum with that specific blue missing. Yet he or she can. To deal with the apparent counterexample to his assumption, Hume in effect dismissed it, saying it was singular and not worth considering modifying his basic view of ‘simple’ ideas, of where they come from, and of how they ground further experience. He likely dismissed it because of his empiricist views. If they are abandoned, it allows for an appeal to ‘instinct’ and thus for the operation of innate internal systems (although he did not believe that these could be investigated, unlike Chomsky and others who pursue the task of constructing theories of the mind on nativist assumptions). For further discussion, see Appendix X.

  Page 99, Some cross-references

  Truth-indications (as opposed to truth conditions) are discussed on p. 273. For an outline of differences between Chomsky's naturalistic views of syntax, semantics, and pragmatics and those found in standard approaches, see Appendix XI. For further justification, see VI.

  Chapter 18

  Page 100, On universalizing moral principles

  Chomsky and I appear to be speaking of different things. I was wondering whether, in denying universality (of a moral principle), someone refuses to treat others as humans. He understood me as asking whether there are people who deny universal application of a moral principle. Of course there are – people like Kissinger, who does it openly, and racists of various sorts, who may do it less openly. Nevertheless, Chomsky points out, even while denying that one must apply the same moral standards to one's own actions as one applies to others, Kissinger is likely to endorse a further universal principle: the USA always acts in the best interests of a majority of humanity, or perhaps even the interests of those against whom aggression is committed. Whether such a universal claim can be justified on moral grounds – or factual ones, for that matter – is a very different question.

  On a related matter, see Chomsky's remarks about John Stuart Mill below and ‘humanitarian intervention’; note also his article (2005b) and, more recently, his address to the UN general assembly them
atic dialogue on the responsibility to protect (2009). On universalizing one's moral assessments and its connection with the responsibility of the intellectual, see Chomsky (1996, Chapter 3) and the reprint of his early article on the responsibility of intellectuals (1987) and elsewhere. Intellectuals for Chomsky include all those who gather, assess, and distribute information; it includes, then, academics and media personnel.

  Chapter 19

  Page 104, Chomsky on ‘faith in reason’

  Apropos what follows, note Chomsky's response to James Peck's (Chomsky 1987) interview question, “Do you have a deep faith in reason?” Chomsky said, “I don’t have a faith in that or anything else.” To Peck's query “Not even in reason?” Chomsky continues, “I wouldn’t say ‘faith.’ I think . . .. it's all that we have. I don’t have faith that the truth will prevail if it becomes known, but we have no alternative to proceeding on that assumption, whatever its credibility may be.” Chomsky apparently sees faith as the abrogation of reason. Its role in maintaining the “state religion” is a continuing theme in his political work.