The Science of Language

by Noam Chomsky

  As for reference of the word–world variety, for Chomsky it becomes a matter of use by people, and thus an issue in pragmatics, not his version of semantics. Can the study of pragmatics be, or turn into, a science? First, let us agree that pragmatics or the study of language use is part of a general ‘theory of language’ (see Chomsky 1999) in some broad sense. But that is not at issue; the issue is whether the use of language by humans can be a naturalistic science that one can investigate using the tools of natural scientific research. In the case of language, those tools assume some degree of biologically based regularity. However, with some exceptions, that is not to be found in the linguistic actions of humans; the creative aspect of language use provides a great deal of evidence to the effect that there are no causal principles that tie environmental or brain stimuli to specific forms of linguistic behavior. It is an empirical question, and cannot be definitely decided now, or ever. But the weight of evidence at the moment is very much against there being such causal principles. For ample discussion, see the main text and Appendices V and VI. However, if one includes under the title of ‘language use’ certain forms of inference that people draw, one can find cases where inferences are at least licensed by the structures and semantic information found in SEMs and the computations that lead to them. From assuming the truth of Jane's brown cow isn’t producing, one can infer that Jane has a brown cow, and from her having a brown cow that she has a cow. An important ‘conjunctivist’ view of why this follows is found in Pietroski (2005). And from you may have cake or ice cream, there is at least an implicature to your being permitted to have one or the other, but not both; see Pietroski and Crain (2005). Are these inferences determined by narrow syntax, the core system of the language faculty and its output at SEM? That is a stretch; assuming so would include these inferences in the computational resources provided by Merge. Still, they are sanctioned by the computations of the language faculty and by the semantic information these make available at SEM in a way that many inferences that people draw are not.

  Appendix XII: An internalist picture of how concepts ‘work’

  In comments on the discussion in the text, I spoke of concepts as ‘configuring’ experience and imagination; the terminology is repeated in earlier appendices. This puts in a different way views expressed by Chomsky (1966/2002/2009): “The strong assumptions about innate mental structures made by rationalistic psychology and philosophy of mind eliminated the necessity for any sharp distinction between a theory of perception and a theory of learning. In both cases, essentially the same processes are at work; a store of latent principles is brought to the interpretation of the data of sense. There is, to be sure, a difference between the initial ‘activation’ of latent structures and the use of it once it has become readily available for the interpretation (more accurately, the determination) of experience” (2009: 102). There is an obvious problem with this statement, apparent in the conjunction of the two sentences. The triggering system that relies on perceptual/sensory input to yield a concept might make a specific concept available, one that is not – however – sensory/perceptual. Whether perceptual or not, it does not fix its employment or use. However, this does not affect the crucial point. The crucial point for present purposes is that innate conceptual, linguistic, sensory, and other forms of internal ‘cognitive’ machineries partially determine experience in that they – not the ‘world outside’ – fix how one can see and understand. Taking this seriously, I suggest that SEMs, which are construed here as complexes of lexically specified innate concepts, do their ‘work’ in an ‘adverbial’ way. They fix – or with other systems contribute to fixing – the ‘hows’ of experience: the various manners or ways in which one can conjecture, understand, imagine, and experience. Interpretation is not a matter of searching for the right concept or right description to fit some ready-formed experience, but a matter of ‘making’ the experience, here understood as participating in a cooperative exercise involving several mental systems, each with its unique form of contribution. This point was anticipated in discussion above, but it needs to be addressed in some detail because it is so easy to start down a hopeless road.

  The basic idea comes from discussion several decades ago by philosophers of what was called an adverbial account of visual sensation. (For the record, it was ignored more than rejected, in part – I suspect – because it conflicted with the commonsense-driven externalist and anti-nativist intuitions that dominate much philosophical discussion of sensation and perception.) This account of sensation was introduced to undermine the grip held on accounts of sensation and perception offered by two incorrect – according to the adverbialists – views of visual sensation. One was the “sense datum” view, the idea that the visual system (“the mind”/“the brain”) produces things called “sensations” or “sensa,” that these sensations (for example, color sensations) are mental objects of some sort, and that their role is to serve as the “immediate objects” of sensation and perception – not the ‘things outside.’ They in turn – on some views – mediate perception of things ‘out there’ or – on others – even constitute the total content of visual experience. The other was the view that sensations can only be classified and individuated by appeal to what they are of or about, so that to say what a sensation of red is, is to say what it is about – generally assumed to be a property of a surface of things ‘out there.’ The danger that lies in the first, sense datum, view is that it seems to depend on what must surely be an incorrect view of the sensory contents of the mind, one that seems to require that when I have a green sensation, that I (or something) sense something green in my mind/brain. The danger in the second is that it seems to support an externalist view of sensory states and events – suggesting that to say what they are, one must speak of things ‘outside’ and their properties, and making the internal states and events mere “re-presentations” of various things out there, with no evidence that anything ‘out there’ corresponds to the way that sensation portrays it. This is one of the points Locke and others were making when they spoke of “secondary qualities” – not that I am endorsing their account of them, nor of the difference claimed between secondary and “primary” qualities.

  The adverbial account offered an alternative account of sensation – that is, of the mind/brain's role in sensory experience. It suggested that instead of thinking of the mind as a theater populated by green sensations at which some internal homunculus (or the person, for that matter) stares, one should think instead of the mind as ‘containing’ various sensory events – ‘sensings’ – and that these events differ from one another in ways that are determined by the nature of the mind/brain's sensory mechanisms. These events might require stimulation of some sort, either external via impingements on the eye (assumed to be the usual case) or internal, but given that stimulation, they participate in building a visual or imaginative scene populated with what the mind makes out as colored surfaces. Intuitively, where a person might be inclined to say that they sense green or see green, he or she should when careful resist and say rather that their visual system/mind/brain senses greenly – that is, that the relevant mental system functions on an occasion in one of the ways that is characteristic of an organism with the relevant kind of mental machinery, machinery that constitutes visual scenes. A minimal visual scene can be thought of as a particular assignment to values of coordinates of a retinocentric six-dimensional volume.1 Each of the points in this volume has a specific set of ‘spatial’ and ‘color’ coordinate values, the spatial coordinates being (visual) depth, altitude, and azimuth and the color coordinates hue, brightness, and saturation. Think of a specific set of stimulus-derived assignments to these coordinates as a representation in the internalist sense. It does not re-present; the organism as a whole ‘uses’ it to do that, virtually automatically in the case of visual experience. It is a representation in the way that a linguistic SEM or PHON is a representation, a complex mental event described in theoretical terminology that in the theory of
mind is treated as a specific configuration of an interface with other mental systems. The advantage for the internalist to this way of construing the matter is that it places colors not ‘on’ sensations, whatever they might be, but makes them out to be specific assignments of color values, specific output values of a subsystem of the mind. Think of these 3-D output values as describing particular complex mental events that momentarily constitute a minimal visual space, where the latter is understood as what the mind produces. They are theoretical ‘objects,’ the varieties of which are specified by a theory that offers a way to describe and explain how a system of the mind/brain works, and its contributions to an organism's mental operations. As for SEMs, a specific assignment in the case of a SEM amounts to a structured set of specific semantic features that help ‘make’ a way to understand and, specifically in cooperation with vision and other relevant systems (for example, object configuration), to perceive things in the world as such-and-such.

  Looked at from the point of view of the organism as a whole and its experience and actions, the science of vision provides a way to think of how the visual system partially constitutes experience, ‘making’ colored, located visual ‘objects.’ Because visual experience in cooperation with other systems such as object- and facial-configuration systems usually proves reliable to an organism and offers partial but not necessary ‘output’ that with contributions from other systems generally proves sufficient to allow the organism to navigate and identify nutrients and enemies in its efforts to get along ‘in the world,’ the ‘things’ that are assembled (in part) through the contributions of an internal sensory system and many others are treated as ‘really out there,’ even though the properties and surfaces and classifications of such things as people and friends, or apples and food, are created by the mind. They can be thought of as ‘projections’ of the mind. Returning to color, the adverbial account claims that from the point of view of the science of color (as opposed to the commonsense conception of color and the entities of the commonsense world), colors are forms of mental event, ways of sensing and perceiving that differ from one another in ways determined by the nature of the visual system. They are not properties of things ‘out there,’ however reliable the objects created in visual experience prove to be for practical purposes, and however tempting the commonsense view of the world with its colored objects might be. Nevertheless, it serves the practical interests of an organism to see and think of the colors as properties of objects ‘out there.’

  The visual system typically contributes to ‘experience’ – normally understood as an organism's reaction to distally caused input. Sometimes it contributes to imagined scenes. The language faculty, in contrast, only sometimes contributes ‘online’ to experience – to conceiving of something seen or heard as such-and-such having such-and-such functions, for example, and thereby constituting ‘it’ as a thing with those functions and other properties. It often contributes ‘offline’ to cases of imagining, speculating, proposing, thinking, etc. – ways of understanding and conceiving. It is a competence system, not an input system. Nevertheless, one can think of SEMs and other forms of mental entities as working in an adverbial way too. So construed, they do not ‘exist in the mind’ as conceptual objects inspected and perhaps manipulated and used as tools by some kind of internal homunculus ‘understander’ and agent, thereby paralleling the way that the sense datum theory construed the entities of the visual system. Rather, they can be seen as specific kinds of ways in which a person can – perhaps with contributions from other systems – understand, imagine, classify, ‘think about’ things and events, and the like. They are mental events that differ from one another in the ways determined by the language faculty and what it provides at SEM. Those ways are described and explained by a theory of the language faculty and its possible SEMs, and given any specific I-language (that is, parameter settings and lexicon). Assuming they contribute along with other systems in some way, the features of ‘interface values’ that the language faculty and other faculties make available at their interfaces contribute to human cognitive capacities.

  Incidentally, one can grant that at least some of the ways that the mind configures experience, thought, etc. are provided – at least in the case of sensory systems that are virtually identical to those found in some other primates, not language – by systems that the organism's mind has available to it as a result of receiving it from a common ancestor or ancestors several millions of years ago, perhaps even from a whole class of organisms that employ rhodopsin in vision, if Gehring is right. One can also grant that the systems that offer these ways to configure experience would not prove as useful for practical purposes as they have without such an origin while, however, denying that the ways that the mind configures experience somehow map the way things ‘out there’ really are, or rather, how they are from the point of view of sciences of the relevant entities. The science of color vision provides a useful reminder of how misleading it can be to conclude that since commonsense objects appear to have the properties our minds assign to them (specifically, some of them appear colored), and since the framework proves so useful to serving our practical interests, that commonsense objects and their properties – not the minds that harbor and assign the properties – are both ‘real’ and must be, at least in part, the targets of a science of color. The point should be even more obvious with the concepts and complexes of concepts expressed in the sentences of human natural languages. As Chomsky points out in discussion in the main text, if you want to find out what the objects of common sense ‘are’ and can be, forget about looking outside. Instead, construct a science of the concepts (here thought of as ways of configuring and constituting thought, imagination, and experience) that we have available and that we employ in our thoughts, speculations, and dealings with the world. To do that, look inside the head. If you want to know what persons are, look at the internal and mental-system-provided concept PERSON with its rich and interest-serving characteristics, characteristics that enable flexible applications by humans when they speak and employ this concept, in a wide range of cases. Do not focus on specific applications of the concept – specific ways in which it configures. The concept, or at least similar versions thereof, may well be universal across the human species because it is fixed by the system(s) involved in bringing it into the mind. However, its uses by multiple cognitive systems and (seen from another point of view) by humans are anything but fixed. And in looking inside the head, you can also avoid the apparently tempting idea (at least to many philosophers – although not Locke or Hume) that the concept of a person is some kind of re-presentation of a person, or persons.

  Why is the idea that our minds represent the world such a tempting view – so tempting that Fodor and others simply default to it, treating it as an axiom that cannot be challenged? An answer is implicit in what is said above: commonsense objects (with their visual-spatial and color properties, all assigned by the mind) prove useful. But they do so for practical purposes only. They do not serve the interests of the scientist.

  Before continuing, a comment on the terminology that Chomsky often employs when he speaks of the systems on the other side of SEM. He speaks of them as “conceptual-intentional” systems to which SEMs “give instructions.” That way of putting it can, I think, mislead; see footnote 2 in Appendix VI. In the discussion and comments on it, I make SEMs out to be complexes of (lexical) concepts themselves, organized collections of the semantic information offered in lexical items. That is a way to avoid the suggestion in Chomsky's term, “conceptual-intentional.” His terminology suggests that SEMs, whatever they are, are items that relate to concepts ‘in’ other systems, or perhaps – as Pietroski sometimes says – instructs other systems to build concepts. This way of talking invites taking seriously notions such as a Fodorian language of thought, the locus of concepts. I suggest it be avoided. For good reasons, I believe, Chomsky explicitly rejects Fodor's view in the main text discussion, and implicitly in (1996, 2000) and elsew
here. The adverbial account offers a way to avoid that suggestion, and to think of SEMs as I suggest, as complexes of lexical concepts – concepts understood as ways of configuring experience, etc. that are articulated in terms of semantic features (here a technical term). More carefully, think of SEMs, as above, as the language faculty's contribution to a human's conceptual (configurational) capacities. The contribution is partial, as noted: the language faculty's semantic information provides “perspectives” from which to view “aspects of the world as it is taken to be by other cognitive systems” (Chomsky 2000). This way of stating it acknowledges that the semantic information that SEMs provide cognition should be seen only as a partial contribution to the ways our minds shape thought, imagination, and direct experience. Occasions on which a single faculty operates in isolation are likely to be rare, at least in actual, not experimental cases. Nevertheless, as the discussion also emphasizes, language's contributions are plausibly unique, and therefore in principle separable – as are the contributions of vision and other faculties. If so, it is possible to divide language's specific contributions to cognition and understanding from those of vision, audition, and so on.

  Rationalist internalists have always thought of the proper location of concepts (one variety of what Descartes called “ideas”) as the mind, not things ‘out there’ and their properties. They have, however, been plagued by the locution (in English and some other languages) “concept of . . . (or “idea of . . .” “thought about . . .,” etc.), a locution which invites and excites externalist intuitions, making it appear that to say what a concept is, one must say what it is ‘about’ or ‘of.’ The adverbial account of a concept and more generally of the contributions of various mental systems, like the adverbial account of a color, undermines that intuition by treating a concept as a way of understanding and experiencing and of its status in the theory of mind as a configuration (stated in theoretical terms) of a system's interface. Undercutting externalist intuitions, this fact is, I think, a reason to adopt the adverbial account, and to think of SEMs as linguistically expressed concepts and ways of understanding. More carefully, the semantic information brought to SEM by linguistic computation is ‘put to use’ by other systems not in the way that people put a language to use to solve practical problems, or anything like that. Rather, the information constitutes a partial – but for humans, crucial (and uniquely human) – contribution to our cognitive resources, resources that – to switch to the agentive mode of speaking ruled out in the science of mind – we use to say and do what we want and need to say and do.