Semantic Externalism, Language Variation, and Sociolinguistic ...

Forthcoming in Mind and Language.
Semantic Externalism, Language Variation, and Sociolinguistic Accommodation*
Daniel Lassiter
Nobody can be forbidden to use any arbitrarily
producible event or object as a sign for something.
G. Frege, ‘On Sense and Reference’
Abstract
Chomsky (1986) has claimed that the prima facie incompatibility between descriptive
linguistics and semantic externalism proves that an externalist semantics is impossible.
Although it is true that a strong form of externalism does not cohere with descriptive
linguistics, sociolinguistic theory can unify the two approaches. The resulting two-level
theory reconciles descriptivism, mentalism, and externalism by construing community
languages as a function of social identification. This approach allows a fresh look at
names and definite descriptions while also responding to Chomsky’s (1993, 1995)
challenge to articulate an externalist theory of meaning that can be used in the scientific
investigation of language.

* Thanks to Alan Musgrave, Colin Cheyne, Charles Pigden, Heather Dyke, Txuss Martín, and an
anonymous Mind and Language reviewer for extremely helpful comments. The bulk of this essay was
written during a year of study at the University of Otago in which I was supported by the Frank Knox
Memorial Fellowship from Harvard University, which I gratefully acknowledge. I am also grateful to the
the Otago Philosophy Department for a very fruitful and enjoyable time. All mistakes are, of course, my
responsibility alone.
Address for correspondence: Department of Linguistics, 726 Broadway 7th Floor, New York, NY 10003
Email: lassiter@nyu.edu
1

---

1. Semantic Individualism and Semantic Externalism
Semantic externalism is rejected by many linguists, as is well known, because it seems to
conflict with the mentalist presuppositions of generative linguistics in the Chomskyan
tradition: see Chomsky (1986, 1995, 2000) for extensive discussion. Even linguists who
reject mentalism and generative linguistics, however, have occasionally objected to
externalism on quite different grounds. Some of the mundane truths of philosophy of
language seem to be in conflict with equally mundane truths of descriptive linguistics: in
particular, descriptive linguists typically see variation and change as a ubiquitous part of
language, while philosophers and formal semanticists, for the most part, view a language
as a static object. Indeed, many philosophers adopt positions that make substantive
predictions about how natural language should work in its everyday use, in particular
those of an externalist persuasion, as I will show. However, insufficient attention has
been paid to the issue of whether these predictions are borne out empirically.
The Chomskyan revolution has received much well deserved attention, but we do
not need to assume his theoretical position in order to derive philosophically interesting
results from linguistics. Rather, undeniable results of descriptive linguistics seem to
militate against dominant trends in the philosophy of language. Many linguists have
taken these problems to indicate that we must choose between externalism and
descriptivism, and this is generally taken to be a sufficient reason to prefer an
individualist approach to language. I will argue that this is a false dichotomy, and that
both sides have misconstrued the terms of the debate. Recent results of sociolinguistic
theory, I will show, suggest a way to bridge the gap between individualist and externalist
2

---

accounts of meaning and language in general without abandoning the basic commitments
of either position.
The broad outlines of the debate are as follows. Individualists believe that the
proper object of the scientific study of language is the language of an individual, his
idiolect or, in Chomskyan terms, his mental grammar, knowledge of language, or I-
language. Individualists typically believe that semantic notions such as reference and
meaning are dependent on such individualistic facts. This does not necessarily mean that
social aspects of language are unimportant or that they do not admit of a scientific
description, though some individualists have made this further claim: cf. Chomsky
(1975). However, most individualists do believe that only individualistic aspects of
language can be formalized and used to make predictions (e.g., about entailment and
grammaticality).
Semantic externalists, on the other hand, hold that a language belongs to a
community of language users, and that common languages or communalects exist above
and beyond individuals. According to this conception, a language has an ontology (e.g.
words and grammatical rules, or social practices and/or conventions) and norms
(standards of correctness) that are in some sense independent of the linguistic competence
of individual speakers. In the words of Michael Dummett, an idiolect is merely ‘a second-
order theory: a partial, and partly incorrect, theory about what the meanings of the
expressions are in the common language, that may be represented as a partial theory of
what the correct theory of meaning for the language is’ (1986, p.469). As Dummett
makes clear, it is possible for speakers to be simply wrong in their use of language
because a language exists independently of its speakers. In contrast, under the
3

---

individualist view, ‘incorrect usage’ is a murky social concept, usually a simple failure of
communication or a faux pas.
The debate is important because the side we choose will determine where we
locate crucial semantic notions such as reference, meaning, and truth. Intuitively, though,
Dummett’s approach makes sense: sometimes an individual’s use of language is just
wrong, and individuals often acknowledge making mistakes upon reflection or correction.
Individualism simply cannot account convincingly for this fact.
Our discussion of these issues will rely on several works that have made this point
clearly. For reasons of space I will merely summarize these well-known arguments of
Wittgenstein, Putnam, Kripke, and Burge very briefly and will not attempt to argue for
their conclusions in detail. I take them to show persuasively that we have strong
intuitions of linguistic correctness that, like intuitions of grammaticality, are an important
empirical bound on the construction of a complete theory of language. Whether or not
this assessment is correct, I hope it will become clear even to the sceptical that
externalism properly construed is not in competition with descriptivism or mentalism and
should not be rejected on these grounds alone.
Wittgenstein argues in Philosophical Investigations (1953) that rule-following
and meaning cannot be explicated by mere description of what an individual is doing: to
say that an individual is following a particular rule already presupposes community
standards about correct application of the rule (though, unlike Wittgenstein, I see no
reason why this could not be a community of one). Kripke’s Naming and Necessity
(1980) undermines the claim that reference depends on a speaker’s knowledge; in
Kripke’s example, I could believe nothing but falsehoods about Gödel, and yet the name
4

---

‘Gödel’ would still refer to Gödel when I say it. Kripke argues instead that names are
rigid designators and that the reference of a name is fixed by a causal chain leading to an
initial baptism. Putnam’s paper ‘The Meaning of ‘Meaning’’ (1975) uses the famous
twin-Earth thought experiment to show that the reference of natural kind terms like
‘water’ is also insensitive to speakers’ knowledge of reference. Putnam argues that their
reference is fixed partly by environmental facts, in this case the actual chemical structure
of water. Finally, Burge argues in ‘Individualism and the Mental’ (1979) that Putnam’s
argument, suitably modified, extends to all terms. Burge shows that, if a speaker of
English were to believe that rheumatism is called ‘arthritis’ and that he has arthritis in his
thigh, the latter belief would be false (since arthritis is an ailment of the joints), rather
than being true-in-his-idiolect as a descriptive theory emphasizing knowledge of language
would predict. Burge concludes that the meaning of ‘arthritis’ is fixed by the word’s use
in a community, whether or not an individual speaker happens to know how his
community uses the word. So, it seems, individual knowledge is not sufficient to
determine reference: environmental and sociohistorical facts are also relevant.1
To repeat, individualism faces a serious problem in accounting for these
systematic intuitions about meaning.2 However, the significance of the arguments for

1 In what follows I am primarily interested in externalism about the linguistic meaning, rather than
externalism about mental content. Burge is primarily interested in mental content, as is much of the
philosophical work that has followed his articles. However, this tradition focuses on the analysis of
language, usually taking for granted that this method can elucidate questions about the nature of mind. I
doubt that this line of research will turn out to be very illuminating about the nature of mind, but I will not
discuss my reservations here.
2 It is not sufficient to reject this normativity as uninteresting, as Chomsky and Davidson do. E.g. Davidson
(2005, p.121): ‘I am not impressed by Michael[ Dummett]’s or Burge’s or Putnam’s insistence that words
have a meaning of which both speaker and hearer are ignorant. I don’t doubt that we say this, and it’s fairly
clear what we have in mind: speaker and hearer are ignorant of what would be found in some dictionary, or
of how people with a better or different education or a higher income use the words. This is still meaning
based on successful communication, but it imports into the theory of meaning an elitist norm by implying
that people not in the right social swim don’t know what they mean.’ In many cases Davidson’s comments
are very much to the point, but we cannot explain all the externalist arguments as relying subtly on elitism.
5

---

externalism has often been misconstrued. Individualists hold that an individual’s
language just is her idiolect; the arguments just mentioned show that this position is too
strong. Environmental and sociohistorical facts are relevant to the determination of
reference, but we have seen no reason to believe that these factors are sufficient for the
purpose. A common approach (perhaps ‘common’ is too strong, since it is rarely made
explicit and less often argued for) is to bite the bullet and hold that individualistic
properties are in fact irrelevant to determining the character of the language. Idiolects,
then, are of interest only to psychologists. Dummett’s quote above exemplifies this claim:
he takes ‘the language’ to be an external entity that is an object of knowledge for
speakers – not like tables and chairs, to be sure, but like social objects such as burial
customs or money, which individuals participate in but do not normally create. A
different but related characterization is due to Lewis (1975, p.6), who writes that the
social aspects of language are mediated by the fact that ‘a given language L is used by, or
is a (or the) language of, a given population P.’ Lewis takes for granted that there is such
an object as the population P to be found (an extremely dubious assumption, as we will
see). I will call this strong form of externalism communitarianism because it holds that a
language is possessed jointly a community and that the identity of an individual’s
language is fully determined by the community to which he belongs (note the analogy
with communitarianism in political philosophy). Communitarianism implies that speech

First, speakers often impose norms on themselves. Second, it is easy to imagine – or indeed locate –
situations in which the ‘correct’ usage is not included in dictionaries or prescriptive grammars, or is
associated with groups outside the elite. Third, our intuitions about linguistic correctness display a
systematicity closely akin to the grammaticality intuitions of theoretical linguistics. Many linguists think, as
I once did, that externalism is no more than a philosophical justification for linguistic chauvinism. My task,
then, is to outline a theoretically interesting position that accounts for the subtleties of the intuitive, pre-
theoretical notion of correctness.
6

---

communities pre-exist individual speakers and are capable of determining a unique
community language with or without their cooperation.3
The upshot of semantic externalism is that there may be a difference between an
individual’s language and what she thinks her language is: there must be some place for
normativity of meaning. This clearly falls short of establishing communitarianism, yet
many externalists rely heavily on an unanalyzed notion of a linguistic community,
effectively making this leap without argument. It is true that communitarianism gives a
simple account of the normativity of meaning and the possibility of error, and we might
prefer it as the best available explanation of externalism for this reason. However, as the
next section will show, communitarianism is fatally incompatible with basic descriptive
facts of human language. Details of this failure point toward a new form of externalism
that is compatible with the insights of semantic individualism, which I will develop in
Section 3. I argue that communalects are neither functions of individual knowledge nor
monolithic objects somehow possessed by a community, but are determined instead by
individuals’ sociolinguistic dispositions. If this is that case, neither the individual nor the
social level of language can be ignored in semantics (or in any other areas of language, I
believe). Rather, a theory of human language that respects the empirical facts must
connect the descriptive and the normative in an illuminating way without disposing of
either.

3 A communitarian can, of course, hold that individualistic properties are relevant to determining which
language is being spoken. Lewis seems to hold this position.
7

---

2. Communitarianism and Descriptive Linguistics
Communitarians often downplay or ignore problems relating to language variation and
change. It is easy to see why: if languages exist apart from the linguistic competence of
individuals, only two analyses of linguistic difference seem to be available – either one or
both of the speakers does not know how to use the language correctly, or they are
speaking different languages.4 In either case, the interesting questions about variation are
not about individuals, but about languages as the joint property of a community. Once we
have given content to the notion of a ‘speech community’, then, variation within the
community can be neatly dismissed as incomplete knowledge of the language. But this
tactic leads us quickly into trouble, for it is extremely difficult to give a descriptively
adequate account of speech communities.
2.1. Individuating Languages
Our ordinary conception of language seems at first to be unproblematic: usually it seems
clear whether two people speak the same language. However, the obviousness of this fact
to well-educated speakers of modern English does not apply to human language in
general. To see this, we need to distinguish two meanings of the term ‘language’. The
first is political: a language such as English or Urdu is defined primarily as the speech of
a certain community of people, often corresponding roughly to political boundaries. The
second notion of ‘language’ involves mutual intelligibility. These two concepts do not
coincide, and only the second has any real interest for philosophers and formal linguists.

4 If we want, we can call this ‘speaking different dialects’ or ‘belonging to different speech communities’.
The terminological difference does not matter as long as it is clear that the only way for communitarians to
avoid analyzing variation as incorrect usage is to invoke distinct non-individualistic standards.
8

---

For instance, standard Swedish and standard Norwegian are mostly mutually intelligible,
but some dialects of Swedish are not. Such situations are commonplace throughout the
world; as a result, politically defined languages are not useful for the externalist.
The mutual intelligibility standard is a more promising way to individuate
languages: two people speak the same language if and only if they can communicate
successfully. An immediate problem is that successful communication is graded and
relative to communicative purposes. In addition, mutual intelligibility is not transitive.
There was presumably an unbroken chain of generations between my distant ancestors
and me such that each generation could speak to its parents and to its children, and yet it
would be absurd to say that I speak the same language as my distant ancestors. Clearly,
then, there is no natural break in the history of English that is sufficient to distinguish
community languages.
Synchronic linguistic variation is often gradual as well, although the facts are less
familiar. Although sharp boundaries between languages sometimes exist due to
geographical or sociopolitical boundaries, in their absence language tends to vary
gradually over geographical and sociopolitical space, simply because people usually talk
more like people they have contact with and less like those they have little or no contact
with. In highly developed and/or recently settled countries this can lead to a high degree
of linguistic homogeneity across vast geographical areas. However, in less developed
countries with poor communication and transport there is usually far more linguistic
variation: for instance, with roughly the area of California or New Zealand, New Guinea
has perhaps 1000 languages (or more or less, depending on your preferred criteria of
9

---

individuation). An interesting effect of these sociological facts is the existence of dialect
continua:
…[S]harp divisions are rare… They normally occur only where there is a
geographical or sociopolitical barrier of such proportions that it severely restricts
communication between subregions. Otherwise, there is likely to be a single
complex of communalects which extends over the entire language area. Any one
communalect will share a few special features with its neighbors to one side, and
a few with those on other sides. No large clearly demarcated dialect regions will
be found. Often communalects spoken at different ends of a dialect chain may
exhibit considerable differences – they may even be mutually unintelligible – but
there will exist intermediate transitional communalects which are mutually
intelligible with both extremes. (Pawley, 1970, p. 3)
An extreme example of this phenomenon is found among the Indo-Aryan languages of
the Indian subcontinent, where ‘[t]he speech of each village differs slightly from the next,
without loss of mutual intelligibility, all the way from Assam to Afghanistan’ (Masica
1991, p.25). This vast area – millions of square kilometres – has hundreds of millions of
Indo-Aryan speakers who use hundreds of mutually unintelligible forms of speech.
The existence of dialect continua shows clearly that mutual intelligibility is not
sufficient to individuate speech communities. Mutual intelligibility is graded and
intransitive, and language variation and change do not proceed smoothly or along well-
defined boundaries. If we were to locate an individual at an arbitrary point on the map of
the Indo-Aryan, Germanic, Fijian, Scandinavian, or Romance dialect continua, the fact
that he shares some feature with certain other individuals in no way entails that he will
10

---