LINGUIST List 7.1208

Sat Aug 31 1996

Disc: Grammaticalization

Editor for this issue: Ann Dizdar <>


  1. Martin Haspelmath, Disc: Grammaticalization
  2. Bill Croft, Grammaticalization
  3. "Larry Trask", Disc: grammaticalization

Message 1: Disc: Grammaticalization

Date: Mon, 26 Aug 1996 18:24:55 +0200
From: Martin Haspelmath <>
Subject: Disc: Grammaticalization
In the discussion of grammaticalization, David Pesetsky writes:

> This remark [by Oesten Dahl] puzzles me. Putting aside Fritz Newmeyer's 
> questions, let's suppose that these studies [on grammaticalization] show 
> what they claim to show. In what way are
> the results incompatible with "the Chomskyan paradigm"?

The results of grammaticalization research are not incompatible with
the Chomskyan program, but they don't fit well with it, as Oesten
notes. The central part of the Chomskyan program is to explain grammar
acquisition (and thus indirectly grammar) on the basis of highly
specific innate mental structures ("universal grammar"). Research in
grammaticalization shows that to a large extent, the way grammars are
structured results from the way they are shaped in diachronic
processes of grammaticalization. (For instance, the fact that
languages with rich agreement are generally pro-drop results from the
generally simultaneous grammaticalization of personal pronouns and
loss of agreement affixes.)
 Grammaticalization changes arise in the process of language use --
it is quite impossible to account for them in a pure competence
perspective. Grammaticalization studies thus show that many
grammatical structures cannot be understood without reference to
language use. This undermines the Chomskyan program in that it reduces
its applicability to those grammatical domains that are unaffected by
language use.
 Ideally, linguists would work together trying to figure out which
properties of grammar are due to innate structures, and which
properties are due to language use. Unfortunately, there seems to be a
widespread tendency among both functionalists and formalists to claim
all of grammar for their approach, and to ignore the results of the
other approach. How can we overcome this tendency?

Martin Haspelmath (Free University of Berlin)
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Grammaticalization

Date: Mon, 26 Aug 1996 14:26:27
From: Bill Croft <>
Subject: Grammaticalization
 I have a few points to add to those already made on

 I think the main problem with Fritz's approach to the
unidirectionality issue is his input-output definition of the change:
if X > Y and Y > X, then the change is not unidirectional. First,
depending on how generally you define X and Y, lots of changes can be
bidirectional. If X and Y are P and V, then yes, P > V as well as V >
P. But if X is "off" and Y is "die" or "go away from", then we find
many examples of "go away from" > "off" but not vice versa, and at
least one example of "off" > "die" but not vice versa.

 A second, more serious objection is that the PROCESSES by which X >
Y and Y > X may be quite different. To take a phonological example,
Ferguson has shown that the process by which d > dh [interdental
fricative] is quite different from the process by which dh > d. In
other papers, Ferguson has also shown that even the "same" process in
the "same" direction is actually two different processes (his work on
s > h). Although I wouldn't go so far as Martin Haspelmath to say that
grammaticalization is "ordinary change", it is certainly a different
process from the word formation processes that Fritz presents as
counterexamples to the unidirectionality grammaticalization.

 If anything, these examples suggest that possibly all diachronic
processes are unidirectional---and demonstrate that one should not
make global claims about language change but instead explore the
specific cases in greater detail. An example of this is Elizabeth
Traugott's arguments (presented at the 1995 ICHL) that the syntactic
"liberation" of certain grammaticalizing elements can be constrained
to elements that become discourse markers. Again, looking at a highly
general level, it seems that anything goes, but in fact that view
overlooks the many clear regularities of specific processes of
language change.

 There is still much that needs to be explained, of course. I do not
share Fritz's optimism that we can already explain all language
changes through principles of speech production and perception. Fritz
alludes to Joseph & Janda's recent presentations, and their analogies
such as the ice cream analogy. While such analogies are entertaining,
they can also be quite misleading. These analogies (they have given
others) mislead one into believing that language change is an easy
problem that has already been solved without needing to appeal to
grammaticalization. A review of the literature indicates that that is
definitely not the case. The nature of language change is a hotly
disputed problem, and most solutions are very general and
unconstrained. At least grammaticalization theorists are trying to
constrain language change, even if particular claims end up being
falsified or revised.

 I also share some of David Pesetsky's puzzlement as to why the
unidirectionality and unity of grammaticalization processes are
believed by Fritz and by Oesten to be such a serious problem for
generative grammar (by Fritz indirectly, in that his critique of
grammaticalization is part of a critique of functional-typological
linguistics in general). Let me make a stab as to why the unity of
grammaticalization might pose a problem. If phonological, syntactic,
and semantic processes are part of a single unified process of
grammaticalization, that would imply that the sign---signifier (form)
and signified (meaning) is a single unit. That unit cuts across the
allegedly autonomous modules or components of phonology, syntax and
semantics in a generative grammar.

 Otherwise, I don't really know. In his 1992 Language paper, Fritz
suggested that grammaticalization fell outside the explanatory domain
of UG (Newmeyer 1992:785). In LGB, Chomsky suggested that phenomena
outside UG are 'idiosyncratic factors, as contrasted with the more
significant reality of UG' (Chomsky 1981:8). It is possible that the
regularity and potential explanatory significance of
grammaticalization is in itself perceived as a challenge to generative

 Finally, it should be pointed out that when historical linguists
decide that word/construction A is the ancestor and not the descendant
of word/construction B, either in a single language or in two
languages in a language family, they use lots of other evidence than
the fact that A > B in other unrelated languages (if indeed they even
have that information). Independently established sound changes, facts
of syntactic distribution, sociolinguistic differences among variants,
etc. also inform a historical linguist's decision, and in fact may be
more decisive than any consideration as to whether A > B or vice versa
in other languages. So while grammaticalization theorists may not use
evidence from reconstructions where the justification of the
reconstruction is that A > B in other languages, it is not
illegitimate to use evidence from reconstructions when the
reconstruction is established on the basis of independent,
language-family-particular evidence.

Bill Croft
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: Disc: grammaticalization

Date: Mon, 26 Aug 1996 11:45:09 BST
From: "Larry Trask" <>
Subject: Disc: grammaticalization
In his response to Oesten Dahl's remarks, David Pesetsky asks "In what
way are the results [from grammaticalization studies] incompatible
with `the Chomskyan paradigm'?"

No doubt it is unwise of me to try to put words into Dahl's mouth, and
I hope he'll be replying himself. But I might venture a couple of

First of all, Dahl does not in fact claim that these results are
"incompatible with" the Chomskyan paradigm. He merely declares (or
rather presupposes) that these results are difficult to fit into the
Chomskyan paradigm, which is not quite the same thing.

Now, what I *think* the point Dahl may be making here is one which has
been made on other occasions by other linguists. The study of
grammaticalization has seemingly revealed powerful regularities in the
way speakers use languages, and, at the very least, the Chomskyan
approach simply has nothing to say about these regularities.
Moreover, it appears difficult to conceive of how a Chomskyan approach
could ever have anything interesting to say about them, since the very
concepts and categories required to characterize these regularities
(or at least many of them) do not exist in Chomskyan linguistics.

Other students of grammaticalization have in fact gone rather further
than Dahl in drawing attention to the perceived inadequacies of the
Chomskyan approach in addressing their findings. The most explicit
such statement known to me is that by Paul Hopper and Elizabeth
Traugott in their well-known textbook (pp. 33-38), though similar
remarks can be found elsewhere. The difficulty that Hopper and
Traugott draw attention to is the typical Chomskyan position of
maintaining that a speaker possesses a more-or-less well-defined
grammar of her language, and hence (as argued at great length by David
Lightfoot) that language change must be interpreted as grammar change.
That is, in language change, one form, rule, or parameter setting is
simply replaced by another.

Now more than thirty years of the study of language change has
demonstrated quite conclusively, I think, that language change is not
like this. Instead, it is frequently gradual in every conceivable
respect: temporally gradual, individually gradual, socially gradual,
and, above all, lexically gradual. In grammaticalization, as
elsewhere, an innovating form typically exists side-by-side with an
older one for quite some time, often over generations; it may be
confined at first to occurring in very narrow circumstances, and be
extended only piecemeal to other circumstances. And such observations
are deeply in conflict with the Chomskyan position.

This last, of course, is not the point Dahl is making, but it's also a
substantial point. But both points are problematic for a Chomskyan
view: grammaticalization (like other types of language change)
proceeds in a *manner* which is inconsistent with a fundamental
Chomskyan assumption (well-defined grammars), and it proceeds in
identifiable *directions* which are inexplicable in Chomskyan terms.

Of course, a proponent of the Chomskyan paradigm might, at least in
respect of Dahl's point, take refuge in a metatheoretical standpoint:
a theory treats what it treats, and we may not reasonably criticize a
theory merely because it fails to treat something else we happen to
find interesting. But an opponent might retort that powerful
regularities in linguistic behavior should reasonably be taken as
evidence of the nature of the human language faculty, and hence that a
putative theory of that faculty which fails to address such
regularities is inadequate. I guess you pays your money and you takes
your choice.

Larry Trask
University of Sussex
Brighton BN1 9QH
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue