grammaticalization.

Grammaticalization – Theory and Data.
Grammaticalization and explanation.
Irene Appelbaum.
University of Montana.

A recurring theme in a special issue of Language Sciences (2001) devoted to
theoretical debates about grammaticalization is that the causal mechanisms
subserving examples of grammaticalization are explanatorily exhaustive and that
the concept of grammaticalization itself is therefore empty. The position seems to
be a straightforward inference from the assumption that explanation must appeal
to causal mechanisms, together with the recognition that grammaticalization
is not itself a causal mechanism. While this position is unobjectionable,
perhaps even unassailable, in addressing questions of the form How did
grammaticalization-examplex occur in languagey? there are other questions that
seem to be better addressed by appealing to the concept of grammaticalization
itself. In particular, questions of the form What makes language-change-examplex
specifically an example of grammaticalization? are best answered by appealing to
the fact that it satisfies the concept or definition of grammaticalization. Satisfying
the definition of grammaticalization, in turn, requires identifying the language
change example specifically as one involving a lexical to grammatical change (or
a change from less grammatical to more grammatical), regardless of the causal
mechanisms involved in that change. That is, it is only under the description of
the phenomenon as a change from lexical form to grammatical form, that the
mechanisms typically adduced to explain this change can be said to be explaining
it as an instance of grammaticalization.
1. Introduction.
It has been well over a decade since Language Sciences (2001) devoted an issue to
theoretical debates about grammaticalization. A recurring theme in the articles
contained therein is that the causal mechanisms subserving examples of grammaticalization
are explanatorily exhaustive and that the concept of grammaticalization
itself is therefore empty.
The position that grammaticalization has no independent status seems
to be a straightforward inference from the assumption that explanation must
appeal to causal mechanisms, together with the recognition that grammaticalization
is not itself a causal mechanism. Proponents of this view characterize grammaticalization variously as “derivative” (Campbell 2001: 113, 116, 151, 154),
“unnecessary” (Joseph 2001: 184); “a cover term” (Newmeyer 2001: 225); “an epiphenomenon”
or “epiphenomenal” (Joseph 2001: 184, 185; Janda 2001: 304, 321;
Newmeyer 2001: 188, 189,190,191). These terms differ somewhat in emphasis and
rhetorical force, but the underlying point is unequivocal: “grammaticalization has
no explanatory power of its own” Campbell 2001: (151).
While this position is unobjectionable, perhaps even unassailable, in addressing
questions of the form How did grammaticalization-examplex occur in languagey?
there are other questions that seem to be better addressed by appealing to
the concept of grammaticalization itself. In particular, questions of the form What
makes language-change-examplex specifically an example of grammaticalization?
are best answered by appealing to the fact that it satisfies the concept or definition
of grammaticalization. Satisfying the definition of grammaticalization, in
turn, requires identifying the language-change example specifically as one involving
a lexical to grammatical change (or a change from less grammatical to more
grammatical), regardless of the causal mechanisms involved in that change. That
is, it is only under the description of the phenomenon as a change from lexical
form to grammatical form, that the mechanisms typically adduced to explain this
change can be said to be explaining it as an instance of grammaticalization. Since
this perspective seems not to have been elaborated in the earlier debate, it seems
worthwhile doing so now. In any case, that is my aim in what follows.
2. A background of consensus.
The debate over the status of grammaticalization is characterized by an enormous
amount of consensus. Both those who affirm and those who deny any independent
status for grammaticalization agree on a working definition of the term, on
prototypical examples, and on a range of mechanisms that subserve instances of it.
Although there is variation and disagreement regarding each of these, the debate
over the status of grammaticalization is not aimed at ironing out differences in
definition, nor at settling disagreements about peripheral examples or mechanisms
of grammaticalization. To a first approximation, the debate is about how to
characterize the relationship between the mechanisms that explain how individual
examples of grammaticalization occur, and the concept of grammaticalization
understood as language change in which a lexical form becomes a grammatical
form (or in which a grammatical form becomes a more grammatical form).
But here too there is widespread agreement: agreement that different or
multiple mechanisms may be involved in a single instance of grammaticalization
(including reanalysis, analogy, phonological reduction, semantic bleaching, borrowing); agreement that no one of these mechanisms is found exclusively in
cases of grammaticalization; and ultimately agreement that the concept of grammaticalization
is not itself an individual mechanism on par with the mechanism of
reanalysis, for example. The debate, then, more specifically, is about the explanatory
power of grammaticalization, given the above relationship between the individual
mechanisms typically appealed to in order to explain how individual examples of
grammaticalization occur and the concept or definition of grammaticalization.
As noted above, proponents of the view that grammaticalization has no independent
status acknowledge that there is agreement on both the definition of
grammaticalization and a set of language-change examples that satisfy this definition.
Campbell notes, “the phenomenon of grammaticalization is interesting and
not really in question (2001: 113–114). Joseph states:
Everyone agrees that the term refers to the phenomenon in which forms that at
one stage of a language have fairly concrete lexical meanings and functions come
to have more abstract grammatical uses and meanings at a later stage (2001: 164).
The catalogue of grammaticalization examples is large and open-ended. Standard
examples include: the development of grammatical future makers from lexical verbs
of volition (English Future < will; Modern Greek θa < thélo: hina); the development
of derivational suffixes from nouns (English -ly < Old English lic, ‘body’; Romance
-ment < Latin mente, ‘with a mind, Ablative’); and the development of prepositions
or adverbs from body part nouns (Ewe ‘behind, back’ < noun ‘back’ ‘megbé’).
There is in addition widespread agreement that cases of grammaticalization
involve a core set of language-change mechanisms (reanalysis, along with analogy,
semantic bleaching and phonological reduction). Moreover, all parties recognize
the fact that each of these mechanisms is at work in language change examples that
do not involve grammaticalization:
We have examined the component parts of grammaticalization and found that
they all are manifested independently (Newmeyer 2001: 202).
[T]he kinds of changes most commonly encountered in grammaticalization…
are encountered commonly also in instances of changes which have nothing to
do with grammaticalization Campbell 2001: (151).
Reanalysis and analogy have been widely recognized as significant for change in
general,… Reanalysis is the most important mechanism for grammaticalization,
as for all change… (Hopper & Traugott 2003: 39, emphasis added).
A number of mechanisms of language change have already been alluded to as
being relevant to grammaticalization…. All these mechanisms make change
possible, but none are restricted to grammaticalization. (Traugott & Heine
1991: 7). Disagreement arises against this backdrop of consensus. Critics of the independent
status of grammaticalization are at pains to point out that once the mechanisms
at work in particular cases are taken into account there is simply nothing
left for grammaticalization to explain. Let us take a closer look at these arguments.
3. Arguments against the explanatory potential of grammaticalization.
The arguments against the idea that grammaticalization has any independent
explanatory power highlight distinct but related aspects of the view. First and
foremost among these criticisms is the claim that appeal to grammaticalization
directly, i.e. not via underlying mechanisms, is simply empty. Grammaticalization,
on this view, is a place-holder for an explanation, not an explanation itself.
This criticism is akin to Molière’s famous parody of accounts which purport
to explain, but do no more than re-label. When asked what causes opium’s ability
to produce sleep, the doctor replies: Its sleep-producing ability (virtus dormitiva).
Citing a process of grammaticalization to explain what causes lexical morphemes
to become grammatical morphemes is similarly hollow, according to this criticism.
It is tantamount to saying that what causes lexical morphemes to become
grammatical morphemes is their grammaticalizing ability. According to this criticism,
what is needed to further our understanding of language changes in which
lexical morphemes become grammatical morphemes is an account of how these
changes occur, an account that can only be provided by specifying the underlying
mechanisms involved.
As this criticism is put by Campbell:
‘Grammaticalization theory’ has no explanatory value because what it claims
to explain is explained already by other well-understood mechanisms which lie
behind it and, as is generally agreed, it cannot ‘explain’ without appeal to these
other mechanisms and kinds of change (2001: 151).
[S]ound change and semantic change apply to many things in addition to
grammaticalizations. These ‘explain’ instances of grammaticalization, but
grammaticalization itself explains nothing without first calling upon these kinds
of changes and the explanations they afford (Campbell 2001: 158).
[T]here can be no such thing as ‘grammaticalization theory’, unless one intends
that expression merely as a convenient way of referring to the set of independent
theories needed to explain the phenomenon (Newmeyer 2001: 192).
A closely related criticism is the charge that grammaticalization is not necessary to
explain the phenomena under question: since it has no causal power as such, and hence no explanatory potential, appeal to it is wholly superfluous. This argument, which appeals to the principle of Ockham’s Razor — choose the theory that makes
the fewest assumptions necessary to account for the phenomena – amounts to
saying that, if the data consisting of lexical morphemes becoming grammatical
morphemes can be equally well explained by (1) ordinary mechanisms of language
change, or (2) ordinary mechanisms of language change, plus grammaticalization,
choose the former. This is a frequently expressed objection:
[W]ell-known forces of linguistic change…would seem to be sufficient to bring on
the results often cited under the rubric of grammaticalization (Joseph 2001: 178).
It is not necessary for the kinds of changes most commonly encountered in
grammaticalization to be present in order for a change to qualify as an instance of
grammaticalization,… (Campbell 2001: 151).
‘Grammaticalization theory’ is seriously flawed and misleading, as well as,
arguably, totally superfluous, since existing mechanisms already suffice to account
for the phenomena… (Campbell & Janda 2001: 108).
[T]here is still no need for a special field called ‘grammaticalization (theory)’ in
order to account for the manifold facts taken in by that view… (Janda 2001: 266).
There is in addition a stronger criticism which, like the previous ones, asserts that
grammaticalization is causally inert. But, while the previous criticisms treat grammaticalization
as an innocuous term – superfluous, but basically harmless – the
present criticism emphasizes that it can be dangerous. “The reason for the danger”
Newmeyer explains “is that it invites one to conceive of such developments as
being subject to a distinct set of laws that are independent of the minds and behaviors
of individual language users” (Newmeyer 2001: 191–192).
The idea here is that using the term ‘grammaticalization’ may lead to positing
events of grammaticalization corresponding to this term, and then to attributing
causal powers to these events to explain them. According to this criticism, the
problem is not simply one of mislabeling what are genuine causal forces, but rather
of postulating causal forces which simply do not exist. Such a charge is evident
in Joseph’s allusion to “higher forces” in his account of the stages of development
of the Modern Greek future marker: “they are not guided by some ‘higher force’
driving them on since, ex hypothesi, there is no process of grammaticalization. …”
(Joseph 2001: 183).
4. In defense of the explanatory potential of grammaticalization.
As noted above, I agree with the above criticisms: grammaticalization is not a mechanism
that explains how a particular lexical morpheme becomes a grammatical one. I agree that without a detailed account of the mechanisms involved in the
transition from e.g. “going to” as a verb of motion, to its function as a future
marker, we have no explanation for how this particular lexical verb became this
particular grammatical marker. Simply citing the fact of grammaticalization – the
fact that a lexical morpheme became a grammatical one – does not explain how
this specific transition occurred.
Nevertheless, I reject the conclusion usually drawn from the above – that the
concept of grammaticalization is therefore empty and that it does no explanatory
work. This conclusion follows only if we assume that the only question we ever
aim to answer in this domain is how a particular grammatical morpheme came
to be from a particular lexical one in some particular language. Explanations of
individual cases will be in terms of underlying mechanisms such as reanalysis,
analogy, semantic bleaching, etc. However, there are additional questions it is
useful to address, such as, what makes an individual example of language change
specifically an example of grammaticalization? To answer this question, citing
the particular complex of mechanisms involved in a particular event of language
change, no matter how detailed, will not explain why the example is classified specifically
as one of grammaticalization. Consider, for example, the development
of “have” as an auxiliary verb in English perfect constructions (Trask & Millar
2007: 176–177). Example (1) illustrates the input stage to this development: Old
English “have” (hæbbe) functions as a main verb with the lexical meaning “possess”,
and the past participle (gefangenne) modifies and agrees with the preceding
object (fisc). In the output stage, illustrated in example (2), “have” functions as an
auxiliary verb having grammatical aspectual meaning (though, of course, the lexical
meaning is retained in its use as a main verb).
(1) Ic hæbbe ðone fisc gefangenne.
I have the fish caught
I have the fish caught.
(= I have the fish in a state of being caught.)
(Trask & Millar 2007: 176)
(2) I have caught the fish.
How did this change occur? Mechanisms cited include: reanalysis (the past participle,
which originally modifies the object, comes to be reanalyzed as being part
of the verb) and semantic bleaching of the lexical meaning “possess”. These mechanisms
explain how “have” came to function as an auxiliary verb with aspectual
grammatical meaning rather than lexical meaning. However, if the mechanisms
involved were different, the explanation for how “have” came to function as an auxiliary
verb would be different, but it would still be described as a case of grammaticalization.
However, when the same mechanisms operate on lexical input and yield lexical output, such cases are not described as cases of grammaticalization. As critics
are at pains to point out, the mechanisms of reanalysis and semantic bleaching
are not mechanisms of grammaticalization per se; they are general language-change
mechanisms. So, while there must be these general language-change mechanisms
at work in each case we identify as grammaticalization, it is not in virtue of these
mechanisms that the change is classified as a one of grammaticalization.
Put differently, we can say that the group of examples we identify as examples
of grammaticalization form a class picked out on the basis of a functional
property – the property specified in our concept of grammaticalization: being an
instance of language change in which a lexical form becomes a grammatical form
(or a less grammatical form becomes more grammatical).
What cases of grammaticalization have in common is not their respective
detailed accounts of which mechanisms led from lexical input to grammatical
output. As the above criticisms emphasize, there is no set or subset of mechanisms
that picks out all and only cases of grammaticalization. If we try to identify
the set of grammaticalization changes in terms of some underlying mechanism or
other, we will inevitably end up with the wrong members in the set: we will either
(1) omit instances that belong in the set (e.g. identifying grammaticalization by
appeal to the mechanism of phonological reduction would, in Joseph’s account
(2001: 166–178), exclude the development of weak subject pronouns in Modern
Greek since, in his view, it depends primarily on the mechanism of analogy) or, (2)
include members that should be excluded (because, for example, they are cases of
reanalysis that do not yield grammatical morphemes).
It is important to emphasize that although I have sometimes identified grammaticalization
as language changes involving a transition from lexical morpheme
to grammatical morpheme, this is a truncated formulation. For as noted earlier,
language changes from a grammatical form to a more grammatical one are also
recognized as being part of the definition or concept of grammaticalization. Joseph’s
discussion (2001: 178–183) of the development of the Modern Greek future marker
θa nicely highlights both these ways of understanding grammaticalization (notwithstanding
the fact that his purpose in detailing this history is quite otherwise).
According to the traditional account, the lexical main verb of volition thélo:
(‘want’), originally used with an infinitive complement and later with finite one
introduced by hina (‘that’), becomes a grammatical future marker eventually being
phonologically reduced and developing affixal properties. The input is a lexical
main verb; the output is a grammatical bound morpheme. Joseph’s account highlights
the fact that there were two separate mappings involved: one from lexical
to grammatical; and a second one from grammatical to more grammatical.
Initially there is a case of grammaticalization consisting of a change from lexical
to grammatical form: the change from thélo: as a main verb meaning ‘want’to thélo: as an auxiliary verb conveying future meaning. Later, a second case of
grammaticalization occurs which takes the previous output as input. This input
thélo: is already a grammatical form in that it functions as an auxiliary verb conveying
future meaning; but as an independent word (followed by hina), it is less
grammatical than the output of this change, the bound morpheme θa.
To explain why a language-change example counts as an example of grammaticalization
we show that is satisfies the definition or concept of grammaticalization.
To explain more generally why we classify the particular subset of language
change events we do – as cases of grammaticalization – we appeal to the fact that
each event maps a lexical morpheme to a grammatical one (or a less grammatical
morpheme to a more grammatical one). Members of the set are identified in virtue
of instantiating the definition or concept of grammaticalization, not in virtue of
the particular causal factors that accomplish this mapping. It is in this sense that
the concept of grammaticalization is not simply a cover term or placeholder for
explanation in terms of underlying mechanisms. They answer different questions:
the latter explains how the change took place, the former explains why the change
is classified as grammaticalization.
5. Unidirectionality, the process question, and reductionism.
In the present section, I consider how the perspective I have been presenting can
be used to respond to additional arguments leveled by critics of grammaticalization’s
explanatory potential.
Unidirectionality.
Criticism of the view that grammaticalization has explanatory power often goes
hand-in-hand with criticism of the thesis of unidirectionality. The charge is put
succinctly by Campbell:
If it [unidirectionality] is taken as a definitional property of grammaticalization,
it is then void of empirical content, since in this view, non-conforming examples
are not counterexamples, but just fall outside the universe of discourse of
grammaticalization altogether (2001: 140).
It is difficult to deny that unidirectionality is built into the definition of grammaticalization.
Although definitions may differ in scope or extent, they do not differ
in the direction of change. It is also difficult to deny that if direction of change is
built into the definition of grammaticalization, then language change examples
exhibiting the opposing direction of change are – by definition- not examples of grammaticalization. But a case of language change which fails to satisfy the definition
of grammaticalization is, as Campbell notes, not a counter-example to grammaticalization.
Given the agreed-upon definition of grammaticalization, it is a
logical error to think that it is. Instances of language change from the grammatical
to the lexical are counter-examples, not to the definition of grammaticalization,
but to the claim that all language change occurs in one direction, namely from the
lexical to the grammatical. (Language-change examples not involving direction of
change on the lexical-grammatical cline at all would also be counter-examples to
this claim.) Newmeyer is explicit on this point:
[G]rammaticalization is defined as a unidirectional process. Suppose that in
some particular case, directionality appeared to be reversed, that is, suppose we
observed a lexical item or construction that had developed a less grammatical
function. Such a circumstance would not bear in the slightest on whether
grammaticalization is unidirectional or not, because nothing could bear on that
question. The point is that definitions make no empirical claims (2001: 203).
The fact that unidirectionality is built into the definition of grammaticalization can
be no more problematic than the definition of grammaticalization itself – a definition
which as we have seen is not in dispute. It is problematic if unidirectionality
too is understood as a causal mechanism alongside e.g. reanalysis or phonological
reduction. But there is no need to understand it this way. In this regard I think
Joseph is mistaken in arguing against the unidirectionality of grammaticalization
by appealing to the fact the causal mechanisms involved in grammaticalization are
not themselves unidirectional. He says:
[I]f grammaticalization is the result of the workings of other mechanisms of
change, then unidirectionality, if a valid generalization to make concerning
grammaticalization changes, would have to fall out from the behavior and nature
of these other mechanisms (2001: 166).
But from the fact that some mechanisms are involved in language changes which
do not exhibit lexical > grammatical direction, it does not follow that these same
mechanisms cannot be involved in other language changes which do. Mechanisms
involved in language changes which are not unidirectional may still be at work in
a subset of language changes which are. Grammaticalization examples are a subset
of language change examples involving ordinary language change mechanisms,
but they are not thereby an arbitrary such subset.
Nevertheless, to the extent that the thesis of unidirectionality is understand
as a claim not about grammaticalization, but about all language change, then it
too is or, may well turn out to be, false. But as with the worry about grammatical
morphemes with no prior lexical history, defenders of the explanatory potential of
grammaticalization need not be committed to this view. If we reject the view that the explanatory potential of grammaticalization hinges on all language change
going in one direction, then exhibiting examples of language change going in the
other direction is not a point against the explanatory power of grammaticalization
per se. From that fact that something does not explain all cases, it simply does not
follow that it cannot explain any.
Similarly, I believe we should detach the claim of irreversibility from the thesis
of unidirectionality. There is nothing built into the definition of grammaticalization
that says that all language change is grammaticalization. So there is nothing in the
definition or claim of grammaticalization to rule out the possibility of “reversibility”.
Examples of such reversals would not count as examples of grammaticalization
but, as above, there is no need to for all language change to count as grammaticalization.
So the status of the irreversibility claim does not bear on the explanatory
power of grammaticalization. The two issues can and should be detached.
Process vs. Processes.
Joseph criticizes defenders of an independent status of grammaticalization for systematically
equivocating on whether to identify grammaticalization as a single,
unitary process or as multiples processes. Calling this issue the Process Question,
Joseph marshals a list of quotations sufficient to quell any doubt about existence of
such equivocation. For Joseph, the problem revealed by this equivocation is that
that both of these claims – it’s a single process; it’s multiple processes – are untenable.
If grammaticalization is identified with multiple processes, then they are
other processes (e.g. reanalysis etc.) and hence grammaticalization is not itself a
process. On the other hand, if grammaticalization is identified as a single, unitary
process, matters are worse, because in this case, one is positing grammaticalization
as a separate causal mechanism in addition to that of the underlying mechanisms.
In Joseph’s view, then, the Process Question (2001: 164–166) is not merely a terminological
one, but an issue that highlights the explanatory dead-end for defenders
of grammaticalization.
On the other hand, having above distinguished the causal mechanisms that
account for how grammaticalization occurs, from the identification of such occurrences
as mappings from lexical to grammatical morphemes, the Process Question
doesn’t reveal a paradox. Instead, it points to the fact that the explanation for how
grammaticalization occurs in individual cases, differs from the explanation of why
all such examples are examples specifically of grammaticalization.
Reductionism.
Much of what I have argued above implicitly assumes that opponents of the independent
status of grammaticalization are ignoring issues of reductionism. As we have seen, they acknowledge the multiplicity of mechanisms at work in different
examples of grammaticalization and the fact that all such mechanisms are
involved in non-grammaticalization language changes as well. Another way to
put these points is to say that there exist many-to-many mappings between the
mechanisms involved in explaining how a particular lexical input yields a particular
grammatical output, on the one hand; and the class of cases picked out under
the description “lexical to grammatical change” on the other: a single instance of
grammaticalization may be underwritten by multiple mechanisms, and a single
such mechanism may give rise to multiple kinds of language change. However,
for the most part, these critics do not acknowledge that these facts have the implications
for whether the concept of grammaticalization is reducible to any subset
of these underlying mechanisms. For Campbell, however, this is certainly not
the case. In one tantalizing passage, he addresses the worry that the view he is
defending is compromised by “an air of reductionism” (Campbell 2001: 151).
While an explanation of how, for example, a window broke, might be given in
terms of the micro-structural properties of glass, Campbell acknowledges that in
some contexts explanation in terms of human action may be appropriate. Citing
an example from Wright, Campbell notes:
To answer the question, ‘why did the window break?’ with ‘because John slammed
it’ is a completely adequate answer/explanation, even if shock waves and the
molecular characteristics of the glass may lie behind the breaking at some other
level (2001: 151).
Similarly, Campbell allows, if there is a level of explanation above that of the underlying
mechanisms in grammaticalization cases, that is “completely adequate” in
some context, then perhaps it is “just too reductionist” to insist that grammaticalization
is wholly derivative:
Thus, one might assert that… while reanalysis, phonological change, and
semantic change, and their interaction, explain the events at a deeper level.
… grammaticalization may itself be explanatory on another level (Campbell
2001: 151–152).
This is the consideration that comes closest to the position I am defending, but
Campbell then dismisses this possibility on the grounds that “almost unanimously
proponents of grammaticalization themselves appeal to reanalysis and other
changes as the explanatory mechanisms upon which cases of grammaticalization
depend” (Campbell 2001: 151). In other words, if there is explanatory potential at a
level of description above that of individual mechanisms, one would surely expect
proponents of this view to appeal to such a level. But they don’t. They invoke the
same set of lower-level causal mechanisms as their critics Against this line of reasoning, though, we must recognize that whether grammaticalization
itself has explanatory potential does not in the end depend on what
arguments supporters of the view recognize. Instead, I have been arguing that it
depends on whether diverse mechanisms underlie different instances of grammaticalization,
and what question we are trying to answer.
6. Concluding remarks.
I have tried to argue that the concept of grammaticalization is not empty because
it is only by appealing to it that we can explain why language-change examples are
examples specifically of grammaticalization. No individual or set of mechanisms
appealed to in order to explain how such language changes occur, explains what
makes these changes cases of grammaticalization. We cannot explain the latter
without appeal to the fact that the input to the change is a lexical form and the
output a grammatical one. Since this fact is captured in the definition or concept
of grammaticalization and not in the account of the mechanisms which effect this
change, the former plays an explanatory role not played by the latter.