EDITORS: Detges, Ulrich; Waltereit, Richard
TITLE: The Paradox of Grammatical Change
SUBTITLE: Perspectives from Romance
SERIES: Current Issues in Linguistic Theory 293
PUBLISHER: John Benjamins
YEAR: 2008
Paul Isambert, Université de Paris 3, Sorbonne-Nouvelle, Paris, France.
SUMMARY
This volume is a collection of papers presented at a workshop of the Deutscher
Romanistentag in September 2005 at the University of Saarbrücken. As its title
clearly indicates, its aim is to tackle the issue of grammaticalization viewed
as a paradox since, while languages obviously change over time, they have no
motivation to do so: they are perfect in the sense that they are fitted for
communication, or at least there is no language more fitted than another one,
and thus there is no reason for a given language to change since the outcome of
this process will fare no better than before. The second goal of the book is to
bring together functional and formal (actually generative) approaches and to
compare their respective solutions to the problem (with Romance as the field of
inquiry).
In their introduction, the editors summarize the four types of solutions
advanced in the generative tradition: imperfect acquisition, language contact,
structural simplification (avoidance of markedness), and Longobardi's (2001)
Inertia Theory, according to which core grammar (i.e. syntax) does not change by
itself but only as a reflex to peripheral (morphological, phonological, lexical)
modifications. The functionalist view, on the other hand, mainly focuses on
usage and frequency effects, emphasizing irregularities over rules. It may be
tempting to reconcile those two approaches by saying that functionalists
describe the external pressure that leads to the fine-grained modifications
modeled by formalists, but this would miss the point: both conceptions are
incompatible and do not simply claim to study different areas of the same field;
on the contrary, they make opposite claims on the same object.
''Syntactic change from within and from without syntax: a usage-based analysis'',
by Richard Waltereit and Ulrich Detges, is a direct charge against the
aforementioned Inertia Theory (Longobardi, 2001). Longobardi's position is that,
first, syntax is notoriously resistant to change, and second, grammar being
perfect (in a Chomskyan sense), there can be no internal mechanism that leads
to change. Only morphology, phonology and the lexicon may evolve for whatever
reason, and grammar changes only to keep in line with those external
components. Waltereit and Detges take languages to be inert as well, but because
they are social conventions and thus can change only in response to what may be
called ''social events''; the change, on the other hand, may be truly syntactic.
They rely on two principles: the principle of reference (''Match the sound string
you hear with what seems to be its function in the situation'') and the principle
of transparency (''Match the sound string you hear with other sound strings in
the same language''). With the first principle, they study the French strong
interrogative particle ''est-ce que'' and show that what was once a compositional
meaning (related to clefting) is now a conventional one attached to a
syntactically simple construction. In essence, this analysis does not contradict
Inertia Theory (except that the locus of change is not language acquisition, but
language use), since the syntactic reanalysis was triggered by
pragmatic/semantic factors. But the authors also examine presentational
constructions in Spanish (''Hay + NP'', where ''hay'' is singular, even if the NP
may be plural) which have to meet conflicting constraints: on the one hand, the
''presented'' NP is focal, and focal information is generally coded as
non-subject; on the other hand, single arguments are usually subjects. When the
NP is singular, thanks to the principle of transparency, the second constraint
may be implicitly respected. And when it comes to plural NPs, it is also
respected with low-frequency constructions, i.e. the same presentational
constructions but in the past tense; in that case, there is agreement (in some
varieties of Spanish) between the verb and what is now its subject (as in
''Habian soldados en el patio'', 'there were soldiers in the courtyard'). In that
case, the change originates in syntax itself, albeit guided by a more general
principle, and Inertia Theory appears to be wrong: crucially, usage and relative
frequency treats language as a whole, not as a modular object with an
inalterable syntactic core and fleeting peripherals.
Andreas Dufter's ''On explaining the rise of 'c'est'-clefts in French'' debunks
some common wisdom about the increase of ''c'est''-cleft sentences in the history
of French. As a focusing construction, this type of sentences has been widely
considered as a repair strategies against the loss of focus accent and flexible
word order. Relying on corpus data, however, Dufter shows that the ''c'est''-cleft
type existed well before those alternative focusing devices began to recede and
continued to increase in frequency well after they had vanished. Thus the
traditional explanation is not the whole story. Dufter notes that the emergence
of ''real'' clefts goes along with the spread of what he calls ''clefting beyond
necessity'', i.e. the use of cleft structures that do not focus the clefted
constituent; but why those structures stepped on stage in the first place
remains largely unexplained.
In ''The role of the plural system in Romance'', Elisabeth Stark investigates the
structure of determiners in Spanish, French and Italian, and their relation to
Latin. The former three show differences with regard to the possibility of bare
nouns and the existence of the partitive article. They also differentiate in the
marking of gender and number on nouns. Taken together, the author argues, those
differences explain themselves with the aid of a syntactic slot called Pl(*)
(such that a noun phrase, which is actually a NumP, has the structure [Num
[Pl(*) N]], where Pl(*) is the head of the internal constituent) with a boolean
feature [LATT] (for ''lattice''), whose value yields the semantic interpretation
of number. To put it simply, the rich nominal morphology of Spanish indicates
that N bears the features Gender and Number, which may be probed (in the
author's terminology) and matched by Num and Pl*; since N is the target of those
operations, it can be subsequently subject to movement, hence bare nominals.
French, on the other hand, has virtually no morphological marking of gender and
number on nouns, and thus N is featureless and cannot move (since it isn't the
target of any operation); consequently, all noun phrases have a determiner.
Italian falls in between, unambiguously marking only gender on noun, and has an
optional determiner for instance in the case of partitives. It is argued that
Latin had all three features Gender, Number and [LATT] on N, and that the loss
of declensional endings and the subsequent reordering of those features led to
the relative need of a determiner, as observed in Spanish, French and Italian.
In ''Morphological developments affecting syntactic change'', Maria Goldbach
studies the evolution of the so-called Accusativus-cum-Infinitivo (ACI)
constructions from Latin to Middle French. The ACI is a complement sentence with
an accusative subject and an infinitive verb (and no complementizer). Latin has
a rich verbal paradigm in the infinitive which indicates a strong
Infl(ection)-category; this strength in turn allows ACI constructions. Old
French, on the other hand, shows a weakened infinitive paradigm in that it
expresses tense and voice to a lesser extent, cannot take a subject nor object
clitics, and cannot be negated. The author thus argues that in Old French
infinitives are not propositional; they have no Infl, and consequently they do
not permit ACI. But in Middle French, object clitics began to reappear with the
infinitive, thanks to their homonymy with unbounded pronouns (which freely
appear in front of the infinitive verb). Since clitics, according to Goldbach,
are inflectional elements, they led to a strengthening of Infl, which paved the
way for the renewed use of ACI.
Susann Fischer's ''Grammaticalisation within the IP-domain'' studies the change in
word order from Old to Modern Romance. In Old Romance, stylistic fronting (the
movement of a head in front of the finite verb), postverbal clitics and negation
mutually excluded each other. To account for this fact, Fischer assumes a new
functional category EP (where E here is meant to iconically stand for an
upper-case sigma, unavailable for technical reasons) that somehow realizes
speech acts: negation (when the negation is inserted), emphatic affirmation (in
the case of stylistic fronting or when the finite verb is moved, leading to
postverbal clitics) and neutral affirmation (when the slot remains empty). To
explain the loss of stylistic fronting and postverbal clitics in Modern Romance,
Fisher takes the view that in a context of pervasive variation, learners will
chose the less marked structure when presented with semantically equivalent
representations of a sentence (movement is more marked than lexical insertion,
which in turn is more marked than empty realization). For some reason, from the
15th century onward, stylistic fronting and verb movement did not denote
emphasis any more, and thus were semantically equivalent to sentences with an
empty EP, but more costly. Consequently, they were dismissed.
As the title indicates, Giampaolo Salvi's ''Imperfect systems and diachronic
change'' does not take languages to be perfectly running systems; the principles
that govern them may lack generality. Consequently, languages have the
possibility to change; more precisely, imperfection causes languages to change,
but change in turn also leads to imperfection. Salvi studies the evolution of
the ''si'' construction, which in Modern Italian may be used to build passive and
impersonal sentences. In the former case, the subject is demoted while the
direct object takes its place; in the latter, this demotion leads to a
subjectless sentence. On the other hand, Old Italian only had passive ''si''.
Salvi shows that a series of small changes beginning in the 17th century leads
to the current situation. But his point is that those changes are not related to
one another. Each step is independent and the overall movement that one seems to
perceive is only a deceptive appearance due to retrospective interpretation.
That there is no long-term change is evidenced by the fact that the system
remains imperfect: both constructions shows restrictions on person (e.g. the
passive construction works only when the direct object is 3rd person).
In ''From temporal to modal: divergent fates of the Latin synthetic pluperfect in
Spanish and Portuguese'', Martin G. Becker investigates the evolution of the
morpheme ''-ara'', a marker of past subjunctive in Spanish and simple pluperfect
in Portuguese, out of the Latin pluperfect morpheme ''-(a)verat''. Becker follows
Reichenbach and defines pluperfect as involving an utterance time (t0), a
reference time in the past (t1) and an event time (t2) prior to t1. He argues
that the event described is true at t2, while at t1 it becomes unspecified and
thus possibly false. The latter case is exploited in the apodosis of
conditionals (where ''-ara'' first extended beyond the limits of simple
pluperfect), if one takes the protasis as the turning point t1 of the validity
of the event. This was the first step toward an irrealis reading, and ''-ara''
extended to the protasis (along with the fade of the competing form ''-ase'') and
then to conditional contexts without explicit conditional structures, especially
to modal verbs like ''poder'', where both the temporal and the modal reading were
allowed. Subsequently, it began to appear in non-veridical contexts (like
temporal clauses with ''until'' or in the scope of a negation); in Portuguese,
where the morpheme had followed a similar path, this latter step was not made,
and the irrealis reading of ''-ara'' vanished, while its temporal reading remained
(albeit in formal registers only). Finally, it lost its temporal value in
Spanish, and was available for clauses subordinated to strong intensional verbs,
with a totally virtual reading, i.e. it turned into a (past) subjunctive. The
author concludes that this evolution shows prototypicality effects, where
marginal values became central over a long period of polysemy.
In ''Non-lexical core arguments in Basque, German and Romance: how (and why)
Spanish syntax is shifting towards clausal head-marking and morphological
cross-reference'', Hans-Ingo Radatz reassesses the issue of presumed empty
subjects and objects. Basque has been said to be a null subject and null object
language, because arguments are always optional in the lexical or pronominal
form, while verbal affixes were seen as mere agreement markers. But Radatz
argues that those affixes are actually arguments; there is no agreement, but
instead cross-reference between those affixes and lexical and pronominal
referents. Moreover, the latter are simply appositions and their appearance
depends on discourse motivations; thus, there is no null argument of any kind.
The confusion stems from the application of a typologically inadequate
terminology, namely describing head-marking languages as dependent-marking ones
(a ''Eurocentric'' perspective). In the former, grammatical relations are marked
on the head, which for the clause is the verb. Empty subjects and objects appear
to be so only if we mistakenly take the language under scrutiny to be of the
dependent-marking type. Radatz next shows that Spanish is on its way from
dependent-marking to head-marking, with for the moment a clear situation of
clitic-doubling. This challenges the traditional view that pronouns
grammaticalize into agreement affixes: they may as well evolve into
cross-reference markers. (I take the opportunity to note that, contrary to what
the author believes, the same tendency in French is not exceptional, and that
the Basque-like example (31) is perfectly grammatical; this remark is a mere
typological clarification and by no means contradicts the author's thesis.)
The last chapter of the book, Esme Winter-Froemel's ''Towards a comprehensive
view of language change: three recent evolutionary approaches'', examines the
influence of the theory of biological evolution in linguistics. Language change
may be metaphorically related to biology (but the factors of change remain
genuinely cultural); it may also be considered as an instance of biological
evolution (it thus obeys Darwinism); or both linguistic and biological
evolutions can be viewed as two instances of a more general evolution process
(they share some mechanisms but have specifics too). The author adds an
orthogonal distinction between adaptive views and two-level models. In the
former, language change is analyzed as an overall, long-term process, as a
response to the needs of language users. The author shows that this approach
runs into troubles, such as the role of the hearer, the question of obviously
non-adaptive changes, and more generally the unreachability of perfection and
even optimization. The two-level views, on the other hand, focus on small-scale
processes, and make a division of labor between an individual innovation and its
social propagation. This approach seems to avoid the problems of the previous
one, but Winter-Froemel argues that both miss the distinction between universal
and language-specific factors in linguistic change. She thus proposes a
tripartite model, carrying over the two-level view, which concerns the processes
of change, and adding a methodological side and a part concerning the factors of
change; the three perspectives are organized from the universal to the
extralinguistic levels and interact so as to avoid the aforementioned
shortcomings. Finally, the author notes that any reference to evolutionary
biology may be unnecessary.
EVALUATION
Since generativism and functionalism are not systematically compared in this
volume (as acknowledged by the editors in their preface), I will try to do so
and finally propose an alternate view, which is not a personal theory but an
appeal to other linguistic frameworks that I would have appreciated to see in
this book. As a beginning, if one was to resume bluntly the debate between the
two streams illustrated here (although some papers, e.g. the excellent
methodological caveat by Radatz and Salvi's careful descriptive study, do not
seem to advocate any particular theory, despite an allusion to Universal Grammar
in the latter case), one could make up a little dialogue where a generativist
would blame a functionalist for not being scientific, while the latter would
answer that generativism has nothing to do with linguistics. Let's review those
two accusations in turn.
The problem with the algorithmic view of language held by generativists is that
algorithms are not supposed to change over time. (An additional problem is that
one may doubt that generativism still preserves any mathematical rigor; at least
it has been proven computationally inefficient, see Johnson & Lappin, 1997.)
That's why the common wisdom is that such change is due to imperfect
transmission. While it is already a highly problematic assertion (e.g. because
of the conservatism of young learners demonstrated among others by Tomasello,
2003; unfortunately, proponents of the imperfect transmission scheme, to the
best of my limited knowledge, do not discuss that fact, as illustrated by
Lightfoot, 2006, who does not even mention Tomasello's work), even if it were
true, there would remain the problem of the existence of the variation necessary
for that change to occur. Pervasive bilingualism may be a solution to that
problem, but then how do speakers become bilingual? We wind up begging the
question: people are bilingual because there is variation, and vice versa.
Another trouble with the generative approach is that two out of the three
clearly minimalist contributions to the volume have to bring additional
syntactic slots into the picture (namely the papers by Stark and Fischer). It is
normal to modify a theory when facing data (although one may wish to know when
facts modify a theory and when they falsify it, the latter case being the
traditional Popperian view repeatedly put forward by generativists), but when
the theory becomes overpopulated with such modifications (and provided they do
not contradict each other), as is clearly the case here, one may fear that the
theory actually overgenerates or is simply trivial. This might not be obvious at
first sight, thanks to empty categories and various movements, but in the end
that's a definition of triviality: when a given concept is in use only when the
facts call for it.
Moreover, generative explanations, in the end, are always in need of a
functional add-on: they focus on some part of a historical development and do
not account for what happened before. For instance, Maria Goldbach honestly
acknowledges that she has ''no answer to the question of why the paradigm
alterations [of Latin infinitive] occur''; ''an answer to this question'', she goes
on, ''would certainly contribute to a more far-reaching explanation'' (p.86).
Indeed, it may be the case that Latin infinitival forms receded because the
syntactic construction where they were mostly used, namely the ACI, began to
vanish. And this would be just the opposite of the conclusions she reaches
(although it would be in direct contradiction to Longobardi's Inertia Theory; as
far as I'm concerned, this does not make me feel uncomfortable).
On the other hand, the strength of generative grammar is to make use of
presumably well-defined and self-consistent notions, while functionalists are
blamed for their inaccurate terminology which may in the end accommodate any
fact. It is true that one may wonder what constraints are put on Waltereit and
Detges' principles or Dufter's ''clefting beyond necessity''. Indeed, we were not
there to witness the speakers' behaviors, and we cannot ascertain that they did
obey those psychological trends, nor can we (or rather I) think of any test that
would falsify them. Moreover, functionalists may be right in their proposals of
external factors, but they do not offer a detailed account of the linguistic
machinery at work (so the generativist goes). For instance, social causes may
hold, but that does not tell us how the linguistic structure itself changes,
just like saying that birds fly because they have wings does not tell us
anything about flight itself.
Moreover, even such a concept as frequency seems ill-defined: a form is accepted
because it has become frequent, but how can it become frequent if it has not
been accepted beforehand? Here Winter-Froemel's paper may help us out of this
dilemma. She distinguishes between the propagation of a form in a linguistic
community and its adoption by an individual speaker. Now, with the aid of any
incarnation of construction grammar (see, among many others, Tomasello, 2003 and
Goldberg, 2006 or, for a formally stringent approach, Sag 2007), which holds
that language is based on memory and generalization (and not algorithm or social
conventions alone), we can say that a language user is able to understand and
even produce new (possibly awkward) constructions, not because there exists a
derivation from common structures to those constructions, but because they fit
nicely into the picture of its linguistic knowledge. Note that one never
encounters, say, a grammatical structure, but only an instance of it, which may
realize it more or less perfectly (this leads to the prototypicality alluded to
in Becker's paper and more thoroughly investigated in Gries, 2003); and this new
construction (a generally very small departure from usual structures) may be
thought of as yet another realization of the targeted ideal (and influence it in
turn). Thus, the change in linguistic structures that generativists try to
investigate comes for free here.
A final desideratum of linguistic theory in general that may be applied to
historical studies is the use of experimental techniques. By no means is history
testable, of course, let alone predictable. But the psychological mechanisms
that are assumed in language change should be amenable to experimental
investigation, as is done for instance by John Ohala in phonetics (note that
this is the minimal requirement of any scientific theory). When it comes to
storytelling, we linguists always show ourselves to be very smart, whatever our
persuasion; thus, if we want reality to kick back, we have to allow it to do so.
Despite the merits of each contribution, I think the book lacks this precise
form of investigation to be fully able to tackle the apparent paradox of
language change (but I understand that its being made out of a conference puts
some limit on its comprehensiveness, for which the editors surely cannot be blamed).
REFERENCES
Goldberg, A. (2006), _Constructions at work. The nature of generalization in
language_, Oxford University Press, Oxford.
Gries, S. (2003), ''Toward a corpus-based identification of prototypical
instances of constructions'', _Annual Review of Cognitive Linguistics_ 1, 1-27.
Johnson, D. & Lappin, S. (1997), ''A Critique of the Minimalist Program'',
_Linguistics and Philosophy_ 20, 273-333.
Lightfoot, D. (2006), _How new languages emerge_, Cambridge University Press,
Cambridge.
Longobardi, Giuseppe (2001), ''Formal syntax, diachronic minimalism, and
etymology: the history of French 'chez''', _Linguistic Inquiry_ 32(2), 275-302.
Sag, I. A. (2007),''Sign-Based Construction Grammar: an informal synopsis'',
Technical report, Stanford University.
Tomasello, M. (2003), _Constructing a language: A usage-based theory of language
acquisition_, Harvard University Press, Cambridge, Mass.
ABOUT THE REVIEWER
Paul Isambert is a PhD student at the University of Paris 3, France. He's
currently working on grammaticalization and discourse structure, especially
concerning topic shifts and anaphora.
|
