theory is a long-standing problem in linguistics, dating from at least
the Neogrammarians. As we have noted, Saussure famously denied the relevance
of diachrony to synchronic linguistic theory, and his position has
been widely emulated for a century. Nevertheless, linguists of all theoretical
camps have been unhappy with this drastic division of the field, and many
have attempted (even if uneasily) to model change within whatever theoretical
framework they favoured. Thus in the structuralist framework, sound
changes were characterised as involving phonemic mergers, allophonic
splits, alterations in phonetic values, and the like (cf. Hoenigswald 1960).
In the generative period, changes were described in terms of rule additions,
losses, reorderings, and so on (cf. King 1969). Recent theoretical developments
such as optimality theory (e.g. Anttila 1997) and exemplar theory (e.g.
Bybee 2001) have often sought to explicitly incorporate accounts of linguistic
change within their models.
Optimality theory has proven to be an exceptionally flexible framework for
modelling change. The theory postulates a universal inventory of constraints,
each stating some desirable phonological state of affairs; where languages differ
is merely in the hierarchical rankings of these constraints (plus, of course,
differing lexical inventories). Since any change in ranking defines a different
grammar, and a different potential ‘language’, both variation and change can
be subsumed into the OT account of language typology. Variable realisation
of a final consonant, such as final€–s and€–r deletion in Caribbean Spanish and
Brazilian Portuguese, and final€–t deletion in English and Dutch, are modelled
as variable rankings of constraints that militate against syllabic codas and those
that favour faithful surface realisations of underlying segments. When faithfulness
constraints are more highly ranked, the segment surfaces, but when
the ‘no coda’ or ‘simple coda’ constraints prevail, surface realisations without
the final segments are preferred. This is the typological difference between
languages with open syllables (e.g. Yoruba), and those with closed syllables
(English, Spanish, etc.), so the same mechanism can be pressed into service
to account for variation and change. Variation is modelled by postulating variable
ordering between the relevant constraints, and change across time is modelled
by postulating a diachronic reordering of the relevant constraints .
Exemplar theory is a recent development that places variation and change
at the core of the model, relying on the naturally occurring variation in the
input as the driving force (cf. Bybee 2001; Pierrehumbert 2002). This theory
eschews abstract representations, postulating instead that speakers remember,
in rich phonetic detail, the tokens of words that they hear pronounced, or produce
themselves. Therefore, speakers have memories of the full range of variants
they have encountered, and use these memories (the ‘exemplar cloud’) as
targets for their own production, which then necessarily varies as well. The
theory emphasises natural phonetic processes such as lenition and assimilation
as the driving force in phonological change; words that are often repeated are
more subject to these processes, altering the exemplar clouds of speakers in the
direction of the change produced by the process. This model emphasises the
importance of lexical identity and lexical frequency in variation and change,
predicting that lexical items may differ (i.e. lexical diffusion), and that frequent
words should lead sound change. Lexical diffusion has long been advocated in
historical linguistics as an alternative to the exceptionless sound change model
of the Neogrammarians , based on a number of empirical cases where lexical irregularities are found in historical changes.
Exemplar theory provides a formal model to account for such facts.
Exemplar theory provides a formal model to account for such facts.
The most widely used theoretical framework in studies of variation and
change, growing out of Weinreich, Labov and Herzog 1968, is the ‘variable
rule’ (VR) model, a broadly generativist model in which optional elements in a
grammar are probabilistically quantified (see Cedergren and Sankoff 1974 and
Sankoff 1978 for further discussion). This is the dominant model in variation
studies, and its extension to modelling change is straightforward, but has subtle
and substantive implications.
The VR model postulates that any variable process may be subject to two
conceptually different quantitative forces. First are contextual conditions:€most
variables have a lumpy distribution across the language, occurring often in
some context and rarely in others:€tensed and raised variants of /æ/ in English
dialects, for example, are more common in pre-nasal contexts, and rarer or
less advanced before voiceless stops. Deletion of final€–t, and€–d in English is
more frequent before a following word beginning with a consonant, and rare
before a following vowel. But second, there are also overall differences in the
rate of use of any given variant, in different speakers, social-class groupings,
speech styles, age cohorts, and so on. A particular dialect may have tensed
/æ/ more frequently or more advanced phonetically than another dialect, even
though both favour tensing in the pre-nasal context. A working-class speaker
may delete /–t,–d/ more than a middle-class speaker, even while both delete
more before consonants than before vowels.
This distinction between overall rate of use and contextual effects is captured
in the VR model by two kinds of factors. Each process is associated with
an ‘input’ probability, or p0, which captures overall rate of use. In addition, a
process may be associated with multiple contextual constraints, capturing the
quantitative effects of favouring and disfavouring environments that promote
or retard the selection of a particular variant. These are factor weights or partial
probabilities associated with contexts, pi, pj, pk, etc.
Given this model of variation, what changes across time is typically the overall
rate of use of the innovative variant. Just as speakers and social groups differ
in overall use while preserving the same constraint effects, and vary their overall
rate in different speech styles while leaving contextual effects unaltered,
successive age cohorts across the course of a change will increment the overall
rate of use, leaving context effects unchanged. Change is change in the value of
p0, while the constraints on variable selection (pi, pj, pk …) do not change.
The constancy of contextual effects across time has been demonstrated in
a number of empirical studies, beginning with the work by Kroch .
Kroch formulates this observation as his ‘constant rate
Kroch formulates this observation as his ‘constant rate
hypothesis’€ – the claim that the rate of change in all contexts is the same.
Kroch shows that the rate rises in English periphrastic do in all the contexts
investigated in the figure are mathematically equivalent; that is, the logistic
transform of each of the curves is a straight line with an essentially identical
slope. Therefore, the most plausible interpretation is not that each context
represents a separate change proceeding at an independent pace, but rather that
there is only one change, following a single time course, governed in variable
rule terms by a single p0.
Syntactically, this single change can be described as a loss of V-to-I (verb
to INFL) raising; in old and early Middle English, in sentences without auxiliaries,
a main verb could move up to the high position (a.k.a. INFL) in the
clause an auxiliary would occupy, thus preceding a negative, for example (e.g.
They know not what they do, with main verb know preceding not, parallel to
They must not know, where auxiliary must precedes not). In Modern English,
however, a main verb cannot occupy that position; instead do is inserted as a
dummy auxiliary just when the main verb becomes separated from that position
by some other material, such as a negative (You INFL not eat fish → You
do not eat fish), or an inverted subject (INFL you eat fish? → Do you eat fish?).
The several contexts of the change show differences in their intrinsic favourability
to the innovative variant that are stable across time. Each successive age
cohort across the 400 years of the change was less and less likely to permit
V-to-I raising, triggering the alternative solution of do-periphrasis at progressively
higher rates.
The constant rate hypothesis follows directly from the VR model, distinguishing
overall rates of use from contextual effects. Indeed, the constancy
of the rate of change across different contexts constitutes important evidence
in favour of VR for both variation and change. Alternative models of change
such as the OT treatment involving constraint re-ranking, lack any overall parameter
comparable to p0. This implies that change in an OT model should
not be a smooth S-curve, but rather, a step function with inconstant contextual
rates:€each time a pair of constraints is re-ranked, the contexts they affect
should show abrupt changes in the rate of occurrence of the variant realisations,
while unaffected contexts would show no change. This is at odds with
the empirical evidence.
8.6 Conclusion
Work on language variation has, since its earliest inception, addressed questions
of language change. Nearly fifty years of research on these problems
has turned up a substantial body of knowledge demonstrating that variation
and change are in essence a single phenomenon viewed from different perspectives.
This discovery requires linguists to develop new methodologies and
theoretical approaches that make possible an integrated understanding of what
the orthodoxy of twentieth-century linguistics treated as belonging to opposed
and unrelated synchrony and diachrony. This is part of a broader integrative
trend in twenty-first-century linguistics, bringing the insights of many disciplines
together to tackle big issues that they were unable to resolve separately.
At the centre of this integration are questions about stability and dynamism in
language:€why do languages change, and why do they (sometimes) remain the
same? These are the questions that research on variation and change is helping
to answer.
Analysing Variation in English- Edited by Warren Maguire and April McMahon
Cambridge University Press, New York
Chapter 8 Variation and change- Gregory R. Guy
8.5- Change and linguistic theory- Pages 195,196,197.
8.6- Conclusion- Page 198.
Nenhum comentário:
Postar um comentário