Professional Documents
Culture Documents
North-Holland 189
SHIFI'ING VARIATION
Erica C. G A R d A *
University of Leiden, Tile Netherlands
The relation of quantitative data on language use to the grammar of the language
remains an important problem in linguistic theory; it has also been the subject of
ambitious claims within sociolinguistics. The paper discusses the evolution of socio-
linguistic analysis to encompass 'syntactic variation', the consequences of confining
analysis to those contexts exhibiting 'variation', and the theoretical implications of
such an approach.
1. Introduction
'(...) the notion "grammatical in English" cannot be identified in any way with the notion
"high order of statistical approximation to English" (...) In the context "1 saw a fra-
gile " the words "whale" and " o f " may have equal (i.e., zero) frequency in the
past linguistic experience of a speaker who will immediately recognize that one of these
substitutions, but not the other, gives a grammatical sentence.'
'(...) variable rules (...) extend the generative notion of "rule of grammar" to account for
a larger range of data (,..) [variable rules] are more than statements that a rule may or
inay not apply; they are lawful co-occurrence relations.'
The new formal device was clearly intended to embody a theoretical claim
(Labor (1969: 759)) :
'It should be (,,.) clear that we are in no way dealing with statistical statements or approxi-
mations to some ideal or true grammar. We are dealing with a set of quantitative RELA'nONS
which are the form of grammar itself.'
define the form of grammar. Now such a claim, if true, would be of the
greatest interest: by making available and relevant to linguistic analysis a
vast reservoir of empirical data hitherto largely untapped, it would span
the opposition perceived by many scholars (cf. Itkonen (1983 : 1 et passim))
between 'autonomous' and 'empirical' linguistics.
Since generativists obviously cannot be asked to demonstrate a negative,
the burden of proof must lie on him who makes the stronger claim. A decade
and a half having elapsed since variable rules were first introduced, we can
justifiably pause to take stock and inquire what, if anything, the analysis
of variation has contributed to the understanding of syntax.
'Tile opening chapters of this study presented the problem of accounting for large scale
variation in the speech of New Yorkers. Inconsistencies and oscillation ranged over a
considerable part of the phonological system, to such an extent that it was difficult to
construct a coherent system for the speech of most individuals (...) [specific analysts]
viewed such variation as deviation from the structure of speech, which had to be eliminated
or disregarded for a systematic presentation of linguistic patterns.
In the present work, this attitude is reversed. We conceive of the variation itself as an
integral part of the structure of New York City speech.'
That is, free variation could be reduced very considerably if the (non random)
distribution of etic variants was described in terms of extra-linguistic para-
meters.
The next step was, quite naturally, allomorph(ophonem)ic variation. In
Labov's (1969) study of the contraction and deletion of the English copula,
variable rules were proposed as such for the first time. The aim was clearly
to do for optionality in (generative) phonological rules what Labov's earlier
192 E.C. Garcia / Shifthtg variation
w o r k h a d d o n e for a l l o p h o n i c free v a r i a t i o n in s t r u c t u r a l i s t p h o n e m i c s ,
of. L a b o v (1969 : 737) :
'(...) if we interpret this notation as meaning no more than the conventional label "optional",
it will hardly allow the facts of systematic variation presented above to be accommodated
in the grammar of NNE. The label "optional" is no more useful in this respect than the
label "free variation" (...) If the data of the preceding sections are to be utilized in
formal rules, it must be shown that the study of variation adds to our knowledge of
linguistic structure, and simplifies the situation rather than reducing the precision of the
rules by uncontrolled and unaccountable notations. To achieve this end, we associate with
each VARIABLERULE a specific quantity tO which denotes the proportion of cases in which
the rule applies as part of the rule structure itself.'
'The data of §4 showed that variation in contraction and deletion is governed by a set
of constraints such as the effect of a preceding pronoun or a following verb. These
variable constraints are features of the environment (...) Thus we may indicate that con-
traction is favored by a following verb and a preceding pronoun by [there follows a rule
where the contexts are qualified by Greek letters indicating the weight of the context in
affecting the implementation of the rule, EG].'
'The variable notation introduced here performs the same function as the other devices
which we use for condensing rules into a single schema: it captures certain generalizations
about the particular language being examined, and tells us something about the structure
of language in general.'
A n d as in the case o f a l l o p h o n i c v a r i a t i o n it w a s a r g u e d in j u s t i f i c a t i o n o f
v a r i a b l e rules t h a t ( L a b o v (1969: 7 5 9 ) ) :
'(...) the construction of complete grammars for "idiolects", even one's own, is a fruitless
and unrewarding task; we now know enough about language in its social context to
realize that the grammar of the speech community is more regular and systematic than
the behavior of any one individual Unless the individual speech pattern is studied within
the over-all system of the community, it will appear as a mosaic of unaccountable and
sporadic variation.'
T h e i n c o r p o r a t i o n o f e x t r a - l i n g u i s t i c p a r a m e t e r s i n t o v a r i a b l e rules t o o k
place next, in a n i m p o r t a n t theoretical p a p e r ( L a b o v (1970)), e n t i t l e d ' T h e
E.C. Garcia / Shifthzg variation 193
'As we explore the relevant details of rules used by the speecb community, it becomes
apparent that many rules will require a matrix of two or more dimensions to describe
their use. One dimension will be the various syntagmatic environments; the other will be
the range of dialects or idiolects which differ in their assignment of categorical, variable,
or prohibited status to a given environment.'
Though the formal implications of this statement are not worked out in the
paper, dialectal conditions are repeatedly (1972c : 788, 789, 790, 814) imposed
on a linguistic constraint.
Now it is a question that deserves the most careful consideration whether
this identification of disparate types of phenomena by means of one and
the , same formal device (the variable rule) is in fact theoretically sound:
does it indeed constitute a significant linguistic generalization? It is striking,
to say the least, that the theoretical implications of the identification should
have been left totally undiscussed in Labov (1970) and (1972c). These
implications only grew in importance, however, as the variable rule became
THE analytic tool with which any kind of (socio-)linguistic variation could
be described, regardless of the nature of the phenomenon in question.
"From the fact (which we have seen conceded though little exploited) that markedness is
not a property which we can assign absolutely to phonetic features, nor even to particular
simultaneous complexes of features (i.e.. segments), it is clearly possible to conclude (...)
that markedness is shown in this way to be merely an observational notion, and that our
attention should in fact be directed towards establishing the constants which conspire to
regulate the occurrence of particular phonetic features in particular environments."
t There has certainly been discussion of the correlations observed among the EXTRA-linguistic
parameters: Trudgill notes that middle-class speech carries (covert) prestige among women
but lower-class speech among men, and suggests (1972: 187-194) a plausible reason why
this should be so. Similarly, van den Broeck (1977) notes a correlation between social class
and formality of speech situation with respect to syntactic complexity, just as Labov (1970:
70-71) early demonstrated a correlation between social class and (phonetic) style shift. What
is at issue in all these cases, however, is a correlation among the extra-linguistic parameters,
which would continue to hold regardless of the particular linguistic phenomena involved
(Itkonen 0983 : 195, 273-274)).
E. C. Garcia / Shifting variation 195
"The phonologist must surely work on the assumption that an explanation for the patterning
which he finds will be discovered in the articulatory and perceptual properties of the
sounds concerned."
And indeed, it was early pointed out that at least some of the frequency
data described by means of variable rules might be no more than the
natural result of a linguistic item's occurring in a particular linguistic environ-
ment. Thus, for instance, Kiparsky (1971: 603):
"An alternative, which is suggested by the cross-linguistic validity of the factors that control
the frequency of application, is the following : the frequencies are not learned by the child,
but predictable from (...) substantive constraints."
This was conceded by the sociolinguists, for instance with respect to the
variability described by the '-t, d deletion rule' (Labov (1972b: 81), Wolfram
(1973 : 9)). Nonetheless, the need for variable rules was insisted on by varia-
tionists (Berdan (1975: 18), Hudson (1980: 17), Fasold (1978: 86--89)) on
the grounds that what remains unprediciable is the actual effectiveness of
particular linguistic parameters among different social groups. This is also
Labov's position (1972b: 81) :
'We must use the -t, d rule in our BE [Black English, EG] vernacular grammar, because
(...) the weighting of the variable constraints is not general - in fact, it varies from group
to group, and from style to style, and from one age level to another in a meaningful
manner."
In short: the rule as such - the fact that certain environments favour one
alternative more than another - may well reflect universal principles: but
what remains arbitrary, and is for that reason claimed to require formulation
in the particular grammar, is the specific (variable) degree to which the
process takes place in correlation with different extra-linguistic parameters,
i.e., the exact relevance of each linguistic parameter for each social group.
This means, however, that a variable rule is, by the explicit admission
196 E. C. Garcia / Sh(/'ting variation
But until a view of language is advanced that makes plausible the uni-
fication of these two disparate types of phenomena, a linguists who value
2 It is perhaps this distinction that Lavandera has in mind when she writes (1978: 174):
'I want to establish a distinction between a difference in frequencies which in itself is
significant, be it socially or stylistically, and a difference in frequency which (...) is simply
the manifestation of a more or less frequent usage of a form in a situational context or
in a social group.'
"~ Despite their theoretical allegiance to generative tenets, variationist studies reflect and
rely on tile transformational formalism of the sixties and mid 70's. Variationist studies have
clearly remained untouched by tile controversies motivating the recent drastic changes within
generative grammar, nor have these changes ill any way been due to insights gained in
variationist studies. The relation between tile two approaches is apparently limited to the
uncritical borrowing of certain expressive devices (Kay and McDaniel (1979: 153), Romaine
(1981 : 96-97)).
198 E.C. Garcia / Shifth~g variation
have opened up new ground, and shown that there is no practical or THEORETICAL
[emphasis mine. EG] harrier to the exploration of syntactic phenomena."
"Speakers do not readily accept the fact that two different expressions actually "mean
the same" and there is a strong tendency to attribute different meanings to them. "s
"Notice where the source of the difficulty lies: units beyond phonology, let us say a
morpheme, or a lexical item, or a syntactic construction, each have by definition a
meaning. They are not like phonemes which, by definition, do not have any "constance
of reference", as Sapir said.'
tiation, by natives and linguists alike. These difficulties, however, were not
recognized as symptoms that 'variation' in syntax differed essentially from
the free variation of allophones.
Now the proper analogue to 'syntactic variation' is not the (socially
significant) free variation of allophones, but rather phonotactics, the non-
random (substance motivated) distribution of distinctive phonological units.
And indeed, just as the frequency distribution of articulatory types is far
from random, so that (within a particular language, as well as across
languages) some types of sounds (or sound sequences) are more frequent
than others (both context-freely and -dependently) so likewise do meaningful
units exhibit distributional skewings.
For instance, Lantolf (1978) reports on the varying degree to which
Spanish-speaking informants will fill in an Indicative or a Subjunctive in
subordinate clauses, depending on the semantic nature of the verb in the
main clause, and Schmidely (1971) has observed a considerable number of
skewings with respect to the choice of Spanish/e (Dative) vs. lo (Accusative)
- for instance, depending on whether the pronoun refers to an animate or
an inanimate entity, whether it is followed or not by an infinitive of which
it can be construed as the subject, etc. - though neither Lantolf nor Schmidely
provide any motivation for the skewings they have observed.
In the case of phonotactics, as pointed out earlier, the explanation for
the observed frequency distribution is a substantive one : universal principles
of articulatory or perceptual ease, rooted in the nature (anatomy, physiology,
or psychology) of the language user. Analogously, in the case of 'syntactic
variation', the motivation of the frequency differences between Subjunctive
and Indicative, or le and lo, must be sought in the greater (respectively
lesser) communicative compatibility between the grammatical units in ques-
tion and the lexical or syntactic contexts in which they occur (cf. Garcia
(1985, to appear)). There is no reason why frequency differences in lexicon
and grammar should not in principle be open to the kind of substantive
explanation appealed to in the case of phonotactics (cf. Schwartz (1980:
330)).
But in that case the very concept of 'syntactic variation' - i.e., 'saying the
same thing in different ways' - is in fundamental error, since it equates the
significant alternation between distinct meaningful units with free variation
among non-distinctive allophones. No wonder that so much criticism (Lavan-
dera (1978:175 fn. 4), Romaine (1984:410-411, 413)) should have been
levelled at the requirement that the syntactic variants 'say the same thing'
200 E.C. Garcia / Sh(fthlg variation
T h e a s s u m p t i o n t h a t d i f f e r e n t f o r m s c a n i n d e e d say the s a m e t h i n g a n d
h e n c e v a r y ( p r e s u m a b l y for social r e a s o n s ) in p a r t i c u l a r c o n t e x t s is f r a u g h t
w i t h great a n a l y t i c a l i m p o r t : the a n a l y s t takes i n t o a c c o u n t o n l y t h a t p a r t
o f the s y n t a c t i c r a n g e w h e r e the alleged m e a n i n g e q u i v a l e n c e is c l a i m e d
to hold.
From the variationist point of view it is only natural that one should
concentrate on that range of use where the alternating forms are felt to be
'linguistically the same' - in traditional terms, in free variation. 8 This
practice can, however, result in significant generalizations being missed,
owing to initial bias with respect to what the relevant data are (Jacobson
(1980: 28)). 9 In this section we shall discuss a series of studies of syntactic
variation, and try to show how:
(i) the very assumption of variation, and
(ii) the consequent restriction of the analysis to only that part of the
range where 'variation' is claimed to occur,
actually results in less than ideal analyses.
We shall begin with the procedure most essential to analyses of 'syntactic
variation', i.e., the restriction of the analysis to those environments where
variation is observed. This can best be exemplified with a concrete - and
classical- example, G. Sankoff and Thibault's (1977) study of the 'syntactic
variation' between avoh" and Etre as Perfect tense auxiliaries in Canadian
French. Sankoff and Thibault are interested in the auxiliary use of Etre,
where Etre + participle combinations 'vary' with avoh" + participle. They
therefore leave out of consideration not only instances of Etre + adjective,
or Etre + complement, but also those cases of Etre + participle that do NOT
vary with avoh'. The operation is saved from circularityt° (but not quite !)
"Characteristic of this kind of study of syntactic variation is that the definition of the
variable requires a series of preliminary steps directed at eliminating all the contexts in
to the "internal constraints of the passive" as do other types. The data would thus have been
messier and the analysis more difficult if these clauses had been included. They were excluded.
E.C. Garcia / Shifting variation 203
which the two alternant forms contrast, i.e. do not say the same thing (...) To return to
the widest possible defining environment as she [B. R. Lavandera] suggests, would reverse
the process and lose the precision of analysis we aim at."lt
'We can infer that there will be contexts where active and passive lead to different
semantic interpretations of states of affairs and ultimately it will be important to locate
these contexts, but these will undoubtedly [sic] be a small subset of the total range of uses
and not likely to affect our search for the general constraints on the choice of active vs.
passive (...) We therefore approach the passive with an eye to a bold simplification of
the problems of meaning."
t t The carving out of only part of the distributional range of a form for analysis in variation
studies had indeed been objected to by Lavandera (1978: 178), who correctly points out:
'It is important to emphasize that this method of gradually getting rid of categorical environ-
ments is not an ad hoc resource in Sankoff and Thibault's [1977] study, but is central to
sociolinguistic analysis."
This is emphatically endorsed by Thibault (1983: 12):
"La d~limitation entre les domaines variable et cat~gorique de la variable constitue un
aspect essentiel de l'analyse variationniste."
204 E. C. Garcia / Sh(fting variation
That the risk is a real one can be shown on the basis of studies where
the assumption of variation - i.e., the equation of distinct expressive means -
stands in the way of a larger, more insightful analysis. Consider, for instance,
G. Sankoff's (1982) study of the circumstances under which French speakers
'duplicate' the subject by adding a proclitic pronoun to either lexical or
emphatic pronominal mention, as in:
'S'agit-il d'un paradoxe? Si l'ajout d'un clitique poss6de une fonction discursive, peut-il
6tre soumis ~. une contrainte syntaxique?'
~ The discussion of yet another study by G. Sankoff should - emphatically - not be under-
stood as reflecting of the quality of her work. On the contrary: it is the caliber of G. Sankoff's
scholarship that brings out most clearly the deleterious effect of variationist analytical practice.
Besides, it was G. Sankoff herself who, in her influential (1973) paper, threw syntax open to
variationist analysis. Most important, she stands out among variationists for daring to tackle
so-called syntactic variation: Sankoff's analysis of bai is in fact singled out by Romaine
(1984:433 fn. 7) as a prime example of a 'pure' syntactic variable. Other studies (Schiffrin
(1981), Ashby (1981)) are not fundamentally concerned with variation per se, the variationist
techniques serving but as a means of presenting large bodies of quantitative data.
206 E. C. Garcia / Shiftbsg variation
Table 1
Bai placement in terms of subject NP (G. Sankoff (1973 : 46)).
I stsg. mi 78 7
2ndsg. yu 52 I
3rdpl. ol 31 I
lstpl, excl. I mil
I stpl. incl. yumi: pela 22 6
2ndpl. t yu /
3rdsg. em Il 47
other NP 22 36
0subject 53
Despite the fact that the order Pronoun-bai-verb is stated to have a com-
municative value of its own, the array of table 1 does indeed suggest that
statement by (variable) rule is the only way to account for the data. Such
a conclusion, however, is heavily dependent on the categorization of the
data, which is itself influenced by the assumption of variation. And indeed,
the data presented in table 1 are open to other classifications.
Consider, for instance, the implicit claim that with zero-expression of a
(third-person singular) subject bai immediately precedes the verb: it is just
13 Sankoff does not make explicit what she means by "emphasis', but we may (plausibly)
assume that under 'emphatic' circumstances the hearer's attention is drawn to the subject
of the sentence - for whatever (discourse) reason.
E. C. Garcia / Shifting variation 207
~4 Sankoff (1973) does not specify whether "0 subject' applies, as table I suggests, exclu-
sively to the third-person singular, or to all persons and numbers. However, in Sankoff and
Laberge (1973:46 fn. 6) we read:
"The status of em as a pronoun which behaves similarly to the others is already dubious
on other grounds [besides the clear difference with respect to placement of bai]. Its optional
character permits its frequent omission, which accounts for the majority of 0 subject
sentences in Table 4."
(Sankoff and Laberge (1973) is the source of the data in Sankoff (1973), table I of the latter
paper exactly reproducing table 4 in the earlier study. Both are here reproduced as table i).
That zero-expression of the subject should be stron~y favoured in the case of third-person
singular subjects is indeed highly plausible, in view of the frequency with which (cross-
linguistically) the third-person singular lacks pronominal expression (Benveniste (1966: 228-
230)).
ts Cf. for instance the "emphatic" value of subject pronouns in Spanish (Rosengren (1974:
234 et passim)). Particularly relevant in this connection is the "emphatic" use of em as an
ADDITION tO lexical subjects in order to mark topic change (Sankoff (1980b: 267-268)). The
possibility that simple em is to 0 as Lexical NP +em is to plain Lexical NP is not discussed
by Sankoff: according to Mihalic (1971 : 15) em is used in Tok Pisin also as a demonstrative.
Sankoff (1984: 114) describes em as a non--clitic subject pronoun.
208 E.C. Garck~ / Sh(fting variation
fact provides evidence that this is what lies behind the use of em (vs. 0)
for object-references in Tok Pisin.
It appears highly likely, then, that the greater frequency of em-bai-verb
with respect to bai-em-verb is due to the fact that the order Subject bai verb
has attention-calling value by itself (cf. the effect it has with personal pronoun
subjects (Sankoff (1973:48)). As such, it is the appropriate order to use
with an attention-deserving form like em. When a speaker wishes to draw
attention to the subject, he will normally do so by all means available:
by the use of a strong, explicit form (em), and by means of the position of
bai. Em-bai-verb, where explicit mention of the subject and position of bai
coincide in attracting the hearer's attention, is semantically more coherent
(it hangs together better) than bai-em-verb. The intrinsic coherence of two
devices that independently work towards the same communicative goal is
the reason, then, behind the greater frequency of em-bai-verb.
In the case of lexical reference to the subject the order subject-bai-verb
is favotired for a different, but just as understandable reason. Lexical
reference to the subject occurs only in the case of third persons which, by
definition, are not given in the speech situation. When, furthermore, these
referents are not recoverable from the context, i.e., are new, the only
alternative is to make use of lexical indications. It follows that 'lexical'
Table 2
Recourse to attention-drawing placement of bai dependent on communicative problem.
3rdsg. 0 None 53 53
3rdsg. Explicit,attention- I1 47 58 81
em drawing reference
to a familiar
referent
E. C. Garcia / Shifting variation 209
The responsibility of the subject for his arrest is transparent in the case
of the sentence with get, not so, however, when be is used. Because the
subject is perceived as the cause of what happens to him, Labor (1975a:
122) speaks of a "causal' meaning in get.
However, the minimal pair presented above contains a purpose infinitive,
which may actually be what 'brings out the causative meaning of get" (ibid.).
E. C. Garcia / Shifthag variation 211
So w h a t a b o u t plain:
I was arrested.
versus:
I got arrested?
' D o these mean the s a m e ? ' L a b o v asks (1975a: 121). Given that:
and :
'If we argue that [I was arrested] and [1 got arrested] mean something different, there is a
strange (...) consequence. It is well known that the "get passive" is more colloquial than
the be passive; in fact, many people deny using it altogether. Though no systematic study
has been made, there is reason to believe that its use is stylistically and socially stratified.
Is it the case that people think more causally when they are talking more informally?'
is However, Labov remains surprisingly silent on the issue of linguistic analysis. He says
( 1975a : 124):
212 E.C. Garcia / Sh(fiing variation
'(1) Get passives generally reflect the speaker's opinion on an event which happens, happened
or may happen to the subject of the sentence (...)
(2) Get conveys in all passive sentences an actional and resultative/s~atal meaning. This
basic meaning of ge t passives assumes different shades o f meaning which depend on
the specific type o f verb and the tense form in which it is used (...),21
'The results of this experiment point towards two distinct modes o f interpretation o f get
with past participles. With a purpose clause, the causative meaning appears to be strong;
without a purpose clause, we do not detect any difference between get and be, and we
would therefore regard the get and be passives as social variants.'
... in all contexts? Or only in the absence o f purpose clauses? In which case, what is the
ANALYTICAL relation between get in the context of purpose clauses, vs. elsewhere?
19 Stein's (1979: 46-67) careful discussion of what get contributes to the communication o f
"passive' messages at no point invokes 'causality'.
20 Stein is able to arrive at her insightful characterization o f the semantic contribution o f get
because she refuses to limit herself to what traditional grammar labels a ~passive'. She rather
takes as her point o f departure the linguistic phenomenon get + past participle, in its entire
lexical range. For perceptive discussion o f this methodological point, cf. Stein (1979: 74-78).
For another analysis o f the get-passive, also within the larger perspective o f the entire range
o f uses o f get, cf. Miller (1985: 181-185).
2t The contrast between get and be (i.e., highlighting 'action' vs. 'statal' traits o f the event)
is, interestingly enough, clearly observed in contexts lacking a purpose clause (Stein (1979: 53))
- precisely there where Labov claims equivalence between the two expressions.
E. C. Garcia / Shifthlg variation 213
Now the fact that get is used to signal the speaker's opinion allows us
to understand why it should be favoured in colloquial rather than in formal
speech. A subjective ('speaker's opinion') presentation of the subject matter
is more likely in emotionally charged speech, and this will be more easily
allowed in a colloquial setting, rather than in formal discourse or in
scholarly prose. Besides, one can advance one's own personal views more
safely in informal speech, because it is here that full responsibility for one's
words can be more easily declined.
The possibility of motivating the favouring of get in informal rather than
in formal speech suggests, then, that an adequate understanding of the facts
can be arrived at if one seeks out the connection among all uses of get,
in all contexts (vs. all uses of be, in all contexts), and that it is unnecessary
to conclude, as Labov does (1975a: 125) that we must 'regard the get and
be passives as social variants'. Such a conclusion is particularly unwarranted
when no correlation is shown to hold between get and be and any social
parameter whatsoever. 22
"(...) the method depends on the existence of an initial two-way cross-cutting classification
of the situations.' (ibid.: 167)
But it is clear that the LINGUISTIC significance of the variation plays second
fiddle :
'Once again, given a way to calculate Or measure x [the variable, EG] in a large number
of situations, the major problem is this: is there some number d, and some meaningful
22 Stein (1979: 164) points out that it is in fact highly unlikely that get should be a colloquial/
stylistic variant of be, since for certain messages get is the only possibility.
214 E. C. Garcia / Shifting variation
d-dimensional classification o f these situations, such that the d-dimensional scaling con-
dition is satisfied?" (ibid.: 166)
and :
"Using statistical and linguistic criteria, we select the best dimensionality and associated
scale." (ibid. : 168)
"(...) main claim is that the syntactic variable is largely recognized through its distributional
properties, mainly its distribution across a speech community." (ibid. : 207)
"(...) quantitative inverse relationship between rate o f usage o f two forms across a com-
munity." (ibid.: 207) z3
"(...) something additional is needed, but this is difficult to pin down (...) In many cases (...)
the most we can say is that the proposed variants can serve one or more generally similar
discourse function. We cannot even require that they be identical discourse functions, the
lack o f rigorous criteria [being] seemingly related to the role o f these variables in the change
process.' (ibid.: 207-208)
And to make quite sure that the point gets across they state:
"In working with these variants we are generally convinced that we are dealing with
bonafide variables and that the particular nature o f the underlying form, or even its
existence is irrelevant. How can this be? Because o f the distributional facts.' (ibid. : 207) 2~
.,a Sankoff and Thibault oppose "weak' complementarity, which is fundamentally quanti-
tative in nature, to "ordinary or strong complementarity o f distribution, which is qualitative'.
By the latter they presumably mean normal linguistic complementary distribution.
:~ The same position is taken in Thibault (1983: 7):
E. C. Garcia / Shifting variation 215
"If we succeed in proving weak complementarity for alld and dtd despite the semantic
distinction on which some would insist it would show the irrelevance or the neutralization
o f this distinction in discourse."
No indication is given, unfortunately, o f how one can decide what constitutes "un taux
anormalement ~lev~ d'occurrences'.
216 E.C. Garcia / Shifting variation
In short, only one 'variant' has been isolated, on purely quantitative grounds
- because terminal tags have forced themselves on the observer's attention.
But, as has long been recognized, one category is no c a t e g o r y : linguistic
units are defined by the oppositions into which they enter. Dines herself
clear-sightedly concludes, with c o m m e n d a b l e honesty ( 1 9 8 0 : 2 9 ) :
'The outcome of the preceding exercise is that it is as yet pointless to undertake quantitative
analysis."
"As far as the synchronic aspect of language structure is concerned, it would be an error
to put for [sic] much emphasis on social factors. Generative grammar has made great
progress in working out the invariant relations within this structure, even though it wholly
neglects the social context of language."
'Linguistic variables or variable rules are not in themselves a "theory of language". They
are all heuristic devices."
Finally, in a recent paper Labov (1982: 34) alludes to the variable rules
of the seventies as 'techniques for multivariate analysis'; he also acknowl-
edges (ibid. : 55) that phonology and syntax work according to very different
principles, and that linguistic sameness is not to be taken for granted (ibid. :
85 fn. 2). He even admits (ibid.) that important issues are raised in Lavandera
(1978) (the object of caustic rebuttal in Labov (1978)). However, as was
seen above, these valuable insights were apparently thrown to the winds in
Weiner and Labov (1983).
How is all this to be accounted for? The failure of the variationist
approach seems to be due to sociolinguists' having adopted the formalism
of generative rules as a purely descriptive device, without seriously con-
sidering its theoretical implications (cf. Labov (1972b: 43), Cedergren and
Sankoff (1974: 353)).26 For instance, the crucial question raised by Kiparsky
' F r o m the point of view of linguistic theory, it would be much more interesting if
patterned behavior in linguistic variation could be shown to be a RESULT of linguistic
competence, rather than just a PART of it, in the same way as the regular behavior of a
coin upon flipping is a result of its physical structure.'
no formal theory of variable d,ata on linguistic tokens than to attempt to graft an account
of these data onto a formal theory that was specifically designed for other and contrary
purposes. It appears that there simply does not exist currently any formal theory that
comes reasonably near to giving a coherent account of the systematicity observed in
respect to linguistic tokens.'
:~ Consider the following remark by Labov (1972b: 81):
'(...) the I-t, d deletion] rule (...) is an excellent candidate for a panlinguistic grammar (...)
the general rule may indeed have escalated out of our grammar into a metagrammar.'
Despite his use of the generative formalism, Labor apparently understands by 'grammar'
something different from Kiparsky (1971 : 645): it is not clear in what sense a 'pan linguistic'
or 'meta' grammar would be learnt, nor by whom. This unclarity with respect to the central
issue of what (and where) grammar is has allowed variationist studies to develop more and
more independently of concerns in theoretical linguistics and - as we have seen - to center
more and more on research techniques.
2a Consider, for instance, the following statement (Labor (1978: 13)):
'(...) it is the explanation of the variable constraints that lead [sic] us to conclusions about
the form of grammar. When we reach those conclusions we will not hesitate to place
probabilistic weights upon our grammatical rules, no matter where they do occur.'
No suggestion is offered as to what might constitute an 'explanation of the variable constraints'.
An even more fundamental question is left unasked, let alone answered: once an explanation
is found for the 'variable constraints', will there be any rule left on which to place probabilistic
weights?
E. C. Garcia / Shifting variation 219
'Before we conclude that given vs. new is the dominant factor in determining the choice
of active or passive (...)' (ibid.: 47)
But:
'(...) the choice of active and passive (...) is conditioned by formal, syntactic factors
(...)' (ibid.: 52)
And again :
"[the data on age and social class correlations] do not indicate that external factors have
a sizeable influence on the choice of active or passive in agentless sentences," (ibid. : 42)
But :
'(..,) tile over-all distribution of this constraint [i,e,, surface-structure parallelism] proves
to be somewhat larger than the given vs, new effect," (ibid. : 47)
Weiner and Labov never do make clear what exactly they understand by
'the choice between active and passive" and, particularly, whether "choice"
?Z20 E.C. Garcia S3~#?ing n~ri,:tion
the contrary, they have done a fine job of vindicating Chomsky's scepticism.
Despite Labov's recent optimistic claims (1980: xiv), the large amounts of
quantitative data produced by variationist studies have not shed any new
light on the structure of language (Itkonen (1983: 258)). Linguistics has
certainly been provided with extremely useful research and analytical tech-
niques, and much very interesting information has been amassed on many
aspects of language use. But the significance and relevance of such data
for linguistic theory has yet to be demonstrated. For this, a coherent view
of language is required.
References
Anttila, R., 1975. The indexical element in morphology. Innsbrucker Beitr~ige zur Sprach-
wissenschaft. Vortrag 12.
Ashby, W.J., 1981. The loss of the negative particle ne in French. Language 57, 674-687.
Benveniste, E., 1966. Structure des relations de personne dans .le verbe. In: E. Benveniste,
Probl/~mes de linguistique g6n6rale, 225-236. Paris: Gallimard.
Berdan, R., 1975. The necessity of variable rules. In: R.W. Fasold and R.W. Shuy (eds.),
Analyzing variation in language, I 1-25. Washington D.C. : Georgetown University Press.
Bickerton, D., 1971. Inherent variability and variable rules. Foundations of Language 7,
457-492.
Bickerton, D., 1972. The structure of polylectal grammars. In: R.W. Shuy (ed.), 23rd. George-
town Round Table Conference, 17-42. Washington D.C. : Georgetown University Press.
Bickerton, D., 1973. Quantitative vs. Dynamic paradigms: the case of Montreal que. In:
Ch.-J. Bailey and R.W. Shuy (eds.), New ways of analyzing variation in English, 23-43.
Washington D.C. : Georgetown University Press.
Bloch, B., 1947. English verb inflection. Language 23, 399-418.
Bolinger, D., 1977. Meaning and form. London : Longman.
Brasington, R.W.P., 1982. Markedness, strength and position. In: D. Crystal (ed.), Linguistic
controversies, 81-94. London: Edw. Arnold.
Broeck, J. van den, 1977. Class differences in syntactic complexity in the Flemish town of
Maaseik. Language in Society 6, 149-181.
Brown, P., 1980. How and why are women more polite: some evidence from a Mayan com-
munity. In: S. McConnell-Ginet, R. Borker and N. Furman (eds.), Women and language
in literature and society, 111-136. New York : Praeger.
Cedergren, H., 1973. On the nature of variable constraints. In: Ch.-J. N. Bailey and R.W. Shuy
(eds.), New ways of analyzing variation in English, 13-22. Washington D.C. : Georgetown
University Press.
Cedergren, H. and D. Sankoff, 1974. Variable rules: performance as a statistical reflection
of competence. Language 50, 333-355.
Chomsky, N., 1957. Syntactic structures. The Hague: Mouton.
Chomsky, N., 1966. Cartesian linguistics. New York: Harper & Row.
222 E.C. Garcia / Sh(fting variation
Dines, E.R., 1980. Variation in discourse- 'and stuff like that'. Language in Society 9,
13-31.
Fasold, R., 1978. Language variation and linguistic competence. In: D. Sankoff (ed.), Lin-
guistic variation. Models and methods, 85-95. New York: Academic Press.
Garcia, E.C., 1985. Quantity into quality: synchronic indeterminacy and language change.
Lingua 65, 295-326.
Gareia, E.C., to appear. Grasping the nettle: variation as the proof of invariance. Pro-
ceedings of the First International Roman Jakobson Conference.
Gazdar, G., 1976. Quantifying context. York Papers in Linguistics 6, 117-129.
Goilo, E., 1953. Gramatical Papiamentu. Curaqao: Hollandse Boekhandel.
Hasan, R., 1973. Code, register and social dialect. In: B. Bernstein (ed.), Class, Codes and
Control, Vol. 2, 253-292. London: Routledge and Kegan Paul.
Hatcher, A.G., 1949. To get/be invited. Modern Language Notes 64, 433--446.
Householder, F.W., 1962. On the uniqueness of semantic mapping. Word 18, 173-185.
Hudson, R.A., 1980. Sociolinguistics. Cambridge: Cambridge University Press.
Itkonen, E., 1983. Causality in linguistic theory. Bloomington, IN: Indiana University Press.
Jacobson, S., 1980. Issues in the study of syntactic variation. In: S. Jacobson (ed.), Papers
from the Scandinavian Symposium on syntactic variation, 23-36. Stockholm Studies in
English LII.
Kay, P. and C.K. McDaniel, 1979. On the logic of variable rules. Language in Society 8,
151-187.
Kay, P. and C.K. McDaniel, 1981. On the meaning of variable rules. Langt, age in Society
10, 251-258.
Kiparsky, P., 1971. Historical linguistics. In: W.O. Dingwall (ed.), A survey of linguistic
science, 576-649. College Park, MD : University of Maryland.
Labor, W., 1963. The social motivation of a sound change. Word 19, 273-309.
Labov, W., 1966. The social stratification of English in New York City. Washington D.C.:
Center for Applied Linguistics.
Labov, W., 1969. Contraction, deletion and inherent variability of the English copula. Language
45, 715-762.
Labov, W., 1970. The study of language in its social context. Studium Generale 22, 30-87.
Labov, W., 1971. Methodology. In : W.O. Dingwall (ed.), A survey of linguistic science, 412-
497. College Park, MD : University of Maryland.
Labor, W., 1972a. The internal evolution of linguistic rules. In: R.P. Stockwell and R.K.S.
Macaulay (eds.), Linguistic change and generative theory, I01-171. Bloomington, IN : Indiana
University Press.
Labov, W., 1972b. Where do grammars stop? In: R.W. Shuy (ed.), 23rd. Georgetown Round
Table Conference, 43-88. Washington D.C. : Georgetown University Press.
Labov, W., 1972c. Negative attraction and negative concord in English grammar. Language
48, 773-818.
Labov, W., 1972d. Sociolinguistic patterns. Philadelphia, PA: University of Pennsylvania
Press.
Labov, W., 1973. The boundaries of words and their meanings. In: Ch.-J.N. Bailey and
R.W. Shuy (eds.), New ways of analyzing variation in English, 340-373. Washington D.C. :
Georgetown University Press.
Labov, W., 1975a. Empirical foundations of linguistic theory. In: R. Austerlitz (ed.), The
scope of American linguistics, 77-133. Lisse: de Ridder.
E.C. Garcia / Shifthtg variation 223
Labov, W., 1975b. The quantitative study of linguistic structure. In: K.H. Dahlstedt (ed.),
The Nordic languages and modern lingt, istics, Vol. 2, 188-234. Stockholm: Almqvist.
Labov, W., 1978. Where does the linguistic variable stop? A response to Beatriz Lavandera.
Sociolinguistic Working Papers, 44. Austin, TX: Southwest Educational Development
Laboratory.
Labov, W., 1980. Introduction. In: W. Labov (ed.), Locating language in time and space,
xiii-xx. New York: Academic Press.
Labov, W., 1982. Building on empirical foundations. In: W.P. Lehman and Y. Malkiel (eds.),
Perspectives on historical linguistics, 17-92. Amsterdam: John Benjamins.
Lakoff, R., 1971. Passive resistance. Chicago Linguistic Society 7, 149-162.
Lantolf, J.P., 1978. The variable constraints on mood in Puerto-Rican-American Spanish. In:
M. Suffer (ed.), Contemporary studies in Romance linguistics, 193-217. WashingtOn, D.C.:
Georgetown University Press.
Lattey, E., 1979. Beyond variable rules. Papers in Pidgin and Creole Linguistics 2, 21-36
(Pacific Linguistics, Series A, N ° 57).
Lavandera, B.R., 1978. Where does the sociolinguistic variable stop? Language in Society 7,
171-182.
Lavandera, B.R., 1982. Le principe de r6interpr6tation dans la th6orie de la variation. In:
N. Dittmar and B. Schlieben-Lange (eds.), La sociolinguistique darts les pays de langue
romane, 87-124. Tiibingen: Gunter Narr.
Lefebvre, C., 1981. The double structure of questions in French : a case of syntactic variation.
In : D. Sankoff and H. Cedergren (eds.), Variation Omnibus, 229-237. Edmonton : Linguistic
Research Inc.
Lyons, J., 1977. Semantics. Cambridge: cambridge University Press.
Mihalic, F., 1971. The Jacaranda Dictionary and Grammar of Melanesian Pidgin. Milton:
Jacaranda Press.
Miller, J., 1985. Semantics and syntax. Cambridge: Cambridge University Press.
Nai'o, A.J., 1978. A study on the origins of pidginization. Language 54, 314-347.
Naro, A.J., 1980. Review of D. Sankoff (ed.). Linguistic Variation. Language 56, 158~170.
Romaine, S., 1981. The status of variable rules in sociolinguistic theory. Journal of Linguistics
17, 93-119.
Romaine, S., 1984. On the problem of syntactic variation and pragmatic meaning in socio-
linguistic theory. Folia Linguistica 18, 409-437.
Rosengren, P., 1974. Presencia y ausencia de los pronombres personales sujetos en espaffol
moderno. Stockholm: Acta Univ. Gothoburgensis XIV.
Sankoff, D., 1978. Probability and linguistic variation. Synth6se 37, 217-238.
Sankoff, D., 1982. Sociolinguistic method and linguistic theory. In: J.J. Cohen et al. (eds.),
Logic, Methodology and Philosophy of Science VI, 677-689. Amsterdam: North-Holland.
Sankoff, D. and H. Cedergren, 1976. The dimensionality of grammatical variation. Language
52, 163-178.
Sankoff, D. and W. Labov, 1979. On the uses of variable rules. Language in Society 8,
189-222.
Sankoff, D. and P. Thibault, 1981. Weak complementarity: tense and aspect in Montreal
French. In : B.B. Johns and D. R. Strong (eds.), Syntactic change, 205-216. Natural Language
Series, 25. Ann Arbor, MI : University of Michigan. (Linguistics Department.)
Sankoff, G., 1973. Above and beyond phonology in variable rules. In: Ch.-J.N. Bailey and
R.W. Shuy (eds.), New ways of analyzing variation in English, 44-61. Washington D.C.:
Georgetown University Press.
224 E.C. Garcia / Shifting variation
Sankoff, G., 1980a. A quantitative paradigm for the study of communicative competence.
In : G. Sankoff, The social life of language, 47-79. Philadelphia, PA : University of Penn-
sylvania Press.
Sankoff, G., 1980b. Variability and explanation in language and culture : cliticization in New
Guinea Tok Pisin. In : G. Sankoff, The social life of language, 257-270. Philadelphia, PA :
University of Pennsylvania Press.
Sankoff, G., 1982. Usage linguistique et grammaticalisation: les clitiques sujets en franqais.
In : N. Dittmar and B. Schlieben-Lange (eds.), La sociolinguistique dans les pays de langue
romane, 81-85. Tiibingen : Gunter Narr.
Sankoff, G., 1984. Substrate and universals in the Tok Pisin verb phrase. In : D. Schiffrin (ed.),
Meaning, form and use in context, 104-119. Washington, D.C.: Georgetown University
Press.
Sankoff, G. and S. Laberge, 1973. On the acquisition of native speakers by a language.
Kivung 6, 32-47.
Sankoff, G. and P. Thibault, 1977. L'alternance entre les auxiliaires avoh" et ~tre en franqais
parl6 en Montr6al. Langue franqaise 34, 81-108.
Saussure, F. de, 1959. Course in general linguistics. New York: Philosophical Library.
Schiffrin, D., 1981. Tense variation in narrative. Language 57, 45-62.
Schmidely, J., 1971. Grammaire et statistique: l'alternance le/Io dans I'expression de I'objet
pronominal 'direct' en espagnol. Etudes de linguistique appliqu~e 6, 37-58.
Schwartz, L.J., 1980. Syntactic markedness and frequency of occurrence. In: Th.A. Perry (ed.),
Evidence and argumentation in linguistics, 315-333. Berlin: De Gruyter.
Silva-Corval~in, C., 1981. Extending the sociolinguistic variable to syntax: the case of pleonastic
clitics in Spanish. In: D. Sankoff, H. Cedergren (eds.), Variation omnibus, 335-342. Ed-
monton: Linguistic Research Inc.
Stein, G., 1979. Studies in the function of the passive. Tiibingen : Gunter Narr.
Thibault, P., 1983. Equivalence et grammaticalisation. University of Montr6al. (Ph.D. Disser-
tation.)
Traugott, E.C., 1982. From propositional to textual and expressive meanings: some semantic-
pragmatic aspects of grammaticalisation. In : W.P. Lehman, Y. Malkiel (eds.), Perspectives
in historical linguistics, 245-271. Amsterdam: John Benjamins.
Trudgill, P., 1972. Sex, covert prestige and linguistic change in the urban British English of
Norwich. Language in Society 1, 179-195.
Weiner, E.J. and W. Labov, 1983. Constraints on the agentless passive. Journal of Linguistics
19, 29-58.
Wierzbicka, A., 1984. Cups and mugs: lexicography and conceptual analysis. Australian Journal
of Linguistics 4, 205-255.
Winford, D., 1984. The linguistic variable and syntactic variation in Creole continua. Lingua
62, 267-288.
Wolfram, W., 1973. On what basis variable rules? In: Ch.-J.N. Bailey and R.W. Shuy (eds.),
New ways of analyzing variation in English, 1-12. Washington D.C. : Georgetown University
Press.