Foreword to: “Sometimes I'll start a sentence in Spanish Y TERMINO EN ESPAÑOL”: Toward a typology of code-switching (1980)

Full text

DOI ./ling--Linguistics ; (Jubilee): –
The second decade (1973–1983)
Shana Poplack, “Sometimes I’ll start a sentence in Spanish Y TERMINO EN
ESPAÑOL”: Toward a typology of code-switching. Linguistics 1980, Volume 18,
issue 7/8, 581–618. DOI 10.1515/ling.1980.18.7-8.581
Introductory comments by the author
What qualies Sometimes I’ll start a sentence in Spanish Y TERMINO EN ESPAÑOL
as the most heavily cited paper in the 50-year history of Linguistics? It did inspire
a substantial and productive research tradition, but it has also generated recur-
rent and ongoing attacks. I’d like to reect on why its main proposals have been
so controversial, and what their current status is today. A remarkable fact is that
despite 33 years of intense research activity since Sometimes was published, there
is still no consensus on the nature or identity of even the major manifestations of
language contact (codeswitching [CS] and borrowing [B]), let alone the linguistic
conditions governing their use. This discord, so characteristic of the eld of con-
tact linguistics, arises not so much from the recalcitrance of its subject as from the
incommensurable perspectives of its practitioners on language, the conduct of
research, the nature of “fact” and evidence, and the principles of scientic proof.
Sometimes represents the rst application of the variationist paradigm to the
study of CS. Foremost among the theoretical and methodological tenets that dis-
tinguish this framework is the recognition that bilingual speech, just like its
mono lingual congeners, is inherently variable, i.e. involves choices during speech
production which incorporate some degree of unpredictability. Key aspects of a
methodology capable of handling such variability include (i) consideration of the
data of actual bilingual interactions situated in the context of the speech commu-
nity in which they were produced, (ii) distinction among dierent manifestations
of language contact, (iii) contextualization of mixing strategies with respect to
each other and to the monolingual benchmark varieties (standard or not) involved,
and iv) systematic quantitative analyses of the data. Taken together, these meth-
odological imperatives can help us determine when (and how) other-language
items may come to assume the linguistic structure of a recipient language (Lr) into
which they are incorporated, or alternatively, retain their donor-language (Ld)
identity. More important, they reveal patterns of language mixing within and
across communities, enabling us to tease out the major strategies from idiosyn-
cratic outliers.
Unauthenticated
Download Date | 1/23/17 8:16 PM

S. Poplack
Sometimes was a nascent attempt to delineate, for quantitative analytical
purposes, the fundamental distinction between CS and B. This distinction is
manifest in the contrast between those Ld items that function like their Lr counter-
parts (i.e., are borrowed), and those retaining Ld grammar (switched). It was intro-
duced in this article in terms of its corollary, the “Free Morpheme Constraint”
(FMC), essentially a preliminary formulation of the claim that B (established or
nonce) and CS (single-word or multiword) could be identied and characterized.
The reasoning behind the FMC was that any apparent word-internal change in
language could be traced to the conversion – phonological, morphological
and syntactic – of an Ld free morpheme into an Lr counterpart, following well-
documented processes of loanword integration. Switching between words
intra-sententially, on the other hand, was found to be mainly conditioned by the
“Equivalence Constraint” (EC), a formalization of the widespread avoidance by
bilinguals of word order conicts at switch points. Subsequent work in the varia-
tionist framework has claried that “nonce” borrowing in running speech is not
restricted to previously attested forms, and that the gradience of phonological
integration renders it a generally poor indicator of loanword (or CS) status.
What have we learned since ?
The last several decades have witnessed the accumulation of a coherent body of
variationist studies on a large number of typologically distinct and similar lan-
guage pairs, using a rich variety of morphosyntactic diagnostics involving such
key linguistic phenomena as determiner expression, adjective placement,
case-marking, word order and verb formation, among many others. These have
conrmed beyond a doubt the original insight of Sometimes that CS and B are
indeed two distinct processes, which are governed by dierent rules, and which
– using the appropriate methodology – may be operationally distinguished as
such (except in inherently language-neutral constructions). They have also con-
verged on two groundbreaking discoveries, correcting assumptions implicit in
the article (as in the eld more generally), while rening the broad lines of the
proposals sketched there.
() Phonological and morphosyntactic integration are independent. Phonology,
of both CS and B, is variable, and thus cannot reliably be used to distinguish
between them.
() Morphosyntactic integration is not contingent on frequency and diusion
(i.e., achievement of loanword status), but rather occurs abruptly at the
nonce borrowing stage.
Unauthenticated
Download Date | 1/23/17 8:16 PM

Toward a typology of code-switching
Only hinted at in Sometimes was what we now know from subsequent community-
based studies to be the quantitatively preponderant categories of nonce and more
frequent, though unattested, borrowings. Other recurrent ndings to emerge
from systematic analyses of tens of thousands of tokens of spontaneous bilingual
production data are:
() The vast majority of all instances of language mixing are single other-
language items, and
() Nearly all of them are immediately integrated morphologically and syntacti-
cally into Lr.
This means that single Ld items tend to display the linguistic characteristics of
established loanwords, independent of their frequency or recurrence, while
single-word CS are exceedingly rare. (Notwithstanding, single Ld incorporations
cannot simply be assumed to be B; the variationist literature oers clear discovery
procedures to determine their status.) This disproportion was already clearly in
evidence in Sometimes, despite the fact that many single Ld items were misidenti-
ed there as CS (largely because the principle in (1) had not yet been recognized).
Crucially, the linguistic properties of CS – of whatever length – have by now been
shown repeatedly to contrast diametrically with those of Bs: both the internal
constituency and positioning of Bs come from Lr; the internal constituency of CS
is that of Ld, but its placement in the sentence tends to respect the word order re-
quirements of both Ld and Lr. This is the basic insight behind the EC.
These dierences, coupled with the severe quantitative disproportions
among single- and multiword Ld items, conspire to render any study treating them
as one and the same in eect a study of B. (They also explain the appeal of inser-
tion models for CS. They do in fact account for the bulk of the language mixing
data, but not for the data of the relatively rare multiword CS.)
Counterexamples?
Over the years, many counterexamples have been oered to the original con-
straints proposed in the article. In assessing the arguments based on them,
we must consider that relatively few were ever actually produced by bilingual
speakers. Others arise from failure to distinguish purported violations from repli-
cations of (variable) Lr structure, oen including null-marking. The bulk of them
concern the EC prediction that (unambiguous) intra-sentential CS will occur at
sites sanctioned by the grammars of both languages. Remarkably, the lion’s share
of these “counterexamples” involves single Ld words. These of course do not re-
spect the EC, but instead pattern syntactically and morphologically with Lr, just
Unauthenticated
Download Date | 1/23/17 8:16 PM

S. Poplack
as predicted by the FMC and its more recent formulation, the Nonce Borrowing
Hypothesis.
Crucially, only rarely have any of these counterexamples been demonstrated
empirically to represent quantitatively meaningful patterns of language use in
any well-dened speech community. Once the data of language mixing have been
accurately circumscribed, and the analytical and methodological imperatives
outlined above applied, (nonce) borrowing and CS under equivalence emerge as
the quantitatively major strategies employed in bilingual communities studied
using accountable methodology.
The preoccupation with isolated counterexamples characteristic of the eld
of language contact highlights the disappointing fact that even where Sometimes
and other quantitative empirical work like it is enthusiastically received, it only
rarely inspires the type of research required to replicate it. To be sure, many schol-
ars of language contact have elicited bilingual production data of varying degrees
of naturalness, and some have even constructed sizeable corpora. But exploita-
tion of this material is too oen restricted to a few selected examples, or declara-
tions in lieu of demonstrations that all the data satisfy some constraint. The real
value of a corpus can be realized only through the principle of accountability,
where every token of all constructions carrying out a certain role is counted and
placed in statistical context. Only then are patterns discernable and general
trends emergent from the complex production of spontaneous discourse.
Shana Poplack: University of Ottawa . E-mail: spoplack@uottawa.ca
Unauthenticated
Download Date | 1/23/17 8:16 PM