Professional Documents
Culture Documents
a historian’s view
JOAN-LLUÍS MARFANY
Abstract
The author takes an outsider’s look at sociolinguistics and aims various criti-
cisms at it. He sees the discipline’s reluctance to let go of proper sociological
and historical questions while at the same time insisting on remaining strictly
a branch of linguistics as its first and foremost problem from which the rest
ultimately derive. More specifically, he criticizes the idealistic and ahistorical
use of such terms as “language” and “dialect” and the failure to see the use-
fulness of “patois”; the failure properly to define “bilingualism” and “diglos-
sia” and to see the fundamental distinction between them; and the unfortunate
confusion created around the idea of “code switching,” which inextricably
mixes together (i) sociologically significant language switching (typical of
situations of bilingualism), (ii) cases of diglossia, (iii) the would-be monolin-
gualism of as yet imperfectly assimilated immigrant or subject populations,
and (iv) a ragbag of banal individual examples too idiosyncratic to be socio-
logically meaningful.
1. Introduction
I am a historian, not a linguist. A few years ago I published a book on the his
tory of the penetration of Spanish in Catalonia (Marfany 2001) or, more pre
cisely, on how Catalonia became increasingly diglossic over time. By the latter
part of the nineteenth century, Catalan society had become a copybook case of
diglossia: Catalans without exception used Spanish for a whole set of functions
and Catalan for another, and the distinction between the two was neat, unam
biguous, and absolute. Catalans wrote (and spoke from the chair, the rostrum,
or the pulpit) in Spanish and spoke in Catalan.
about the ragbag that goes by the name of “social history,” for example. They
do, as it happens, but that is the subject for a different article.1
Let me begin, then, with the uncertainty surrounding the name itself. Why so
ciolinguistics? Why not the sociology of language? It is something which some
sociolinguists also appear to feel uneasy about: Fishman (1970, 1972) no less
changed the title of his very much reprinted book from Sociolinguistics to The
Sociology of Language, while leaving the contents untouched, and this special
ized journal he has been editing since 1974 is called the International Journal
of the Sociology of Language. Other sociolinguists, the majority, in fact, do
dismiss the matter as a red herring (for a recent example, see Boyer 2001: 16)
and, if pressed, wearily reply that we are talking about two different things. At
the root of the distinction lies the idea, best articulated by Labov (1972), that,
language being essentially social, linguistics is either sociolinguistics or it is
nothing at all. If I have understood it rightly, then, sociolinguistics would in
essence be an antistructural conception of linguistics centered on the idea of
variation. The sociological and historical study of language would be an en
tirely different affair, not concerning sociolinguists, but belonging strictly to
the fields of sociology and history and operating with strictly sociological and
historical concepts and methods — and “sociology of language” and “history
of language” would be the aptest names for it. A most explicit statement of this
view can be found in López Morales (1989: 26 –27), but most sociolinguists
nowadays, it is my impression, would go along with that in principle.
This seems eminently reasonable to me, who, needless to say, would only
be professionally interested in the second of these two quite distinct sets of
disciplines. For it is plain to see that such a conception of sociolinguistics au
tomatically subtracts from its field of endeavor a whole set of questions per
taining to the relation between language and society, foremost among which
are those of the how and why of the birth and the death of languages. Socio
linguistic variation may show how certain varieties of speech actually evolve
out of others, but the emergence of a language out of a group of varieties
is the historical result of a socioeconomic and political process, not of a
(socio)linguistic one. Likewise, (socio)linguistic analysis of the contact be
tween two languages may show how one is transformed by the other, but a lan
guage does not die because it gradually fades into another, but because people
stop speaking it, and (socio)linguistics cannot explain how and why that hap
pens. Nor can sociolinguistics thus defined have anything to say about how and
why Catalonia became increasingly diglossic between 1600 and 1870 and de
creasingly so thereafter, not even if we qualify it with the adjective “historical”
are less blunt, but their position is basically the same: yes, sociolinguistics and
the sociology of language are two different things, but the distinction is really
a matter of emphasis rather than anything else, there is great overlap between
the two, . . . and the overlap insensibly spreads in the direction of the latter,
until only a small residue of “linguistically extrinsic” — and not particularly
interesting, it is implied — matters is left out of it.
But one cannot have one’s sociolinguistic cake and eat it, and the result of
such greed is conceptual confusion. For “sociolinguistics” does not really mean
the same in “linguistic sociolinguistics” as in “nonlinguistic sociolinguistics,”
but the sociolinguist, by adopting precisely such a perspective, is badly placed
to see this, that is to say to be fully conscious of the fact that the same word
can designate two very different realities according to whether it is used as a
(socio)linguistic or as a sociological/ historical term. Take the most basic and
obvious of such terms, that of language itself. Sociolinguists on the whole tend
to take it rather for granted. For all their criticisms of structural linguistics they
tend to slip as a matter of course into a structural conception of individual lan
guages as determined by their internal, ideal, linguistic characteristics. Wolof
and Occitanian are just as much languages as French. Of course, sociolinguists
are not blind to the fundamental differences between them nor to the fact that
those differences are of a social and historical nature (though, it seems to me,
they are less aware of the latter). Their answer to this would be that they tend
to adopt a flexible, pragmatic approach to the question, fitting the term to the
particular reality in question. Which is fine, but in so doing they risk losing
sight of the fact that, from the sociologist’s and the historian’s point of view,
those differences are the only thing that matters. This is exactly what happens
to Le Page and Tabouret-Keller (1985: 188–193) in the act itself of warning
about the need to distinguish between “Four meanings of the term Language
X.” It is what happens to Calvet (1998: 57–58), too, when discussing his old
lady who is unaware that she speaks Occitanian and thinks she speaks patois.
From a strictly (socio)linguistic point of view, Calvet is undoubtedly right to
point out that what the lady and her neighbors speak is the one same language
and that this applies to the whole area some people call “Occitania.” But when
he goes on to analyze the situation in sociohistorical terms and sees in it simply
proof of the low status of Occitanian and of the consequent lack of linguistic
self-awareness on the part of its speakers, he badly misses the point. In those
terms, he is wrong and his old lady is right — by definition: if she does not
know she speaks Occitanian, then she does not speak Occitanian. In fact, what the
old lady knows and Calvet apparently does not, is that, when she speaks patois,
she speaks patois, not a language; when she wants to do that, she speaks French.
What is more, the historical record shows that no one has ever spoken Oc
citanian, that there is not and there never has been a language called Occita
nian. There have been and to a certain extent there still are, over a certain area
vernaculars that never developed into a language, a patois. But what is of para
mount importance to the sociologist and the historian of language is the dis
tinction between the two in social terms. Thus, in his comparison between the
situations arising where Spanish and Portuguese face each other across the
Uruguay–Brazil border and where Catalan meets Spanish in the Iberian Penin
sula, Trudgill (1986: 83–84) only just alludes in passing to what, to the soci
ologist and to the historian of language, is the gist of the question. Even then,
his allusion that “speakers of dialects from the Catalan part of the continuum
are generally [ . . . ] regarded as speaking a language separate from Spanish/
Castilian” somewhat misses the point.2 That there is a dialectal continuum in
that particular area of contact is, on the (socio)linguistic level, an undeniable
fact. On the Western side of the political border between the now autonomous
regions of Catalonia and Aragon, however, what is essentially the same (Cata
lan) dialect as spoken on the other side is known, to its own speakers, by the
derogatory term txapurriau, in other words “patois” ( patués being in fact the
name given to it further north). Accordingly, it is only (less and less) spoken in
a very limited range of circumstances, outside of which people prefer to use
what they consider to be their language, that is to say Spanish. East of the bor
der, the same speech, which is used in all circumstances, is regarded — and I
do not mean “generally,” but by its speakers themselves — not as a patois, but
as their language, Catalan — dialectal Catalan perhaps, but Catalan none the
less. And the real difference between this case and the situation arising at the
Uruguay–Brazil border is not to be looked for in the respective processes of
convergence or divergence, but in the fact that, in this case, it is on both sides
that speakers are in the position of thinking of themselves as speaking not a
patois, but a language. To put it, as does Woolhiser (2005: 262) in relation with
a very similar case on the border between Poland and Belarus, as “the fact that
the border delineates the geographical limits of two distinct sociolinguistic
hierarchies and their associated linguistic ideologies” seems unnecessarily
convoluted and opaque, but I think he means what I have just written. To at
tribute it to the heterogeneity or homogeneity between the dialect and its “roof
language,” as Chambers and Trudgill (1980: 25–29) would, I imagine, have us
do, thus introducing yet another layer of unnecessary terminological com
plication, might be an attempt at bringing the issue within the purview of
(socio)linguistics, but it would be no explanation at all. The only possible ex
planation lies in the way those “sociolinguistic hierarchies” have come about
and those “linguistic ideologies” have developed, and that is a historical, not a
sociolinguistic, question.
Let me insist, using now a historical example, the case of Venetian. Venetian
is not unlike Provençal: the homogenized form of a group of related vernaculars
which, toward the end of the Middle Ages, came very close to becoming a lan
guage, but never did (Ferguson 2007). We all know the story: it was a similar
We thus come now to the question that most directly affected me: diglossia.
The “focus article” in the already-mentioned issue of the International Journal
of the Sociology of Language devoted to the subject began with the admission
that, although the “3,000 or so entries in Mauro Fernández’s bibliography
[ . . . ] might well lead one to the conclusion that everything that needed to be
said [on diglossia] had already been said,” the truth was that sociolinguists
were as far away as they had ever been from agreeing on “a theory of diglossia”
(Hudson 2002: 1). I am not sure that would be something we would need to
worry too much about, but, unfortunately, it is on a definition of diglossia that
sociolinguists still cannot agree.
As I see it, Ferguson’s (1959) original definition would have been perfect
had he not restricted it to the use of one or the other of two distinct varieties of
the same language. He was right, in my view, in not wanting to apply the con
cept to the relationship between any language and its dialects, which would
dilute it to the point of uselessness. Sociologically and historically, there is a
clear and fundamental distinction between the Cockney who writes in English
because to him Cockney (if he recognizes it as such) is simply the way he
speaks English and the Catalan who wrote in Spanish because Spanish, not
Catalan, was the language of writing. But, as my own example suggests, Fer
guson should have contemplated the possible application of the term to the
relationship between two unrelated varieties of speech, i.e., two languages or
one language and an unrelated (or not-directly related) patois. Worse still, he
also went on to add to his sufficient and succinctly precise definition a whole
series of unnecessary additions which lay the concept open to objection on the
grounds of its not quite fitting some specific case or another and which conse
quently led to a whole series of instances of the classical operation of throwing
the baby out with the bath water. (This is particularly the case with sociolin
guists directly concerned with Catalan, Occitanian, and Swiss German, who,
almost to a man or woman, either reject the concept altogether or deny that it
properly fits their particular subject [see, for a summary, Boyer 2001: 50 – 62],
for reasons that are entirely ideological.) Then the inevitable followed: along
came Fishman (1967) who, on the strength of those objections, proceeded to
complicate the whole issue beyond redemption. Though he has been criticized
from the very beginning, sometimes most severely, as by Timm (1981), Fish
man’s model has been the basic frame of reference of, and been reproduced
almost verbatim by, practically everyone dealing with the question, with the
exception of a minority who have remained staunchly loyal to Ferguson’s re
stricted conception. And even a Fergusonian like Britto (1986: 27– 49), who
provides probably the most complete catalogue of Fishman critiques up to that
date, believes his contribution to have had a positive effect on balance. It must
also be said that, though duly covered in all general books on sociolinguistics,
the concept has never had much success as an analytical tool. Understandable
as this is, the concept, in my historian’s experience, is absolutely crucial and it
must be released from the horrible entanglement in which Fishman’s model
has landed it.
The real problem with the Fishmanian idea of diglossia, the one from which
all the others derive, lies in his misunderstanding of what constitutes the social
nature of the phenomenon. The problem lies in his distinction between bilin
gualism as a “psycholinguistic” concept and diglossia as a “sociological” con
cept. Psycholinguists have every right to claim bilingualism as their own if
they so wish and are welcome to do with it as they see fit. But bilingualism is
as much of a social phenomenon as diglossia and therefore as much of a socio
logical concept, too. By the same token, diglossia can only be rationally defined
as a form of individual linguistic behavior, just like bilingualism. Unlike Mrs.
Thatcher and friends, I do believe that there is such a thing as society. I know
there are even such things as collective social phenomena. Speaking, however,
whether in the broad or the narrow sense of the word, is most emphatically not
one of them. With that in mind, we might most productively regard bilingualism
as having two meanings, a general one and a restricted one. The general one,
of which both bilingualism in the restricted sense and diglossia are specific
forms, is an individual’s habitual use of two languages. Bilingualism in the
restricted sense is the linguistic behavior whereby an individual uses one or the
other of two languages depending on strictly personal circumstances and there
fore in a totally unpredictable way. Diglossia is the linguistic behavior whereby
an individual uses one or the other of two languages according to function in
obedience to rules accepted by the whole of the social group to which he be
longs and therefore in a totally predictable way. The two concepts are neatly,
beautifully I would even say, complementary: the bilingual person uses one
language or the other (more often, in fact, a language and a patois) depending
on whom he is addressing in all circumstances and for all functions; the diglos
sic uses one language or the other depending on the function no matter whom
he is addressing.
This means, of course, that both concepts can only be applied to social
groups or whole societies as the expression of aggregate individual behavior:
such groups will be bilingual or diglossic according to whether their members
or a significant number of them behave bilingually or diglossically. Any other
use of the terms, and very particularly that of referring to national states as bi
lingual or plurilingual simply because more than one language is spoken within
their confines, is inane and obfuscating — see for example Romaine (1995:
25–30), not to mention Calvet (1998: 31– 40). To say that Switzerland is a
quadrilingual country or that Brussels is a bilingual city is at best a trivial state
ment hardly worth making, at worst an entirely misleading one. Hardly any
body in Switzerland is quadrilingual. What we find in Switzerland is (Schläp
fer 1985), unequally distributed over the territory and varying hugely in size,
two monolingual populations (French and Italian) and two (or rather two sets
of ) diglossic ones (Italic patois/German, in some valleys of the putatively
Italian region, and Romansh patois/German), plus another population, the
German-speaking one, which it is probably best to regard as diglossic, too,
given the strength of the conviction that Schweizerdeutsch, not used in writing
but always spoken, is a distinct language in its own right. Similarly, the popula
tion of Brussels is made up of French monolinguals and Flemish bilinguals
who used to be diglossic in a not all that distant past. Let me insist, then: in
“macrosociolinguistic” terms, as sociolinguists like to say, only populations,
groups of individuals, can be bilingual or (never “and,” by definition rather
than for the reason given by, for instance, Berruto 2003: 195–196) diglossic;
never, ever, territories, cities, or states.
There is another important point to be made regarding diglossia as I have
defined it. What about Ancien-Régime societies, where the vast majority of the
population was monolingual while an important minority habitually used two
languages? How are we to describe them? Well, if we want to be pedantic, as
largely monolingual societies with a diglossic upper class. But in fact one
4. Code switching
Take the much-quoted example from that classic of the genre, Gumperz
(1982: 30 –37), the one in which a black student in an American college, after an
exchange with a white lecturer in perfectly standard American English, turns to
his friends and says, affecting a ghetto accent and intonation: “Ahma git me a
gig!”. To say, with Gumperz himself (1982: 32), that this particular case of
“code switching” lends itself to “a multiplicity of interpretations” would be a
gross understatement: it is in fact liable to any number of interpretations, and
that is precisely what is wrong with it. Gumperz, and — whether in his footsteps
or not, I do not know — others, e.g., Myers-Scotton (1983) or Woolard (1995),
have devised, or perhaps borrowed from anthropology, a way of circumventing
the problem which seems to have become an important tool of the trade. Instead
of directly interpreting the raw material of the spoken exchange themselves,
they present it to “a panel of listeners,” “subjects,” or “consultants,” and ask
them for their comments. I am not sure that anything is really achieved by ask
ing other people to speculate on something rather than do one’s own speculat
ing, but insofar as there is some kind of outcome it can only be a self-fulfilling
one: the interpretations offered will be those expected from the various types of
people chosen to be on the panels, in accordance with the criteria for selection.
Of course, such speculation may sometimes be rather plausible: it is reason
able to conclude, for instance, that the Kenyan worker in Myers-Scotton (1993a:
477) who, in rebuffing the farmer asking him for money, repeats his rebuttal in
Swahili and English as well as in the shared native vernacular, is trying, among
other things, to put distance between himself and his interlocutor. It remains
speculation none the less. Repeated, systematic observation of similar cases,
even prepared experiments, can perhaps yield meaningful results for those,
such as the anthropologist or the psychologist, who may be interested in look
ing at oral exchanges in terms of “discourse strategies” dictated by the “nego
tiation of identities.”4 For the social study of language such endeavors are, it
seems to me, not only irrelevant, but positively damaging in that they lead it up
the garden path.
There are also times, however, when the reason for the switch in conversa
tion from one language to another is clear beyond doubt and when we can also
discern a true “choice” in the switch. Such is the case of the exchange also in
Myers-Scotton (1993a: 484) in the course of which a Luo trying to cadge a
favor from a fellow-Luo post-office clerk in Nairobi switches, as the last re
source, from Swahili to Luo. I am not sure how worthwhile the analysis of such
a case is in terms of “discourse strategy,” but what I do know is that, in terms
of the social study of language, the relevant datum in this instance is not the
switch to Luo, but the fact that non-personally acquainted Luo speakers should
address one another, in the first instance, in Swahili rather than in the shared
native vernacular, in spite of the immediate mutual identification as Luos. We
are here in one of the three categorizable types I mentioned earlier. This type
As far as the sociology and the history of language are concerned, what we
have here are situations along a sliding scale that goes from the language one
does not yet sufficiently master to the language one no longer sufficiently
knows. They are typical of the transitional stages in the linguistic assimilation
of immigrant or of politically subjected or absorbed populations. As such, one
of the reasons why code-switching studies fail to recognize them for what they
are is the deep-seated ahistorical bent of the specialism, as of sociolinguistics
in general, and the consequent inability, acknowledged too, in passing, by Ro
maine (1995: 176), to see the cases subjected to analysis precisely as stages in
a developing process. It is this inability which explains the inordinate attention
paid to, and the misguidedly elaborate interpretations of, phenomena such as
the so-called parler “melandjao” of certain Spanish immigrant communities
in France (Lagarde 1996), speech varieties bound to disappear in the not very
long term — over one generation, in fact — as surely as the sun is going to set
down this evening. Gumperz (1982: 64) has tried to deny this essential transi
toriness of many of the cases students of code switching work so hard at de
crypting with the fallacious argument that there are always people somewhere
doing exactly the same sort of thing. Perhaps, but we are then talking about a
new, different set of people engaged in another, new process, no matter how
similar, destined to run the same swift course to oblivion.5
Not all cases of language switching or mixing by default, however, are of
this kind. Some, as I hinted at before, are related to diglossia, in the sense that
it is at the root of the speaker’s limited competence which causes the switch.
Thus, in my childhood and adolescence in diglossic Catalonia, in Franco’s
time, when Spanish was the only language of education, whenever the conver
sation with my school friends turned to learning-related matters (if we dis
cussed, say, the math problem or the parsing exercise we had been set for
homework) our Catalan kept lapsing into Spanish. It is the same type of
diglossia-related difficulty which explains many of the instances analyzed by
Gardner-Chloros (1991) as well as those in Bentahila and Davies (1992), the
differences observed by the latter between the two age groups under study
being attributable to the fact that the effects of the diglossia were only residual,
though still present, in the younger subjects, brought up in a post-Protectorate
society. Similarly, the rapidly diminishing frequency of code switching in
Brussels detected by Treffers-Daller (1992) is essentially the result of the re
jection of diglossia by Flemish speakers (the only bilinguals and therefore the
only ones ever in a position to switch codes). This very simple reality tends to
be obscured by a misleading concentration on the idea of limited competence,
which is not really the problem with the speakers in these cases, who are mostly
not just entirely competent, but even fluent in both languages. But such com
petence is perfectly compatible with very specific, repeated glitches: there are
certain things one is not used to saying in one or the other language. One does
know how to say them or could easily work out how to do so, but it is much
easier to switch, without a thought. Unless, as a Flemish in Brussels, you are
determined to avoid it.
For the third type of code switching I can offer again a personal illustration:
the situation which very often arose at home, when I was a child, whenever,
say, my mother, her brother, and his wife were in conversation. My mother and
my uncle were the children of Spanish-speaking parents and Spanish was their
mother tongue and the only one they used between themselves, but, having
lived in Barcelona since the age of five or, in my uncle’s case, since birth, they
both spoke perfect Catalan and spoke it always, as a matter of course, with
Catalan speakers — such as my aunt, my uncle’s wife. Consequently, in their
conversations, she remained monolingual — in Catalan, of course. My mother
or my uncle, looking at each other, would start a sentence in Spanish, but then,
as their gaze might shift toward my aunt, so would their tongue, with the same
unconscious ease, slip from Spanish into Catalan, then back again to Spanish
as they moved their eyes toward their sibling — all this in the middle of a sen
tence. This, by the way, was not something peculiar to my family, but it is
something that happens all the time, now as then, in Catalonia (see some ex
amples, though certainly not their convoluted interpretations, in Nussbaum and
Tusón 1995). This, but really only this, is identifiable as a form of behavior for
which the term “language switching” seems an apt name: this really is chang
ing from one language to another — not for any subtle psychological or an
thropological reason that must be skillfully teased out by a practiced specialist,
nor in accordance with some recondite “discourse strategy,” nor because of an
insufficient mastery of one or the other language, but simply because in a bilin
gual society individuals speak one language or the other depending on whom
they are addressing, for entirely personal and aleatory reasons, and always
stick to the same language with any given person. Just as the first of the three
categorizable types I have discussed boils down to diglossia, this last one is
bilingualism in action. Bentahila and Davies (1992: 444) sum up fairly accu
rately, if somewhat in spite of themselves, how to tackle so-called code switch
ing in terms of the social study of language when they write: “In particular, we
feel that more attention should be paid to potentially influential aspects of the
users of code-switching, which might include their degree of proficiency in
each of the languages, the extent to which and domains in which they use each
language, their attitudes towards their languages and towards mixing them,
and the functions each language tends to fulfill in their everyday life and dis
course.” Take away the misplaced tentativeness, the residue of misleading jar
gon (code switching is not something people use, but something they do), and
the unnecessary repetitions (extent and domains/functions, which I much pre
fer), and you have the whole thing in a nutshell: people switch or mix lan
guages because either they try to speak in a language they do not know well
enough and keep lapsing into the one they know, or they are diglossic, or they
are bilingual. And that, it seems to me, is that.
University of Liverpool
Notes
1. This article started life as a valedictory talk, upon retirement, to my colleagues of the School
of Cultures, Languages and Area Studies at the University of Liverpool, later repeated in a
somewhat changed form in front of an audience of mostly linguists at the University of Lan
caster. This is an expanded and much amended version. I am grateful to Paul Kerswill, Florian
Coulmas, and an anonymous referee for their very useful comments. The usual disclaimers
apply.
2. The latter part of Trudgill’s sentence (“whereas the acceptance of Galician as a separate lan
guage is much more controversial”) is ambiguous. The status of Galician as a separate lan
guage is only questioned in relation to Portuguese, not in relation to Spanish. If, on the other
hand, what is meant is that, in reality, it is a patois — and there is a direct relationship between
that and the reluctance to be assimilated to Portuguese — then I could not agree more.
3. Under the general heading “Norme in materia di tutela delle minoranze linguistiche storiche,”
Law num. 482, of 15 December 1999, art. 2, subsection 1 recognizes 12 linguistic minorities:
Albanian, Catalan, Germanic, Greek, Slovene, Croatian, French, French-Provençal, Friulian,
Ladin, Occitan, and Sardinian.
4. I have nothing to say, of course, about the study of the linguistic aspects of code switching,
as surveyed for instance in Romaine (1995: 122–161) and exemplified by, for instance too,
Myers-Scotton (1993b).
5. Another, even more basic, problem posed by code-switching studies is that the exchanges they
analyze are very often extremely short snippets of conversation, insufficiently contextualized
and without enough background information, and authors usually launch themselves eagerly
into their interpretation without even asking, let alone answering, the many previous questions
that come rushing to the reader’s mind.
References