Gumperz 1977

The Sociolinguistic Significance of
Conversational Code-Switching
John J. Gumperz
University of California, Berkeley
By conversational code-switching, I refer to the juxtaposition of pas-

sages of speech belonging to two different grammatical systems or sub-
systems, within the same exchange. Most frequently the alternation takes
the form of two subsequent sentences, as when a speaker uses a second lan-
guage either to reiterate his message or to reply to someone elses statement.
The following examples are taken from natural talk recorded in bilingual
communities. The language pairs in question are Spanish and English (Sp-E),
Hindi and English (H-E) and Slovenian and German (Sl-G). Speakers are
fluent in both languages and regularly use both in the course of their daily
routines.
1. Chicano professionals in California, exchanging goodbyes (Sp-E).

A. Well, Im glad I met you.
B. andale pues (O.K. SWELL).
2. A college student in India, telling an anecdote (H-E).

A. mdi gaya jodhpur me (I WENT TO JODHPUR). There is
one professor of Hindi there, he is a phonetician.
to us-ne pronauns kiyd apne vais-se (SO HE PRONOUNC-

ED IT IN HIS OWN VOICE).
3. Family conversation in Slovenian village in Austria, talking about

a visiting peddler (Sl-G).
A. toca kuarbca ya mewa (SHE HAD SUCH BASKETS)

B. no najinyan (NO I DONT BELIEVE IT)
C. ya ya di mit di ke.rbalan (THE ONE WITH THE BASKETS)
B. vinarca ya : ~ioa (SHE WAS VIENNESE)
C. na! di mit di kirbalan? (NO! THE ONE WITH THE
BASKETS?)
Each of the above exchanges forms a single unitary interactional whole.
Speakers communicate fluently, maintaining an even flow of talk. No hesita-
tion pauses, changes in sentence rhythm, pitch level or intonation contour
mark the shift in code. There is nothing in the exchange as a whole to indi-
cate that speakers dont understand each other. Apart from the alternation
itself, the passages have all the earmarks of ordinary conversations in a
single language.
Downloaded from rel.sagepub.com at Bobst Library, New York University on April 27, 2015
RELC JOURNAL, VOL. 8 NO. 2, DEC 1977
Often code-switching also takes place within a single sentence as in

the next set of examples:
4. Go and get my coat aus dem Schrank da (OUT OF THE CLOSET .
THERE) (English-German)
5. uZ:ma&scaron;ti kafe (WILL YOU TAKE COFFEE) oder te? (OR TEA?)
(SI-G)
6. jo wo accha ticar hota (ANYONE WHO IS A GOOD TEACHER)
hell come straight to Delhi (H-E)
7. That has nothing to do con que hagan ese (WITH THE FACT
THAT THEYRE DOING THIS) (E-S)
8. Those are friends from Mexico que tienen chamaquitos (WHO

HAVE LITTLE CHILDREN) (E-S)
Here phrases with the internal characteristic of two distinct grammatical

systems enter into regular syntactic constructions of the topic-cornment,
noun-noun complement, predicate-predicate complement type. They com-
bine to form one message, the interpretation of which depends on under-
standing both parts.
Metaphorical and Conversational Usage
The conversational switching described here clearly differs both lin-
guistically and socially from what has been characterized as diglossia in the
sociolinguistic literature on bilingualism (Ferguson 1964). In diglossia, code
alternation is largely of the situational type (Blom and Gumperz, 1972).
Distinct varieties are employed in certain settings (such as home, school,
work) associated with separate bounded kinds of activities (public speaking,
formal negotiations, special ceremonials, verbal games, etc.) or spoken with
different categories of speakers (friends, family members, strangers, social
inferiors, government officials). Although speakers in diglossia situations
must know more than one grammatical system to carry on their daily affairs,
only one code is employed at any one time.
To be sure, in some cases of situational alternation, passages in the two
varieties may follow one upon the other within a relatively brief timespan.
In the old catholic mass, for example, Latin was interspersed with the local
languages. Or in some tribal societies the etiquette of public address may
require that something said in one language be translated and repeated in
another. Yet the alternation always corresponds to structurally identifiable
stages or episodes of a speech event. Both members and linguists agree in
assigning each sentence or group of sentences to one code or another. There
is a simple, almost one-to-one, relationship between language usage and
social context, so that each variety can be seen as having a distinct place or
function within the local speech repertoire.
Where such compartmentalization of language use occurs, norms of
code selection tend to be relatively stable. The rules of etiquette which
govern their use are often taught and breached and are subject to overt
comment. There is some justification in these cases for dealing with code
selection in traditional sociological terms, as a matter of conformance or non-
conformance to contextually or situationally determined norms or usage rules
(Fishman 1972). Information on such norms can be elicited through ques-
tionnaire surveys, language usage diaries or similar self-report methods and
compared to frequency counts of code incidence in actual texts.
In conversational code-switching, on the other hand, where as in our
example, the items in question form part of the same minimal speech act,
and message elements are tied by syntactic and semantic relations apparently
identical to those which join passages in a single language, the relationship
of language usage to social context is much more complex. While linguists,
concerned with grammatical systems as such, see the code alternation as
highly salient, members, immersed in the interaction itself, are often quite
unaware which code is used at any one time. Their main concern is with the
communicative effect of what they are saying. Selection among linguistic
alternants is automatic, not readily subject to conscious recall. The social
norms of rules which govern language usage here, at first glance at least, seem
to function much like grammatical rules. They form part of the underlying
knowledge which speakers use to convey meaning. Rather than claiming that
speakers use language in response to a fixed, predetermined set of prescrip-
tions, it seems more reasonable to assume that they build on their own and
their audiences abstract understanding of situational norms, to communicate
metaphoric information about how they intend their words to be understood
(Blom and Gumperz, 1972, p. 425, Gumperz & Hernandez, 1971).
Language Usage and Members Reports
To ask a bilingual to report directly on the incidence of particular
switched forms in a conversational passage is in fact equivalent to, and
perhaps no more effective than, asking an English-speaking monolingual to
record his use of - for example - future tense forms in messages referring to
something that is about to take place. Attempts to elicit such self-report
information on bilingual usage regularly show significant discrepancies be-
tween speakers descriptions of their own usage and empirical studies of tape-
recorded texts.
When residents of a small North Norwegian town were asked to recall

which codes they had used in an informal tape-recorded conversation, they
categorically claimed that they had spoken only the local dialect and not used
standard Norwegian, since as they said everyone in our town speaks only
village dialect, except in school, church or in some formal meetings. Yet
when tape recordings were examined sentence by sentence, they revealed
frequent conversational switching into standard Norwegian. On further
questioning, participants referred to their own metaphorical switching as
lapses of attention, or failures to live up to village norms, and promised that
only the village dialect should be used in subsequent discussion sessions. Yet
tape recordings of these later sessions showed no significant decrease in the
amount of switching.
In thesame vein, some Spanish-English bilinguals living in a Puerto
Rican neighborhood in Jersey City consistently claim that they speak only
Spanish at home and English at work. Yet tapes of their informal conver-
sations showed a great deal of metaphorical switching. In interview sessions
where conversational code-switching is discussed, speakers tend to express
widely different attitudes. Some characterize it as an extreme form of lan-
guage mixing or linguistic borrowing attributable to lack of education, bad
manners or improper control of the two grammars. Others see it as a legiti-
mate style of informal talk. For the most part members have no readily avail-
able words or descriptive terms to characterize the process of switching as
such. Whatever words exist take the form of stereotypical labels which vary in
meaning with changing attitudes.
In Texas and through the American South West, where code-
switching is common among Mexican Americans, the derogative term Tex-
Mex is widely used. In French Canada the word joual has similar stigma-
tizing connotations. Montreal buses some time ago carried the slogan Bien
parler est bien penser (To speak right is to think right), reflecting the official
attitudes, which according to local linguists are by no means shared by all
sectors of the population.
When political ideology changes, attitudes to code-switching may

change also. In California and elsewhere in the South West, pocho or cal6
served as a pejorative term for the Spanish of local Chicanos. But with the
awakening of ethnic consciousness -and the growing pride in local folk tra-
ditions, these speech styles and the code-switching they imply have become
symbolic of Chicano ethnic values. Pocho or calo are now increasingly and
quite effectively used in the modern Chicano poetry and prose which seeks
to depict the California experience.
Until quite recently pejorative attitudes to code-switching were also

found among many students of folk culture. Barbara Kirschenblatt-Gimblet,
in an article on narrative performances among Jews in Canada, analyzes several
instances of dialect humor in which the juxtaposition of Yiddish and English
phrases serves to create humorous effect. Yet she notes that many folklorists
refuse to recognize this type of material as a legitimate form of Yiddish.
Nathan Asubel, author of the well-known A Treasury otlewish Folklore,
refers to the large body of Jewish dialect jokes which are not Jewish at all
but which are the confections of anti-semites who delight in ridiculing the
Jews. (1972)
In the linguistic literature on bilingualism, conversational code-switch-
ing tends to be treated primarily as a marginal or transitory phenomenon, a
type of linguistic interference which accompanies the learning of a new
grammatical system. Existing studies are for the most part concerned either
with language change or second language acquisition and tend to concentrate
on identification of the type of structures that can be exchanged and on the
linguistic and extralinguistic factors that trigger the switch (Haugen 1973).
That code-switching serves to convey semantically significant information
in verbal interaction has not been systematically considered. The purpose of
the present study is to focus on these semantic aspects of code-switching; to
show how speakers and listeners utilize both social and grammatical know-
ledge in interpreting bilingual conversations.
Some Social Uses of Conversational Code-Switching
Existing descriptive and historical information on bilingualism provides
littlesupport for the contention that code-switching is either historically
transitory or a mere matter of individual choice. A recent survey by Timm
(1975), which reviews much of the recent literature on the subject, cites
evidence going back to the early Middle Ages. During the last few centuries
the practice has been noticed throughout the world in many situations of
language and culture contact. Literary histories of seventeenth century
Germany, nineteenth century Russia and Edwardian England describe the
speech habits of upper class speakers whose German, Russian or English is
interspersed with French phrases. In our own time many urban residents of
the ex-colonial countries of Asia and Africa freely alternate between their
own tongue and one of the world languages.
Code-switching is perhaps most frequent in the informal speech of those

speakers of minority languages in modern metropolitan centers who use the
majority language at work and when dealing with members of groups other
than their own. All these are cases of rapid transition where traditional bar-
riers are breaking down and norms of interaction are changing. Eventually
such traditional situations tend to lead to the displacement of one language
variety by the other. Yet bilingualism in any one population often persists for
several generations. Furthermore, as old populations assimilate, new foreign
language speakers move in and other types of bilingualism arise. Thus there is
little indication that code-switching will soon disappear. On the contrary,
with the increasing displacement of formerly stable populations and the
growing ethnic diversification of metropolitan centres the communicative
uses of code-switching are more likely to increase than to decrease.
The bilingual exchanges we have examined furthermore show that code-

switching does not necessarily indicate imperfect knowledge of the gramma-
tical systems in question. Only in relatively few passages is code alternation
motivated by a speakers inability to find words to express what he/she wants
to say in one of the other codes. In the great majority of cases, the code-
switched information could be equally well expressed in either language.
Something may be said in one code and reiterated in the other code elsewhere
in the same conversation. Considerations of intelligibility, lucidity or ease of
expression, important as they are in some cases, can therefore not be the
main determining reasons. Nor is educational inferiority an important factor.
The bulk of our examples come from the everyday talk of urbanized
professionals, students and other educated speakers, who know both lang-
uages well. The individuals in question live in ethnically and culturally diverse
settings and spend much of their day interacting with others of different lin-
guistic backgrounds. To be effective at work or in business, they must have
near native control of the majority language. Yet at the same time they also
actively participate in functioning, ethnically-based peer, friendship or
kinship networks, which stress separate values, beliefs, communicative norms
and conventions.
overtly marked separation between in-and out-group standards
It is this
which perhaps best characterizes the bilingual experience. The problem is not
merely one of cultural differentiation such as one finds among geographically
separated societies. What distinguishes the bilingual from his monolingual
neighbor is the juxtaposition of styles: the awareness that his own mode of
behavior is only one of several possible modes, that interpretation of what a
speaker intends to communicate depends on the style of communication, that
there are others with different communicative conventions and standards of
evaluation which must not only be taken into account but can also be imita-
ted or mimicked for special communicative effect. The juxtaposition of cul-
tural standards is most evident in in-group activities where participants are
bilingual. While in relation with outsiders, of necessity, only the majority
style prevails, in bilingual situations the participants awareness of alternative
communicative conventions becomes a resource, which can be built on to
lend subtlety to what is said. Much conversation in such settings, as Claudia
Mitchell-Kernan points out (1971), is in fact marked by explicit or implicit
allusions to what the others do or think.
The Empirica~ Study of Conversational Switching

At the most general level it can be said that grammatical distinctions
which mark the bilinguals two codes directly reflect or signal the contrasting
cultural styles and standards of evaluation which they encounter in daily in-
teraction. The tendency is for the ethnically specific, minority language to
be regarded as the we code and become associated with in-group and infor-
mal activities, while the majority language serves as the they code, associa-
ted with the more formal, stiffer and less personal out-group relations. But it
must be emphasized that, in situations such as those discussed here, this asso-
ciation between communicative style and group identity is a symbolic one;
it does not directly predict actual usage. There is no one-to-one relationship
between the occurrence of a particular set of linguistic forms and a certain
extralinguistic context. Only in relatively few interaction situations, such as
for example in contacts with older monolinguals, with very small children, or
for certain highly ritualized activities, is only one code appropriate. Elsewhere
a variety of options occur and interpretation of messages, as with conversa-
tions in general, is in large part a matter of discourse context, social presuppo-
sitions and, speakers background knowledge.
Because of this,. there are a number of empirical difficulties in describing

members perceptions of what count as instances of we and they codes
which the analyst must face. To begin with, code-switching must be separated
from loan word usage or borrowings. Borrowing consists of the introduction
of single words or short, frozen, idiomatic phrases from one language into the
other. The items in question are incorporated into the grammatical system of
the borrowing language. They are treated as part of its lexicon, take on its
morphological characteristics and enter into its syntactic structures. Code-
switching, by contrast, relies on the meaningful juxtaposition of what
speakers must process as strings formed according to the internal syntactic
rules of two distinct systems.
Loan words differ from native words primarily m etymological origin.
Existing studies of borrowing are somewhat unclear as to the degree to which
a word must be integrated into a grammatical system to count as an item in
that system. There are always some items, such as the English nice (from
Latin nescius) which, although clearly identifiable as loans on historical
grounds, can for all intents and purposes count as regular parts of the native
vocabulary. Others continue to be seen as foreign, either because they are re-
cent in origin or because they retain some perceivable non-native characeris-
tics. The word ticar (teacher) in example 6 above, for instance, has a recog-
nizably English phonological shape and, when quoted in isolation, might be
classed as an English word. Yet in this example it functions as a Hindi term
since it obeys Hindi number and gender concord rules. Some other seemingly
marginal cases are:
9. Er hat das gefzxt. HE FIXED IT. (German)
10, usne fix kiya. HE FIXED IT. (Hindi)
11. yas grin mit. I GO ALONG. (Slovenian)
12. hice klaim HE CLIMBED. (Spanish)
13. na hiy-me, ni &scaron;iyo-me tha. HE WAS NEITHER WITH THE MEN
NOR THE WOMEN. (Hindi)
In item nine the underlined borrowed verb stem takes on German pre-
fixes and suffixes, while in item ten the same English stem fix forms a com-
pound verb with the Hindi kiy serving as the inflected auxiliary. Notice
that all our loans obey the grammatical rules of the borrowing language. Eng-
lish words as datum (plural: data) and cactus (plural: cacti) are only ap-
parent exceptions to this, since although they take a borrowed plural suffix,
they nevertheless obey English agreement rules.
Item thirteen odder. Here the English pronouns he and
seems even
she are borrowed, become Hindi nouns and take on regular Hindi case
endings (Khubchandani, 1974). It is interesting to note that Hindi personal
pronouns do not show gender distinctions. In examples eleven and twelve
loan items participate in what itself is a borrowed syntactic construction.
But such borrowed, separable prefix constructions can also occur with native
lexical items. The mit acts as a separable prefix on the pattern of German
ich gehe mit (I go along). The similar case of borrowing like hice klaim
is discussed in a recent paper by Reyes (1974).
While the loan words discussed so far are always phonetically assimila-
ted into the borrowing system, there are some cases where phonologically un-
assimilated items from a high prestige foreign language are inserted as marked
expressions into an otherwise monolingual passage. Examples are: She is a
grande dame ; He has great savoir faire . Here speakers may pronounce
grande with French-like nasalization or emphasize the fricative r in savoir
and thus make conscious use of foreign sounds to suggest refinement or
ridicule. The semantic effect here is similar to metaphorical code-switching.
For the most part, however, these examples are isolated cases, quite different
from the constant alternation which marks code-switching in bilingual com-
munities.
A further descriptive problem arises from the fact that, as our discuss-
sion of attitudes to language usage suggests, norms of grammaticality vary
greatly with respect to both borrowings and code-switching. Bilingual speech
is highly receptive to loans. Items which are in general use in bilingual com-
munities are often unknown and unacceptable in the respective monolingual
regions of origin. Even within the bilingual community itself, norms of appro-
priateness vary. In a relatively small Puerto Rican neighborhood in New
Jersey, some members freely use code-switching styles and extreme forms of
borrowing in everyday talk as well as in more formal gatherings. However,
they spoke Spanish with very small children and older family members. Other
local residents spoke Spanish only on formal occasions, reserving code-switch-
ing styles for informal talk. Others again spoke mainly English, using Spanish
or code-switching only with small children or with some neighbors.
Depending on such factors as region of origin, local residence, social

class and occupational niche, each communicating subgroup tends to establish
its own conventions with respect to both borrowing and code-switching. To
judge a bilingual by any a priori standards of grammaticality can therefore
hardly be satisfactory. The best that can be done is to establish a range of
interpretable alternations or communicative options and thus to distinguish
between errors due to lack of grammaticae knowledge and meaningful dis-
course. Rules of productive control within this range of options are always
context-bound and difficult to generalize. They are learned through constant
practice by living in a group and vary just as control of lexicon and style
varies in monolingual groups. ,
There is evidence to show that most bilinguals have at least a compre-

hensive knowledge of usage norms other than their own and that they can use
this knowledge to judge speakers, social background and attitudes in rnuch
the same way that monolinguals use dialect in assessments of social status. Re-
sidents of such large Spanish-English speaking communities as San Francisco
or New York, which include immigrants from many Latin American regions,
in fact claim that they can tell much about a persons family background and
politics from the way he code-switches and uses borrowings. What the out-
sider sees as almost unpredictable variation, becomes a communicative re-
source for members. Since bilingual usage rules must be learned by living in
a group, ability to speak appropriately is a strong indication of shared back-
ground assumptions. Bilinguals, in fact, ordinarily do not use code-switching

styles even with other bilinguals unless they know something about the
listeners background and attitudes. To do otherwise is to run a risk of serious
misunderstanding.
Conversational Code-Switching as a Sociolinguistic Phenomenon
Code alternation among bilinguals shows some similarity to alter-
nation among dialect variables in the urban speech community studies in
recent sociolinguistic surveys. Here also, selection of variants is in large part
due to subconscious processes. Furthermore, attitudes to speaking as well
as members self-reports differ systematically from actual usage (Trudgill
1974). It seems that in both cases social factors outweigh actual usage as
predictors of message form.
Yet there are some significant differences between the two types of
problems. Variable distribution rules are statistical abstractions relating the
incidence of certain items of surface form in the speech of significant samples
of speakers to macro-sociological categories, such as social class, ethnic
identity, education and the like. With bilingual groups which are deeply
divided with respect to code-switching rules, these relations do not necessarily
hold.
In fact Labov (1971), who first formalized the notion of variable rule,
explicitly cites the following Spanish-English passage as an instance of non-
rule governed variation. (In the transcription, Spanish passages are italicized
and followed by capitalized English translations. English passages are given
directly in lower case.)
Por eso cada, (THEREFORE) you know its nothing to be proud of,
porque yo no estoy (BECAUSE I AM NOT) proud of it, as a matter
of fact I hate it, pero viene Vieme y Sbodo yo estoy, tu me ve haci
[sicl a mf, sola (BUT COME FRIDAY AND SATURDAY I AM, YOU
SEE ME LIKE THIS, ALONE) with a, aqul solita a veces que Frankie
me deja, (HERE ALONE, SOMETIMES FRANKIE LEAVES ME) you
know a stick or something, y yo aqu solita, queces [sic] Judy no sabe
y yo estoy haci [sic], viendo televisibn, (AND I HERE ALONE, PER-
HAPS JUDY DOESNT KNOW AND I AM LIKE THIS, WATCHING
TELEVISION) but I rather, y cuando estoy con gente yo me ...
borracha porque me siento mas, (AND WHEN I AM WITH PEOPLE

I ... DRUNK BECAUSE I FEEL MORE) happy, ms (MORE) free,
you know, pero si yo estoy con mucha gente yo no estoy, (BUT IF I
AM WITH MANY PEOPLE I AM NOT) you know, high, more or less,
I couldnt get along with anybody.
Labov describes the switching in this passage as idiosyncratic behavior,

not covered by the regularities which determine the occurrences of socio-
linguistic variables. His argument hinges on his implicit definition of the term
social as limited phenomena that show statistically predictable distributions
within extralinguistically defined human groups. Labov does not attempt to
account for the listeners ability to assign speakers to social categories, to
place them within the spectrum of known social categories and to assess
shared social background. If we extend our definition to these latter pheno-
mena, which clearly fall within the scope of sociological role theory, then
code-switching cannot be dismissed as merely a matter of idiosyncratic
behavior. To be sure, code-switching occurs in conditions of change, where
group boundaries are diffuse, norms and standards of evaluation vary, and
where speakers ethnic identities and social backgrounds are not matters of
common agreement. Yet if it is true that code-switching styles serve as
functioning communicative systems, if members can agree on interpretations
of switching in context and on categorizing others on the basis of their
switching, there must be some regularities and shared perceptions on which
these judgements are based.
Perhaps a more fruitful way to visualize the issue of social regularities

in code alternation is to set aside the assumption that speakers either do or
do not conform to the norms of one or another clearly defmed ethnic group
and to consider speaker participation in various networks of relationship.
Network analysis focuses on the various kinds of social contacts that actors
establish in the course of their regular routines and on the norms governing
behavior in these contacts. It makes possible the empirical examination of
the relationship of ethnic group membership to everyday behavior. Exami-
nation of speech variation in these terms has shown that within what on the
surface seems like a single speech community (Gumperz 1971) it is often
necessary to distinguish two categories of individuals: (a) individuals in-
terested in closed network relationship, where family, friendship and occu-
pational and other contacts are carried on mainly with others of similar
network characteristics and(b)open network speakers, for whom such regular
relationships do not overlap and interaction is carried on with others of dif-
ferent background.
A number of scholars have noted that the speech of closed network

groups is marked by an unusually large number of truncated, idiomatic stock
phrases and context-bound deictic expressions (Sapir 1922, Bernstein 1971).
Although some tend to see this as evidence for socially based differences in
language ability, it is more reasonable to assume that exclusive interaction
with individuals of similar background leads to reliance on unverbalized and
context-bound presuppositions in communication and that formulaic speech
reflects this fact. Speakers who have little experience to the contrary often
fail to account for the fact that others who do not share their communicative
experience may also not have the background knowledge to interpret their
speech as they themselves do.
Open network situations, by contrast, are marked both by diversity of
norms and attitudes and by diversity of communicative conventions. To be
effective here speakers must be aware of differences in interpretation pro-
cesses. They cannot expect that their unspoken communicative conventions,
characteristic of their own peer group, are understood by others, and must be
flexible with respect to speech style.
Note that network position is only partly a matter of ethnic group.

It is afunction of actual communicative experience and also varies with
education, occupation, generation cohort, political values and individual
aspiration for mobility. Accordingly members of the same family and neigh-
borhood background group may show different language usage practices.
The bilingual speakers we have described show the social and attitudinal
characteristics of open network situations. One would expect their language
usage practices to reflect this also. Yet because of its reliance on unverbalized
shared understandings code-switching is typical of the communicative con-
ventions of closed network situations. Our observation that switching strate-
gies serve to probe for shared background knowledge suggests an explanation
for this apparent contradiction. Since the appropriate usage can be learned
only through actual communicative experience, if a speaker in a situation of
social diversity can appropriately employ these strategies as part of the give
and take of a longer conversational exchange, this is in itself socially signifi-
cant. Regardless of the attitudes members may express elsewhere, regardless
of how they would rate on conventional social scales, their control of the
10
relevant communicative strategies is prima facie evidence for the existence
of shared underlying assumptions which differentiate them from others who
cannot use these strategies.
It is not necessary therefore to turn to larger samples of text of extra-

linguistic indices in order to determine whether bilingual alternation is more
than idiosyncratic behavior. What we need are detailed investigations of
speakers use of code-switching strategies, in actual conversational exchanges,
to show that they exhibit some form of linguistic patterning, that they con-
tribute to the meaning of constitutent messages and that participants in the
interaction agree on their interpretation. Even a casual examination of the
above text in these terms reveals a number of regularities in the speakers
use of the two codes. Many of the seemingly English items clearly count
as borrowings in terms of criteria. Examples are television; happy in
me siento mas happy; free in mas free. Proud of it in yo no estoy proud
of it is marginal, but probably also a borrowing because of its position
within the Spanish phrase.
The rest of the passage, if we disregard the initial sentence, which is

complete, is mainly in Spanish. The code-switched phrases consist largely
of interjections, dependent clauses or verb complement phrases. The speaker
describes her loneliness at home while her husband is away. English serves
mainly to amplify or to qualify information already introduced. This is not
random language mixture, yet motivation for code-switching seems to be
stylistic and metaphorical rather than - grammatical. The process by which
meaning is conveyed must be studied in terms of the stylistic interrelationship
of sentences or phrases within the passage as a whole, not in terms of the
internal structure of particular sentences.
Code-Switching in Three Language Situations

To explore the mechanisms by which code-switching conveys meaning
and their relationship to grammar and speakers and listeners social pre-
suppositions in more detail, code-switched passages, isolated from a number
of conversational exchanges, were examined. Examples derived mainly from
three linguistically and socially distinct situations.
The first is an Austrian village of farmers and laborers located along

the Austrian-Yugoslavian border. The population here has a history of one
hundred and fifty to two hundred years of bilingualism. Speakers use Slo-
venian at home but they are educated in German and live in close proximity
to German-speaking villages and shopping centers. German is the exclusive
language of most business and work relations. The second group consists of
Indian college students from urban Delhi. All students are native speakers of
Hindi who have had all their secondary education in English. Some members
of the group are teachers of English, some have published poems and short
stories in Hindi. In situation three participants are members of a group of
Chicano college students and urban professionals who were bom in the
United States and are largely of economically deprived backgrounds. They
speak Chicano Spanish at home, especially at home with their elders, but
speak English in many of their work and friendship relations. The conver-
11
sations studied were recorded for the most part by participants themselves
and interpretations of meanings in each case were checked with participants
and with others of similar social and linguistic background.
Knowledgeof cultural values and social factors affecting language use

are a starting point for any study of code-switching, but, as we have
necessary
argued above, this information is only one of the factors which enter into
the speakers interpretation process. When interviewed about their language
usage, speakers in all three situations readily identify Slovenian, Hindi and
Spanish respectively as the we code, suitable with kin and close friends.
German and English serve as they codes to be used with outsiders or for
special types of formal discourse. Beyond this, however, opinions about
language usage norms vary and can be interpreted and understood only in
relation to the background conditions that shape each language situation.
The Austrian village is part of that area of Slovenia which in the elec-
tions held shortly after the first world war voted to remain with Austria,
rather than become part of the newly-formed Yugoslavian State. Slovenian
speakers, although numerically in the majority in their own villages, were
largely marginal farmers and landless laborers, who until the end of the nine-
teenth. century had been looked down upon and effectively excluded from
commerce and urban middle class occupations. In modern Austria ethnic
minorities have been legally emancipated and their linguistic rights are pro-
tected by law. This policy was only briefly reversed under the German occu-
pation during the second world war, when Slovenians were officially declared
to be of German descent. They were forbidden to speak Slovenian and any-
one caught using it was subject to denunciation or arrest. At present Slovenian
instruction is again available where the demand is sufficient to warrant it.
Since the second world war, with the advent of tourism as a major
source of income and with the growth of the timber industry, the economic
differential between German- and Solovenian-speakers has largely disappeared
and with it ethnic prejudice is also waning. The norms of language usage,
however, continue to reflect the history of inter-ethnic relations. For all
practical purposes German is accepted by all as the only official and literary
language. All village residents agree, for example, that it is impolite or even
crude to speak Slovenian in the presence of German-speakers, be they for-
eigners or German-speaking residents. So strong is the injunction against
speaking Slovenian in mixed company that tourists can live in the village for
weeks without knowing that any language except German is spoken. Further-
more, although Slovenian continues to be spoken in most homes and is
positively valued as a sign of village in-group solidarity, young people are
encouraged to learn proper German lest they have difficulty in school or
employment.
In urban North India, English has since the nineteenth century been
the main symbol of urbanization and Western technology. Until quite recent-
ly secondary and higher education were almost entirely in English. Hindi is a
literary language with a written tradition going back to the Middle Ages.
Hindi literature has flourished during the last decades; poetry, novels and
short stories have been widely read since Indian independence. Furthermore,
Hindi has become the official language of administration and has replaced
12
English as an important medium for business in much of North India. There
has been a great deal of effort by language reformers and government
planners to replace English altogether. Yet English continues to be widely
used, especiallyin those metropolitan centers where large sectors of the
population come from non-Hindi-speaking areas.
By the time they go to college most students in these larger cities have
a functional reading and speaking knowledge of English and use it along with
Hindi. The use of English in informal conversations is deplored by many
critics, who see the tendency to use foreign loan words and to mix languages
as a threat to the purity of Hindi and a threat to the preservation of tradi-
tional values. Yet among students and young intellectuals of the type re-
corded here, knowledge of English serves as a mark of sophistication. The
individuals in question pride themselves on their knowledge of modem Hindi
literature and on their sense of Hindi and their use of English in everyday
talk.
Speakers of Chicano Spanish in California are in part descendants of

Mexican immigrants to the Southwestern United States who came as farm
laborers or industrial workers, and in part descendants of indigenous Spanish-
speaking populations. They tend to live in ethnically segregated Spanish
neighborhoods where Spanish-speaking natives of the United States inter-
mingle with recently arrived monolinguals. Until quite recently they ranked
lowest among Californian ethnic minorities in income or education. Middle
class occupations were not open to those individuals who retained obvious
signs of ethnic distinctness. Spanish-speakers who entered the middle class
felt obliged to assimilate to middle class American culture and this meant
giving up ones ties to ones Spanish-speaking background.
As elsewhere in the case of minority language settlements, residents
of Spanish-speaking neighborhoods have developed their own dialect of
Spanish. This language has many features in common with the dialects of
Mexican farmers. It has also incorporated some of the features of Calb, the
slang of urban youth groups, and incorporates large numbers of borrowings
from English. Residents of Mexico tend to use the term pocho to Tefer both
to the Americans of Mexican descent and - in a derogatory sense - to the
urban dialect the latter speak. As we pointed out before, with the recent
awakening of ethnic consciousness the terms pocho, calb and chicano have
been adopted as symbols of the newly-found self pride. Urban professionals
and intellectuals consciously affirm their tie to their low income ethnic
brothers and symbolize this by deliberate adoption of pocho speech along
with English and literary Spanish.
All three societies represent situations of change, characterized by both

inter- and intra-group diversity of aspirations and standards of evaluation.
This is reflected (a) by the symbolic associations of the respective we and
they codes and (b) by the lack of general agreement about the linguistic
realizations of these codes. In abstract, purely grammatical terms the linguistic
repertoires are perhaps most easily described as dialect continua in which
the gap between the extremes is bridged by intermediate varieties, marked
by various degrees of grammatical convergence borrowing and code-switching.
13
However, as the discrepancy self-report and recorded texts suggests,
between
members do not attempt to establishexplicit relations between usage and
grammatical facts. When members talk about language, the term language
serves as a metaphor for talk about interpersonal relations, political values
and behavioral etiquette. Yet, on the other hand, members also use criteria of
linguistic appropriateness to judge others and make inferences based on what
they say. These latter judgments are always context-bound and depend on
conventions governing appropriate co-occurrences between context and con-
textualization cues, which are difficult to state in the abstract. Yet since in
any one context members can agree on what is meant, examination of the
relevant conversational processes can elucidate the underlying bases for this
agreement.
The Conversational Functions of Code-Switching
As an initial step in our analysis, an attempt was made to study the

conversational functions that code-switching has. Illustrative brief exchanges,
just long enough to provide a basis for context-bound interpretation, were
extracted from tape-recorded conversations in all three situations. It was
found that switching serves roughly similar functions in all three situations, so
that a single preliminary typology can be set up which holds across languages.
A. Quotations
In many instances the code-switched passages are clearly identifiable
either as direct quotations or as reported speech. Here are some examples
(Slovenian, Hindi and Spanish sentences are italicized and followed by trans-
lation in capitals. German is not italicized but is also translated in capitals):
10. Slovenian-German From an informal business discussion among

neighboring farmers, called to discuss the sharing of farm ma-
chinery. The speaker is reporting on a conversation with a
German-speaking businessman:
pa prabe THEN HE SAID wen er si nit colt gib i si nit
IF HE DOES NOT PAY FOR IT, I WILL NOT GIVE IT.
11. Elsewhere in the same discussion a speaker reports as follows on

what a fellow villager, who is a potential participant in the shar-
ing arrangement, has said:
pa waguta ya tudi reku mana ucera AND VAGUTA HAS
ALSO SAID TO ME YESTERDAY: also a hektar hab i gel
SO I HAVE ABOUT A HEKTAR also i bin gewilt SO I
AM WILLING.
12. Slovenian German Village woman talking with neighbors about

her conversation with the German-speaking doctor:
tadei ya viu tola tudi tol ya THEN THERE WAS ALSO ...
THERE WAS .. , prou vauda ya mou HE ACTUALLY
HAD WRINKLES pa yas varawa sn rainarya yas sn reakua
AND I ASKED (DR.) RAINER, I SAID is etwas k svol n
IS SOMETHING SWOLLEN. praba: HE SAYS nain er is
14
guternert er hot kain vosar unt guar niks NO HE IS WELL
NOURISHED HE HAS NO WATER OR ANYTHING.
13. Hindi-English From a conversation among young Hindi-speaking

college teachers. The speaker is talking about his visit to the
doctor:
He says: ye hi medsin kantinyu karo bhai CONTINUE
TAKING THIS MEDICINE FRIEND.
14. Hindi-English From a conversation among Hindi-speaking college

students and writers in Delhi:
I went to Agra, to maine apne bhaiko bola ki THEN I SAID
TO MY BROTHER THAT, if you come to Delhi you must
buy me some lunch.
1 S. Spanish-English. From a conversation among two Chicano pro-
fessionals. The speaker is talking about her babysitter:
She doesnt speak English, so, dice que la regafian: sl se
les va olvidar el idioma a las criaturas SHE SAYS THAT
THEY WOULD SCOLD HER: THE CHILDREN ARE
SURELY GOING TO FORGET THEIR LANGUAGE.
16. Spanish-English. Later in the above situation, the speaker is

reporting on what her father said about her childrens inability
to speak Spanish: .
To this day he says that uh its a shame that they

... ...
dont uh
speakSpanish. estan como burros. Les habla
... ...
uno y THEY ARE LIKE DONKEYS. SOMEONE TALKS

TO THEM AND: What he say, whats he saying?
B. Addressee Specification
In a second set of examples the switch serves to direct the message
to oneof several possible addressees. This occurred very frequently in the
Austrian village when a speaker turned to someone standing aside from a
group of conversationalists:
17. Slovenian-German. Informal conversation about the weather in

a village home. A strong wind is blowing and there is a danger
of rain and of the fruit being blown off the trees:
A (speaking to B): nceaba pri&scaron;u, vo ki &scaron;u vaitar IT WILL
NOT COME, IT WILL PASS BY.
B (speaking to A): ya ki taka na{3asan aapkama pa ya zia
ciu &scaron;t.Jm ya pastrana IT IS SO OVERLOADED WITH
APPLES AND THE ENTIRE TREE IS BENT ALREADY.
B(continues, turning to C sitting apart): regan vert so ain
vint is drausan IT WILL RAIN IT IS SO WINDY OUTSIDE.
18. A group of Hindi-speaking graduate students are discussing the

subject of Hindi-English code-switching:
A: Sometimes you get excited and then you speak in
Hindi, then again you go on to English.
15
B: No nonsense, it depends on your command of English.
B: (shortly thereafter turning to a third participant who
has just returned from answering the doorbell): kau hai
bhai WHO IS IT? (Note the discrepancy here between
actual usage and talk about usage).
19. A Hindi-speaking student couple is talking to a Hindi-speaking

visitor in their home:
Wife: piparmint piyeggi ap WILL YOU HAVE SOME
.
PEPPERMINT?
Visitor: piye1)gi? DRINK?
Wife: pinekihi ~zz hai THAT IS WHAT ITS FOR, DRINK-
ING.
Visitor: ye kaise piya jiti hai HOW CAN I DRINK IT?
Husband: But she doubts us, ki isme koi alcohol to nahi
THERE MIGHT BE SOME ALCOHOL IN IT.
Husband (turning to his wife): Put it in a glass for her.
C. Interjections
In other cases the code switch serves to mark an interjection or
sentence filler. Item one above is a good example. The complete sequence is:
20. Spanish-English. Chicano professionals saying goodbye, and after
having been introduced _by a third participant, talking briefly:
A: Well, Im glad I met you.
B: Andale pues (O.K. SWELL) And do come again. Mm?
21. Spanish-English. A is talking to someone else later on in the same

situation. Here the main message is in Spanish and the switch to
English:
pero como BUT HOW you know ... la Estela y la Sandi ...
relistas en el teltono (STELLA AND SANDI ARE VERY

PRECOCIOUS ON THE PHONE.)
22. Slovenian-Gennan Austrian village conversation.

B replies to A prior to continuing on in Slovenian:
A: grtayt GO THERE
B: Ya so ist das
D. Repetition
Frequently a message is repeated in the other code, either literally
or in somewhat modified form. In some cases such repetitions may serve to
clarify what is said, but often they simply amplify or emphasize a message.
23. Spanish-English. Chicano professionals:
A: The three old ones spoke nothing but Spanish.
Nothing but Spanish. No hablaban ingles THEY
DID NOT SPEAK ENGLISH.
16
24. Spanish-English. Later in the same conversation:
A: I was ... I got to thinking vacilando el punto ese
MULLING OVER THAT POINT you know? I got
to thinking well this and that reason ...
25. Hindi-English. Father in India calling his son, who was learning
to swim in a swimming pool: .
baju me jdo betd, ndr mt GO TO THE SIDE, SON, NOT

INSIDE.
Keep to the side.
26. English-Hindi Father calling his small son while walking through
a train compartment:
A: Keep straight
Keep straight (louder)
sdh jo beta
27. Spanish-English. Puerto Rican mother in New York calling to

her children who are playing on the street:
A: ven aca COME HERE
ven aca COME HERE
Come here, you.
28. Slovenian-German Austrian village family conversation. Talking

about a woman peddler wha had come by some time ago.
Father: tota ka ya uana mewa kuarbc THE ONE WHO
LAST YEAR HAD BASKETS
Daughter: ka ya usa mewa THE ONE WHO HAD LICE
Father: koi ya mewa WHAT DID SHE HAVE?
Daughter: taca kuarboa pa u&scaron;a ya mewa SUCH BASKETS
AND SHE HAD LICE
Father: no na zinian NO I DONT BELIEVE IT
Mother: ya ya di mit kerba lan YES THE ONE WITH
THE BASKETS
Father: vinarca ya (3oa SHE WAS FROM VIENNA
Mother: na di mit di kerb alan NO THE ONE WITH THE
BASKETS
Father: ya vinarca YES FROM VIENNA
Daughter: ya YES
Mother: a fon vin vor si? FROM VIENNA SHE WAS?
The father here istalking about a peddler who had come to the house
to sell baskets. When thedaughter claims that this peddler had lice, the father
disputes her claim by asking in Slovenian, &dquo;What did she have? The daughter
answers in Slovenian, She had baskets and she had lice. When the father
says, I dont believe it, the mother breaks in in German, supporting her
daughter. The father replies in Slovenian. The mother repeats her statement
in German. The father replies once more in Slovenian. The daughter replies
in German. Finally, the mother ends by questioning the fathers statement
in German.
17
E. Message Qualification
Another large group of switches consist of such qualifying con-
structions as sentences and verb complements or as predicates following a
copula. Items six, seven and eight cited above illustrate this. Other examples
are:
29. English-Spanish.
A: Weve got all all these kids here right now. Los
...
que est4n ya criados aqul No los que estan recien

venidos de Mxico THOSE THAT HAVE BEEN
BORN HERE, NOT THE ONES THAT HAVE JUST
ARRIVED FROM MEXICO. They all understood
English.
30. English-Spanish.
A: The oldest one, la grande la de once anos THE.BIG
ONE WHO IS ELEVEN. YEARS OLD.
31. Hindi-English. College student conversation:

A: bina vet kiye itp d gae WITHOUT WAITING YOU
CAME?
B: nal NO,I came to the bus stop nau bis paccis par
ABOUT NINE TWENTY-FIVE.
3 2. Hindi-English.
A: nai, no, aegi zarur SHE WOULD CERTAINLY COME
because she said ki yadi mai n5i augi to IF I SHOULD
NOT COME THEN Ill ring you up and she hasnt rung
me up.
F. Personalization versus Objectivization

Lastly, in a very large group of instinces function is somewhat
more difficult to specify in purely descriptive terms. The code contrast here
seems to relate to such things as: the distinction between talk about action
and talk as action, the degree of speaker involvement in, or distance from, a
message, whether a statement reflects personal opinion or knowledge, refers
to specific instances or whether it has the authority of generally known fact.
Perhaps the best way to illustrate this is through more detailed discussion of
examples.
33. Slovenian-German Farmers discussion in the Austrian village.
Making plans for sharing machinery and dealing with problems
that might come up:
A: ala mormaya taka nadrit OK LET US DO IT LIKE
THIS dann von etwas is, noguat THEN IF SOME-
THING HAPPENS, OK FINE pa tola gax wikalna IF
SOMETIMES (THE MOTOR MUST BE) REWOUND
kost sibn zit hundart siling IT COSTS SEVEN OR
EIGHT HUNDRED SHILLINGS
18
B: ja ja paya danar tau OK, OK THEN THE MONEY IS
THERE
[Later on in the same discussion]:
A: yas sak leta dian oli ntpr I PUT IN EVERY YEAR OIL
kost virzen suing IT COSTS FOURTEEN SHILLINGS
A begins with a personalized statement, suggesting what the group

should do. He shifts to German upon mentioning a possible problem with
the arrangement, as if to imply that such things may happen without anyone
being at fault. Later on the cost of the repur is given in German, as is the
cost of the oil in the last statement. Perhaps the shift to German gives the air
of objective factuality to cost figures quoted.
34. Austrian village. Same situation as above. The discussion now

concerns origin of a certain type of wheat:
the
A: vigala ma ya z america WIGELE GOT THEM FROM
AMERICA
B : kanada prida IT COMES FROM CANADA
A: kanada mus i sogn nit I WOULD NOT SAY CANADA
Here B disputes As statement and A counters in German, as if to lend

his statement more authority.
35. Hindi-English. College student conversation:
A: anuradha ai? DID-ANURADHA COME?
B: She was supposed to see me at nine-thirty at Karol Bag.
A: Karol Bag?
B: or mai no baje gharse nikld AND I LEFT THE HOUSE
AT NINE
Bs English response to As Hindi question here treats the appointment

as an objective fact. B shifts back to Hindi in explaining his own actions.
36. Hindi-English. College girls talking about what a male friend said:
A: tera nam liya, nitiiktinizm liya HE MENTIONED YOU,
HE MENTIONED NITA
B: ahd kya kahne AH WHAT SHOULD I SAY shell be
flattered aj mdi leke a rahi thi na TODAY I WAS
GOING TO BRING HER YOU SEE
Here Bs shift to English in talking about Lipas feelings suggests that

the statement is a casual one, not implying personal involvement. B shifts
back to Hindi in talking about what she personally intended to do.
37. Spanish-English. Chicano professionals. A talks about her attempt

to cut down on smoking:
A: ... Id smoke the rest of the pack myself in the other
two weeks.
B: Thats all you smoke?
A: Thats all I smoked.
B: An how about now?
19
M: estos ... los hallb
me estos Pall Malls me los halla-
...
ron FOUND THESE PALL MALLS I

(THESE ...)
THEY WERE FOUND FOR ME No, I mean thats
all the cigarettes ... thats all. Theyre the ones I buy.
[Later in the same conversation]
A: they tell me How did you quit, Mary? I didnt
...
quit I I just stopped. I mean it wasnt an effort

...
that I made que voy a dejar de fumar por que me hace

dafio o THAT IM GOING TO STOP SMOKING
BECAUSE ITS HARMFUL TO ME OR this or that
uh-uh. Its just that I used to pull butts out of the
waste paper basket, yeah. I used to go look in the ...
se me acababan los cigarros en la noche MY CIGA-

RETTES WOULD RUN OUT ON ME AT NIGHT
Id get desperate y ahi voy al basurero a buscar a sacar,
AND THERE I GO TO THE WASTEBASKET TO
LOOK FOR SOME, you know.
Note how the code contrast symbolizes varying degrees of speaker

involvement in the message. Spanish statements are personalized while
English reflects more distance. The speaker seems to alternate between
talking about her problem in English and acting out her problem through
words in Spanish.
The above list, although by no means exhaustive, illustrates some of

the common uses of code-switching. The variety of interpretations which
result is much greater than one would expect from speakers descriptions
of language usage in terms of the simple we and they dichotomy. In-
formation conveyed varies greatly with context and discourse content- Yet
the same kinds of uses or functions tend to recur in what on. both linguistic
and social grounds are quite distinct situations.
The fact that it is possible to isolate such conversational functions

constitutes a convenient initial step in our analysis of code-switching. It
opens up the possibility of examining code-switching directly and discussing
the relevant problems of interpreting with members who ordinarily have no
words for referring to the phenomenon. If members agree on an interpre-
tation of a code-switched passage, one can assume that this agreement is
based on similar linguistic perceptions and then proceed to investigate code-
switching as part of the contextualization cues which give rise to these
perceptions.
Yet a list of functions cannot by itself explain what the linguistic bases
of listeners perceptions are and how they affect the interpretation process.
It is, of course, always possible to postulate extralinguistic social factors or
items of background knowledge which affect the incidence of switching, yet
to attempt to set up language usage rules which predict or reliably account
for the incidence of code-switching proves to be a highly complex task.
Consider the problem of quoted or reported speech. It is clear that not

all speakers quoted in the language they normally use. In item ten a Slovenian
20
bilingual quotes a German monolingual in German, but in eleven a Slovenian
speaker quotes a Slovenian-speaking neighbors remarks in German. One
might attempt to formulate a rule such as the following: Quote a message
in the code in which it was said. Examples ten, eleven, twelve and thirteen
might support this rule. Example sixteen, where a speaker reports on a con-
versation in English and shifts to Spanish for the direct quote, would seem
to illustrate the signalling value of the rule. But note that in item sixteen
Spanish is used both for reported speech and for the direct quote. In item
fourteen, moreover, the speaker tells about his trip to Agra in the they code,
switching to the we code to state that he talked to his brother, and switching
back to the they code for the quote. It might be said that this last item
reflects the actual language used but this does not explain why the we code
was used to introduce the quotation.
A detailed examination of the conversation from which example

twelve was excerpted reveals some further difficulties.
38. Speaker A begins with a Hindi question:

aur kya bola aukt r ne AND WHAT DID THE DOCTOR
SAY?
B replies, starting in English and shifting to Hindi for the
quote: _
he says ye hi medisin kantinyu- karo CONTINUE THIS

MEDICINE
B then goes on in English:
Hell see me on Monday. But he told me: You con-
tinue with this medicine.
A then counters in Hindi:
injek&scaron;-an usne lagd di DID HE GIVE AN INJECTION?
Whatever patterning there is in this type of code-switching cannot be
explained by simple rules relating conversational functions to instances of
code use.
The issue becomes even more complex when we consider that not all
functions listed are equally readily identified. Quotes, address specification,
interjections and to some extent repetitions, are for the most part marked
through both surface form and overt discourse content. But categories such
as message qualification, personalization and objectification are only rough
labels, at best, for a large class of stylistic or semantic phenomena. As we will
attempt to show later, members are likely to interpret we code passages as
personalized or reflecting speaker involvement and they code passages as
indicating objectification or speaker distance in examples such as items
33-37. But this does not mean that all we code passages are clearly identi-
fiable as personalized on the basis of overt content or discourse context
alone. In many cases it can be argued that it is the choice of code itself in
a particular conversational context which determines this interpretation.Thus,
rather than attempting to refine our classification of functions, so as to be
able to predict code occurrence, it seems more useful to take a more semantic
approach to code-switching and to examine how code-switching constrains
the process of conversational inference.
21
Code-Switching as a Pragmatic Phenomenon
The fact that in all three socially and linguistically distinct language
situations code-switching is used for roughly the same ends in similar dis-
course contexts, suggests that its contribution to the interpretation of
messages is independent of the internal grammatical structure of constitutent
codes. In all the examples we have collected it is the juxtaposition of two
alternative linguistic realizations of the same message that signals inforrnation,
not the propositional content of any one conversational passage. The contrast
may be overt as when the same phrase is followed by a linguistically different
but semantically similar string. In all cases speakers associate one of the two
alternative expressions with the casualness of intimacy of home or peer
group relations and the other with the formality or public or out-group re-
lations. The ultimate semantic effect of the message, however,derives from a
complex interpretive process in which the code juxtaposition is in turn eva-
luated in relation to the propositional content of component sentences and
to speakers background knowledge, social presuppositions and contextual
constraints.
Speakers Notions of Code

What is the linguistic knowledge that speakers must have to distinguish
the meaningful code juxtapositions from mere random alternations or idiosyn-
cratic alternations and to draw appropriate conversational inferences? We
have made a distinction between conversational code-switching, which builds
on participants perception of two contrasting systems, and borrowing, where
elements associated with one code are assimilated into the grammatical
system of the other. Once a passage has been identified as involving switching,
there remains the problem of determining the boundaries of switched pas-
sages, or of chunking speech in such a way as to assign constituent elements
to one or another code.
One might assume that this ability to delimit contrasting codes is pri-
marily a function of the grammatical distance between the two systems.
Everything being equal, it should be easier to identify code contrasts where
bilingualism involves two grammatically distinct and historically unrelated
languages than in cases of bidialectalism. But, as we have pointed out, the
bilingual phenomena we are concerned with are usually accompanied by
extensive convergence and structural overlap. Examples eleven to thirteen
show that this overlap may affect all levels of grammar. The perceptual
distinctness of potential code contrasts is further reduced by the fact that
we and they codes share a significant portion of their vocabulary. Since
code-switching styles are much more tolerant of borrowing than their mono-
lingual equivalents, items like gefixed and hiyo-me in examples nine and
thirteen, which, when judged in isolation, are usually found unacceptable,
nevertheless pass without notice as part of the borrowing code in conver-
sational context.
As a result of convergence and borrowing, sentences in language pairs

which, when seen from the monolingual perspective, seem quite distinct, may
on the surface appear to be almost identical. The following sentence discussed
22
in an earlier study of Hindi and Punjabi, as spoken by some college students
in Delhi (Gumperz 1971), is an extreme example:
39. Delhi Hindi: o nai kha-t-a HE DOESNT EAT

Delhi Punjabi: o nai kha-nd-a HE DOESNT EAT
The two codes here appear indistinguishable phonetically and almost

identical both in syntax and lexicon. To juxtapose such sequences in natural
conversation contexts, members must be sensitive to what, to the outsider,
may appear as quite subtle perceptual cues. What seems to happen in these
cases is that members respond to the presence or absence of certain key,
surface grammatical elements such as morphs participating in core inflectional
paradigms, as for example the Hindi participial affix -t- and its Punjabi equi-
valent -nd- in item 39 case endings, and agreement markers, or of certain
function words or phonological variables with very high frequency. Codes
are identified on the basis of expectation concerning the co-occurrence of
such key symbols of code separation with other aspects of message form.
These co-occurrence expectations have some similarity with syntactic

selection restrictions but what is involved here are not abstract semantic
features such as animate, inanimate or stative, non-stative, but surface charac-
teristics of the sentence such as the stylistic associations of particular lexical
items, the phonetic realization and prosodic and rhythmic characteristics of
message sequences. Example six where the English borrowing teacher serves
as part of the Hindi we code illustrates the seeming arbitrariness and conven-
tionality of co-occurrence judgements. There are two perfectly good Hindi
alternates which are quite frequent in other colloquial contexts and could
ideally have been used instead: the slightly more formal or literary adyapdk
or mastar (itself originally a borrowing, which by now has become an in-
tegral part of some forms of urban speech). Yet neither of these forms is
considered appropriate in this particular context. Similarly, in the second
English part of the message, the name of the city appears in its English-like
phonetic form Delhi; the local pronunciation dilli would have been in-
appropriate. Identification of codes, therefore, is only partly a matter of
what the linguist would consider grammatical knowledge. Only a subset of
the total inventory of grammatical elements and syntactic rules of a code
are critical; others are optional. Furthermore, the co-occurrence expectations
which tie criterial elements to surface style are matters of subcultural con-
ventions not covered in the ordinary grammatical description.
Syntactic and Pragmatic Constraints on Switching

Apart from knowing how to identify and keep apart we and they
codes, speakers must also have the ability to distinguish between meaningful
and non-meaningful code contrasts. We have shown that although most in-
stances of switchingcoincide with sentence boundaries, there are also a
number of common cases where elements of what on linguistic grounds
counts as a single construction are contrasted. What are the constraints which
govern this kind of intrasentential juxtaposition? Clearly, if code-switching
is meaningful, it must be subject to some forms of linguistic regularity and
we should be able to isolate instances of switching which for linguistic reasons
are not meaningful.
23
To investigate thenature of these constraints, a series of appropriate-
ness tests were devised. Some experimentation was necessary to work out
reliable testing procedures, in view of the difficulty of obtaining information
on code-switching in interview situations and the diversity of attitudes to
language usage. The usual linguists practice in which the investigator asks
for native judgements of hypothetical isolated sentences illustrating key
grammatical points clearly cannot work here, since what is judged are con-
versational inferences and the investigator and his subjects do not share the
same cultural presuppositions. On the other hand, to find sufficient naturally
occurring examples would have required much larger samples of free infor-
mation than can be conveniently collected. To avoid these difficulties key
passages were isolated from natural conversation and used as the basis of sub-
stitution frames in which certain test elements were systematically altered.
To provide an appropriate context for elicitation purposes these sentences
were pot quoted in isolation, but were first reintroduced into conversational
exchanges similar to those in which the originals had appeared. Substitutions
were then listed in sets and members were asked to rank them in order of
appropriateness.
The results were grouped in accordance with the syntactic relationship
among the juxtaposed construction. The test frames are given in the form
of their English equivalents with the, code-switched sequence italicized.
Sentences are arranged in order of acceptability. A double star indicates
that an item is completely unacceptable, a single star that items are simply
odd, unmarked items are acceptable. Unless specially noted all examples
hold for all three language pairs.
40. Subject-predicate constructions

Item My uncle Sam es el mas agabachado IS THE MOST
AMERICANIZED is a striking naturally occurring
example of this construction.
My uncle Sam from San Jose is the oldest
*
My uncle Sam
*
My uncle
That one
**
He
Any noun phrase can be switched here, except for the non-emphatic
personal pronoun he, which is clearly unacceptable throughout. The em-
phatic pronoun is marginally acceptable as is a simple noun phrase. On the
whole the longer the noun the more natural the switch.
41. Noun complement constructions

Thats the book the one that was lost
that was lost
**
Thats the lost book
**
Thats the book lost (E-Sp)
Here again acceptability decreases with decreasing length of the con-

trasting phrase. The simple adjective cannot be switched even if as in -
Spanish - it is postponed.
24
42. Object-embedded relative clauses
Thats the big car that I saw yesterday
**
Thats the big car I saw yesterday
When an English relative pronoun refers to the object of an embedded
sentence, it can optionally be deleted. In the English part of code-switched
sentences only the full form is possible. Pronoun deletion is clearly un-
acceptable.
43. Subject-embedded relative clauses.
The man who was here yesterday he didnt come today
** The man who was here yesterday didnt come today
When a relative clause is embedded in a subject phrase, this phrase

cannot stand alone; it must be followed by a personal pronoun.
44. Verb-Verb complement constructions
You should go to the field
**
You should to the field.
Verb complements can be freely switched. In Hindi and in the Austrian

villagedialect of German, complements can either precede or follow the main
verb. In code-switched sentences only post poned complements are possible.
45. Conjoined phrases .
I was reading a book and she was working

**
I was reading a book and she was working
I wanted to stop smoking but I couldnt
**
I wanted to stop smoking but I couldnt
John stayed at home because his wife was at work
**
John stayed at home because his wife was at work
Both co-ordinate and subordinate conjoined sentences can be freely

switched, but the conjunction always goes with the second switched phrase.
46. Verbs of propositional attitude
When a message is preceded by a phrase like think ....believe .. , etc., -
the switch can occur only after the performative verb.
He went to the field

I think he went to the field
**
I think he went to the field
My father is the oldest
I think that my father is the oldest
** I think that my father is the oldest
47. Two verbs of propositional attitudes
When a sentence contains two of the above phrases a switch can occur
either after the first or after the second phrase. Sequences after the switch
must all be in one code.
25
I think that he believes that my father is the oldest
I think he believes that my father is the oldest
**
I think he believes that my father is the oldest
48. When a phrase is seen as an idiomatic whole it cannot ordinarily

be a broken-up switch.
**
**
They like bread and butter
They like eating and drinking
but
They saw John and his brother is quite acceptable.
49. Gapping
Frank ordered beer, John ordered wine, Eric ordered
*
brandy
Frank ordered- beer, John wine, and Eric brandy
Switch phrases like the above, in which the main verb is not repeated,
are only marginally acceptable at best.
It is evident that, although code-switching can cut across many com-
mon syntactic relationships, it is nevertheless subject to at least some syn-
tactic constraints. This is also confirmed in a recent independent study of
Spanish-English code-switching (Timm 1975) which points to the ungramma-
ticality of switching in pronoun-verb constructions such as those in example
41, as well as in verb-pronoun object constructions (i.e. xx mira LOOK AT)
him), verb plus infinitive complement, auxiliary-verb and negative-verb con-
structions and others.
Our data suggests, however, that such syntactic constraints are in turn
motivated by underlying factors which depend more on certain aspects of
surface form and of pragmatics than on structural or grammatical charac-
teristics as such. For example, the pronoun-verb constructions which cannot
be switched and the noun phrase-verb construction which can be are both
of the NP VP type. Furthermore since emphatic pronouns like that one
are switchable, the pronoun-noun distinction does not alone account for the
constraint. The ease with which a sequence can be switched is most closely
related to the following factors: a) the length of the phrase or perhaps its
stressability or contrastability (see examples 41, 40) b) sequential unity.
Discontinuous sequences cannot be switched (44). c) Semantic or pragmatic
unity. Idiomatic units cannot be. broken (48). Conjunctions go with the
phrase they conjoin (45). Pronoun-verb sequences are more unitary than
noun-verb sequences (40). When a phrase has both an expanded and con-
tracted form only the former appears as part of .code-switched sequences
(42, 43, 49). When a sentence is dominated by a performative verb unit the
main clause acts as a single unit (46). d) The total number of switches
within any message sub-unit cannot be more than one (47).
To generalize, then, although the production of appropriate switches

in any one situation requires knowledge which is subculturally specific and
acquired only through practical experience, the process of switching at a
26
more governed by perhaps universal underlying con-
abstract level is also
straints, which bear similarity to the grammatical phenomena discussed
some
in the recent work on pragmatics. Switching in other words is a pragmatic
or perhaps stylistic phenomenon in which verbal sequences are chunked
into contrastable units. Any feature of the context or of the message itself,
such as, for example, the presence of a conjunction or an adverbial or a per-
formative phrase, that sets off one sequence from preceding or following
segments, favours the process or makes the alternation sound more plausible.
On the other hand, constraints on switching bear some similarity to the island
phenomena discussed by Ross in his analysis of syntactic movement rules
(1967).
Switching is blocked where it violates the speakers feeling for what
on syntactic or semantic grounds must be regarded as a single unit.
Some additional support for the pragmatic nature of switching pro-

cesses comes from longer conversational sequences, where there are no
syntactic constraints. In talking about his activities one morning, a speaker
in the Austrian community said:
50. yasn yutra stawa I GOT UP IN THE MORNING ya sn traktar

startawa I STARTED THE TRACTOR npa sn wizn nawarawa
AND THEN I TURNED THE GRASS
The three statements here all build up to a single activity. When the
passage was -repeated using various combinations of Slovenian (S) and German
(G) for the three sentences, the following combinations were judged marginal-
ly acceptable: SGG, SSG, GSS, and GGS. As one member put it: One could
say it that way, but it would sound funny. There is no real reason to say it
that way. Combinations such as SGS or GSG, where a speaker shifts twice
in what is seen as a single narrative sub-unit, on the other hand, are judged in-
appropriate. One speaker in fact exclaimed: Thats language mixing, no one
would do that.
In another conversation in the Austrian village a speaker talking about

the events of the previous evening said.
51. yasn &scaron;u lo kaisarya I WENT TO KAISERS pa ya siadu erix

cesaryu pa karli yanaxu mpa frantsi mimayu AND THERE SAT
ERICH OF THE CESAR FAMILY, KARL OF THE X JANACH
FAMILY AND FRANZ OF MIMIS FAMILY pa sma vina pila
AND WE DRANK SOME WINE.
The people mentioned here are all local village residents and are seen
as forming a single group. Any form of the list in which two members of the
group are listed in German is judged as odd. One speaker again referred to
this type of switching as language mxiture. If seems that longer passages are
subject to constraints on switching which are similar to those that hold within
sentence boundaries.
27
The Conversational or Situated Interpretation of Switched Passages
While the above examples show that conversational code-switching
contributes to meaning through stylistic juxtaposition of message elements,
they do not explain what this semantic effect is and how the various situated
interpretations relate to the speakers identification of the two varieties as
we and they codes. Conversations where the same message is said first in
one and then in the other code throw some light on this issue. Consider once
more the repetitions cited in examples 23-26:
52. Father talking to his five-year-old son, who is walking ahead of

him and wavering from side to side in the first class compartment
of a train in India:
A: Keep straight
Keep straight
sidha jao beta
WALK STRAIGHT SON
Adult talking to a ten year old who is practicing swimming in the
swimming pool:
B: bju-me beta ON THE SIDE SON
SON bju-me ON THE SIDE stay on the side!t
The two sequences were reversed so that the shift was from the we
code to the they code in A and from the they code to the we code in B.
Both sets of sequences were played and members were asked if the reversal in
direction of the code switch changed the meaning of the message. There was
clear agreement that the reversal does make a difference. The shift to the
we code was seen as signifying more of a personal appeal, paraphrasable as
wont you please whereas the reverse shift suggests more of a waming or
mild threat.
53. A Spanish-English sequence taken from a mothers call to

children:
Ven ach
Ven aca
Come here, you
was similarly interpreted as a warning by Spanish-English bilin-
guals, whereas the reverse:
Come here
Come here
Ven aca
was seen as a personal appeal.
Interpretation processes were analyzed in more detail with college
student Hindi-English conversations. The procedure followed was to isolate
key passages, change the English switches back to Hindi and ask members
familiar with the relevant rhetorical strategies to judge the two versions.
When the Hindiized versions were judged to be inappropriate, or at least
different in meaning from the original versions, judges were asked for more
detailed explanations of exactly what they thought speakers intended to
convey in each case. These descriptions were used as the basis for construct-
28
ing alternative paraphrasesor verbal expansions for each of the possible inter-
pretations. The conversations were then presented to a second group of
judges who were asked to choose an appropriate paraphrase. Our predictions
of what choices would be made proved right in all cases.
In the following example, a conversationalist who recently applied for

a job talks about his job interview. The preceding conversation is in Hindi.
54. A: ipkd intarvyu kaisd huwa? HOW DID YOUR INTERVIEW

GO?
After a hardly noticeable pause, when there is no immediate
answer, the same speaker repeats
A: How did your interview go?
Two possible interpretations for the question are: a) Tell me frankly,
how did the interview affect you? b).Give me a general impersonal account of
what went on. Members agreed on the second alternative. In other words
they interpreted the shift to English as signalling that what was wanted was
a casual reply rather than an indication of personal feelings.
In another case a speaker reports on a missed appointment with a

female acquaintance, who was to have accompanied him on a trip to town.
Talking in Hindi, he says that she had called him to say she would meet him
at the bus stop, but when he arrived she was not there. He then goes on as
follows: -
55. A: tiamarpur k~ bas sdmne khari thi THE TIMARPUR BUS

WAS STANDING BEFORE ME. Then I thought I might
as well take it.
Here both the all-Hindi and the Hindi-English version were judged as
potentially appropriate, but their meanings were seen as quite distinct. The
English version was interpreted as implying that the appointment in question
was a casual one, that there was no personal involvement. The Hindi version
on the other hand is seen as suggesting that the appointment was more in
the nature of a date and that he was annoyed at his friends not turning up.
Elsewhere a speaker is attempting to persuade his friends to change

their college program. He begins with a personalized statement:
56. A: t aplae kar de YOU SHOULD APPLY. iiiai bhi aplae kar
d. I WILL ALSO APPLY. di e es next year-me beth rahd
AM I WILL TAKE THE LA.S. EXAM NEXT YEAR.
Shifting to English he then continues.
A: Tell Rupa that Ashok is next year I.A.S. officer in any case.
The shift from Hindi to English signals that the last sentence reflects a
generally known fact and not merely personal opinion. The semantic effect
here is quite similar to that illustrated in examples 34 & 37.
29
Code-switching is thus more than simply a way of contrastively em-
phasizing part of a message. It does not simply set off a sequence from pre-
ceding or following ones. The direction of the shift also carries clear semantic
value. The oppositions warning/personal appeal; casual remark/personal
feeling; decision based on convenience/decision based on annoyance ; personal
opinion/generally known fact can be seen as metaphoric extension of the we//
they code opposition. What at the societal level are seen as norms of language
usage or symbolic affirmations of ethnic boundaries are transformed here and
built upon in conversation to affect the interpretation of speakers intent
and determine effectiveness in communication.
Sociolinguistic norms in this sense are more than just rules that are
either obeyed violated. They are an integral part of the knowledge that
or
a speaker must have to achieve his ends in interpersonal relations.
Perhaps the closest parallel in the recent linguistically oriented work on

conversational analysis to our code-switching phenomena is found in Searles
discussion of indirect speech acts (1975). Searle is concerned with messages
such as Do you want the salt?, which have the linguistic form or proposi-
tional content of one act, but can carry the elocutionary force of another. His
model of interpretation suggests that a listener in such cases notes that the
literal meaning does not seem to apply, and then turns his knowledge of
felicity conditions which determine the successful enactment of an act as a
source for further conversational inferences. Thus, in our example, the
addressee decides the message cannot be interpreted as a question about his
ability to perform the act. Knowing that to be able to do something is a
preparatory condition for fulfilling a request, he concludes that the question
may be meant as a request.
Searle goes on to point out that idiomaticity, or expectation about

the proper form and mode of enunciation, play an important part in the
interpretation of speakers intent. Yet the examples of indirectness he gives
all take the form ot interences based on semantic judgements. What we
would like to suggest is that the phenomena we have discussed here are
examples of idiomaticity judgements at work. Determination of speakers
intent when seen in these terms is always a matter of evaluating both the
content and the form of what is said. In monolingual situations optionality
at the level of form is in large part a matter of prosody, rhythm and voice
quality. In code-switching situations optionality extends to matters of gram-
mar and phonology and this increase in the range of options carries greater
potential in communicating indirect meanings.
To say that code-switching is meaningful, however, does not mean that
a switch can be assigned a single interpretation in any one case. Indirect in-
ferences differ from the processes involved in the communication of simple
factual information. What is generated are preferred or possible interpre-
tations, not clear and uncontestable meanings. Judgements of intent are
situated, subject to change as additional information becomes available
and thus never invariable. The idiomaticity judgements a speaker makes
suggest possible lines of interpretation, or favor certain chains of inferences
over others. To the extent that the communicative conventions which govern
30
these judgements are learned through socialization into particular networks
of relationship and into ethnically specific traditions, a persons ability to
interpret what others say and to be effective in communication with them
is also a function of his social background.
Conclusion
We have argued that the incidence of code-switching in human groups,

its persistence over time, its linguistic characteristics and its function in
human interaction can best be explained by treating it as a communicative
phenomenon, in which speakers build on others knowledge of alternative,
grammatically distinct, subsystems and of the social assumptions these imply,
to generate indirect conversational inferences. If this position is valid, it has a
number of important implications, for our understanding of the functioning
of human signs in communication, of social symbols in verbal interaction and
of the role of speech variation in human groups.
The classic views of linguisticdecoding, diverse as they are in detail, all

assume a basically undirectional, step by step process of abstraction in which
the sound stimuli received are interpreted in turn by phonological, syntactic
and semantic rules to arrive at meaning. What grammar does, in other words,
is to enable speakers to separate meaningful from non-meaningful verbal
signs. Only those aspects of the vocalizations a speaker emits are semantically
significant, which can be shown to have a place in the system of rules that
make up the grammatical process. If a certain phonetic feature, an intona-
tional contour, or a string of words is subject to formal linguistic analysis,
it is said to contribute to the interpretation of message content. If not, it is
either disregarded as irrelevant or accounted for by other non-linguistic types
of explanation.
In theory, to be sure, it is the goal of modern linguistics to account for

all the perceptual distinctions which speakers regard as significant. Yet limi-
tations of currently available elicitation procedures, the difficulties, such as
those illustrated in this paper, inherent in finding out what members really
believe, as well as the constraints imposed by accepted analytical paradigms,
act to limit the range of verbal phenomena that are taken into account. Thus,
in practice, at least, meaningfulness tends to become associated with a gram-
matical analyzability. The result is that speech and code alternation, certain
aspects of sentence rhythm and prosody, as well as speakers reliance on idio-
matic or stock phrases in conversation, tend to be treated as different in kind
from ordinary grammatical phenomena. Their significance in communication
is discussed by resorting to analytical concepts such as the notions of socio-
logical variable or social function, which were originally developed in the
study of human population aggregates and do not necessarily apply to inter-
action (Gumperz 1973), or by relying on psychological attitude theory.
The scope of grammars has been extended greatly in recent years and
many of the earlier notions regarding the separation of linguistic from non-
linguistic phenomena have been abandoned. It has been shown that social
presuppositions play an important part in understanding speech. Most im-
portant of all, the definition of meaning has been broadened from reference
31
to speakers intent. Yet the basic view of linguistic decoding as a unidirec-
tional progression from sound through syntax to meaning remains. Even the
work in conversational analysis which is directly concerned with indirectness
in conversation seems to regard indirectness as resulting almost entirely from
lexical or grammatical ambiguity inherent in the message. It does not account
for semantic phenomena of the type discussed here.
Our analysis suggests that theories of conversational inference will have

to be modified to take care of the fact that aspects of sentence form (such as
the code in which a sentence is spoken in the case of bilingual speech or its
phonetic realization in the
case of monolingual styles) can directly affect its
interpretation.In other words, such features of surface structure play a role
in communication which is equivalent to that of performatives discussed in
the recent work on pragmatics, in signalling the illocutionary force of mes-
sages. Our models of verbal communication must reflect this fact.
The assumption that at the level of conversational inference linguistic
signs serve both to communicate contextual presupposition as well as content,
provides a relatively simple explanation. Contextual information, in other
words, is not given, but communicated as part of an interpretation process
which has as its input linguistic contextualization cues along with extra-
linguistic setting, knowledge of participants and related factors. This inter-
pretation process then generates the presuppositions in terms of which
speakers decode the details of what is said. In cases of situational usage the
role of language in this signalling mechanism is relatively simple. Whenever
a code or speech style is regularly associated with a certain class of activities,
it comes to signify or connote them, so that its very use can signal the enact-
ment of these activities even in the absence of other clear contextual cues.
Component messages are then interpreted in terms of the norms and symbolic
associations that apply to the signalled activity.
The case of metaphorical usage is much more complex. The signalling

mechanism involved is a shift in contextualization cues, which is not accom-
panied by a shift in topic, and in the other extralinguistic context markers
that characterize the situation. This partial violation of co-occurrence ex-
pectation then gives rise to the inference that some aspects of the conno-
tations, which elsewhere apply to the activity as a whole, are here to be
treated as affecting only the illocutionary force and the quality of the speech
act in question. Whereas situational usage relies largely on lexical knowledge,
metaphorical usage, as we have shown, requires additional syntactic and
conformance to co-occurrence expectations. While this knowledge is grammar-
like, it nevertheless differs from what we ordinarily mean by grammatical
knowledge in that a) it operates at the context-bound level of conversation,
b) it is subculturally variable and thus available only to some speakers of the
languages in question and c) it generates possible inference rather than ex-
plicit and overtly discussable meanings. It is in this sense that it is perhaps
best to treat it as sociolinguistic knowledge.
The fact that metaphorical switching presupposes the ability to identify

code distinctions has important bearing on the discussion concerning Saus-
sures well known concepts of langue LANGUAGE and parole SPEECH*.
32
Although there is general agreement about the usefulness of the distinction,
what is meant by the two terms, especially by the second langue, continues
to be highly controversial. Theoretical linguists tend to see langue as a highly
abstract set of rules while other more socially oriented scholars see it in
Durkeimian terms as the aggregate or perhaps vector sum of the processes
of change in a statistically significant sample of speakers. (Labov 1973)
In both views language is regarded as a separate system, independent

of overt individual behavior, at any one time. The study of code-switching
exchanges leads to the conclusion that members have their own socially
defined notions of code or style. Although such notions are often sub-
stantially different from those derived through linguistic analysis or taught
in standard grammars, it is nevertheless clear that, in situations such as we
have discussed, effective speaking presupposes sociolinguistically-based
concepts of system. Speakers rely on these notions to categorize and lump
together sets of grammatical rules at various levels of structure, to relate
speech to non-linguistic environment and to generate indirect conversational
inferences.
Turning finally to the issue of social symbolism. Analysis of code-

switching exchanges provides evidence for the existence of underlying un-
verbalized assumptions about social categories, which differ systematically
from overtly expressed values or attitudes. It suggests empirical methods
for studying the working of such symbols and the role they play in per-
suasion and rhetorical effectiveness. More detailed studies of conversational
processes along these lines might provide important new insights into the
functioning of broader social concepts in interpersonal relations.
Note
Field work for this paper was supported by National Science Founda-
tion grant No. GS 30546. Arpita Agrawal and Sarah Wikander assisted in the
collection of conversational examples. I am grateful to Julian Boyd, Susan
Ervin-Tripp, Celia Genishi, Eduardo Hemandez-Chavez and Paul Kay for
comments on earlier drafts. Many of the Spanish-English examples derive
from an earlier study by Gumperz and Hernandez-Chavez (1971).
References
Bernstein, Basil. 1971. Class codes and control. London: Routledge and
Kegan Paul.
Blom, Jan Petter and John J. Gumperz. 1972. Social meanings in linguistic
structures. In John J. Gumperz and Dell Hymes (eds.) Directions in
sociolinguistics. New York: Holt, Rinehart and Winston.
Ferguson, Charles. 1964. Diglossia. In Dell Hymes (ed.) Language in culture
and society. New York: Harper and Row.
33
Fishman, J. 1967. Bilingualism with or without diglossia. In J. McNamara
(ed.) Problems of bilingualism, Journal of Social Issues 23.2: 29-38.
Gumperz, John J. 1971. Language in social groups. Stanford: Stanford

University Press.
Gumperz, John J. and Eduardo Hernandez-Chavez. 1971. Bilingualism,
bidialectualism and classroom interaction. In Courtney Cazden, Vera
John and Dell Hymes (eds.) The functions of language in the classroom.
New York: Teachers College Press.
Haugen, Einar. 1973. Bilingualism, language contact and immigrant languages

in the United States. In Thomas Sebeok (ed.) Current trends in lin-
guistics, Vol. 10, pp. 505-591. Mouton: The Hague.
Khubchandani, Lachman. 1974. Personal communication.
Labov, William. 1971. The notion of system in creole lan -ges. In D. Hymes
(ed.) Pidginization and creolization of languages. Cambridge: Cambridge
University Press.
Mitchell-Kernan, Claudia. 1969. Language behavior in the black community,
Monograph II. Language Behavior Research Laboratory, University of
California, Berkeley.
Reyes, Rogelio. 1974. Studies in Chicano Spanish. Typescript.
Ross, John R. 1967. Constraints on variables in syntax. Bloomington, Indiana:
Indiana University Linguistics Club.
Sapir, Edward. 1949. Language. In David Mandelbaum (ed.) Language,

culture and personality. Writings by Edward Sapir. Berkeley: University
of California Press.
Searle, John R. 1974. Indirect speech acts. In Peter Cole and Jerry L. Morgan
(eds.) Syntax and semantics, speech acts. New York: Acadernic Press
Inc.
Timm, L.A. May 1975. Spanish-English code switching: El Porque y How-

not-to. Romance philology.
34

Gumperz 1977

Uploaded by

Copyright:

Available Formats

Gumperz 1977

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gumperz 1977

Uploaded by

Copyright:

Available Formats

The Sociolinguistic Significance of

By conversational code-switching, I refer to the juxtaposition of pas-

1. Chicano professionals in California, exchanging goodbyes (Sp-E).

2. A college student in India, telling an anecdote (H-E).

to us-ne pronauns kiyd apne vais-se (SO HE PRONOUNC-

3. Family conversation in Slovenian village in Austria, talking about

A. toca kuarbca ya mewa (SHE HAD SUCH BASKETS)

Often code-switching also takes place within a single sentence as in

4. Go and get my coat aus dem Schrank da (OUT OF THE CLOSET .

8. Those are friends from Mexico que tienen chamaquitos (WHO

Here phrases with the internal characteristic of two distinct grammatical

When residents of a small North Norwegian town were asked to recall

When political ideology changes, attitudes to code-switching may

Until quite recently pejorative attitudes to code-switching were also

Code-switching is perhaps most frequent in the informal speech of those

The bilingual exchanges we have examined furthermore show that code-

The Empirica~ Study of Conversational Switching

Because of this,. there are a number of empirical difficulties in describing

Depending on such factors as region of origin, local residence, social

There is evidence to show that most bilinguals have at least a compre-

ground assumptions. Bilinguals, in fact, ordinarily do not use code-switching

borracha porque me siento mas, (AND WHEN I AM WITH PEOPLE

Labov describes the switching in this passage as idiosyncratic behavior,

Perhaps a more fruitful way to visualize the issue of social regularities

A number of scholars have noted that the speech of closed network

Note that network position is only partly a matter of ethnic group.

It is not necessary therefore to turn to larger samples of text of extra-

The rest of the passage, if we disregard the initial sentence, which is

Code-Switching in Three Language Situations

The first is an Austrian village of farmers and laborers located along

Knowledgeof cultural values and social factors affecting language use

Speakers of Chicano Spanish in California are in part descendants of

All three societies represent situations of change, characterized by both

As an initial step in our analysis, an attempt was made to study the

10. Slovenian-German From an informal business discussion among

11. Elsewhere in the same discussion a speaker reports as follows on

12. Slovenian German Village woman talking with neighbors about

13. Hindi-English From a conversation among young Hindi-speaking

14. Hindi-English From a conversation among Hindi-speaking college

16. Spanish-English. Later in the above situation, the speaker is

To this day he says that uh its a shame that they

uno y THEY ARE LIKE DONKEYS. SOMEONE TALKS

17. Slovenian-German. Informal conversation about the weather in

18. A group of Hindi-speaking graduate students are discussing the

19. A Hindi-speaking student couple is talking to a Hindi-speaking

21. Spanish-English. A is talking to someone else later on in the same

relistas en el teltono (STELLA AND SANDI ARE VERY

22. Slovenian-Gennan Austrian village conversation.

baju me jdo betd, ndr mt GO TO THE SIDE, SON, NOT

27. Spanish-English. Puerto Rican mother in New York calling to

28. Slovenian-German Austrian village family conversation. Talking

que est4n ya criados aqul No los que estan recien

31. Hindi-English. College student conversation:

F. Personalization versus Objectivization

A begins with a personalized statement, suggesting what the group

34. Austrian village. Same situation as above. The discussion now

Here B disputes As statement and A counters in German, as if to lend

Bs English response to As Hindi question here treats the appointment