See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/249840228

From codeswitching via language mixing to fused lectsToward a dynamic

typology of bilingual speech

Article in International Journal of Bilingualism · December 1999

DOI: 10.1177/13670069990030040101 · Source: OAI

CITATIONS READS

715 7,084

1 author:

Peter Auer
University of Freiburg
367 PUBLICATIONS 10,380 CITATIONS

SEE PROFILE

All content following this page was uploaded by Peter Auer on 26 March 2014.

The user has requested enhancement of the downloaded file.

InLiSt No. 6
Interaction and Linguistic Structures

From Code-switching via Language Mixing

to Fused Lects: Toward a Dynamic Typology
of Bilingual Speech
Peter Auer

Freiburg i. Br.
September 1998

This paper is based on a lecture given at the University of Constance in honour of Prof.
Dr. Aldo di Luzio on the occasion of his retirement (February 1998) and on an invited
lecture given at the First International Symposium on Bilingualism, Newcastle, April 10-
12, 1997. I would like to thank Li Wei, one of the organizers of the conference, for his
encouragement to turn the lecture into writing. Previous versions have also been
presented and discussed at the Universities of Berlin (FU, JFK-Institute) and Barcelona
(Univ. Autonoma). Thanks also to Elizabeth Couper-Kuhlen, Monica Heller and Yaron
Matras for critical and helpful comments.
 In recognition of the enthusiasm he has brought to all
aspects of the study of spoken verbal interaction,
we dedicate this series to Professor Dr. Aldo di
Luzio, University of Konstanz.

_____________________________________________________________________________
Prof. Dr. Elizabeth Couper-Kuhlen Prof. Dr. Margret Selting Prof. Dr. Peter Auer
Dr. Susanne Günthner Universität Potsdam Albert-Ludwigs-Universität Freiburg
Universität Konstanz Institut für Germanistik Deutsches Seminar I
FG Sprachwissenschaft Postfach 60 15 53 Postfach
PB D 180 D-14415 Potsdam D-79085 Freiburg i. Br.
D-78457 Konstanz
_____________________________________________________________________________
Additional copies may be ordered from:
Universität Konstanz, Sekretariat LS Angl. Sprachwissenschaft, InLiSt
PB D 180, D-78457 Konstanz, Tel.: +49/7531/88-2552, fax: +49/7531/88-4157
http://www.ling.uni-konstanz.de/home/couperku/inlist-index.html

II
 Dedicated to Aldo di Luzio on the occasion
of his 65th birthday

Introduction

Over the last few decades, a wide range of phenomena have been described in which
two languages are juxtaposed in discourse and/or within a sentence, variously called
language alternation, code-switching, code-mixing, etc. It seems worthwhile (and
possible) at this stage of research to consider the ways in which these phenomena may
be subject to a typology. The present paper aims at such a typological approach. A
continuum of language alternation phenomena will be presented which spans out
between three well-documented cases (conceived as prototypes) which will be labelled
code-switching (CS), language mixing (LM) and fused lects (FLs), with CS and FLs
representing the polar extremes of the continuum and LM a point inbetween. Since
these three prototypes have been amply documented in the literature,the continuum
rests on relatively secure empirical grounds. However, I will also suggest an
interpretation of it which is somewhat more tentative, i.e., to see the continuum CS —>
LM —> FL as a case of structural sedimentation which some might call
”grammaticalization”. Particular attention will therefore be given to the transitions, CS —>
LM and LM —> FL. The possibility of such transitions has been hinted at, in particular,
by Scotton 1988 who suggests that ”overall switching as an unmarked choice seems to
be the first step to what has been called the development of a semi-autonomous ‘Mix’”
(165), for which she also uses the term ”fused variety” (158). Generally speaking,
however, particularistic, differential and historical studies of bilingualism have often been
hidden under the more universalistic interests dominating the past decades of research.
For this reason, little is known about the dynamic aspects of speech in individual
bilingual communities over a period of time.

The terms CS, LM and FLs will be used in the following way: CS will be reserved for
those cases in which the juxtaposition of two codes (languages) is perceived and
interpreted as a locally meaningful event by participants. The term LM, on the other
hand, will be used for those cases of the juxtaposition of two languages in which the
use of two languages is meaningful (to participants) not in a local but only in a more
global sense, i.e. when seen as a recurrent pattern. The transition from CS to LM is
therefore above all an issue to be dealt with by interpretive sociolinguistic approaches
since it is located on the level of how speakers perceive and use the ‘codes’ in question.
Stabilized mixed varieties will be called fused lects. The transition from LM to FL is
primarily an issue for grammatical research; essential ingredients of this transition are a
reduction of variation and an increase of rule-governed, non-variable structural
regularities.1

1
1. Code-switching

(Conversational) code-switching defines the pragmatic pole of the continuum to be

discussed. In CS, the contrast between one code and the other (for instance, one
language and another) is meaningful, and can be interpreted by participants, as indexing
(contextualizing) either some aspects of the situation (discourse-related switching), or
some feature of the code-switching speaker (participant-related switching). Discourse-
related code-switching is part of the realm of everyday rhetorics, not part of grammar (the
scholastic grammarians´ ars obligatoria). It is one of the available strategies used b y
bilinguals to convey meaning. As a contextualization strategy, it represents a
metapragmatic comment2 on the on-going interaction which marks it it as bilingual.
Participant-related code-switching covers instances of diverging language preferences
and competences.

Code-switching (understood in this sense) has been dealt with by many scholars (cf.
the pioneering work by Gumperz 1982:59ff, as well as Li Wei 1994, 1998; Alfonzetti
1992, 1998; Auer 1984, 1995; Sebba/Wootton 1998, among many others). Its
significance must not be equated with the ‘social meaning’ of the various languages
within a multilingual repertoire, as it is often believed to exist in diglossic situations (such
as language A being the ‘we code’, ‘informal code’, ‘solidarity code’, etc., and language B
being the ‘they code’, ‘formal code’, ‘deference code’, etc.3). Although the languages
involved in code-switching may index some kind of extra-conversational knowledge,
code-switching can never be analysed as a mere consequence of such indexing,
without taking into account the sequential position in which it occurs and from which it
receives its meaning. The following illustrative example is taken from di Luzio´s analysis
(1984:67) of reproaches and teasing among children and youngsters in the
Italian/German group of speakers also investigated in Auer (1984):

Ex. (1) (Italian underlined; transcription follows GAT4)

01 Ag.: noja am spielplatz

(well in the playground)
02 Cl.: am (spielplatz)
(in (the playground))
03 Ag.-m: wissen Sie wo der spielplatz isch?
(do you know where the playground is?)
04 m: das is:[:
(it is: :)
05 Cm.: [der kann ja h[ingehe
(he can pass by there)
06 m: [äh::
07 Ag.-Cm.: se: mo va ccamená propri´ [llá
(oh I´m sure he will pass by there)

2
The switching in line 07 changes the ‘footing’ of the interaction; while the adolescents
Ag., Cm. and Cl. have been trying to explain to adult m. the location of a playground - a
purely ‘technical’, reference-establishing sequence without any emotional overtones - ,
Ag.´s Italian dialectal utterance in this line is directed at Cm.´s and responds to his
previous suggestion that m. should ´pass by´ there in order to see himself. In retorting in
Italian dialect, Agostino marks his reply as ironical (‘there is no reason why an adult
should spend his time looking for playgrounds’), but also as reproachful and as a mild
critique of Camillo´s utterance. In his analysis, di Luzio shows that such switching into
Italian dialect occurs regularly in this group of friends. It cannot, however, be derived
from whichever ‘meaning’ one may want to attribute to this variety within the speakers´
repertoire.

The prototypical case of (discourse-related) code-switching can be portrayed as

follows: (a) it occurs in a sociolinguistic context in which speakers orient towards a
preference for one language at a time; i.e., it is usually possible to identify the language-
of-interaction which is valid at a given moment, and until code-switching occurs; (b)
through its departure from this established language-of-interaction, code-switching
signals ‘otherness’ of the upcoming contextual frame and thereby achieves a change of
‘footing’. The precise interpretation of this new footing needs to be ‘filled in’ in each
individual case, although previous episodes may also be brought to bear on the
interpretation of the case at hand; (c) it seems possible to describe the mechanisms b y
which code-switching relates to the two codes and to the context in which it occurs in
very general ways. Contexts are theoretically innumerable, of course, as are the
interactional meanings of code-switching; however the ways in which these meanings
are construed remain constant from one community to the next; (d) code-switching may
be called a personal or group style. As a group style, its use may be subject to
normative constraints valid within a speech community; however, it certainly is not a
variety in its own right; (e) most code-switches occur at major syntactic and prosodic
boundaries (at clause or sentence level). Since switching serves to contextualize certain
linguistic activities, the utterance units affected by the switch must be large enough to
constitute such an activity. For this reason, code-switching does not provide much
interesting data for syntactic research; (f) although code-switching bilinguals may be
highly proficient in both languages, balanced proficiency is by no means a prerequisite.
Indeed, code-switching is possible with a very limited knowledge of the ‘other’
language.5

In order to ascertain that in a particular case of the juxtaposition of two co-occurring sets
of structural parameters we are dealing with code-switching, it is essential to show that
speakers orient towards this juxtaposition. Therefore the question of what counts as a
code must refer to participants’, not to linguists’ notions of ‘code A’ and ‘code B.’ An
‘objective‘ statement (i.e., one exclusively informed by the ‘linguistic facts’, such as (the
absence of) phonological or morphological integration, or frequency) that a given
3
arrangement of signs constitutes a combination of elements of two systems is not only
very difficult to make at times (cf. the Castilian/Galician example discussed in Auer
1998), it is also irrelevant. There may be cases in which the two codes juxtaposed are
‘objectively speaking‘ very similar, but regarded by the members of a bilingual
community as completely separate (as in some cases of dialect/standard switching; cf.
Alfonzetti 1998 for Sicily), just as there may be codes which are ‘objectively speaking‘
very distinct but nevertheless seen as non-distinct by the speakers (see section 2).6

A methodology to prove participants´ orientation at the juxtaposition is precisely to show

that it is used as a contextualization cue (i.e., that it is ‘functional’). Further evidence is
provided by self- and other-corrections of language choice (cf. Gafaranga, MS) as they
may also be found in the data from which ex. (1) is taken:

ex. (2): (same participants as in (1); Italian in italics; transcription according to GAT)

01 m: di ch’ di che cosa parlate general[mente=

(what do you usually talk about together)
02 Ag.: [<pp>tutt´ cos´ =
(all kinds of things)
03 <mf> della scuOl´ =
(about school)
04 m: <f>della [scuola
(about school)
05 Ag.: [was wir WERde wolle (-) [alor cos devo: (-)
(what we want to be I mean what I need to)
06 Al.: [compagn
((about) our mates)
07 Ag.: <p>devo (1.0)
(I need to)
08 m: cosa (d )
(what (n ))
09 Ag.: <p>devenDAre;
(become)

In this part of the conversation, the adolescents have temporarily accepted adult m´s
language choice (standard Italian); this is demonstrated by the fact that Ag’s momentary
‘excursion’ into German in line 05 is immediately self-repaired, although the Italian version
is not without linguistic difficulties for the speaker (who runs into trouble finding the
standard Italian infinitival form for dial. devendá ~ std. diventare ‘to become’). Together
with language negotiation sequences in which participants (try to) find a common
language-of-interaction (cf. Auer 1984: 13ff, 1995: 128ff , in prep.), such repairs prove
participants´ orientation to a preference for one language at a time.

To summarize, although we are used to approaching conversational code-switching from

the presumption that there are two codes (languages) which are used alternatedly, and
then proceeding to investigate the function switching between them may have, a strictly
interpretive approach forces us to state the question the other way round. We need to

4
start from the observation that there are two sets of co-occurring variables between
which participants alternate in an interactionally meaningful way, and then proceed to
ask whether we can see them as belonging to or constituting two varieties or languages
(cf. Alvarez-Cáccamo, 1998, in prep.). The ‘codes’ may turn out to be ‘languages’, but
they may also be ‘dialects’ or other varieties, or even sets of prosodic contextualization
cues.

One final point needs to be made. The prototypical case of code-switching sketched
above represents the alternational type: one in which a return after the switch into the
previous language is not predictable. There is another type of code-switching where
this is not the case and which may be called insertional.7 In this type of switching, a
content word (noun, verb, rarely adjective/adverb) is inserted into a surrounding
passage in the other language. As in alternational switching, participants show an
orientation towards the ‘other-languageness’ of the insertion, either by deriving some
particular interactional meaning from it, or by relating it to the speaker´s (momentary)
incompetence in the established language-of-interaction. In both cases, prosodic cues
(extra emphasis, preceding pause) and verbal markers (metalinguistic comments,
hesitation) may serve to underline the juxtaposition and turn it into a locally noticeable
phenomenon.8 Note that the insertion may be morphosyntactically fully integrated; or it
may carry over grammatical elements into the receiving language.9 The communicative
function of insertions (and their status as CS) does not depend on its grammatical
format.

2. Language mixing

The range of phenomena covered by the term CS according to the interpretive

approach chosen here is considerably smaller than often suggested by the pervasive
usage of the term in the literature. From this, it follows that there are many cases of the
juxtaposition of two languages other than code-switching. One such case of language
contact is ‘mixing’ (of languages, dialects, etc.).

From the early times of research onward data for the conversational juxtaposition of two
languages have been presented which clearly do not fit the prototype of CS as
presented in section one. This is true, for example, for the ”frequent code-switching”
(particularly of the non-emblematic type) investigated in the pioneering work by Poplack
(1979 [1981]) on Puerto Rican bilingual language use in the USA, but also for most of
the data collected in Africa (for instance, Scotton´s ”code-switching as the unmarked
choice”, 1993a).

5
Ex. (3) is taken from an European context and from the same language pair as ex. (1),
(2), which were discussed in the preceding section as examples of code-switching.
Nevertheless, we are clearly dealing with a case of LM here:
Ex. 3: (Preziosa Di Quinzio 1992: X, quoted from Franceschini 1998:59f)
[Italian immigrants in Switzerland, Swiss German dialect and Italian (underlined); author´s
transcription conventions]

p11: perchŽ meinsch che se tu ti mangi emmentaler o se tu ti mangi una fontina isch au
(Ôbecause, you mean, if you eat Emmental cheese or if you eat Fontina cheese, there is also

en unterschied, oder? schlussŠndlich • sempre dentro l“ per˜ il gusto isch andersch.

there is also a difference, isnÕt there? Actually, itÕs still there, but the taste is differentÕ)

p6: • vero!
(ÔthatÕs right!Õ)
.
. ((ommission))
.

p11: es git verschiedeni fondue aso die heisset verschiedŠ, aso ja das isch en
(Ôthere are different kinds of Fondue, they have different names, well thereÕs a

himmelwiitŠ unterschied se prendi questo o se prendi il chŠs normal.

huge difference whether you take that one or whether you take the ordinary cheese.Ô)

p6: ehrlich! beh , zum biispil io raclettechŠs lo prendo sempre fresco. raclettechŠs
(Ôreally! well, for instance me, Raclette cheese I always get it fresh. Raclette cheese

hol ich immer im dings... Šs git au im migros cos“ implasticato gits au.
I always get at what's-its-name... they also have it at Migros, wrapped in plastic they have it, too«)

In this case of the juxtapositions of Italian and Swiss German dialect it is difficult if not
impossible to say whether the language of interaction is Italian or Swiss German dialect;
rather than one of the varieties involved, it seems to be their alternating use which in
itself constitutes the ‘language’-of-interaction. It is equally difficult to argue that the
juxtaposition of the two languages triggers a change of footing or is related to the
competences or preferences of the speakers on each occasion; these juxtapositions do
not seen to have local meaning, i.e. from an interpretive point of view, they cannot be
called code-switching (although speakers will be well aware of the bilingual language
mode in which they converse).

The fact that in LM of the type exemplified by (3) individual turns cannot be labelled
language A or language B is mainly due to the frequency of turn-internal language
juxtaposition. Since LM does not contextualize linguistic activities, such juxtaposition
may affect units of any size, typically not only at clause boundaries but also below. LM
is therefore much more intricately linked to syntax than CS. However, most researchers
agree that it is not the case that ‘anything goes’ in LM; rather, the ways in which the two

6
languages in play may be intertwined are subject to certain constraints. It is these
constraints that most syntactically oriented research on bilingualism has focused on.

As a consequence of the frequent intrasentential juxtapositions of the two languages it is

often difficult to maintain the distinction between insertional and alternational
juxtapositions in LM. In fact, it is a typical feature of LM that alternational and insertional
strategies converge. Yet, the distinction does not always collapse completely. Rather, it
is often possible to identify mixing styles of a more insertional kind and those of a more
alternational kind. Extract (3) above is clearly of the alternational type (and, incidentally,
obeys the equivalence contraint suggested by Poplack 1979 [1981] which may be
typical for alternational LM in general), and so are the bilingual data presented b y
Zentella on Puerto Rican bilingual speech (1997:15ff, cf. in particular her example on p.
117). On the other hand, an example such as the following is clearly on the insertional
side:10

Ex. (4) (from Bentahila & Davies 1995:83)

[younger generation of Moroccan speakers; French in italics; clause numbers added;
authors´ transcription conventions]

(1) hadu les cousins djali z&ajjin men la France w ¿andhum

(these (the) cousins of mine were coming from France and they had)

la voiture ...(2) m¿a la voiture djal xali ... (3) merra la

(a car with the car of my uncle one day the)

plage, merra z&z&ebel, merra la forêt, kul merra w fin.

(beach, one day the mountains, one day the forest, every time
somewhere different)

(4) ¿andna fih des photos, derna les photos bezzaf. (5) On a
(We have photos taken there, photos a lot We even)

même filmé, (6) hakka wlad ¿ammi z&abu la caméra.

(made a film. So my cousins brought the camcorder.)

(7) filmaw.(8) filmana bbahum f lbhar. (9) filmana même fe

(they filmed Their father filmed us in the sea. He filmed us even)

ttriq f la ville. (10) z&abna ma¿na xir rrabbi djal les

(in the street in the town. We brought back with us a lot of)

souvenirs. (11) wahed ssuq ¿andhum, (12) men daks&i ¿andhum,

(souvenirs.They have a market there what a market they have)

(13) il est immense. (14) ¿andhum ... (15) huwwa un seul

(it´s enormous They have it´s all one)

souk walakin divisé en parties, (16) bhal fih la partie djal

(market but divided into sections for example there is the )

7
 Vi les poissons. (17) haduk ... (18) kajnin les pêcheurs.
(section only for fish those there are fishermen there)

(19) Les pêcheurs enfin makajketru la journée djalhum f

(The fishermen in fact don´t spend the day of them at)

lbhar, (20) kajz&ibu dak ... (21) duk les poissons frais,
(sea, they bring that those the fresh fish)

(22) w katelqaj l¿jalat Vi kajs&riw ... (23) Vi les poissons,

(and you find women buying only only the fish)

(24) Vi lfenn.
(only the best)

It is intuitively clear that the ‘matrix language’ (to use a term coined by C.M. Scotton) of
most clauses in this passage is Moroccan Arabic.11 Into these grammatical frames, single
French words, particularly nouns (and occasionally discourse particles/adverbs such as
enfin, 19, and même, 9) are inserted, which are not part of this variety of Arabic (i.e.,
they are nonce borrowings, not integrated borrowings). Nouns take with them certain
grammatical elements (in particular, the definite and indefinite articles), sometimes also
their modifiers such as adjectives (cf. clauses 21, 15). French verbal stems may be
transferred as well (cf. clauses 7, 8, 9) and can be integrated into Moroccan Arabic
morphology. For all these reasons, the extension of an insertion is not necessarily
restricted to its lexematic (V, N) ‘core’, but may affect larger (NP) or smaller units (stem).
The only cases of alternational LM (or perhaps CS) are in clauses 5, 13 and 15. Only in
the latter case is it difficult to assign a matrix language to the clause.

Most researchers on LM in the speech of the African elites agree that in their data, a
matrix language can be identified; this implies a dominantly insertional mode of LM (cf.,
among many examples, Scotton 1996b, Haust 1995, Swigart 1992, Goke-Pariola 1983,
Gafaranga, MS). Speakers may look upon insertional mixing as a variety of the
language into which elements are inserted, i.e. the ”matrix language” (as argued b y
Gafaranga MS, Swigart 1992), although this variety may be looked down upon in some
cases. LM of the insertional type is also claimed by Backus 1996, writing on LM in
Dutch-Turkish bilinguals.

Not all examples are unambiguous, however; a more complex case of both insertional
and alternational LM (combined with CS) are, for instance, Luther´s table conversations.
As is well known, switching between Latin and the vernacular language (here: Early
New High German) was a wide-spread practice among intellectuals in the 16th century:

8
Ex. (5): (Luthers Tischreden , Veit Dietrich’s notes, Dec. 1531, No 122)

[Early New High German in italics]

(1) Qui autem vexantur Spiritu tristitiae, inquit, debent

(but those who are troubled by the spirit of sadness, so he [sc.
Luther] said, should be)
summe cavere, ne sint soli, (2) denn Gott hat societatem
(most careful that they are not alone for God created the society of)
ecclesiae geschafft et fraternitatem gebotten, (3) sicut
(the church and called for brotherhood as)
scriptura dicit: ‘Vae homini soli, quia cum cediderit etc.’
(it is written: Woe to the lonely man, for when he dies etc.’)
(4) Tristitia quoque cordis coram Deo non placet ei,
(For before God, sadness of the heart is (something) not pleasing to
him)
quanquam tristitiam coram mundo permittat,
(even though he may permit sadness before the world)
(5) sed non vult, das ich gegen yhm betrubt sol sein, (6)
(but he does not want that I should be distressed before him)
sicut dicit: ‘Nolo mortem peccatoris’; item: ‘Laetamini in
(he says: I don´t want the sinner´s death; and: Rejoice in )
Domino.’ (7) Non vult servum, der sich nit guts zu yhm
(God. He does not want a servant who is not on good terms )

versehe, (8) haec scio, (9) sed wol zehen mal in einem tag
(with him. This I know but surely ten times a day)
wurd ich anderst zu sinn, (10) et tamen resisto Satanae.
(my mind changed and still I resist Satan.)

In one of the first and most brilliant studies on the syntax of bilingual speech, Stolt
(1964) shows that insertional mixing (which she calls Einschaltung) such as in clause 2
follows different regularities from alternational mixing (her Umschaltung) such as in
clauses 5 or 7; Luther freely inserts Latin nominal and verbal elements into his German
sentences, while German insertions into Latin sentences are rare and restricted in type
(cf. clause 9, where the adverbial/discourse particle sed is preposed to a German
sentence). Alternational switching on the other hand can go in either direction.

The very fact of selecting a mixing mode from the repertoire (to the exclusion of other,
more ‘monolingual’ modes) can of course be of social significance; for instance, it may
signal group identity.12 Since LM is ”such an integral part of the community linguistic
repertoire that it could be said to function as a mode of interaction similar to monolingual
language use” (Poplack 1988:217), speakers may also contrast this mixing mode with
other (mixing or ‘monolingual’) modes within the repertoire; we then find locally meaningful
alternation, i.e. in code-switching, of a second order (cf. Meeuwis and Bloomaert 1998 for
discussion and an example).

The fact that LM, other than CS, sometimes has a folk name can be attributed to its
identity-related functions; for instance, the LM mode employed by the speakers of ex.
(2) is called Italoschwyz by its users (Franceschini 1998), Gibraltarians call their
9
English/Spanish mixing Yanito (cf. Moyer 1988), the Gurindji Aborigines of the Victoria
River District in the Nothern Territory of Australia their mixing of Gurindji (a Pama-
Nyungan vernacular) and Creole/Aboriginal English mikijimap (‘mix-im-up’) (McConvell
1988:97), younger aboriginal people in the Western Torres Strait in Australia their LM of
Torres Strait Pidigin and Kalaw Lagaw Ya ap-ne-ap (‘half and half’) (Bani 1976), etc.
Bilinguals and monolinguals often have very strong positive or (more often) negative
attitudes towards mixing (cf., e.g., Hill & Hill 1986 for the second case).13 Even among
linguists, LM seems to provoke strong reactions, usually of the negative kind (Singh
1996:71ff may be the most recent example).

Compared to CS, LM seems to require a higher bilingual competence; in addition, there

is some evidence that alternational LM requires more proficient bilinguals than insertional
LM (Poplack 1979 [1981], Backus 1996, Bentahila & Davies 1996). On the continuum
from CS to FL, the most balanced bilinguals are likely to be found here.

3. From CS to LM

Having sketched the prototypes of CS and LM, we can now proceed to look at the
transition from CS to LM. The hypothesis is that both on the level of the individual and
on that of the community, there is a tendency to move from CS to LM, but not in the
opposite direction.14 (Note that I do not claim that CS —> LM is the only way in which a
mixed speaking style may emerge, but only that LM cannot develop into CS.) For
instance, the Italian/German mixing found in ex. (3) represents as a later stage of
Italian/German language contact after migration than the switching style in ex. (1) and
(2).15

In the transition, phases occur in which the ‘older’ CS pattern and the ‘newer’ LM pattern
co-exist; these phases may be quite prolonged. For this reason, one should be careful
not dismiss too early the possibility of discourse-functional switching in interactions in
which mixing is observed: one does not exclude the other.16 ”Campus Swahili” in Dar es
Salaam (Tanzania) seems to be one such speaking style:

Ex. (6) (from Blommaert 1992:61)

[English in italics; transcription converted into GAT]

A: manake (.) shule hata CHAkula hawapati siku hizi

(because the schools don´t even get)

cha (.) chakula standard hawapatikani kule (-) basi

(standard food it can´t be obtained there so)

10
 imebidi WAwithdraw (-) wanasema <↑>ooohh nijilimie
(it was necessary for them to withdraw they say ”oh!”)

mwenyewe (-) na kila mwanafunzi wanacholima hakitoshi

(I´ll work in the field myself. But with all the students
farming, even that isn’t good enough)

<↑>even for TWO months (-) basi (.) mambo

(so things)

yame deteriorate (-) basi unakuta (.) ndiyo (-)

(have deteriorated and so there you are, yes)

B: sasa hivi wana (.) wanaanza kuaminii (.) lakini sasa

(right now they they start to believe but now)

haija pickup (.) haija pick (.) ni hali ambayo kwa kweli
(it hasn´t picked up yet, it hasn´t picked yet; it´s a situation
which really)

it is STILL deteriorating (-) lakini kwenye

(but in a)

situation kama hiyo hata kama umepata nafasi ya kusoma

(situation like this, even if you get the oaccasion to study)

nafikiri it´s just low (-) they <↓>can´t go on (-)

(I think)
((etc.))

As Blommaert argues, this bilingual mode of speaking is made up of two different ways
of juxtaposing Swahili and English. On the one hand, there are many cases of
insertional mixing (which he calls borrowing) (cf. standard, withdraw, deteriorate, pick up,
situation) which do not seem to carry any local meaning. In many ways, they
correspond to the ‘unmarked’ way of speaking among the Tanzanian élites for which
”speaking ‘pure’ Swahili seems to require special attention and effort” (Blommaert
1992:61), although corresponding terms in Swahili partly exist and are used on other
occasions (cf. B´s Swahili hali in line 8 and English situation in line 10). On the other
hand, there are repeated cases of CS (in lines 5, 9, 11) which Blommaert analyzes as a
conversational strategy used to add special emphasis to certain turn components, or to
conclude a statement.

The Tanzanian case represents a fairly wide-spread way in which CS and LM co-
occur: insertional contact phenomena are LM, alternational phenomena are CS. More
generally speaking, insertions often seem to precede alternations on the way from C S
via LM to FL. The opposite case (i.e., insertional switching co-occurring massively with
alternational mixing) seems to be very uncommon.

11
How can a locally meaningful contextualization strategy such as code-switching lose its
pragmatic force and turn into mixing? It is clear that the frequency of the juxtaposition of
two codes within a speaker´s turn plays a role here: frequent juxtaposition weakens the
contextualization value of this cue. In gestaltpsychological terms, the figure of code-
switching is most salient against a ground which is monolectal. The more frequently
code-switching occurs, the less salient it becomes; as a consequence, the potential for
using it in locally meaningful ways is diminished.

This, in turn, raises the question of why intra-turn code-switching should become more
frequent at a certain point in the development of a bilingual community. From a
sociolinguistic point of view, at least two answers suggest themselves, one couched in
more negative, one in more positive terms. The more negative one is that the transition
CS —> LM may start when speakers feel obliged to resort to strategies of neutrality in
an increasing number of cases; i.e., when they have social reasons to produce turns
which cannot be assigned to one language or the other unambiguously. Heller (e.g.,
1988) has drawn attention to this ”strategic ambiguity” achieved by code-switching and
analysed its social functions in various publications on French-English switching in
Canada, in which she argues that intra-turn code-switching ”can allow the simultaneous
accomplishment of tasks through conversation and the management of conversation and
of personal relationships through the avoidance of the conflict which categorical language
choice would entail” (1988:82).

A more positive reason for frequent code-switching is that at a given point, the identity-
related purposes of this style may become more important than the discourse-related
tasks code-switching has served so far. The prevalent scenario for such a reevaluation
of functions is one in which a bilingual group needs to define its own identity vis-a-vis
both contact groups (for instance, in the case of the Italian migrants in Germany or
Switzerland, both against Italian and against Swiss/German society). The transition is
complete as soon as participants no longer actively avoid speaking one language or the
other in an attempt to be ambiguous with respect to the language-of-interaction, but
rather positively orient towards (i.e., accommodate) the mixing style of the other
speaker.

Typical alternational switches spearheading the transition from CS to LM are those that
serve to structure turns internally, such as CS on emphatic repetitions and summaries
(cf. ex. 6), CS for different ‘voices’ (including reported speech), CS in order to
contrapose material of high or low relevance (such as asides or parenthetical remarks),
CS for marking personal vs. objective passages, etc. (Many examples for such a
transitional stage may be found in Sebba and Wootton´s 1998 account of switching
between London English and London Creole.) Trigger words (Clyne 1967) such as
diamorphs of the two languages (including established borrowings) may play a role in
the development of a non-functional, mixing style as well.
12
Insertional switches typical for the transition from CS to LM are those that are not
restricted to lexemes (nouns, verbs) but spread to larger constituents in which this
lexeme is embedded (such as, typically, a noun or verbal phrase, a verb plus its
´incorporated´ object, etc.; cf. the examples in ex. 4). Thus, the distinction between
alternation at a point in time and insertion of a unit of speech dissolves, on the one
hand, because alternational switching on its way to alternational mixing no longer
questions the language of interaction (which is ambiguous anyway), on the other hand,
because insertional switching spreads out to larger constituents.

4. Fused Lects

The difference between mixing and fused lects17 is mainly a grammatical one;
pragmatically speaking, neither type of language contact is locally meaningful. On the
surface, a FL may look similar to LM. Often, the difference becomes visible at a deeper
grammatical level only. While LM by definition allows variation (languages may be
juxtaposed, but they need not be), the use of one ‘language’ or the other for certain
constituents is obligatory in FLs; it is part of their grammar, and speakers have no
choice. Thus, structural sedimentation (grammaticalization sensu Givón 1979) of ML into
a FL presupposes a loss of variation and the stabilization of function-form relationships.
Comparing the FL grammar with that of the two languages involved, this means
simplification, since alternatives are lost. But in addition, structures from language A and
B which are more or less equivalent in monolingual use may develop specialized uses in
the fused lect AB. Also, fused lects may have to adapt structurally to the massive
combination of elements from A and B by developing new structures identical neither to
those of A nor B (see below, section 5).

Speakers of a fused lect AB may but need not be proficient speakers of A and/or B.

Good candidates for full-fledged FLs (but certainly not the only ones) are so-called
”mixed languages” (cf. Bakker & Mous, eds., 199418), at least those which develop in
second and third generation speakers after first generation mixed marriages between
colonizing men and indigenous women have occurred. The best-known example of such
a mixed language is probably Michif (Bakker 1997), the language of the Métis Buffalo
hunters in Canada, in which almost all nouns are said to be taken from French, while
Cree provides most of the verbal structures (with the exception of the French copula).
Other examples of fused lects (which presumably emerged as closed-group languages,
certainly not as dual ancestry languages) are many of the European Romani dialects.
The following extract is from the Rómanes dialect (Sinti) spoken by the Hamelner Sinte
(Gads&kene Sinte, Germany), who have been biligual in Romani and German for many
generations:19

13
Ex (7): (from Holzinger 1993:324f)

[author´s transcription conventions and translation into German; German underlined]

((extract from a longer narrative))

Ap jek kopo del ir kamlo dad ap mande gole: ‘Tŝava, dŝa vri,
o grai hi los!’
Auf einmal ruft mir mein seliger Vater zu: ÔBub, geh raus, das Pferd ist los!Õ
(All of a sudden my blessed father calls out at me: Son, go out, the horse is loose!)

Na, rivom i xoleb an, dŝajom vri.

Na, ich zog die Hose an, ich ging raus.
(So, I put on my trousers, I went out.)

Taprom i graies, pandevom es pale an.

Ich packte das Pferd (und) band es wieder an.
(I grabbed the horse (and) I tied it up again.)

Me homs noch nicht an o vurdi dre, his o grai ŝon pale los.
Ich war noch nicht im Wagen drin, (da) war das Pferd schon wieder los.
(I was hardly back in the car, the horse was loose again.)

Tŝava, hoi kerdal denn, pandal i graies gar richtig fest? Dŝa
vri, pande i graies fest!’
ÔBub, was hast du denn gemacht, hast du das Pferd nicht richtig festgebunden? Geh raus, binde
das Pferd fest!Õ
(Son, what have you done, didn«t you tie the horse up properly? Go out, tie up the horse.)

Me pre, taprom pale i graies, pandevom les pale an, kerdom

knote dre, menschenunmögliche knote.
Ich (stand) auf, packte wieder das Pferd und band es an, ich machte Koten hinein, všllig
unmšgliche Knoten.
(I (got) up, grabbed the horse again and tied it up, I tied knots, completely impossible knots.)

Dŝajom dre an o vurdi, tek pantŝ minute, o grai pale los.

Ich ging rein in den Wagen, keine fŸnf Minuten, (war) das Pferd wieder los.
(I went back into the car, no five minutes, the horse was loose again.)

((...))

Und koi his ŝ vaiso ap leste.

Und da war Schwei§ auf ihm.
(And there was sweat on him.)

Draus i naseloxe phudehs o grai.

Aus den Nasenlšchern schnaubte das Pferd.
(The horse snorted out of its nostrils.)

O khas tel leste his vegŝlaifedo, futer grabrehso.

Das Heu unter ihm war weggeschliffen, so sehr hatte es gegraben.
(The hay had been grouond away from under him, so much he had dug.)

Ach tŝave, pandel miro dad kote fest und me kate fest.
Ach Jungs, mein Vater bindet es dort fest und ich hier fest.
(Oh boys, my father ties it up there and I tie it up here.)

14
((etc.))

Apart from numerous German nominal, verbal and adjectival stems (wegschleifen,
graben, Nasenloch, Schweiß, Knoten, menschenunmöglich) which are partly integrated
into an older Romani morphology, there are many German particles, conjunctions and
adverbials inserted into Romani (noch, denn, und, ach); for these, there are no
equivalents of genuine Romani origin in this variety. In addition, this Romani dialect has
borrowed various converbs from German corresponding to the separable verbal
prefixes of that language (an, fest, los ), various prepositions (such as draus, used as a
preposition here but an exophoric adverbial - daraus - in standard German) and modal
verbs (not documented in the extract; but cf. Holzinger 1993:82ff). What looks like a
case of insertional LM is in fact quite different from ‘mixing’, for the ‘German’ elements in
this (and other Sinti) dialect(s) are an obligatory part of its grammar and lexicon.

As outlined before (section 2), the distinction between alternation and insertion is far less
clear in LM than in CS; nevertheless it often remains possible to distinguish more
insertional from more alternational types of LM. In FLs, all the available evidence
suggests that they can only be of the insertional type. (This not only applies to
example (7), in which Romani clearly is the matrix language, but also to Michif and other
”mixed languages”; the fact that no verbs are taken over from French is easily explained
by the agglutinating grammar of this language which makes verbal insertions into Cree
matrices virtually impossible.)

5. From LM to FL

In this section it will be argued that, given the appropriate sociolinguistic context, there is
a tendency in communities of speakers who code-mix to further constrain the
possibilities of juxtaposing the two languages and to develop functional specializations.
LM therefore usually does not develop towards a loss but rather towards an increase of
linguistic structure. Note again that we do not claim that in each and every case FLs can
be traced back to LM; there are certainly other origins (such as relexification). The
contention is, rather, that the development FL —> LM is not possible.

In order to investigate the transition LM —> FL one can either look at FLs and uncover
where they originate, or look at LM moving towards a FL and try to establish the
mechanisms which prompt this transition. Starting from established FLs such as the so-
called ”mixed languages”, it is revealing that most specialists in this field hypothesize
prior mixing. For instance, Bakker (1994: 22) notes that the fused patterns of Michif are
very similar to LM between other Algonquian languages such as Montagnais and

15
French.20 He concludes from this observation that Michif may have originated among
bilingual children with a slight dominance of Cree, ”as a humorous extension of code-
mixing in the community” (1994:23). Golovko (1994: 118), who investigates Copper
Island Aleut (a fused variety of Russian and Aleut), also believes that ”code-mixing is a
necessary but not sufficient condition for the emergence of a mixed language”. (The
sufficient condition for him is the sociolinguistic need of the Creoles to delimit their own
group against the Aleuts.). Scotton (1993:220ff) assumes that the mixed Bantu
language Ma´a (Mbugu) goes back to code-mixing21 between a Kushitic matrix language
and Bantu insertions with subsequent change of the matrix language.

These statements are, of course, plausible guesses rather than empirically based; in
FLs such as Mbugu, Copper Island Aleut or Michif the transition from whatever prior
stage they may have originated from is already complete. For this reason, it is also
useful to look at incipient transitions of the type LM —> FL. There are certain indications
as to where processes of this type start.

One of these starting points are relatively unbound elements of grammar, such as
discourse markers, conjunctions and certain adverbials which typically serve the function
of modalizing the utterance (”utterance modifiers” in the sense of Matras 1998), i.e. they
have a discursive function rather than a referential one. Ex. (8) demonstrates this
development for a variety of American German as spoken by immigrants from Germany:

Ex. (8) (Sabine Behling, Hamburg, unpublished interview data)

[Low German, Northern Standard German (italics), American English (boldface); A =

interviewer, B = interviewee]

B: see, wir sind ausnandergegang, see.

(‘you see, we went to different places, you see’)
du gehst hierhin, du machst dein lebn (-)
(‘you go this way, you make your living’)
now Kurt, der is ungefŠhr so hunnert miles from Chicago,
(‘Now Kurt, he is about hundred miles from Chicago’)
wo er gewohnt hat da und hat sein geschŠft da (.)
(‘where he lived and had his business’)
und Erwin auch, see.
(‘and Erwin as well, you see.’)
Erwin kann auch nich viel PlattdŸtsch schnacken.
(‘Erwin can´t speak much Low German either.’)
A: nee, aber hŠi versteht dat.
(‘No, but he understands it.’)
dat [hŠtt hŠi mi sŠ[cht.
(‘So he told me.’)
B: [ ja/ [ja, er versteht, [because er/
(‘yes yes he understands because he’)

16
 A: [aber hŠi kann dat nich selbst
schnacken.
(‘but he can´t speak it himself.’)
B: see, Šh (.) sin mudder un vadder de hebt ja Ÿmmer tosomm platt/
plattdŸtsch schnackt.
(‘you see, ehm (.) his mother and his father always spoke Low/ Low German
together.’)
A: jo.
(‘yes’)
B: see, de hebt Ÿmmer tosomm plattdŸtsch schnackt.
(‘you see, they always spoke Low German together.’)

Salmons’ (1990) analysis of Texas German, shows that 3rd-6th generation speakers
almost exclusively use American English discourse markers and conjunctions (such as
see, now, because in ex. (8)). At the same time, all German particles, discourse markers
and conjunctions have been lost, i.e. one system of discourse and text organisation has
been replaced by another. It does not seem to be of any relevance whether German
and American particles/conjunctions/discourse markers are equivalent on a one-to-one
basis or whether they function quite differently (cf. because/weil (denn) for the first
case, see and the German modal particles for the second case).

Another example for a complete replacement of a particle subsystem of one language

by another (and thus a case of fusing) can be taken from Unserdeutsch (Rabaul Creole
German), the German ‘creole’ that originated in Papua New Guinea during German
colonial times in a missionary school among the children of indigenous women and
German colonialists. Rabaul Creole German is best described as a simplified variety of
German with influences from English, German dialect/vernacular, and, above all, Tok
Pisin (cf. Volker 1982). In some cases, code-mixing between ‘German’ and English, and
subsequent structural adaptation and sedimentation took place. For instance,22 RCG
borrowed the English answering particles yeah and no , which, however, are used
according to the Tok Pisin system which diverges both from German and English usage;
thus, there is no special particle for negative answers to negative questions (such as
German doch ), and the answer paralleling the question is (unlike in English) always
yeah :

Ex. (9): Rabaul Creole German [Volker 1982:38]

Question: du weiß ni wir?

‘kennst Du uns nicht?´
you don´t know us?

agreeing anwer: yeah, i weiß ni euch.

‘nein, ich kenne Euch nicht’
no, I don´t know you
disagreeing answer: no, i weiß eu
‘doch, ich kenne Euch’

17
 yes, I know you

In order to decide if in a given community or speaker certain discourse markers,

adverbials or conjunctions have become ‘grammaticized’, the following questions need to
be asked: (a) Are the insertions the only resources available for the speaker(s) or do
they compete with corresponding ones in the surrounding language? (b) If there are
competing resources provided by the surrounding language, are they used
interchangeably (i.e., without a difference in meaning), or have they rather specialized in
function? (c) Have both the insertions and the competing resources of the surrounding
language retained their meaning and/or function when compared to monolingual usage?
In the examples from Texas American and Rabaul Creole German, question (a) clearly is
to be answered positively, and there can be little doubt that fusion has occurred.23 A
more subtle development is implied when the answer to questions (b) and (c) is
negative. For instance, various investigations of bilingual communities have found
functional specialization of markers/conjunctions going back to language A and those
originally from language B.

One such case is reported by Oesch Serra (1998) and concerns the
connectives/markers mais, ma and però in the bilingual speech of Italian migrants in
French-speaking Switzerland. Oesch Serra demonstrates that the discourse functions of
(Italian) ma and però and those of (French) mais in bilingual speech differ from
monolingual usage. The differences point to a functional specialization and therefore a
reduction of possible contexts of usage. One of Oesch Serra´s findings is that, due to
the fact that there are now three connectives available (while Italian has only two and
French just one), an ordering principle has developed which could not exist in the
monolingual varieties; this principle organises the hierarchy of arguments and places ma
on the lowest level of a scale of argumentative strength, followed by mais and finally
però, which is used to introduce the strongest argument.

A similar case of a functional specialization was found by Maschler (1998) in a study on

Hebrew/English bilingualism. Maschler, too, analysed the emergence of a fused variety
in the structural field of discourse markers and connectives. Her main finding is a
separation of conjunctions and discourse markers: discourse markers overwhelmingly are
in Hebrew, conjunctions overwhelmingly (i.e., with the exception of structural markers) in
English.

Borrowing of content words (nouns, verbs, adjectives) of course represents the best-
known of the continua from a bilingual into a monolingual mode. It is easy to follow the
route of integration into the receiving language system from purely discourse- or
competence-related insertional CS, meaningful on each single occasion of occurrence,
via nonce borrowings for which no such local interpretation can be given, to established
borrowings. The final step on this route is the use of the inserted materials to the

18
exclusion of the equivalent forms in the surrounding language. But obviously, such
borrowing alone is not enough to speak of a FL since it only superficially affects the
language system (i.e., its lexicon) but as such has no consequences for grammar. (In
fact, it may be argued that integration of borrowings provides evidence for the intactness
of the receiving language´s grammatical system and the absence of fusion.) However,
massive lexical borrowing can have repercussions that go beyond the adding of certain
lexical elements to the lexicon of the receiving variety; this restructuring may represent a
step towards a FL. For instance, the insertion of other-language materials often requires
the development of special grammatical formats. Backus (1996) presents an in-depth
analysis of such a format which has developed in the variety of Turkish spoken in the
Netherlands24 and based on the Turkish verb yapmak (‘to make/do’) which is used in
order to accommodate Dutch infinitival verbs in a Turkish sentence frame (cf. ex. 10).
(Because of the agglutinating morphology of Turkish, the insertion of bare Dutch stems
or words is avoided.)

Ex. (10) [Backus 1992/1996:278]

bir sŸrŸ taal larõ beheersen yapõyorken

many language-PLUR master (INF) make/do-while
Ôwhile s/he spoke many languagesÕ

The Dutch infinitival form beheersen (together with its direct object, taal ‘language’) is
preposed to the Turkish verb yapmak which carries the necessary affixes. In this
variety, so Backus argues, yapmak has been grammaticalized into an auxiliary (or even
suffix). Evidence for this grammaticalization comes from a comparison with Turkey
Turkish where yapmak is (almost) exclusively employed as a nativization strategy for
nouns but not for verbs, and where other verbs or affixes are used for adapting foreign
verbs (such as the verb etmek, as in stop etmek ‘to stop’, or the suffix -lE as in
faxlamak, ‘to fax’); thus, variation is reduced in the Netherlands. Semantically, Dutch
Turkish yapmak is bleached; it may be combined with practically all full verbs of Dutch,
including non-perfective ones (despite the fact that Turkey Turkish yapmak is a
perfective verb).25

A third phenomenon which may indicate transition to a FL is (obligatory) double marking

of certain grammatical functions by functionally equivalent but structurally divergent
strategies. Such double marking is a well-known exception from the generalization that
grammatical elements should always be taken from the matrix language (cf. the
discussion in Myers Scotton 1993b: 110f, 132f). Ex. (11) presents a case from bilingual
language use in the 16th century:

19
Ex. (11): (Luther’s table conversations; LbTb. 3854; Vol. III, p 660)

mit den ecclesiis haben sich ... geschlagen

with Art.DEF.DAT. PL. churchesDAT.PL. have ... battled
Ôwith the churches ... they foughtÕ

Although the matrix language of this sentence is clearly German, and although the
inserted Latin noun ecclesia receives case and number marking from the definite German
article which precedes it, this marking is doubled by the Latin suffix -is which agrees with
the article.

In Luther´s table conversations, such double marking is frequent but not obligatory, and
should therefore be taken as an indicator of fusing. In other cases of language contact,
however, double marking has become compulsory, i.e. it is part of fused grammar. A
case in point is a Uzbek-Tajik ”mixed variety” spoken in Samangan (Afghanistan) and
investigated by Boeschoeten (1983):

Ex. (12) [Boeschoten & Backus 1997:57]

tå sinp-i s&as& gac&a

until grade six TERMINATIV(until)
Ôuntil grade sixÕ

Here, a case relation is marked both by a Turkic suffix and a Tajik preposition. Although
grammaticalization is involved, the fused variety is not simplified, but rather more
complex than the non-fused languages.

Other than in the case of the replacement of one system of ”utterance modifiers” b y
another, whether or not fusional processes such as double marking occur depends on
the grammatical type of the two languages in contact. For instance, double marking for
case seems more likely in a pre-/postmodifying language pair than in a language pair in
which both languages express case through prepositions.

20
 alternational code-mixing fused lects
code-switching (alternational)

+
- preference for one language
at a time code-mixing - as before
- functional qua alternation - additionally
- involves renegotiation of (insertional) positive
language grammatical
of interaction constraints
- functional as group
- usually at syntactic clause - no choice for
style
boundary speaker
- not functional qua
- rhetorical/stylistic device for (obligatory)
+ alternation
speaker - partly within sentence - adaptive changes
but towards new
- not restricted to words overall system
insertional
- no preference for one
switching language at a time
- no language of
- as above but interaction
- on small constituent - variable but negative
- does not threaten grammatical con-
language straints
of interaction

pragmatics grammar

21
7. Conclusion

I have argued in this paper that three forms of the juxtaposition of two varieties or
languages in bilingual speech should be distinguished: code-switching, mixing and
fused lects. The diagram on the following page summarizes their main features.

In addition, I have argued that there is a ‘cline’ from pragmatics/discourse to grammatical

structure, which links the three prototypes. CS presupposes liberty of the individual
speaker, it is a contextualization device which can be used in creative ways b y
participants; FLs, on the other hand, presuppose positive structural regularities. Code-
mixing is a frequent type of bilingual speech between these two extremes in which the
juxtaposition of the two languages lacks pragmatic-stylistic function and in which
grammatical structure not-yet sedimentated (as shown by variation, non-obligatory
regularities, and negative constraints).

Although moving from from CS in the direction of FLs is the ‘natural’ tendency (while the
opposite movement FL —> LM —> CS is prohibited), such a move does not
necessarily take place; rather, a bilingual community may stabilize on a certain point on
the continuum. In order to find out about the sociolinguistic correlates of movement or
non-movement on the continuum, more differential studies will be needed particularly on
differences within one bilingual community (but cf. Backus 1996, Li Wei 1994, Bentahila
& Davies 1991/1995; Blommaert 1992, Poplack 1988 for steps in this direction).

22
References
Agnihotri, Rama Kant, 1980, Processes of assimilation: a sociolinguistic study of Sikh children in
Leeds. Ph.D. Thesis, University of York, UK.
Alfonzetti, Giovanna, 1992, Il discorso bilingue. Pavia: Francoangeli.
Alfonzetti, Giovanna, 1998, The conversational dimension in code-switching between Italian
and dialect in Sicily, in: P. Auer (ed.) Code-switching in conversation. London:
Routledge, pp 180-214.
Alvarez-Cáccamo, Celso, 1998, From 'switching code' to 'code-switching': Towards a
reconceptualisation of communicative codes. In: P. Auer (ed.) Code-switching in
conversation. London: Routledge, pp 29-50.
Alvarez-Cáccamo, Celso, in prep., Communicative codes and speech varieties in Galizan-
Portuguese/Spanish conversation. Paper presented at the First International
Symposium on Bilingualism, Vigo, 21-25 October 1997.
Auer, Peter, 1984, Bilingual Conversation. Amsterdam: Benjamins.
Auer, Peter, 1990. A discussion paper on code alternation. In: European Science Foundation
(ed.), Network on Code-Switching and Language Contact. Papers for the Workshop on
Concepts, Methodology and Data, Basel, January 12-13, 1990, pp 69-89.
Auer, Peter, 1995, The pragmatics of code-switching: a sequential approach. In: L. Milroy and P.
Muysken (eds.), One Speaker, Two Languages. Cambridge: CUP, pp 114-135.
Auer, Peter, 1998, Introduction: "Bilingual Conversation" revisited. In P. Auer (ed.) Code-
switching in conversation. London: Routledge, pp 1-24.
Auer, Peter, in prep., Why should we and how can we determine the ”base language” of a
bilingual conversation? Paper presented at the First International Symposium on
Bilingualism, Vigo, 21-25 October 1997.
Backus, Ad, 1996, Two in One. Bilingual Speech of Turkish Immigrants in The Netherlands.
Tilburg University Press.
Bakker, Pieter 1994, Introduction, in: Bakker & Mous (eds.) 1994.
Bakker, Pieter & Mous, Maarten eds., 1994, Mixed Languages. 15 Case Studies in Language
Intertwining, Amsterdam: IFOTT.
Bakker, Pieter, 1992, A Language of Our Own. The Genesis of Michif. Ph.D. Thesis, University
of Amsterdam. (= A Language of Our Own. The Genesis of Michif, the Mixed Cree-
French Language. New York: OUP, 1997)
Bani, E., 1976, The language situation in the Torres Strait, in: P. Sutton (ed.), Languages of
Cape York. Canberra: Australian Institute of Aboriginal Studies, pp 3-6.
Bentahila, Abdelâli & Davies, Eirlys E., 1991, Constraints on code-switching: A look beyond
grammar. In: ESF (ed.), Network on Code-switching and Language Contact, Papers for
the symposium on code-switching in bilingual studies: Theory, significance and
perspectives, Barcelona, 21-23 March 1991, Vol. II, 369-404. Strasbourg: ESF.
Bentahila, Abdelâli & Davies, Eirlys E., 1995, Patterns of code-switching and patterns of
language contact. Lingua 96, 75-93.
Blommaert, Jan, 1992, Codeswitching and the exclusivity of social identities: some data from
Campus Kiswahili, Journal of Multilingual and Multicultural Development 13, 1/2, 57-70.
Boeschoten H., 1983, Ösbekisches aus Aibak (Samangan). Utrecht, Dept. of Turkology.
Boeschoten, Hendrik & Backus, Ad (1997), Code-switching and ongoing linguistic change,
Turkic Languages 1, 41-62.
Boretzky, Norbert & Igla, Birgit, 1994, Romani mixed dialects, in: Bakker/Mous (eds.), 35-68.
Clyne, Michael, 1967, Transference and Triggering. The Hague: Mouton.
Clyne, Michael, 1987, Constraints on code switching: how universal are they? Linguistics 25,
739-764.
Di Luzio, Aldo, 1984, On the meaning of language choice for the sociocultural identity of
bilingual migrant children. In: P. Auer & A. di Luzio (eds.), Interpretive Sociolinguistics,
Tübingen: Narr, 55-86.
Franceschini, Rita, 1998, Code-switching and the notion of code in linguistics: Proposals for a
dual focus model. In: P. Auer (ed.) Code-switching in conversation. London:
Routledge, pp 51-75.
Gafaranga, MS (1997), Code-switching/code-mixing or Kinyarwanda for all practical purposes:
The base language issue. MS, Univ. of Lancaster, Linguistics Department.
Givón, Talmy, 1979, From discourse to syntax: grammar as a processing strategy. In: Givón
(ed.), Discourse and Syntax (= Syntax and Semantics Vol. 12), New York, Academic
Press, pp 81-111.
Goke-Pariola, Abiodun, 1983, Code-mixing among Yoruba-English bilinguals. Anthropological
Linguistics 25, 39-46.
23
Golovko, Evgenij V. , 1994, Copper Island Aleut, in: Bakker/Mous (eds.), pp 113-121.
Gumperz, John, 1982, Discourse Strategies. Cambridge: UP.
Haust, Delia, 1995, Codeswitching in Gambia. Köln: R. Köppe Verlag.
Heine, Bernd, Claudi, Ulrike & Hünnemeyer, Friederike, 1991, Grammaticalization: A
Conceptual Framework. Chicago: University of Chicago Press.
Heller, Monica, 1988, Strategic ambiguity: codeswitching in the management of conflict. In: M.
Heller (ed.), Codeswitching: Anthropological and Sociolinguistic Perspectives. Berlin:
Mouton de Gruyter, 77-98.
Hill, Jane H. & Hill, Kenneth, C., 1986, Speaking Mexicano. Tucson: University of Arizona
Press.
Holzinger, Daniel, 1993, Das Rómanes: Grammatik und Diskursanalyse der Sprache der Sinte.
Innsbruck: Inst. f. Sprachwiss. der Universität (Verlag des Instituts)
Lehmann, Christian, 1985, Grammaticalization: Synchronic variation and diachronic change.
Lingua e stile 20, 202-18.
Li Wei, 1994, Three Generations, Two Languages, One Family. Clevedon: Multilingual Matters.
Li Wei, 1998, The 'why' and 'how' quesions in the analysis of conversational code-switching. In:
P. Auer (ed.) Code-switching in conversation. London: Routledge, pp 156-179.
Luther = D. Martin Luthers Werke (Kritische Gesamtausgabe). Weimar: Böhlau, 1914.
Matras, Yaron, 1996, Prozedurale Fusion: Grammatische Interferenzschichten im Romanes.
Sprachtypologie und Universalien-Forschung (STUF) 49, 1, 60-78.
Matras, Yaron (1998a), Utterance modifiers and universals of grammatical borrowing. Linguistics
Matras, Yaron (1998b), Para-Romani revisited. In: Matras (ed.), The Romani element in non-
standard speech. Wiesbaden: Harrassowitz, pp 1-27.
McConvell, Patrick, 1988, MIX-IM-UP: Aboriginal codeswitching, old and new. In: M. Heller (ed.)
Codeswitching. Anthropological and Sociolinguistic Perspectives. Berlin: Mouton de
Gruyter, pp 97-150.
Meeuwis, Michael & Blommaert, Jan, 1998, A monolectal view of code-switching: Layered code-
switching among Zairians in Belgium. In: P. Auer (ed.) Code-switching in conversation.
London: Routledge, pp 76-100.
Moyer, Melissa, 1998, Bilingual conversation strategies in Gibraltar. In: P. Auer (ed.) Code-
switching in conversation. London: Routledge, pp 215-236.
Muyksen, Pieter (1997), Codeswitching processes: alternation, insertion, congruent
lexicalization. in: M. Pütz (ed.), Language Choices: Conditions, Constraints and
Consequences. Amsterdam: Benjamins, 361-380.
Nait M’barek, M. & Sankoff, David, 1988, Le discours mixte arabe-français: emprunts ou
alternances de langue? Canadian Journal of Linguistics 33, 143-154.
Poplack, Shana, 1979, Sometimes I'll start a sentence in Spanish Y TERMINO EN ESPANOL:
Toward a typology of code-switching. Language Policy Task Force, New York, Paper No
4 (= Linguistics 7/8, 581-618, 1981).
Poplack, Shana, 1988, Contrasting patterns of codeswitching in two communities, in: M. Heller
(ed.) Codeswitching. Anthropological and Sociolinguistic Perspectives. Berlin: Mouton
de Gruyter, pp 151-186.
Poplack, Shana, 1993, Variation theory and language contact, in: D.R. Preston (ed.), American
Dialect Research, Amsterdam: Benjamins, pp 251-286.
Preziosa di Quinzio, I., 1992, 'Teoreticamente la firma fa indietro'. Fammistione di italiano e
schwyzertütsch nella conversazione di figli di emigrati. Unpubl. M.Phil. Thesis,
University of Zurich.
Romaine, Suzanne, 1989, Bilingualism, Oxford: Blackwell.
Salmons, Joe, 1990, Bilingual discourse marking: code switching, borrowing, and convergence
in some German-American dialects, Linguistics 28, 453-480.
Scotton, Carol Myers, 1988, Codeswitching as indexical of social negotiations. In: M. Heller (ed.)
Codeswitching. Anthropological and Sociolinguistic Perspectives. Berlin: Mouton de
Gruyter, pp 151-186.
Scotton, Carol Myers, 1993a, Social Motivations of Code-switching. Oxford: Clarendon Press.
Scotton, Carol Myers, 1993b, Duelling Languages. Oxford: Clarendon Press.
Sebba, Mark, 1998, A congruence approach to syntax of codeswitching. IJB 2, 1, 1-20.
Sebba, Mark & Wootton, Anthony, 1998, We, they and identity: Sequential versus identity-
related explanation in code-switching, in: P. Auer (ed.), Code-switching in conversation.
London: Routledge, pp 262-289.
Selting, Margret, et al., 1998, Gesprächsanalytische Transkription (GAT). Linguistische Berichte
34 (173), 91-122.

24
Silverstein, Michael, 1992, Metapragmatic discourse and metapragmatic function. In: J. Lucy
(ed.), Reflexive Language: Reported Speech and Metapragmatics, Cambridge: UP, pp
33-58.
Singh, Rajendra, 1996, Lectures against Sociolinguistics. New York P. Lang
Stolt, Birgit, 1964, Die Sprachmischung in Luthers ‘Tischreden’. Stockholm: Almqvist and
Wiksell.
Stolz, Christel & Stolz, Thomas, 1996, Funktionswortentlehnung in Mesoamerika. Spanisch-
amerindischer Sprachkontakt (Hispanoindiana II). STUF 49 (1), 86-123.
Swigart, Leigh, 1992, Two codes or one? The insiders´ view and the description of
codeswitching in Dakar. Journal of Multlingual and Multicultural Development 13, 1/2,
83-102.
Volker, Craig, 1982, An Introduction to Rabaul Creole German (Unserdeutsch). Unpubl. M.A.
Thesis, University of Queenland.
Voßler, Karl, 1904, Positivismus und Idealismus in der Sprachwissenschaft. Heidelberg.
Zentella, Ana Celia, 1997, Growing Up Bilingual. Oxford: Blackwell.

Endnotes

The term grammaticalization is used roughly in the sense of Givón (1979, etc.), not in the more
restricted one of, e.g., Lehmann (1985) or Heine/Claudi/Hünnemeyer (1991). Parallels can also
be found in the idealistic tradition of German linguistics, e.g. in Spitzer´s or Voßler´s contention
that grammatical structures always have a precedent in ‘stylistic’ strategies (see Voßler 1904).

1
Those of my readers who happen to be familiar with my previous publications on language
alternation and code-switching (from Auer 1984 onwards) I should warn that the terminology
introduced here is not completely identical with my own earlier usage (although certainly similar
in spirit). In particular, ”language alternation” is no longer used as a cover term . The ‘new’ notion
of code-switching (but cf. Auer 1990) owes much to discussions with Celso Alvarez-Cáccamo
(see references in the bibliography).

2
In the sense of Silverstein´s ”metapragmatics”; cf. 1992.

3
But see Auer 1984b as well as the contributions by Jørgensen, Sebba & Wootton, Stroud,
Rampton in Auer (ed.) 1998 for a critical discussion of the ‘social meaning’ of CS and, in
particular, the distinction between ‘we-code’ and ‘they-code’.
4
Cf. Selting et al. (1998).

5
Cf. note 10 for an example.

6
Earlier research (e.g. Gumperz 1982: 84f) has already stressed the fact that the languages
involved in code-switching in a bilingual community may be considerably different from their
monolingual relatives due to convergence as a consequence of long-standing language
contact (also cf. Clyne 1987). For this reason, a definition of language alternation as proposed in
Poplack 1993:255 (my italics) is problematic: ”Code-switching is the juxtaposition of sentences

25
or sentence fragments, each of which is internally consistent with the morphological and
syntactic (and optionally, phonological) rules of the language of its provenance”). Note that my
distinction between CS and LM/FLs is independent of structural considerations such as
convergence or integration.

7
This terminological opposition is also used by Muysken 1997; Nait M ‘barek & Sankoff 1988
speak of ”constituent insertion”; Haust 1995:63 of ”Insertionen” vs. ”ML-Wechsel”; ”insertional
code-switching” roughly corresponds to ”transfer” in Auer 1984.

8
For examples of competence-related insertions, cf. Auer 1984: 55ff, for examples of discourse-
related insertions, cf. Auer 1998: 6f as well as Hill & Hill 1986:387ff.

9
For instance, the English verb to fax (< fax (N) < facsimile (N)), nowadays an established loan in
German and integrated into its verbal morphology (faxen Sie mir das mal!, faxst Du mir das? habs
Dir schon rübergefaxt), was used in the same grammatical formats until some ten years ago in
order to achieve a comical effect (being a pun on Faxen (N), ‘sillyness’ as in Faxen machen ‘to
fool around’). The discourse-functional effect was reached precisely because the word (which
was still preceived as a nonce borrowing from English at that time) was morphologically
integrated (the non-comical variant would have been ein Fax schicken ‘to send a fax’). The
example also shows that code-switching does not presuppose any kind of deep knowledge of
the language from which the insertion comes.

10
The authors argue that this insertional LM is typical for younger generation speakers while the
more balanced bilinguals of the older generation mainly employ the alternational style; cf. in
particular the examples in Bentahila & Davies 1991:383f.

11
A general definition of the matrix language and a generally valid methodology of establishing it
in each and every given case is quite difficult however; cf. the discussion in Auer, in prep.

12
Cf. Blommaert 1992 on Tanzania and Hill & Hill 1986 on Malinche Indians in Central Mexico for
the complex relationship between LM and identities.

13
A notable exception from the often negative attitudes towards mixing are the mixed varieties
used by the African elites - showing, among other things, that it is not varieties or ”codes” that
are evaluated but their speakers.

14
Of course, higher level CS between mixed varieties may develop from LM.

15
The Swiss data are from adult speakers and were recorded in the mid-80s, while the
Constance data are from youngsters and were recorded 15 years earlier. The same
development holds for the German-Turkish mixed style which has emerged in the large German
26
cities in recent years.

16
See my re-analysis of Scotton´s data in Auer 1995:130f.

17
The term is taken from Matras 1996.

18
The precise relationship between ”mixed languages” according to Bakker´s definition (1994:
5) and FLs is unclear. Bakker does not require a mixed language to show surface juxtapositions
of the two underlying languages, i.e. there may be ”mixed languages” which are no FLs. On the
other hand, not all fused languages are mixed languages.

19
Romani dialects such as Sinti which have undergone massive convergence with the
respective contact language (here, German) should not be equated or confounded with so-
called para-Romani varieties (often called “mixed varieties of Romani”); according to Boretzky &
Igla (1994) the latter may be the result of conscious though incomplete relexification at a stage
of imminent language shift, and have lost more or less all Romani syntax and most morphology.
The reader is referred to Matras 1998b for an overview. Both “converged” Romani and Para-
Romani may qualify as “fused lects” in the present sense (but certainly the former).

20
Like Michif, Montagnais/French LM only allows the insertion of French nouns or noun
phrases, but there is variation between Montagnais and French nouns according to semantic
domains (greetings, calendrical reckonings, time expressions, numerals, expressions of
quantity and measures, conjunctions, exclamatives, interjections, adverbs and discourse
particles are taken over from French). In Michif, practically all NPs are French.

21
She uses the term ”code-switching” to designate LM.

22
Further examples are the ‘borrowing’ of various prepositions such as Engl. by (Unserdeutsch
bei which is only graphically matched with the German preposition) or Engl. around
(Unserdeutsch rund, again only graphically adapted to look like the German adjective for
‘round’).

23
Many more cases are discussed in Stolz & Stolz 1996, Hill & Hill 1986 and Matras 1998.

24
The same holds for Turkish in Germany.

25
The same strategy of nativization can be observed in other mixed codes on the way to fused
lects; cf., e.g., Agnihotri 1980: 287ff and Romaine 1989:123ff on Panjabi compound verbs
used to syntactically accommodate English verbs, nouns and adjectives. For a general
discussion, also cf. Sebba 1998.

27
 Previously published in this series:

No.1 Elizabeth Couper-Kuhlen, Coherent Voicing. On Prosody in Conversational

Reported Speech (1998)
No.2 Peter Auer, Zwischen Parataxe und Hypotaxe. 'Abhängige Hauptsätze' im
gesprochenen und geschriebenen Deutsch (1998)
No.3 Susanne Günthner, Polyphony and the "Layering of Voices" in Reported
Dialogues. An Analysis of the Use of Prosodic Devices in Everyday Reported
Speech (1998)
No.4 Margret Selting, TCUs and TRPs: The Construction of Units in Conversational
Talk (1998)
No.5 Helga Kotthoff, Irony, Quotation, and Other Forms of Staged Intertextuality:
Double or Contrastive Perspectivation in Conversation (1998)

28

View publication stats