The Uniformity of the
Matrix Language in
Classic Code-switching
Alberto Rosignoli
ESRC Centre for Research on Bilingualism Seminar Series
Bangor 6th April 2009
A definition
• The alternation of two languages within a single
discourse, sentence or costituent (Poplack 1980,
The concept of Asymmetry
• The languages involved in CS do not contribute
equally to form bilingual utterances.
• The language source of different types of
morphemes is constrained.
• ! Not a universal feature of CS theories.
Early work: Joshi 1985
• “Despite extensive intrasentential switching
speakers and hearers usually agree on which
language the mixed sentence is “coming from”.
• Matrix Language vs Embedded Language
Existence of a control structure which allows shift
from the ML to the EL .
• Nonswitchability of closed-class items.
The Matrix Language Frame Model
• Myers-Scotton (1993, 1997, 2002)
• Framework for the analysis of intrasentential
• A model of bilingual production
• Based on Levelt’s 1989 Speaking Model
Asymmetry in the MLF
• Asymmetry in the degree of participation of the
languages involved
▫ Matrix Language (ML)
▫ Embedded Language (EL)
• Asymmetry in the retrieval procedures of
▫ Content morphemes
▫ System morphemes
4-M Model
Late bridge
morpheme morpheme
Theta role
outside max
The MLF Unit of Analysis
• In 1993: the discourse as a whole
• After 1997: rejection of the discourse in favour of
the CP (projection of complementiser)
The Matrix Language Principle
• There is always an analyzable or resolvable frame
structuring the morphosyntax of any CP. This frame
is called the Matrix Language. In bilingual speech,
the participating languages never participate equally
as the source of the Matrix Language.
2002: 8)
Classic CS
• In Classic CS only one of the participating
languages is the source of the abstract morphosyntactic frame of the bilingual CP.
• In other types of CS, both languages may
contribute abstract structure.
Ex. Mi oedd
drws nesa pobl yn gwneud sloe gin
PRT be.3s.past door next people PRT do.nonfin sloe gin
‘The next-door people made sloe gin’
(Deuchar and
Davies, 2009)
Identifying the ML: criteria
• Morpheme Order Criterion
In Matrix Language + Embedded Language constituents
consisting of singly occurring Embedded Language lexemes and any
number of Matrix Language morphemes, surface morpheme order
(reflecting surface syntactic relations) will be that of the Matrix
(Myers-Scotton, 1993a: 83; 2002: 59)
fath-a rhedeg
you need.NONFIN do.NONFIN like
run.NONFIN exercise
“You need to do, like, running exercise”
(Deuchar, 2006)
Identifying the ML: criteria
• System Morpheme Criterion
In Matrix Language + Embedded Language constituents, all
system morphemes which have grammatical relations external to
their head constituent […] will come from the Matrix Language.
(Myers-Scotton, 1993a:83; 2002: 59)
Ex He sells me their prenotazioni for their seats but not my
A Critique of the MLF (MacSwan 2000,
2005a, b)
“Nothing constrains code switching apart from the
requirements of the mixed grammars.”
• CS is the union of two lexicons
• No constraints specific to bilingual speech (no
• Grammaticality of mixed utterances can be
ascertained through checking of features.
• The case of Spanish-English DPs (Moro 2001)
a. D, phi = {person, number, gender}
N, phi ={person, number, gender}
b. D, phi = {person, number}
N, phi ={person, number}
c. D, phi = {person, number, gender}
N, phi ={person, number}
d. D, phi = {person, number}
N, phi ={person, number, gender}
A comparison
Matrix Language Frame
Minimalist Program
• Asymmetry between
participating languages.
• No asymmetry between
participating languages.
• In bilingual CPs where
Spanish is the ML DETs will
come from Spanish.
• In bilingual CPs DETs will
always come from Spanish
(see feature mismatch).
• English DETs will occur in
bilingual DPs where English is
the ML.
• English DETs will not occur in
bilingual DPs.
Feature checking cont.
• The non-occurrence of NPs of the type the casa
in available data could be due to factors other
than feature mismatch.
• The argument relies partially on grammaticality
judgments by simultaneous bilinguals.
A MLF case study: Smith 2006
• Spanish-English community in the US
• 56 speakers (10-20 mins per conversation)
• ‘The asymmetry between the ML and the EL in a
single utterance is replicated in the speech of an
entire community in which the community ML is
[…] Spanish and the community EL is English.’
E> Sinsert
• maestro y a ónde vamos
a ir al
swimmin’ onde onde?
teacher and to where to go to+the swimming where where
‘Teacher, and where are we going to go swimming where where?’
S> Einsert
• I don’t want those (NO’S?/NOSE?)* como three horns
‘I don’t want those (no’s?/nose?) like three horns’
The ML as a dynamic construct
• Because the ML is defined at the level of the CP, it is assumed
that the language providing the source of the ML could change
(as an extreme case) even within the same sentence, from one
CP to the following.
• This, however, rarely seems to be the case in the available data
Myers-Scotton 1993 (Swahili-English)
Finlayson et al 1998 (Zulu-Sotho-English)
Boussofara Omar 2003 (Standard/Tunisian Arabic)
Owens 2005 (Standard/Nigerian Arabic-Hausa-English)
Smith 2006 (Spanish-English)
Deuchar 2006 (Welsh-English)
Myers-Scotton (2002) “ML of the discourse”
Smith (2006) “Community ML”
ML is the same
for every CP
ML changes at
every successive
Not accounted for in the MLF
Most data show far less variability as regards the source of the ML than the
model allows.
Problems with the MLF
• The ML as a dynamic
• !Uniformity of the ML
• The CP as the unit of
• !ML beyond the CP
• Analysis of well-formed
• !Naturalistic data
A shift of perspective?
• Uniformity in ML assignment in bilingual CPs is
the factor that justifies the extended use of the
ML construct.
• Rather than imprecise applications, these uses
are capturing a generalisation that the model
does not explain.
• Motivation is normally found within a
sociolinguistic framework (e.g. ‘Markedness’)
CS studies on a continuum
A conversation analysis perspective
• Interest in the issue of the ‘base language of the
▫ Part of the overall organisation of the discourse
• Different coverage of naturalistic data
▫ No well-formedness requirements
• Problematising the notion of ‘code’ in language
▫ What counts as a ‘code’ for participants?
A typology of code-alternation (Auer)
Change of the base
*SAR: vabbé no io è un po’ che non vado al cinema .
(well no I haven’t been to the cinema in a while)
*ANT: hmm@ 0 [>] .
*SAR: però [<] # come sai non ho tempo &=sigh di fare niente xxx [>] .
(but as you know I haven’t got time to do anything)
*ANT: <come va il phd> [<] ?
(how’s the PhD going?)
gonna [: going to] go finish it ?
*SAR: ah PhD eh ho avuto un momento brutto # mercoledì [>] .
(ah PhD I had a bad moment on Wednesday)
*ANT: perché [<] ?
*SAR: perché: Laura Layton era incazzata con tutti mercoledì .
(because Laura Layton was pissed off with everyone on
&e: hmm &e: stava per scoppiare.
(hmm uuh and she was about to burst)
infatti ieri ho parlato con altra gente mi fan+"/.
(actually yesterday I spoke to some others and they were like)
+"/ è arrivata a un punto che neanche lei non ne può più .
(she’s come to the stage where she herself can’t take it
*ANT: in che senso [>] ?
(how do you mean?)
*SAR: perché ha le sue deadline per <la fine del> [<] mese .
(because she’s got her own deadlines at the end of the
*ANT: hmm .
*SAR: e: ha tutti (que)sti studenti # ehm dieci st(udenti) [//] dieci PhD
students # e: [>1] .
(and she has all these students ehm ten st(udents) ten PhD
*ANT: <non ce la fa più> [<1] <anche lei> [>2] .
(she too can’t take it anymore)
*SAR: e non ce la fa più perché ognuno ha unproblema .
(she can’t take it anymore because everyone has got a
*SAR: e allora gli ho detto che appunto <ho inv(itato)> [//] avevo ricevuto
questo invito da Catherine .
(so I told him right that I inv(ited) that I received this
invitation from Catherine)
Catherine <era la:> [/] era la supervisor di: ehm Paul [>] .
(Catherine was Paul’s supervisor)
*ANT: ah [<] sì sì .
(ah yes yes)
*SAR: hmm.
che abita qua intorno [>] .
(who lives around here)
*ANT: hmm [<] .
*SAR: e allora fa:
(so he goes like)
+”/ma ha detto che ha [//] fa il compleanno perché ha raggiunto: quell'età
in cui c'è zero no .
(so she said she’s having her birthday because she got to that
age in which there is zero, right)
mhmm <# zero> [>] ?
<e: allora fa> [<] +”/.
(so she goes like)
*SAR: +”/ zero # c'è lo z(ero) [/] zero in it .
(zero # there’s zero in it)
e: e allora [>] .
(and and then)
*ANT: forty . [<]
*SAR: four # e allora <mi fa però> [?] +"/.
(four # and so she goes but)
+"/ you have to guess [>] .
*ANT: hmm [<] .
*SAR: mi ha detto +"/.
(she said to me)
+"/ e: you have to guess ehm quanti anni ho .
(uuh you have to guess uhm what my age is)
ehm: allora ho pensato io +"/.
(uhm and so I thought)
+"/ sarà quaranta .
(she must be forty)
*ANT: hmm [>] .
*SAR: non [<] penso che ne abbia cinquanta # e neanche trenta .
(I don’t think she’s fifty or even thirty)
The future…
• Comparing the two frameworks, with particular
reference to data covered by both
(using English-Italian data)
• Assess whether the contribution of a CA-type
approach can offer a satisfactory explanation for the
regularities encountered in CS that the MLF cannot
readily account for.
• Can a CA approach reveal whether a switched item
counts as such for participants themselves?
Thanks to
Present (and past) members of the ESRC Bilingualism
Centre Corpus Based Research Group and AHRC project
Margaret Deuchar
Maria del Carmen Parafita Couto
Dirk Bury
Peredur Davies
Jon Herring
Sian Lloyd
Elen Robert
Jonathan Stammers
Thank you

The Uniformity of the Matrix Language in Classic Code