in cultural or other characteristics of the communities themselves. This dif-
ference between the two approaches has been noted by for example Gardner-
Chloros (2009: 59), who states that ‘variation in CS can be divided for
practical purposes into variation between communities and variation within
communities’.
Using the recorded materials from our two corpora and the sociolinguis-
tic questionnaires that were collected after each recording, we pursue the
following objectives:
(i) to analyse production data in order to assign a matrix language (ML) to
each bilingual clause in a sample of data from each corpus;
(ii) to perform a quantitative analysis of the results of (i) to describe the
distribution of the ML in the two corpora;
(iii) where variation in the ML is found, to investigate the relation between
the ML as a dependent variable and independent speaker-based variables
(education, proficiency, age of acquisition, social network and
identity);
(iv) to use our questionnaire data to provide an overall characterisation of
the two communities, and to relate these to our findings regarding the
distribution of the ML;
(v) to weigh up the relative impact of speaker-based and community-based
factors on the patterns found in our data.
This chapter is structured as follows. In the section ‘The Matrix Language
Frame (MLF) Model’, we explain the main ideas behind the MLF, which we
use to classify our CS data. The section ‘The Bilingual Communities’
describes the two communities. The section ‘Previous Research on
Uniformity vs Variability in CS Patterns’ summarizes previous research
using the MLF in the language pairs under study and the effect of extra-
linguistic variables on CS. In the section ‘Methodology’ we outline the meth-
odology of data collection and analysis. The analysis is presented in the
Results section. The discussion of our findings and the conclusion bring the
chapter to a close.
The Matrix Language Frame (MLF) Model
In this section we give a brief overview of the MLF model (e.g. Myers-
Scotton, 1993, 2002), which has been both very influential and controversial
in the field (cf. MacSwan, 2005). For a rationale of why we selected the MLF
model over other CS frameworks, see Carter et al. (2011).
The MLF proposes that in ‘classic code-switching’ (Myers-Scotton, 2002:
8) there exists an asymmetrical relationship between the two languages
involved: one being the main or ‘matrix’ language (ML) and the other being the
112 Part 3: Bilingual Language Use