either the ML or the EL. To show how identifying the outside late system
morphemes in a clause help identify the ML of that clause, consider example
(2) below, which is taken from Myers-Scotton’s Swahili-English corpus.
(2) Ile m-geni hata si-ku-COMMENT
DEM/CL9 Ch/S-visitor even 1S-NEG-PST-NEG-comment
‘That visitor, I did not even comment.’ (Myers-Scotton, 2002: 89)
All the morphemes are from Swahili, apart from the verb stem comment,
which is English, and it is also the main verb of the clause. Nevertheless, the
first person marking on the verb comment comes from Swahili. As subject/
verb agreement is an example of an outside late system morpheme, and this
comes from Swahili, Swahili is identified as the ML according to the SMP.
English is identified as the EL. If the finite subject–verb agreement morphol-
ogy in this clause had been English instead, then English would be identified
as the source of the ML by the SMP.
The MOP states that word order will be sourced from the ML, with the
exclusion of what Myers-Scotton (2002: 139) calls ‘EL islands’ – strings of EL
morphemes which are well-formed according to the grammar (which
includes word order but also for example morphological inflection) of the EL.
An example of using morpheme order to identify the ML can be seen in the
Welsh example in (3) below:
(3) mae raid bod
be.V.3S.PRES necessity.N.M.SG+SM be.V.INFIN
gynno fo ryw fath
with_him.PREP+PRON.M.3S he.PRON.M.3S some.PREQ+SM type.N.F.SG+SM
o attitude problem (yn)de.
of.PREP be.IM+SM
‘He must have some kind of attitude problem, right’ [Davies 6]
In (3) the finite verb mae precedes the subject raid ‘necessity’ in VS word
order, which is a predominantly Welsh morpheme order. As the third person
singular form of the Welsh verb bod ‘to be’, mae agrees with its third person
subject raid ‘necessity’. Welsh is thus identified as the source of the ML. We
analyse the English NP attitude problem, as an EL island with its English mod-
ifier-head order. (In Welsh NPs the head usually precedes the modifier.)
We selected the MLF as a means of classifying our data because it had
previously been used successfully to identify the ML on a clause-by-clause
basis in both our Welsh-English and Spanish data (cf. Deuchar, 2006;
Deuchar & Davies, 2009; Smith, 2006). The central question in comparing
CS patterns in different communities is the relation between the linguistic
and the extra-linguistic variables, but we need to first have a clear classifica-
tion of the range of CS patterns. Before we move on to discuss how the MLF
114 Part 3: Bilingual Language Use