they did not consider whether any speaker-based variables could be related
to the choice of ML by a given speaker. They did, nevertheless, predict that
in a speech community where four different languages are spoken (Zulu,
Sotho, English and Afrikaans), individual speakers would use just one of
these languages to provide the morphosyntactic frame (or Matrix Lang-
uage) of all the bilingual clauses in the same turn. Their prediction was
supported by the data. In addition, they studied the relation between one
speaker-based variable, that is educational level, and the type of English code-
switched constituents produced, which range from individual English word
insertions to whole English phrases. Their results showed that those with a
higher level of education (Grade 10 or above) produced significantly more
English ‘Embedded Language (EL) Islands’, (constituents consisting of English
multi-word phrases inserted into the morphosyntactic frame of another
language), than those with a lower level of education (Grade 9, i.e.14–15 year
olds, or below).
Previous studies have suggested that the role of proficiency and age of
language acquisition influence various aspects of CS behaviour, although
their effect on the choice of the ML has not been studied. Poplack (1988) took
into account the bilingual ability of the speakers in her comparative study of
Spanish-English Puerto Ricans in New York and the French-English com-
munity in the Ottawa-Hull region of Canada. In her study Poplack argues
that it is essential to include information about the informants’ competences
and linguistic backgrounds when exploring the language behaviours of bilin-
gual communities. The study found that there was a correlation between CS
types (tag, sentential and intrasentential) and bilingual ability among the
Puerto Rican speakers, with ‘the most highly bilingual speakers switching
mainly within the bounds of the sentence’ (1988: 219). She also found that
the bilingual speech of the Puerto Rican speakers could be described as con-
taining ‘smooth’ switches. For the French-English community, however,
‘smooth’ CS was not a community-wide mode of discourse, but instead
tended to be flagged, presumably to call attention to the switch. Poplack
proposed that the differences in CS patterns in the two communities could
be attributable to differences in communicative strategies which arise from
a difference in their attitudes to English. In particular, Poplack suggests that
bilingualism may be seen to be emblematic of the Puerto R ican New Yorkers
in a way that is not the case for the Ottawa-Hull speakers.
Sayahi (2004) explores the relationship between identity and CS in three
communities of bilinguals: Spanish-Valencian in Spain, Spanish-English in
New York, and Arabic-French in Tunisia. He used questionnaires to assess
the bilinguals’ own notions of their identity and either observation or self-
report of their CS to assess the relationship between identity and CS. For
example, he found that 68% of the Spanish-English bilinguals considered CS
to be a sign of their identity and concluded that this and other studies show
that ‘code-switching by Spanish/English bilinguals in New York goes beyond
Factors Influencing Code-Switching 119