the Spanish-English speakers. While in the recordings Sastre 1 and Herring 10,
all four speakers alternate between English and Spanish ML, in Zeledon 1
both speakers choose Spanish as the ML in bilingual clauses. Looking at
these 12 speakers (and also observing our entire corpus) it seems that there
may be more speaker-based variation in the choice of the ML in Miami. This
led us to conduct a Goldvarb analysis as described in the next section.
Goldvarb analysis
Given that 100% of the clauses from the Welsh-English data had Welsh
as the ML, and therefore were completely invariant, the analysis was only
conducted with the Spanish-English data set. After the data had been
entered into the Goldvarb X program and an initial analysis was conducted,
the following factor was removed because of its invariability: age of acquisi-
tion of Spanish. (All had reported this as being 2 years old or younger.)
Cross-tabs were used to check the distribution of the remaining variables,
and revealed an uneven distribution of the data (cf. Table 6.4 for ML distri-
bution by speaker). Only five clauses were provided by the Venezuelan
speaker AME, all of which had Spanish as the ML. AME had finished an
undergraduate degree (or equivalent) and reported having a high proficiency
in both English and Spanish. However, she had only acquired English during
her time at secondary school. The other speaker from the same conversa-
tion, CAR, only produced one clause with English ML. She had finished
high school, acquired both English and Spanish from two years of age, and
reported a high proficiency level in both languages. Interestingly, she
reported her national identity as being American but 14 out of the 15 clauses
she produced had a Spanish ML. This could be attributable to an accom-
modation effect in the dialogue.
The results of the statistical analysis did not reveal any of the factors as
being significant. This outcome suggests that individual speaker-based vari-
ables may not be able to account for the patterns in our Spanish-English
data, and a larger sample size of speakers may be required for further analy-
sis. For both corpora, Welsh-English and Spanish-English, we therefore
looked next at community-wide factors based on the results from our ques-
tionnaires. The purpose of this next step was to compare the two communi-
ties and to see whether differences in the constellation of the factors listed
previously might be related to differences in the choice of ML in the two
communities.
Community-wide factors analysis
We now report the questionnaire results for the two communities in
relation to the eight extra-linguistic variables listed on p. 125.
Age of acquisition of Spanish or Welsh
We expected that the languages learned earliest in each community
might be those that supplied the ML most often. The questionnaire results
Factors Influencing Code-Switching 129