Welsh speakers, with 52% of speakers proficient in the language according to
the 2001 census (http://www.statistics.gov.uk/census2001, see also
Gathercole & Thomas, 2009; Jones, 2010). The participants were assigned to
one of seven age cohorts ranging in 6-monthly intervals from 2;6 to 6;0. The
sample was further stratified in terms of language dominance. Children from
Welsh-only homes were classified as Welsh-dominant, children from English-
only homes as English-dominant. Following Gathercole et al.’s (20 08)
approach, children from homes in which the dominant home language is
used most of the time, albeit not exclusively, were also included. On the other
hand, children from homes in which both languages are used regularly in
everyday interactions were excluded as the study sought to capture maxi-
mally distinct types of linguistic experience. Information on home language
use was obtained via a language background questionnaire sent out to the
children’s parents (Baker, 1988; Munro et al., 2005).
Materials
Welsh, like English, allows complex phonotactic patterns with up to three
consonants in syllable onsets (see Ball & Williams, 2001; Awbery, 1984 for
Welsh; Yavaş, 2006 for English). Both languages permit two-element clusters
consisting of an obstruent and a sonorant as well as a sequence of two obstru-
ents. Note that, with the exception of /s/-clusters,^1 these conform to the SSP.
Three-element onsets in both languages may consist of /s/ + stop + sonorant.
However, only Welsh distinguishes three-element onsets of the type
/g/ + /w/ + sonorant (e.g. /ɡwrandɔ/ listen, /ɡwnɛɪd/ do, /ɡwlɑd/ country).
These clusters fail to conform to the SSP because the glide constitutes a more
sonorous element than the following liquid or nasal. Note, however, that
some represent them as two-element clusters with concomitant lip-rounding,
that is, [ɡwr],[ɡwl],[ɡwn] (cf. Ball & Williams, 2001: 16).
Twenty-nine English onset clusters and thirty Welsh onset clusters were
included in the study (cf. Table 1.1). They capture all phonotactically admis-
sible patterns, with the exception of rare clusters that only occur in low-
frequency items, such as /gn/, /sn/ and /sbl/ in Welsh and /sf/, /θw/ and /skl/
in English. Clusters containing /j/ as second or third element, for example,
/fj/, /dj/, /spj/, were also excluded as the glide tends to be eliminated in the
English varieties of West Wales.
Of the onset clusters used in the study, some are language-specific,
whereas others occur in both languages. Note that clusters containing /r/
were included in the latter category, despite the fact that rhotic consonants
may have different realisations in the two languages (Ball et al., 2001; Ball &
Williams, 2001). Thus, according to descriptive accounts, English /r/ is realised
as a post-alveolar approximant [ɹ], whereas Welsh /r/ is realised as a voiced or
voiceless aspirated alveolar trill, that is, [r] or [r
°
h] respectively. However, theserealisations may not necessarily be kept distinct by Welsh-English bilinguals.
Thus, some speakers, in particular Welsh-dominant bilinguals, may never
8 Part 1: Bilingual Speech