/gr/ (r = 0.628, p < 0.001), /fl/ (r = 0.352, p = 0.024), /fr/ (r = 0.505, p = 0.001),
/skr/ (r = 0.718, p < 0.001) and /str/ (r = 0.492, p = 0.001). No significant rela-
tionship between the two languages was found for /pl/, /bl/, /dw/, /kw/,
/sb/ (/sp/), /sg/ (/sk/) and /skw/. Note, however, that these clusters were
all produced with near-perfect accuracy from the outset. In a similar vein,
correlations could not be computed for /gl/, /sm/, /sw/ and /sl/, as at least
one of the variables in these exhibited 100% accuracy across all age groups,
thus showing ceiling effects.
Substitution patterns
Tables 1.4 and 1.5 depict the substitution patterns for the Welsh and
English clusters, respectively. Note that the figures denote the number of
tokens for a given pattern, with a maximum of 12 in each cell (i.e. 3 chil-
dren × 4 instances). The ticks, in turn, indicate that all tokens in a given cell
were produced accurately.
As the tables indicate, cluster reductions predominate in younger chil-
dren, whereas cluster simplifications are more common in older ones. Some
clusters only had one type of substitution pattern, for example, [t] for
/tw/; [d] for /dw/, others exhibited a large number of different ones, with
English /ʃr/ the most heterogeneous with ten different realisations, that
is, [swrʃsɹtrswhɹɬɹʃw].
Cluster reductions
Double clusters with a stop or fricative as the first element followed by
/l/, /r/ or /w/ were reduced to a stop or fricative, respectively, for example,
/pl/ → [p]; /fr/ → [f], with the exception of /xw/, which the younger chil-
dren tended to realise as /w/, the older ones as /x/. In double clusters consist-
ing of a stop or nasal preceded by /s/, in turn, the second element tended to
be preserved, for example, /sp/ → [p]; /sm/ → [m].
Three-element clusters of the type /s/ + stop + sonorant were occasion-
ally reduced to a single stop or [s], but more typically to a double cluster
consisting of /s/ + stop. Less common reductions include /s/ + sonorant and
stop + sonorant. The Welsh-specific clusters /g/ + /w/ + sonorant, in turn,
showed a number of patterns: /gwl/ was predominantly realised as [gl],
/gwr/ as [gw] and /gwn/ as [n].
Cluster simplifi cations
In stop + /r/, stop + /l/ and fricative + /r/ clusters, the second element
was commonly substituted by /w/, for example, /pr/ → [pw]; /fr/ → [fw],
in conformity with the process of liquid gliding. Rhotics were also replaced
by [ʋ] or [l] in some cases, however only in the English word set, for example,
/kr/ → [kʋ]; /kr/ → [kl]. English /tr/, in turn, was realised as /tʃ/ in two
instances, showing coalescence, although it could be argued that this substitu-
tion pattern is, in fact, target-like as it is a wide-spread connected speech
phenomenon in the adult language.
Speech Learning in Bilinguals 15