Each session commenced with a brief interaction between the participant
and the second author, a native speaker of Western Welsh. This was done so
as to set the participant into a monolingual language mode (Grosjean, 2001b),
to the extent that this was possible with a bilingual experimenter. Upon
familiarisation with the task, the children were shown 120 pictures in the
Welsh session (two repetitions of the 60 target words) and 116 pictures in the
English session (two repetitions of the 58 target words). This yielded 120
(pictures) × 41 (children) = 4920 productions of Welsh words and 116
( pictures) × 41 (children) = 4756 productions of English words. In the few
instances where words could not be elicited spontaneously from the children
they were modelled by the experimenter. This was not considered problem-
atic, as previous studies have shown that error patterns do not differ across
imitated and spontaneous productions (Munro et al., 2005; Siegel et al., 1963).
On average, the recording sessions lasted approximately 10–15 min.
The experimenter made initial auditory transcriptions during data collec-
tion. Subsequently, all tokens were re-analysed on the basis of the audio
recordings, and transcriptions were adjusted in the few instances where
divergences occurred across the two sets of transcriptions. All data were
transcribed in broad phonetic transcription using the symbols of the
International Phonetic Alphabet. Tokens were deemed correctly produced if
they conformed to the patterns used in the adult language.
Results
Tables 1.2 and 1.3 depict the participants’ mean percent correct scores on
the 30 Welsh consonant clusters and the 29 English ones, broken down by
age and language dominance.
In order to investigate the effects of age and dominance on cluster acqui-
sition systematically, a series of 7 (age) × 2 (dominance) mixed plot ANOVAs
(repeated measures) was conducted separately on the Welsh and the English
data sets. The results revealed significant main effects of age (Welsh:
F(6,16) = 15.945, p < 0.001; English: F(6,16) = 52.956, p < 0.001), dominance
(Welsh: F(1,16 ) = 14.332, p = 0.002; English: F(1,16 ) = 11.884, p = 0.003) and
cluster (Welsh: F(29,464) = 45.188, p < 0.001; English: F(28,448) = 66.581,
p < 0.001).
Significant interactions for the Welsh clusters include clusterage
(F(174,464) = 4.049, p < 0.001), clusterdominance (F(29,464) = 4.579, p < 0.001)
and clusteragedominance (F(174,464) = 1.477, p = 0.001). Significant interac-
tions for the English clusters include clusterage (F(168,448) = 6.639, p < 0.001),
clusterdominance (F(28,448) = 7.11 7, p < 0.001), clusteragedominance
(F(168,448) = 2.98, p < 0.001) and age*dominance (F(6,16) = 3.583, p = 0.019).
This analysis was followed up by a series of one-way ANOVAs that
investigated the effects of age and dominance on the production of each of
the clusters.
10 Part 1: Bilingual Speech