180-260 ms in L2). These results suggest that lexical access begins approxi-
mately at 180 ms after the onset of a picture and phonological encoding
commences between 240 and 350 ms. These estimates are compatible with
the meta-analysis on speech planning by Indefrey and Levelt (2004), in
which lexical access and phonological encoding have been hypothesized to
occur between 150 and 275 ms and between 275 and 400 ms, respectively.
Again, the presence of the cognate effect in the P2, N3, and P3 ranges sug-
gests that the non-target language is active at the phonological level as well
as the lexical level.
Conclusions
In this chapter, we reviewed some of the behavioral and ERP research
on bilingual word production. Although behavioral research suggests that
the locus of language selection varies depending upon experimental para-
digms and materials, proficiency, and language dominance, recent ERP
research appears to provide more converging evidence for late language
selection rather than early language selection during picture naming. That
is, both languages are active all the way down to the phonological level and
language selection occurs after phonological representations have been
retrieved. The activation of the non-target language during production sug-
gests that the cognitive processes involved are not organized in a purely
serial fashion (e.g. Goldrick et al., 2010). However, it is not clear to what
extent lexical access and phonological encoding overlap in time. In future
research, ERPs recorded during production and analyzed millisecond-by-
millisecond (e.g. Costa et al., 2009) will be a valuable tool to examine the
flow of information in each of the bilingual’s languages in a more fine-
grained manner.
The next chapter discusses the use of ERPs to examine the co-activation
of two grammars.
Notes
(1) When naming an object, it is assumed that not only the semantic representation of
the intended concept ‘CAT’ but also those of semantically related concepts such as
‘DOG’ and ‘ANIMAL’ are activated to some degree (e.g. Bock & Levelt, 1994). Then,
the activation of the semantic representations in the conceptual system spreads to
the lexical system, activating proportionally the corresponding lexical nodes (but see
La Heij, 2005 for a different view). Therefore, multiple lexical candidates are activated
at the lexical level.
(2) However, a subsequent study on language switching has shown that even balanced
bilinguals produce an asymmetrical pattern of switch costs if one of the languages
in the task is extremely weak such as their fourth language (L4) or a recently learned
language during the experiment (Costa et al., 2006).
212 Part 5: The Bilingual Brain