vice versa). A critical manipulation in this type of studies is to include pic-
tures whose initial consonant or vowel status is identical in the two lan-
guages of the bilingual (e.g. consonants in both languages) and whose status
is not (e.g. a consonant in one language and a vowel in the other). If bilin-
guals activate the phonological form of the picture name in both languages,
they should experience a greater difficulty when the status of the initial
letter is different across two languages. Rodriguez-Fornells et al. (2005)
found this effect of overlap in two languages, which was reflected in
increased negativity from 300–600 ms (N2) for ‘go’ trials but not for ‘no-go’
trials. This result suggests that alternative candidates from both languages
are active and compete for selection at the phonological level.
Other ERP studies on bilingual production have adapted delayed naming
paradigms in which participants are asked to name pictures immediately for
some trials but withhold their response longer in others (e.g. Guo & Peng,
2006; Jackson et al., 2001; Misra et al., 2012). The logic behind delayed
naming is that ERPs recorded in the case of delayed responses reflect the
process of speech planning before articulation commences. Trials requiring
an immediate response are discarded from analysis because of heavy motor
contamination, while keeping the participants actively engaged in the task.
However, it is important to note that go/no-go and delayed naming para-
digms require more cognitive control over actual speech planning than
immediate naming because speakers have to withhold their responses and
little is known about interactions between motor inhibition and speech plan-
ning mechanisms. Given that ERPs are typically artifact-free until partici-
pants articulate their response, it is possible to study early ERP effects
associated with speech planning in a context of immediate naming. However,
it is plausible to assume that observed differences between conditions in
immediate naming experiments might have resulted from differential motor
preparation (i.e. prior to speech onset) in each condition. The question then
is whether such differential activities are solely driven by cognitive differ-
ences between experimental conditions or also affected by differences
between stimulus sets used in the different conditions. In the former case,
we are still measuring differences between conditions triggered by the cogni-
tive processes under scrutiny. The latter case is unlikely when stimuli in the
different conditions have been adequately controlled for lexical properties
such as length, frequency, and initial phoneme, because there should be little
baseline shifts introduced by the articulatory programming of stimuli with
comparable characteristics. If the stimuli used in all the conditions are the
same (which is only possible for specific experimental designs), no baseline
shift can theoretically be accounted for by spurious differences between
stimulus sets (e.g. Hoshino & Thierry, 2011).
Immediate naming paradigms have been adapted recently not only to
within-language production research (e.g. Costa et al., 2009; Koester &
Schiller, 2008), but also to bilingual production research with ERPs (e.g.
Language Selection During Speech Production in Bilingual Speakers 209