| |
Abstract:
The language ERP literature has witnessed much debate in
recent years over the existence and interpretation of left anterior
negativity (LAN). One view is that LAN is associated with
(morpho)syntactic ill-formedness, the other that it is associated
with verbal working memory. We present evidence that both
interpretations are correct: our German-language materials elicited
separate LAN effects of both types.
We compared grammatical subject and object wh-questions into
complement clauses without complementizers ("Wer/Wen meinst du
soll...?"), and ungrammatical subject wh-questions into complement
clauses with complementizers (i.e. that-trace violations: "Wer
meinst du dass...?"). Object questions elicited sustained left
anterior negative potentials relative to grammatical subject
questions at two points in across-sentence averages: at the subject
gap position (the auxiliary 'soll') and after the earliest possible
object gap location in the embedded clause. A time slice of this
sustained negativity also appeared in single-word averages at the
auxiliary 'soll'. This is consistent with previous findings
associating LAN with the greater working memory load of filler-gap
configurations in object wh-questions and relative clauses, and
equating LAN effects seen in single-word and across-sentence
averages.
Ungrammatical subject questions (that-trace violations) elicited
no such reliable across-sentence effects relative to grammatical
subject questions in the first half of the sentence. They
nevertheless elicited a LAN effect in single-word comparisons of
the complementizer `dass' and the auxiliary 'soll'. This is
consistent with previous findings associating LAN with the
detection of (morpho)syntactic violations. This LAN was followed by
a late positivity to the next two words, and a sustained right
posterior negative potential during the second half of the
sentence.
We thus believe we have dissociated two separate LAN effects, one
local, (morpho)syntactic, and unrelated to slow potentials, and the
other a global slow negative potential seen in both across-sentence
and single-word averages in the context of increased working memory
load.
Likewise, we suggest that it is time to ascertain whether
sustained right posterior negative potentials following grammatical
violations can be equated with or dissociated from the N400 effects
seen in single-word averages, an issue that has remained largely
unresolved in the literature.
Finally, while the ERPs to that-trace violations in our data
(local LAN followed by local late positivity) were consistent with
a two-stage parsing model, we argue for alternative interpretations
of these same effects. In particular, we question the validity of
equating effects of late positivity with a second stage of
syntactic reanalysis.
|
