| |
Abstract:
Language requires a mental lexicon of the memorized sounds
and meanings of words, and a mental grammar of rules that underlie
the predictable composition of lexical forms into larger words,
phrases, and sentences. On a modular view, the two language
capacities are separable, with the lexicon dependent upon left
temporal-lobe structures, and aspects of grammar upon left frontal
structures (1, 2). A contrasting perspective holds that a single
associative memory with broad anatomic distribution subserves all
words and rules (3, 4).
These opposing viewpoints have been difficult to resolve, in
part because the investigation of grammar has focussed on syntactic
computations in sentence processing, which are difficult to match
against tasks probing lexical processes. For example, distinct
Event-Related Potential (ERP) components are associated with
syntactic processing (an early Left Anterior Negativity [LAN] and
later positivity [P600]) and with lexical-semantic processing (a
central negativity [N400]). However, the conditions by which they
are elicited are quite different, for example word-order violations
as compared to conceptually unexpected words.
We therefore investigated the domain of morphology, which is
well-suited for a matched comparison of lexical and combinatorial
processes. On a modular view, highly productive morphophonological
processes such as regular -ed-suffixation are computed by grammar
in left frontal regions, whereas largely unproductive patterns,
such as those found among irregular past-tenses, are memorized and
are dependent upon left temporal-lobe structures (1, 2). A
competing view holds that all forms depend on associative memory
(3).
We recorded ERPs while right-handed men viewed sentences with
and without violations of lexical-semantics, of syntactic
phrase-structure (after 5), and of regular and irregular past-tense
morphology (e.g., Yesterday I walk over there; Yesterday I dig a
hole.) (see 6, 7).
In an early time window (300-500 ms), violations of regular
morphology and phrase-structure yielded LANs. Violations of
irregulars and lexical-semantics yielded negativities that were
significantly more posterior than those associated with regulars
and phrase-structure, respectively. The negativity associated with
irregulars had a left temporal-occipital focus that was
significantly more left-lateralized than the N400 associated with
lexical-semantics.
In a later time window (600-800 ms), violations of regulars,
irregulars, and phrase-structure, but not of lexical-semantics,
yielded P600-like positivities. Intriguingly, the positivity
associated with irregulars was significantly more anterior and
laterally right-shifted than that associated with phrase-structure
violations. This finding, in conjunction with the left posterior
focus of the earlier negativity associated with irregulars, appears
consistent with the idea that violations of irregulars lead to the
temporally overlapping activation of N400 and P600 neural
generators, and that, due to additivity effects (see 8), the
associated scalp distributions of these negative and positive
components are largely canceled out, leaving spatially distinct
maxima.
These data (i) indicate that productive (regular) and
unproductive (irregular) morphological processes are associated
with distinct electrophysiological correlates; (ii) suggest that
the processing of regular (but not irregular) morphophonology and
of syntax share a neural (and functional) substrate, as revealed by
LANs; (iii) are consistent with the view that syntactic violations
of tense (for both regular and irregular verbs) as well as of
phrase-structure activate the generator(s) of the P600, suggesting
common processing mechanisms; and (iv) appear to support the view
that violations of irregular morphology and lexical-semantics
activate generator(s) of the N400, suggesting a shared substrate,
which is rooted in left temporal-lobe structures (9).
|
