| |
Abstract:
One of the most effective tests for distinguishing a
morphological structure from a syntactic one is based on the
observation that only the latter allows its elements to bear long
distance dependencies. The anaphoric island constraint (Postal,
1969) is one case of this distinction. For example, in a sentence
like (1), the co-reference between the noun "devil" and the pronoun
"him" is licensed because the co-referrers are both free elements
of the syntax. The same co-reference in a sentence like (2) is
forbidden because "devil" is a constituent of the compound word
"daredevil" but not of the syntactic structure itself. In other
words, the compound "daredevil" behaves as an atomic unit with
respect to syntax. This account of the contrast between (1) and (2)
directly reflects a desirably constraining hypothesis about
grammatical computation: at any given level, representations and
processes specific to previous levels are inaccessible. On the
other hand, a number of psycholinguistic studies have provided
evidence that multiple levels of grammatical structure interact
with each other during the course of sentence processing (see
Tanenhaus and Trueswell, 1995, for review). These approaches have
mainly explored interactions between syntax and lexicon, syntax and
semantics, and syntax and pragmatics. Here, I describe a study,
using Italian V+N compounds of the type "daredevil", which explores
a new domain of potential interaction: the interaction between
syntax and morphology.
The two sentences in (3) contrast with respect to whether the
pronoun (le/li) agrees in gender with the second subconstituent of
the preceding Verb-Noun compound (porta-letter-e, 'letter
carrier'). In a self-paced word-by-word reading paradigm,
participants read through the region between word +3 and word +4
significantly faster when the pronoun agreed with the noun
"letters" in gender and number ("le") than when one of the two
features did not agree ("li"). The effect cannot be due merely to
the formation of a conceptual connection between the pronoun and
its antecedent because, in a semantically identical control case
(4), the effect completely disappeared [Interaction: F1(1,31)=5.9
p=.021, F2 (1,19)=4.87 p=.04].
This differential sensitivity suggests that the syntactic
processor is affected by the possibility of a "non-atomic" reading
of the sentence parsed. The presence of a subtle "false" agreement
is sufficient to distort the activity of the parsing system pushing
it toward a "partial crossing" of the morphology-syntax barriers.
The compound "portalettere" [carry-letter-F/PL] is read as if it
were a syntactic construction, perhaps something like a relative
clause "che porta le lettere" [who carries the letters]. The result
is consistent with other evidence that subtle violation of
grammatical constraints by the sentence processor can be detected
using sensitive psycholinguistic measures (e.g., Tabor and
Richardson, 1999.)
(1) Mary, who used to dare the devil(j), is now scared by
him(j).
(2) *Mary, who used to be a daredevil(j), is now scared by him(j).
(3) Un porta-letter-e disonesto l-e/l-i apre con la speranza di
rubare qualcosa.
-3 -2 -1 0 1 2 3 4 5 6 7
A carry-letter-PL&F dishonest PL&F/PL&M open with the
hope of stealing stuff
'A dishonest mailman opens them in the hope of stealing
something.'
(4) Un postino disonesto l-e/l-i apre con la speranza di rubare
qualcosa.
-3 -2 -1 0 1 2 3 4 5 6 7
A mailman dishonest PL&F/PL&M open with the hope of
stealing stuff
'A dishonest mailman opens them in the hope of stealing
something.'
|
