| |
Abstract:
In the present study, the role of the lexical item in the
processing of Subject-Verb agreement is examined through the lens
of agreement attraction errors (e.g., "The sign on the colorful
cardboard boxes were difficult to read"). Models of agreement
processing normally account for attraction errors by one of two
means: through the action of autonomous grammatical features, or
through the competition that arises between a local noun and the
appropriate agreement controller. On the former model, a feature
that is copied onto (or checked against the features of) a verb
normally originates with the lexical head of the subject NP phrase.
Before the copying/checking process takes place, the feature must
become affiliated with the entire subject NP by means of some
percolation process. Attraction errors arise when a feature that
originates with a local noun ("boxes" in the example) is mistakenly
copied to the entire subject NP by this same percolation process.
In contrast, accounts of attraction errors based on lexical
competition attribute attraction errors to mechanisms for
identifying lexical sources for agreement: If a local noun is
sufficiently active (or accessible) when the agreement properties
of the Subject are being calculated, then this may cause the local
noun itself to interfere with the agreement process. One prediction
of this competition based explanation is that the production of
agreement errors may depend on properties of the local and head
noun that make a lexical item a viable agreement controller. For
example, attraction may depend on whether a local noun is ambiguous
(as a lexical item) with respect to grammatical case, even when its
case status is disambiguated by the preposition that governs the
local noun in the immediate syntactic context (Badecker &
Kuminiak, 1999).
In the experiments presented here, we provide further support
for the competition based account of attraction (and for the role
of the lexical item in calculating agreement) by examining the
effect of conjoined local NPs on the calculation of agreement. In
our first experiment we demonstrate that conjoined NPs with
singular NP constituents (e.g., "the color of the skirt and the
blouse") exert a strong attraction effect in comparison to both
singular and plural local NPs of comparable length (e.g., "the
color of the very fashionable blouse(s)"). We take these results to
indicate that the conjoined NP structure bears a highly visible
mark for number.
In our second experiment, we find that lexical plurals inside
the conjoined NP (e.g., "the color of the skirts and the blouses")
adds to the local NP's ability to attract agreement from the head
of the subject NP. Percolation based accounts would equate all
three cases of plural local NPs, because the attraction error
arises not with NP internal percolation, but with the percolation
of [+plural] from the root of the local NP to the superordinate
phrasal node. At that juncture, though, all three types of plural
local NP will be identical in that all three will have the same
number specification. In contrast, the competition account predicts
this effect for lexical plurals internal to conjoined NPs.
Badecker, W., and F. Kuminiak. 1999. The role of the lexical
item in processing agreement: Evidence from Slovak gender concord.
Unpublished manuscript, Johns Hopkins University.
|
