exclusively depended on linear information, we should expect to see agreement
attraction shown in (2a) and (2b) in equal proportion:
(2) a. ∗The keys to the door is missing.
b. ∗The key to the doors are missing.
In comparison to (2b), however, agreement attraction as in (2a) is rarely induced
in production (Eb erhard, Cutting and Bock 2005), nor do speakers fail to notice
the unacceptable subject-verb agreement in comprehension (Wa gers 2014, Le wis
and Phillips 2015). Only intervening plural nouns attract agreement as in (1) and
(2b). Second, agreement attraction is selective (Ph illips et al. 2011): agreement
attraction results in illusions of grammaticality in cases like (1) but does not
induce illusions of ungrammaticality in cases like (3):
(3) The key to the doors is missing.
If linear distance were the only relevant factor in language production, speakers
would judge (3) as unacceptable since the plural noun doors should induce plural
agreement on the verb. However, illusions of ungrammaticality do not occur
in grammatical sentences like (3). Third, agreement attraction is elicited inside
a “plurally headed relative clause”:
(4) ∗The squirrels that the cat are methodically stalking... (Wa gers 2014)
In (4), the plural head noun squirrels induces the plural number agreement on the
verb, which indicates that the linear proximity of the noun with respect to the verb
is not responsible for agreement attraction. Fourth, Fr anck, Vigliocco and Nicol
(2002) show that when the subject contains two stacked PPs as in (5), a plural
noun hierarchically closer to the head of the subject as in (5a) induces more agree-
ment attraction errors than a plural noun linearly closer to the verb as in (5b):
(5) a. ∗The inscription on the doors of the toilet are...
b. ∗The inscription on the door of the toilets are...
This again suggests a role of hierarchical structure in language processing. Fifth,
Fr anck, Vigliocco and Nicol (2002) conduct an elicitation experiment and report
that English native speakers make agreement attraction errors in interrogative
sentences like (6a) at almost the same rates as in declarative sentences like (6b):
(6) a. ∗Are the helicopter for the fl ights safe?
b. ∗The helicopter for the fl ights are safe.
The polar interrogative changes the relative word order of the attractor and the
verb, but elicits agreement errors at almost identical frequency. This suggests
Syntax in the brain 219