Previous research has suggested that attraction errors are not due to the proximity of the local noun and verb, as a more distant local noun can result in more errors than a nearer one (e.g., *The helicopter for the flights over the canyon are vs. *The helicopter for the flight over the canyons are; Franck, Vigliocco, & Nicol, 2002). However, the verb tends to agree in number with the nearer noun of a disjoined subject, suggesting that linear order can indeed play a role in agreement computation (e.g., The horse or the clocks are vs. The horses or the clock is; Haskell & MacDonald, 2005). In the present study, two experiments using a two-alternative forced-choice production paradigm and one experiment using eyetracking during reading directly compared agreement computation in the classic attraction configuration and when the subject is a coordinate phrase. The 2AFC experiments replicated both the lack of a linear distance effect in classic attraction and the presence of a linear order effect in disjunction agreement, which was also extended to conjunction agreement; when the second conjunct was singular, subjects frequently selected a singular verb. This order effect was also modulated by the presence or absence of additional material between the subject and verb. In the eyetracking experiment, a singular second conjunct both facilitated processing of a singular verb and inhibited processing of a plural verb. These results suggest that variable agreement with coordinate subjects is not a form of agreement attraction and that distinct theoretical treatments are required for two distinct phenomena.