|1a.||Which customer did the secretary call on the office phone?|
|1b.||Which article did the secretary call on the office phone?|
|2a.||Which child did your brother remind to watch the show?|
|2b.||Which movie did your brother remind to watch the show?|
The sentences in 1 and 2 differ in that the verb remind in 2 is typically used with both an object and an infinitival complement (here: to watch), while call in 1 does not typically take an infinitival complement. 1b stops making sense at call (the secretary called the article would be nonsensical), and 2b stops making sense at to. In 2b the filler which movie is not plausible as the object of remind, but it could be the object of an infinitival complement, and so the reader is not forced to consider it as the object of remind. But the sentence does require an object for remind and once to is read it becomes obvious that it does not have one.
In the original experiment reported in Tanenhaus et al. (1989), the investigators asked their participants to take part in a word-by-word reading task in which they were required to press a “Yes” button to get the next word or a “No” button as soon as the sentence stopped making sense. They found that for call-type sentences such as 1, there was a plausibility effect beginning at the verb, so that sentences like 1b started getting “No” responses from the verb onwards, and for those sentences that didn’t get a “No” response till later in the sentence there was a clear reading time difference, with words from the verb onwards taking longer to process in the b versions. For remind-type sentences, the plausibility effect started at to, when syntactic information makes it clear that the filler must be the object of the verb, and the reading time differences started one word later.