The time course of inferences in language comprehension

We often communicate much more than we actually say when make an utterance. For example, saying "Dave has a nice personality" when asked whether he is physically attractive can communicate that Dave is not terribly handsome, even though the words say nothing of the kind. As human beings we are able to augment linguistic input with inferences, thereby speeding up communication to the extent that many fewer words are needed than might otherwise be necessary. However, it is not clear how we know which particular inferences to generate after hearing a particular expression, nor how we incorporate the inferences in real time while hearing sentences. To continue the previous example, how do we know to infer that Dave is not attractive (rather than not having red hair, or not being tall, or not being from London etc)? And what kind of procedures operate on the sentence to generate this inference? Different theories make different predictions about the length of time it takes to understand sentences with inferences. To learn which is the most appropriate theory therefore, we will measure the time it takes to understand the sentences.