ABSTRACT
It is well known that speakers and writers cannot make explicit everything that they want to communicate. Nor do they need to. Instead, they rely on their listeners or readers to fill in whatever gaps may exist in the message; in other words, they rely on their audience to make inferences. The question confronting researchers of language comprehension is: Which inferences can people be counted on to reliably draw? This question is particularly important for reading research because the reader, unlike most listeners, is not in a position to ask for assistance in gap-filling from the author. If we can answer this question of which inferences are likely to be drawn, not only will writers (like us) be able to write more comprehensible texts, but we will be well on our way to a general theory of language comprehension. And it is only when this question is answered that we will be able to solve such practical problems as designing machines that we can converse with.
