If everyone knew in advance exactly what a person was going to say, he wouldn't need to say it. Communication occurs only when his message is, at least to some degree, unpredictable. When we talk about communication, therefore, we are talking about unpredictable events. The way we talk about more or less predictable events, of course, is to use the theory of probability. And that is how the theory of probability gets involved in our theories of communication.
This chapter discusses some of the probabilistic aspects of linguistic communication. In particular, it reviews some of the psychological research on language that was stimulated during the 1950s by the realization that linguistic communication really is a probabilistic affair. I will point out some of the reasons people became dissatisfied with this line of research. And I will suggest another way to look at the relation between language and probability -- a way that fits better with what we now know about how people produce and interpret linguistic messages.
Much of the psychological research on language during the 1950s was stimulated by the ideas of a man who was neither a psychologist nor a linguist, but a communications engineer -- Claude Shannon. In his famous monograph, A Mathematical Theory of Communication, which appeared in 1948, Shannon