From a theoretical point of view, these ideas are very powerful and
general, although of course they run into all kinds of difficulties as a
theoretical model, and these are well known. However, what we are concerned
with here is their appropriateness as a model for understanding
foreign language learning, and for producing coherent and interesting
language teaching materials. They relate well, for example, to the influential
view of reading as a psycholinguistic guessing game (Smith, 1973), with all its
implications for teaching reading. Leaming a foreign language is seen,
therefore, as acquiring preditive competence. Halliday {1978, p. 200) points
out that there are certain things which are particularly difficult for a speaker
of a foreign language. These include: {1) saying the same thing in different
ways; (2) hesitating and saying nothing much; and (3) predicting what the
other person is going to say. These aspects of linguistic competence are all
closely related; they all have to do with understanding and producing
language in discourse under the constraints of real time. They all also
concern ways of exploiting the redundancy of natural language in use.