Listening to people reading 221//p he COULDn’t //r have been misTAken //
//r the BOOKS //p were still lying on his TAble //
//p it must have been PEter //r they had SEEN //If these three sentences are presented as a text, and the reader’s decisions
are made on a sentence-by-sentence basis, taking in the situation that has
been created by earlier sentences, we might expect something like:
//p he COULDn’t //r have been misTAken //p the BOOKS
//p were STILL lying on the TAble //r it MUST have been
//p PEter they had seen //In this version, which is probably only one of several that could be invented,
the books are treated as the reason for saying that he couldn’t have been
mistaken and are therefore proclaimed; and the judgement made by must
have been has referring tone since it reiterates that implied by he couldn’t.
It follows from this kind of argument that the first sentence retains its Stage
3 form. It would seem, indeed, that someone beginning to read a so-far
unseen text, whose relationship to the known interests of the hearer is not
yet determined, has little option but to adopt temporarily a Stage 3 stance.
At this point, two other areas suggest themselves as ones which might
be fruitfully investigated. One arises directly from what was said at the
beginning of this section. It seems not unlikely that, under the pressure of
real-time decision-making that reading evidently involves, readers will
sometimes find themselves in a state of indecision as to whether they are
processing a text on a sentence-by-sentence basis—an essentially short-
term business—or processing it as a continuing entity. When Stage 4 engagement
favours a form which differs radically from that which would be favoured
by Stage 3 involvement, they might well have difficulty in deciding between
the two. Supposing there were such a problem, we might ask what effect
it would have on the phonological shape of the outcome. One possibility
would be a temporary movement down the scale of engagement to Stage
2 or even Stage 1. This, we have already suggested, is what commonly
happens when either speakers or readers encounter cognitive problems. In
the circumstances we have envisaged, such recourse would relieve readers
of the necessity for choosing and allow them time to re-establish their
bearings in the present state of speaker/hearer convergence. Of course,
only an extensive study of intonational features in the neighbourhood of
errors, hesitations, self-corrections and so on in a large amount of data
would give empirical substance to this speculation. I include it here, not
because I can prove it, but rather as an example of the kind of question
the present approach enables us to bring into focus.
Much the same is true of the other area. In the first of the examples cited
above, the intonation of one tone unit is related to a feature of the state of
speaker/hearer convergence that is articulated in the tone unit immediately