Listening to people reading 217A study of this activity (Brazil 1984b) led to two observations, the interpretation
of which does have an important bearing upon present concerns. When
subjects were asked to read out lists of sentences presented to them simply
as sentences:
(a) there was a high measure of conformity among the performances of
various readers when they were confronted with certain sentences, but
great variation when they were confronted with others;
(b) when readers did agree, their performances could often be manipulated
in an equally predictable way by making some small grammatical or
lexical change in the item they were asked to read.
As an instance of (b), compare the preferred readings:
1 //r WHERE it CAME from //p is a MYStery //
2 //p WHERE it CAME from //r is the MYStery //
Differences in tone choice seem here to be dependent on whether a definite
or an indefinite article is used. A similar alternation can be elicited by
varying the adjunct:
3 //r preSUMably //p the WEATHer improved //
4 //p adMITTedly //r the WEATHer improved //
The initial temptation to ascribe the differences between (1) and (2) simply
to the reader’s apprehension of a grammatical change, and to seek correlations
between intonation and grammatical constituency, has to be resisted for at
least three reasons:
1 A similar difference can be elicited by a lexical change as in (3) and (4).
2 The choice of either definite or indefinite article does not actually
compel a particular phonological treatment; discourse contexts can easily
be invented in which the alternative versions of both (1) and (2) are
perfectly natural.
3 Both these pairs and an indeterminate number of other pairs in which
different kinds of alteration are made can be given identical explanations
if we make direct reference to the context of interaction.
The explanation takes the following form. For readers, the material to be
read is part of the context of interaction. When that material comprises
uncontextualized sentences it is the only basis they have on which to construct
a conversational setting for what they read out. Some details of the organization
of the sentence can powerfully suggest—in terms of probability—what that
setting would be like, and these suggestions may be pertinent to the choice
of particular intonation features. Thus, the occurrence of the in (2) suggests
that the idea of a mystery has already been introduced into the conversational
setting that the reader shares with some (unidentifiable) hearer, and is thus
available to be referred to. The change to a, on the other hand, suggests a