54 A Grammar of Spoken English Discourse
argue that speakers only need to check back three levels. Mutual know-
ledge is secured if:
A speaker knows that the hearer knows that the speaker knows that t is R.^6There are two main diffi culties with this approach. The fi rst is that it is pos-
sible to construct scenarios^7 where such limited regression cannot guarantee
mutual knowledge (Clark and Marshall: 1981: 13). The second diffi culty is
that restricting the speaker’s regression to three levels is arbitrary. Lee
(2001: 35) notes that Bach and Harnish fail to produce any evidence
indicating that individuals engaged in communication limit mutual beliefs
to three levels. Indeed other scholars, again without producing much
evidence, have theorized that people engaged in communication make
replicative assumptions to four or fi ve levels (Harder and Kock 1976: 62)
or to six levels (Kaspar 1976: 24). In any case, Clark and Marshall (1981: 14)
produce a scenario which demonstrates that regression to fi ve levels is
not always capable of securing mutual knowledge and argue that it is
possible to produce scenarios which demonstrate that regression to six
or more levels does not always secure mutual knowledge.
The second purported fi x is co-presence heuristics (Clark and Marshall 1981:
32–43) which argues that speakers do not engage in either an infi nite or
limited regression of checks in order to secure the mutuality of their know-
ledge. Instead they assure themselves of the mutuality of their knowledge
by seeking independent confi rmatory evidence. This evidence is of three
kinds: community membership; physical co-presence; and linguistic co-presence.
In other words, by utilizing independent evidence communicators are able
to infer mutual knowledge.
There are, however, a number of potential problems with this view.
First, an inference of mutual knowledge is not itself a hundred per cent
certain and thus, by defi nition cannot itself amount to mutual knowledge
(Sperber and Wilson 1995: 19). Second, speakers can have stronger or
weaker supporting evidence of mutual knowledge (Wilks 1986: 268). A
direct experience such as physical co-presence is stronger evidence than
an indirect experience such as linguistic co-presence or a background
experience such as community membership. But as mutual knowledge
requires one hundred per cent certainty, the evidence securing it cannot
be stronger or weaker. Third, Clark and Marshall’s fi x does not appear to
be applicable to all types of language use. Consider:
(3) A man and a woman are both looking at a painting.
She says as she points at the painting: ‘That is a Picasso’.