Our attempt to model conversations led us naturally to make use of logic. Contrary to traditional use of logic in linguistic semantics, we never try to establish the truth of any proposition by reference to a given world. For us logic will simply be a language to represent part of the meaning, and a tool to check consistency. The question will not be to know if a given utterance expresses a truth, but rather to represent this utterance using a logical syntax and to see if it is consistent with other statements, as shown in the following example taken from [Tannen 1984:62]:
[ex_Goffman] (from [Tannen 1984])
context: A,B and C were speaking about sociology, and B showed a fairly good knowledge of Erving Goffman's books. A and C are surprised, since they thought this author was known only among specialists.
A1- But anyway. ... How do you happen to know his stuff?
B1- Cause I read it.
C1- What do you do?
A2- [??] are you in ... sociology or anything?
B2- Yeah I read a little bit of it. [pronounced reed]
A3- Hum?
B3- I read a little bit of it. [pronounced red]
A4- I mean were you... uh studying sociology?
B4- No.
A5- You just heard about it, huh?
B5- Yeah. No. I heard about it from a friend who was a sociologist, and he said read this book, it's a good book and I read that book 'n
A6- I had never heard about him before I started studying linguistics.
B6- Really?
A7- Yeah.
In the preceding conversation, B appeared to have a very good knowledge of Erving Goffman's books, which was surprising since they are intended for sociologists. To quote D. Tannen, who is A in this excerpt: "Both C and I expected B to tell how his life - and more likely his work or education - led him to Goffman's books". We know enough to express the " content " conveyed (directly or indirectly) by these utterances, first in English and then with a logical representation:
reply
|
contextual
knowledge
|
| A1,
A2, A4:
|
if
somebody knows E.Goffman's books, then he must be a sociologist. B knows
E.Goffman's books. Is B in sociology?
|
| B1,
B2, B3:
|
B
read some of E. Goffman's books and then B knows them.
|
| B5:
|
B
has a friend, he is a sociologist, and he recommended E. Goffman's books and
then B read these books.
|
| A6:
|
As
long as A was not a sociologist, A did not know E.Goffman's books
|
This simplified version of the excerpt can be represented using a logical formalism[1]:
A1 : knows( X, Goffman's_books) => sociologist ( X )
B1, B2, B3 : read(B, Goffman's_books) => knows(B, Goffman's_books)
B5 : sociologist(friend)
[friends(Y, B) & recommends(Y, B,
Goffman's_books)]
=> read(B, Goffman's_books)
A6 : not sociologist(A) => not knows(A, Goffman's_books)
This kind of logical representation suggests three remarks:
- it is not unique
- it does not capture all of the meaning
- some of its elements are not present in the utterances as they are worded
For instance, the above representation does not make the distinction between all Goffman's books ("his stuff") and the single book mentioned in B5; the three replies B1, B2 and B3 are considered as equivalent; no distinction was made between sociology and linguistics in A6, etc. The context of A1, as given here, is not expressed as such by A. So how can we consider this representation as objective?
The answer is that it is not objective, but that it is possible to reach an intersubjective agreement in each case (as soon as the context is known), like for a usual translation. As a consequence it will be necessary to verify that any interpretation based on a logical translation of a given conversation will remain unchanged with another acceptable translation.
From a technical point of view, such a logical translation of conversations makes use of a logical formalism that is fully in conformity with propositional logic or first order logic. The meaning of connectors like &, =>, not is rigorously the meaning they are given in formal logic. For instance [Peter drinks alcohol => Peter is over 18 years old] only means that a situation where Peter drinks alcohol and where Peter is not over 18 years old would be inconsistent. Names of symbols are mnemotechnic and define an interpretation of symbols (predicates, domains of variable assignments, constants) in the " real " world in which the conversation takes place.
The choice of symbols depends on the precision wanted. For example, in the preceding excerpt, "B knows Goffman's books" can be represented by :
B_knows_Goffman's_books
knows( B, Goffman's_books)
knows( B, L) & books(L) & author(Goffman, L)
etc.
It is very important to understand that the use of logic is not a mere technical convenience, and that it cannot be avoided if we want to study relevance. There is indeed no symbolic representation system which is able to represent the puzzle expressed by A and C in the preceding excerpt, and which is not equivalent to logic (or contains logic as a subsystem).
We must now comment upon the fact that our logical representation of the preceding excerpt is more a representation of what interlocutors " have in mind " rather than a representation of what is actually said. The question is not whether interlocutors do in fact have some context present in mind when they are in conversation. The existence of a shared knowledge is acknowledged by most researchers involved in the field (see [Coulthard 1977:79], [Clark & Schaeffer 1989], [Moeschler 1990], [Taylor & Carletta 1994]). The question will rather be (1) Is the shared knowledge available for conversation analysis ? (2) What are the limits of shared knowledge ? Our answer to (1) is straightforward : any analysis of relevance in conversation is impossible if the shared knowledge is not available. We always work on excerpts for which this shared knowledge is also shared by us (this may be the case if we know the situation perfectly well or if we were present during the interaction). The answer to (2) requires more development.
[1] Here we make use of first order logic to represent this excerpt. Predicates are in italics, variables are in bold. [knows( X, Goffman's_books)] represents the fact that X knows Goffman's books.