conventional implicature?

[George Lakoff]

Dear Marta,

        To understand my response to your question, a bit of history is in
order.

In the late 1950's and 1960's, there was a split in British analytic
philosophy between the ordinary language philosphers (following Austin and
Strawson) and the formalist analytic philosophers, mostly followers of
Russell. The Russellians claimed that "scientific philosophy" could only be
done using symbolic logic, which they claimed characterized reason. The
ordinary language philosophers Argued that that ordinary language was fine
for doing philosophy and sought to show its regularities.

Grice thought both camps were right and that there was a way to keep
Russellian logic
and also keep the insights of ordinary language philosophers (like
himself). Grice's William James Lectures at Harvard in 1967 were titled
"Logic and Conversation."
I was fortunate enough to be present, since I was teaching linguistics
there at the time.

Grice's idea was that meaning could be split up into two parts: (1)
Symbolic logic as Russellian had described it, including his classical
theory of descriptions; logic was to characterize logical inferences. (2)
Principles of conversation (conversational implicatures) that were to
characerize other, nonlogical inferences which (so Grice claimed) arose
when language was used in conversational context. Grice called them called
"implicatures" to distinguish them from "implications." (1) was to define
semantics; (2) was to characterize an major aspect of pragmatics.

At this point, the history of manuscript of Grice's Harvard lectures
becomes important.
He refused to publish the lectures for two decades. I managed to
distribute, through the linguistic underground in the 1960's and 70's,
about  1,000 photocopies of the full book manuscript.Most of the most
prominent people who have written on the subject-Steven Levinson, Georgia
Green, Larry Horn, and so on-got into the subject during that period
and had access to the full manuscript.

The most popular version of Grice's ideas were distributed in a chapter of
Logic and
Conversation in Cole and Morgan's "Speech Acts" volume, which Grice has
refused to publish until he got drunk at a party in 1973 at a conference in
Austin. I suggested to Cole that he have a contract ready, which he did,
and Grice signed on the dotted line.  The entire manuscript has since been
published in Studies in The Ways of Words, Harvard University Press, 1989.
But there was a big gap between 1975 (when the chapter in Cole and Morgan
appeared) and 1989 when the whole manuscript appeared (to little fanfare;
it was largely ignored). Grice's maxims entered the published linguistic
literature without their context.

If you read the entire manuscript, it becomes clear that Grice's aim is
very conservative: to preserve Russell's outmoded theory of symbolic logic
in the face of the counterexamples presented by the ordinary language
philosophers. One of those counterexamples had to do with presuppositions,
brought to the fore by Strawson in the 1950's and put into logic in
nonRussellian terms by Bas Van Fraassen in 1968.
Grice tried bravely to preserve Russell's theory of descriptions, but he
got seriously hung up on a real linguistic phenomenon: lexical
presupposition.

Grice discussed only one example: "but." He brought up "but" because it was
a sentence connective that did not work according to Russell's theory.
Russellians translated it into symbolic logic as "and". But "but" does not
mean "and" and Grice knew it! (The previous sentence is an example of why.)
Grice's only way to deal with cases outside of Russellian logic was his
four types of implicature. But "but" did not fit any of them. Moreover, the

 counter-to-expectation pragmatics of "but" was not a feature of
conversational context, but was part of the the conventional meaning of the
word. This
did not fit the neoRussellian paradigm, since the meanings of words were
supposed to be
in the realm of semantics, not pragmatics.
        So Grice resorted to what can best be referred to as a "fudge" or a
"kludge," that is, an adhoc nonsolution: conventional implicature. This was
a name for a problem that,
strictly speaking, did not fit the theory Grice was proposing. What
"conventional implicature" does is break the conventional meaning of "but"
into two parts: a logical part that means "and" (so it will fit Russell's
theory) and a pragmatic part that has the counter-to-expectation meaning.
Since this pragmatic part is not detachable in context, as implicatures are
supposed to be, it is called a "conventional implicature" and
conventionally associated with the word "but". The move was made strictly
to preserve Russellian logic.
        The reason that generative semanticists got involved with the
Gricean program
in 1967 and in successive years is that we too were trying to keep a
version of formal logic for semantics, while trying to deal with the real
facts of language, which included an enormous amounted of pragmatics. By
1969, a huge range of presuppositional phenomena, both grammatical and
lexical, has been discovered by linguists, and now, almost thirty years
later, the range is astronomical.
        Some linguists are still trying to keep to a theory in which
semantics is characterized by formal semantics and pragmatics is done by
different, conversational principles. Many of those scholars are still
trying to make sense of a Gricean notion of "conventional implicature."
Those of us who gave up on that paradigm long ago (or never had it) simply
study presuppositional and other pragmatic phenomena as part of cognitive
and/or functional linguistics. For us, the notion "conventional
implicature" is an odd anachronism, a holdover from analytic philosophy of
the 60's. But for those still trying to carry out a neoGricean program, the
term is very much alive.

I hope this helps.

George Lakoff




>        Dear 'funknetters':
>
>        I am a teacher of linguistics at the Universidad Complutense,
>Madrid, and one of my subjects is pragmatics, of which my students are 5th
>year undergraduates. Now I am teaching cooperation and implicature, as in
>Grice's proposal, and soon I will come across the notion of 'conventional
>implicature'. I am beginning to think that what comes under the label
>'conventional implicature' (for instance, BUT and HOWEVER carry the
>'implicature' that what follows will run counter to expectations) could
>well be included in the LEXICAL MEANING of these items, since these
>'implicatures' are independent of context and persist in all the uses of
>these words. I would appreciate it very much if some of you could send me
>messages about your views on this subject.
>
>        Thank you very much in advance.
>
>
>Marta Carretero
>Departamento de Filologia Inglesa
>Facultad de Filologia - Edificio A
>Universidad Complutense
>28040 - Madrid.