[LMB] OT: Implicature (second and last part)

Jim Parish jparish at siue.edu
Sat Jun 10 01:24:43 BST 2006

Okay, so how does conversational implicature work? The basic idea is 
that ordinary conversation is a cooperative activity, and there are 
certain - unspoken, even unconscious - conventions which are normally 
in operation, which serve to ease the process of conveying information. 
(At a certain level, people think faster than they talk, so reducing the 
amount of speech by requiring additional processing by the listener is, 
within limits, a good tradeoff.) My understanding is that these 
conventions seem to be pretty much the same in all languages, though 
I have a vague memory that some researchers think Malagasy does 
something differently - what, I don't know.

There've been various formulations of the conventions; H. P. Grice 
made the first attempt in 1957, and has offered several variants since 
then. Stephen Levinson's version, the one I'm familiar with, offers the 
following heuristics; I'll state them first before expanding on them.

Q: What isn't said, isn't.
I: What is expressed simply is stereotypically exemplified.
M: What's said in an abnormal way is abnormal.

The Q-heuristic basically means that, if I make statement A, and I could 
have made statement B, which is "stronger" than A in some sense, then 
my failure to say B implicates that B is not true. For example, if I say, 
"Some of the boys came to the party", the fact that I didn't say "All of the 
boys came" suggests that this is false: not all of them came. I say, 
"John tried to climb the mountain", and I don't say "John climbed the 
mountain"; the implicature is that he failed to climb the mountain. Or 
again, "I have a red-headed sister" implicates that I don't have two - 
because if I did, I should have said so. In each case, the implicature is 
cancellable: "Some of the boys came to the party. Come to think of it, 
all of them did." "John tried to climb the mountain, and eventually 
succeeded." "I have a red-headed sister; in fact, I have two."

The I-heuristic means that, if there's a situation which is more typical, in 
context, than any alternative, that's the one which should be assumed 
unless the speaker says otherwise. For example, if I make a reference 
to a knife without specifying what kind, the listener will assume that I 
mean whatever kind of knife "fits" the discussion - if I'm talking about 
cooking, any reference to a knife will be taken to mean a kitchen knife, 
not a machete, and contrariwise if I'm talking about a safari. Again, if I 
say that I'll give you $5 if you mow the lawn, you'll probably assume that 
if you don't, I won't; the "if-then" becomes an "if-and-only-if", because, 
most typically, one doesn't give money except in exchange for goods or 

The M-heuristic is sort of the flip side of (I); if I say something in an 
unusual way, it suggests that something unusual is intended. "John got 
the car to stop" - what does this mean? If "John stopped the car" takes 
the stereotypical interpretation, that he used the brake pedal, then the 
wordier form suggests that that's *not* how he did it; maybe he turned 
onto an upgrade and took his foot off the accelerator. I say, "The 
corners of Sue's lips turned upward"; since I don't make the simpler 
statement, "Sue smiled", I'm suggesting that her expression was 
something different, not quite a smile.

So, to get back to Pete's original point: doesn't "last" mean "last out of 
more than two"? Well, perhaps, but I'd be inclined to think that it's a 
matter of implicature, rather than part of the inherent meaning of "last". 
(Remember the Cold War joke, about the competition between a 
Russian and an American car? The Russian press trumpeted the result, 
saying that the Russian came in second, and the American was next-to-

I want to make one other comment here. There are certain situations in 
which the conventions of implicature don't apply. First, young children 
often violate them, because they haven't internalized them yet. Second, 
there are contexts in which "conversation" is recognized as *not* being 
cooperative; Bill Wenrich mentioned courtroom dramas, which involve 
deliberately antagonistic "conversation"; press conferences and other 
forms of interrogation also fit here, and the conventions don't apply.

Third - and this is the one which I, as a mathematician, find most 
interesting - mathematical and, to some extent, scientific prose often 
violates the conventions. (I could make some guesses as to why this 
developed, but I'll forebear.) A lot of the differences between the 
conventions of formal logic and those of ordinary speech can be 
explained in these terms. The fact that the formal "or" is inclusive - 
what, in ordinary speech, is sometimes presented as "and/or"; the 
sharp distinction between implication ('if-then") and equivalence ("if-
and-only-if"); the fact that the existential doesn't exclude the universal 
or the singular ("some", to a mathematician, means "at least one, and 
possibly all", rather than "more than one and less than all") - all of these 
can be viewed as the failure of the conventions of implicature to apply.

At any rate, that's a general overview of conversational implicature. 
There's a lot more that could be said; Levinson's _Presumptive 
Meanings_ devotes nearly 400 pages to its ramifications, not counting 
the footnotes. I think, though, that I've said enough for now, unless 
someone has questions or comments.

Jim Parish

More information about the Lois-Bujold mailing list