Now for a sketch of Grice's reduction of sentence meaning to speaker-meaning, and indications of how he would have approached Obstacles 1--4 had he been fully aware of them.<9>
    He first concentrates on the narrow notion of: sentence meaning for a particular individual, that is, the meaning that the sentence has in that individual's personal, distinctive speech or idiolect.  (No two English speakers' idiolects are exactly alike.)  And he restricts his initial target further, distinguishing structured utterances from unstructured ones.  A structured utterance has meaningful parts, such as individual words, which contribute to the utterance's overall meaning; any declarative English sentence is an example of this, since it contains words that are individually meaningful and it means what it does in virtue of those words' meaning what they do.  An unstructured utterance is a single expression or nonverbal gesture, such as "Ouch" or a beckoning motion that means "This way," whose meaning is not compositional in that sense.  (Note that Grice uses the term "utterance" very broadly, as including nonverbal communicative acts.)
    After some backing and filling, Grice hypothesizes that x [an unstructured expression] means that P in S's idiolect, just in case (roughly) S has in her/his repertoire the following procedure: to utter x if, for some audience A, S intends A to believe that S believes that P.  (That last clause is a simplified version of "S speaker-means that P"; Grice argues that the simplification is harmless here.)
    Now he expands this analysis to cover utterance meaning for a group of speakers:  x [unstructured] means that P for group G just in case (a) many members of G have in their repertoires the procedure of: uttering x if, for some A, they want A to believe that they believe that P, and (b) the retention of that procedure is for them conditional on the assumption that at least some other members of G have that same procedure in their repertoires.
    I think what are supposed to overcome Obstacle 1 are (a) and (b), that the relevant procedure is widespread in the community and that individual members of the community rely on the other members to maintain that procedure as well.  This seems exactly right.
    But now the trick will be to go from the analysis of unstructured-utterance meaning to ordinary sentence meaning, since ordinary English sentences are all structured.  Grice brings in the notion of a "resultant" procedure.  At this point Grice's article becomes dense and obscure, but I think the idea is this:  Just as English sentences are made up of smaller meaningful parts -- words and phrases -- in virtue of which the whole sentences mean what they mean, an individual speaker will have in her/his repertoire a complex, abstract "resultant procedure" made up of the concrete procedures attaching to its respective composite parts.  Thus, a sentence's meaning will not be directly a function of speaker-meaning, but rather a function of the individual utterance meanings of its ultimate parts.  Only then will the core Gricean idea, and (crucially) his analysis of utterance meaning for a group, be invoked as explicating the utterance meanings of the parts.
    I emphasize "abstract resultant procedure," because very few of those "abstract" procedures will ever actually occur.  And it is that feature that will help Grice with Obstacles 2 -- 3.  For the theme of those obstacles is that unuttered and novel sentences do not correspond to any actual speaker-meanings.  But at least arguably, they do correspond to the hypothetical speaker-meanings that would be generated by Grice's abstract resultant procedures.
    The appeal to abstract procedures may also help to overcome Obstacle 4: Even though a certain sentence's literal meaning is never matched by any actual speaker-meaning, it may still correspond to a hypothetical resultant speaker-meaning.
    Yet I believe that this absolutely necessary appeal betrays the spirit of the Gricean program.  In effect, it gives the game away to a competing theory of meaning; I shall argue that in Chapter 9.
 

Footnote


9.  Schiffer (1972, Chapters V-VI) pursued a different method, employing Lewis (1969)’s theory of conventions.