"Many sentences of English have never been uttered,
and no one has ever meant anything by uttering them.
Therefore their meaning cannot be determined by speaker's intentions."


The quote in the essay title is a criticism (frequently attributed to Mark Platts(1)) that has been levelled against Paul Grice's theory of sentence meaning, as Grice laid out in his "Meaning" article in 1957(2).  Grice's theory starts with the concept of "speaker's-meaning".  According to Grice, a speaker's utterance has meaning according to what the speaker intends to communicate to the audience.  It is an intentionally mentalistic theory -- placing the foundation of linguistic meaning firmly in the realm of mental phenomena.  Grice's theory then uses this concept of "speaker's-meaning" to generate "sentence-meaning" -- making sentence meaning depend on the "standard" speaker's intentions in employing the sentence in question.  The criticism in the essay title argues that because English is a compositional language, a possibly infinite number of English language sentences have never been uttered, and can therefore not have a "standard use" to provide them any meaning.  The argument is that the meaning of such never uttered sentences cannot, therefore, depend on any "standard" speaker's intentions.  Obviously, if meaning primarily rests with sentences, and not words, as it would have to if the basis of meaning is the speaker's intentions, then Grice's early theory has a problem with never before uttered sentences.  

In his 1968 article "Utterer's Meaning, Sentence Meaning, and Word Meaning"(3), Grice responded to this criticism by modifying his theory.  He posited that the literal meaning of an unuttered sentence is built recursively out of basic meanings associated with words.  In essence, Grice employs the concept of a "standard use" of a word in the same way that his previous article posited a "standard use" of a complete sentence.  A word (or a phrase) has a "standard meaning" if, within a given language community, it is conventionally expected that when a speaker uses this particular word (or phrase), his intention is to communicate the standard conventionally associated meaning.  The sense of "conventional" intended here is the sense described by Lewis(4).  A word would come to have a "conventional" meaning associated with it through the coordinated efforts of people using that word to communicate their intentions.  Obviously, communication between speaker and audience will fail unless they come to some agreement on the meanings associated with the words being employed.

Thus, in this revised theory, "sentence-meaning" is not directly a function of speaker's intentions.  Instead it is a complex function of the conventionally agreed meaning of its parts (together, of course with the rules of syntax).  Although, of course, the foundational link to speaker's intentions is maintained because those sub-sentential parts (either individual words, or complex phrases) acquire their conventionally agreed meaning by being used by speakers according to their intentions to communicate specific beliefs, desires, and other propositional attitudes.  Hence, the criticism voiced in the essay's title carried weight only against Grice's earlier theory of sentence meaning.  

To properly understand Grice's theory of sentence meaning, it is necessary to divide the concept of "meaning" into three sub-concepts.  First there is "speaker's-meaning".  This is the concept of meaning that is expressly psychological in nature.  Utterances, or sentence tokens, are generated by a speaker according to that speaker's beliefs, desires and so forth.  As stated in his 1968 article, more formally this is stated as --

(G)       "By uttering x, S meant that P" is to be analyzed as  "S uttered x intending that A form the belief that P"

Where S is the speaker, A is the audience, x is the utterance, and P is the proposition.  "Speaker" and "utterance" are to be understood quite generally.  Any production of a sign intended to communicate is regarded as an "utterance" generated by some "speaker", regardless of the nature of the sign or the form of the "utterance".  The pictorial black silhouette denoting the men's room is an "utterance" of a "speaker".  I should also add that neither Grice nor I mean anything mysterious by "proposition".  We use this word as a general term for the things people assent to, reject, find doubtful, accept for the sake of argument, attempt to verify, deduce things from, and so on.(5)

I have provided here, of course, just the first clause of the whole formalized statement of the theory.  Grice goes on to add a number of complicating conditions to the basic idea, involving second and third order intentions, in order to deal with unusual counter-examples.  But for the purposes of this essay, this simplified version will do and I will ignore the complications.

The second sub-concept of "meaning" is "audience-meaning".  Whatever the intentions of the speaker in his/her effort to communicate to the audience, those intentions will be frustrated unless the audience can gain from the utterance produced enough evidence (given the context) to reach some reasonably accurate conclusions as to the nature of P.  Regardless of what x consists of, A must be able to interpret that utterance as indicating that S intends A to form the belief that P.  The first part of this (that S intends A to form a belief) is straight forward given the minimal context of communication.  But the problem arises in figuring how A comes to know what P is.  So in parallel with the utterance having a "speaker's-meaning" reflecting the intentions, beliefs, desires, and other propositional attitudes of the speaker, the utterance also has an "audience-meaning" reflecting the conclusions that the audience makes with regards to what the speaker intends for him to form a belief about -- the contents of P.

As a speaker, you cannot (normally) expect to utter just any random utterance, and expect the audience to intuit what you mean.  In order for the audience-meaning to come out a reasonably close approximation of the speaker's-meaning, the utterance x must involve some elements that both speaker and audience have previously agreed (explicitly or implicitly) on a common meaning.  (In Paul Ziff's example(6) of George uttering "Ugh ugh blugh blugh ugh blug blug", George is counting on the examiner using the context of discourse and that George's utterance carries no conventional meaning to reach the audience-meaning that he is being dissed.  Given the mythical intelligence of Army examiners, I have my own doubts that George would be successful in his effort to dis the examiner.  So I do not regard this as a counter-example of conventional word meanings.)

In Grice's first article in 1957, he left the necessary common speaker-audience understanding to whole sentences.  He argued that sentences have a "sentence meaning" (also called a "literal meaning") derived from their conventional use to express particular propositions P.  But the quoted criticism in the essay title pointed out the obvious flaw in that approach, and caused him, in his 1968 article, to propose that the necessary common understanding between speaker and audience rests with the sub-sentential parts of sentences -- words or phrases.  Hence, the "sentence-meaning" or "literal meaning" of never uttered sentences can be built up from the conventionally agreed meaning of the individual parts (along, or course, with the conventionally agreed rules of sentence syntax).  And for the audience to determine the audience-meaning, they can start with that built up literal meaning of what was uttered.  The audience can then use information from the context of utterance to infer the speaker's intentions, thus arriving at an interpretation of the audience-meaning of the statement.  (Thus Grice's theory amalgamates the areas of linguistics known as "semantics" with what is called "pragmatics".)

This resting of sentence meaning on the conventionally agreed meaning of words (or common phrases) meshes nicely with the way that we all learn new words in our own language, or in another language.  We learn new language elements either ostensibly (these rough looking rocks here we call "painite") or by using other words in a description (as in the dictionary definition of "kainotophobia" as a fear of change).  Both in acquiring our first language, a second language, or merely extending our vocabulary, we learn the meaning of new sub-sentential parts.  We do not learn the meaning of whole sentences.  As long as the speaker (S) builds the utterance (x) using words or phrases that the audience (A) is familiar with, there will be little opportunity for a significant disconnect between speaker's-meaning and audience-meaning.  But there is an obvious opening for such a disconnect.  The audience is attempting to infer the speaker's intentions without empirical access to those intentions.  But the availability of conventional agreement on the literal-meaning of sub-sentential parts, and the empirically available evidence on the context of utterance goes a long way to alleviate that difficulty.  Grice's theory does not guarantee the success of communication.  If, as the course of conversation continues, there does appear to be such a disconnect between speaker and audience, the audience can become speaker with a ready-made utterance with a recognized conventional meaning to clear up the confusion -- "Say what?"

If I use a brand new, never before uttered word to communicate my intentions, I would have no hope of successfully communicating unless I have first made the audience aware of what I mean by that word.  (I should expect a "Say ahat?" response from my audience.)  On the other hand, by using words (or common phrases) that I expect already have that commonality of understanding between speaker and audience, I can communicate any novel intention I please. 

So sentence meaning -- considering the sentence en bloc -- cannot be determined by speaker's intentions, as argued by the titled quote.  But the meaning (understood as the literal meaning, divorced from the context of a particular utterance of some speaker) of a never before uttered sentence can be determined by the speakers intention's conventionally signified by the words (or phrases) that make up that sentence -- as long as those words (or phrases) have a conventional meaning known to both speaker and audience.  So it becomes obvious that the criticism quoted in the essay title does not translate to words from sentences. 

The specific criticism quoted in the essay title has thus been successfully defeated.  Grice's  modified theory explains how speaker's intentions in communicating can result in conventional meanings for sub-sentential parts.  Thereby providing literal meaning for unique unuttered sentences.  However, it may be worth the effort to investigate the background that motivates the critics of a Gricean approach to meaning. 

In "General Semantics," David Lewis wrote

"I distinguish two topics: first, the description of possible languages or grammars as abstract semantic systems whereby symbols are associated with aspects of the world; and, second, the description of the psychological and sociological facts whereby a particular one of these abstract semantic systems is the one used by a person or population. Only confusion comes of mixing these two topics."(7)

Echoing Lewis, there clearly is a distinction between the questions "What is the meaning of this symbol (for a particular language group)?" and "In virtue of what facts about the world does the symbol have that meaning?"  This gives us two different notions of "meaning".  One notion, employed by a "semantic theory" of meaning, is a specification of the meanings of the sentences of some (object) language in terms of another (meta-)language.  This is the truth-conditional Davidson/Tarski approach to linguistic meaning.  The other notion, employed by a "foundational theory" of meaning, is a specification of what about the world of some language group gives the symbols of their language the meanings that they have.  This is the notion behind Gricean approach to linguistic meaning.  Obviously, the two kinds of theory place constraints on each other.  But that does not change the fact that semantic theories and foundational theories are simply different sorts of theories, designed to answer different questions.(8)

The main difficulty that some philosophers have with the (revised) Gricean theory of linguistic meaning is that it is based on what is considered to be unobservable mental phenomena.  Following Wittgenstein, one of the constraints on an acceptable theory of meaning proposed by these philosophers is what is called the manifestability of meaning thesis.(9)  The manifestability of meaning thesis asserts that the meaning of any linguistic expression must in principle be empirically observable, must be in principle be open to objective (inter-subjective) determination.  In the words of Donald Davidson, the fact that meanings are decipherable in this way "is not a matter of luck; public availability is a constitutive aspect of language."(10)

The manifestability thesis stipulates that the internal states of speakers, such as private mental states, cannot be what linguistic meaning is grounded upon.  Otherwise communication would be based on inferences about unobservable associations of linguistic expressions with language-independent contents (mental phenomena).  Any such association, it is presumed, could not be checked by the language community.  As Wittgenstein pointed out with his private language argument(11), there would be no objective standard of correctness of such associations.

But the manifestability thesis is not equivalent to behaviorism.  Generally speaking, meaning-theoretic behaviorism is the approach to meaning that draws upon purely behavioral evidence.  Mentalistic or intentional idiom is precluded in a behaviorist approach.  As is well known, Quine considers his view of meaning behavioristic(12).  But Davidson, the arch-typical truth-conditional theorist, accepts the manifestability thesis without adopting behaviorism.

"The requirement that the evidence be publicly accessible is not due to an atavistic yearning for behavioristic or verificationist foundations, but to the fact that what is to be explained is a social phenomenon. [...] [L]anguage is intrinsically social. This does not entail that truth and meaning are defined in terms of observable behavior, or that it is "nothing but" observable behavior; but it does imply that meaning is entirely determined by observable behavior, even readily observable behavior."(13)

The objections to the Gricean approach come from the assumption that it bases linguistic meaning on what is considered to be an unobservable realm of speakers' intentions.  But Dummett, for example, already relaxes the requirement beyond the strictly empirical behaviorism of Quine.  He is ready to admit that thoughts and reasons can be attributed to a speaker by the interpreter.  Is it really so problematic to include in the category of manifestable evidence the responses that the speaker might provide when questioned about his intentions, concepts, and the rest of that sort of mentalistic thing?  If sub-sentential parts can acquire empirically knowable conventional meaning in the manner described by Lewis, and the context of utterance is empirically knowable (even approximately), then a Gricean approach to meaning can be considered to satisfy the manifestability thesis.  If such conversational evidence can be considered as part of the manifestability of meaning, then the Gricean approach to an intention-based theory of meaning starts to become less objectionable.  And meaning can be determined in an unproblematic way by speaker's intentions.


Notes & References

(1)  Platts, Mark de Bretton; Ways of Meaning, 2nd Edition, The MIT Press, Cambridge, Massachusetts, 1979, ISBN 0-262-66107-1.

(2)  Grice, H. Paul; "Meaning" in The Philosophical Review, Vol 66 (1957), pp 377-388. (Reprinted in his Studies in the Way of Words.)

(3)  Grice, H. Paul; "Utterer's Meaning, Sentence Meaning, and Word Meaning" in Foundations of Language, Vol 4 (1968), pp 225-242. (Reprinted in his Studies in the Way of Words.)

(4)  Lewis, David;  Convention: A Philosophical Study, Wiley-Blackwell, Oxford, England, 1969. ISBN 978-0-631-23257-5. Kindle Edition.

(5)        "It is thus uncontroversial that there are propositions. The only question that could arise is: What are propositions? Many philosophers apparently think that propositions are sentences, since they think that sentences are what is true or false, and it is evident that those things that are true or false are just the things that are the objects of the activities and states listed above. But I can make no sense of the suggestion that propositions are sentences, and I shall not discuss it further. It is true that I am willing on occasion to speak of sentences being true or false, but this is only shorthand. When I say that a given sentence is true, I mean that the proposition - a non-sentence - that that sentence expresses is true. (To say that an English sentence expresses a given proposition is to say, roughly, that the result of concatenating 'the proposition that' and that sentence denotes that proposition.) Similarly, when I say that a name is honorable, I mean that the individual or family that bears that name is honorable. I can no more understand the suggestion that a sentence might be true otherwise than in virtue of its expressing a true proposition that I can understand the suggestion that a name might be honourable otherwise than in virtue of its being borne by an honorable individual or family."  Peter Van Inwagen, "Ontology, Identity, and Modality".

(6)  Ziff, Paul;  "On H.P. Grice's Account of Meaning" in Analysis, Vol. 28, No. 1 (Oct 1967), pp. 1-8.

(7)  Lewis, David;  "General Semantics" in Synthese, Vol 22 (1970), Pp 18--67.

(8)  Speaks, Jeff;  "Theories of Meaning" in The Stanford Encyclopedia of Philosophy (Summer 2011 Edition), Edward N. Zalta (ed.), URL=<http://plato.stanford.edu/archives/sum2011/entries/meaning/>.

(9)  Glock, Hans-Johann;  Quine and Davidson on Language, Thought and Reality, Cambridge University Press, Cambridge, England. 2003. ISBN 9780511487514.  Pp 268-293.

(10)  Davidson, Donald;  "The Structure and Content of Truth" in The Journal of Philosophy, Vol 87 (1990), pg. 314.

(11)  Wikipedia contributors;  "Private Language Argument" in Wikipedia, The Free Encyclopedia. URL=<http://en.wikipedia.org/w/index.php?title=Private_language_argument&oldid=516667600>

(12)  Quine, W.V.O.; Word and Object, The MIT Press, Cambridge, Mass. 1959. ISBN 0-262-67001-1. Kindel Edition.

(13)  Davidson, Donald;  "The Structure and Content of Truth" in The Journal of Philosophy, Vol 87 (1990).  Pg 314.

Grice, H. Paul;  Studies in the Way of Words, Harvard University Press, Cambridge, Massachusetts, 1991. ISBN 978-0-674-85271-6.

Van Inwagen, Peter;  Ontology, Identity, and Modality: Essays in Metaphysics, Cambridge University Press, Cambridge, England. 2001. ISBN 0-521-79548-6.


[Up] [Home] [Next]