Philosophy 305                                                                                                                        W. Lycan
Spring, 2002
Vs. Unger



    Ted's assimilation of Unger's claim of semantic relativity to Quine's famous doctrine of the indeterminacy of translation was apt, because in Unger's general argument for the claim, he, like Quine, gestures behavioristically:

In discussions of language, few things may be taken as  even relatively basic.  On the one hand, there are certain people...making marks or sounds.  On the other hand, there are certain effects achieved on people as  regards their conscious thought, their experiences, and, most important, their behavior.  Everything linguistic,
in between, is an explanatory posit.
      When such posits are made, observable phenomena-- and even all objective (concrete) facts--get left behind.   Then, we might expect a certain latitude, or room for  descriptive maneuver, where alternative formulations may have equal claims to propriety....  Different total  explanations of behavior each allow for a different semantic approach.    [p. 246]
Unger is hinting what Quine would put more forthrightly:  The behavior and only the behavior is real.  Any linguistic "theory" (really just an abstract formal apparatus) that makes the right behavioral predictions is as "true" as any other, i.e., none is correct to the exclusion of any other.  Some formalisms may be more convenient to use than others, but that is the only way in which they would be superior.
     As Dorit knows better than any of us, there is a gigantic literature on Quine's indeterminacy thesis.  I think it's fair to say that Quine made very few converts.  To put the leading objection very crudely:  Psychological Behaviorism isn't true.  Cognitivism is.  Cognitive psychology offers and justifies hypotheses about the information processing that takes place in subjects' brains--which processing is every bit as much a part of the real world as is the outward behavior that results.  Granted, a linguistic theory is an abstract formalism, but it is an abstract and idealized functional hypothesis as to the syntactic and semantic processing that go on in the heads of speakers and hearers.  If one linguistic theory's formalism is in fact realized in the heads of language users and the competing theories are not, the first theory is true and the others aren't.  So Unger-Quine's general argument fails.  (I do not think Unger really meant it seriously, because surely he knew the standard objections.)
     Fortunately, Unger gives a more specific argument regarding the special case of absolute terms and contextualism vs. invariantism.  It is not a crushing argument, but it is not obviously a loser and he may be right.  Consider again the two theories of absolute terms.  The contextualist holds that what someone says in predicating "flat" or "certain" is literally true, because of one or more hidden parameters whose values are fixed by contextual factors.  The invariantist holds that predications of "flat" or "certain" are always literally false, but what the speaker "merely suggested, informally implied, or whatever" (p. 248) is true.  (We may beef this up by supposing that the informal implication is what is primarily meant to be conveyed to the hearer, and what is "speaker-meant" in Grice's sense.  In conversation, it is entirely normal for what is primarily conveyed to differ from the literal meaning of the actual sentence used.  Grice called this "implicature.")
     Now, Unger's argument starts with the premise that each of the two theories makes one thing simple but another thing complex.  The contextualist story is simple in that according to it, what is meant to be conveyed to the hearer is just the literal meaning of the sentence used; no extra reasoning is needed to bridge a gap between the two. But the contextualist story complicates the recovery of the sentence's logical form, containing hidden parameter etc., from its simple surface form (just four words: "The field is flat").  The invariantist sees no gap between surface form and logical form; "The field is flat" is a simple subject-predicate sentence, of the form "Ff."  What the invariantist view complicates is the recovery of speaker- or conveyed meaning from the sentence's literal meaning plus contextual factors.
    Unger now argues that it's a wash.  One simple process + one complicated process = one complicated process + one simple process.    There is nothing to choose between the two theories.
     The argument could be put more strongly than Unger does put it.  Recall my diagram from class:  All parties agree that the hearer takes in an acoustic signal, processes it extensively, and arrives at a conveyed meaning.  According to the contextualist, the major part of that processing is syntactic and semantic, with "semantic pragmatics" fixing the contextual parameters but no further pragmatic reasoning required.  According to the invariantist, only a tiny bit of the processing is syntactic/semantic; the major part is pragmatic reasoning.  Now, mightn't it be true that all that's real here is the processing itself?  "Syntactic," "semantic," and "pragmatic" are just labels, and there is nothing to determine where in the processing lines should be drawn to mark where those labels apply.  The issue is merely verbal.  (Those of you who know Dennett's work on the temporal anomalies of consciousness will recognize the style of this argument from his discussion of "Stalinist" vs. "Orwellian" theories of the anomalies.)
     It's lucky that that way of putting the argument is not Unger's, because it won't do.  (It fails for the same sort of reason as does Dennett's Stalin/Orwell argument.)  The terms "syntactic," "semantic" and "pragmatic" are not otherwise empty labels.  Each has a prior meaning in linguistic theory, and each comes with fairly well-understood theoretical baggage and  attachments.  (Eventual utopian) brain research might well show that there is a semantic component in the speech center and that the relevant processing takes place within it, or that the syntax and semantics are indeed very simple and the heavy processing goes on in the part of the brain that is responsible for computing implicatures.
     Now notice that the latter objection applies, not just to the extreme version of the argument, but also to Unger's actual version.  There is already known to be a speech center in the brain, and research is being done on its architecture.  The further, eventual utopian research may turn out as just mentioned.  So the issue between the contextualist and the invariantist is empirically verifiable.
     (Of course, that theoretical verifiability does not refute semantic relativism.  For the eventual research might turn out to reveal that there is nothing in the brain but marshmallow, and that the alleged linguistic "processing" is a complete mystery.  Or, semantic relativism might be true for a reason different from the one Unger gives.)
     Unger seems to anticipate and reply to this appeal to future cognitive science, in section 4 ("This semantic relativity cannot be denied, I believe, by an attempt to locate some objectively telling psychological processes" (p. 51)).  But so far as I can tell, all he does in that section is to reiterate his tradeoff argument for psychology.  He is right; the contextualist and invariantist views considered as cognitive psychology will differ in that way.  But that does nothing to show that the issue between them is not empirically verifiable in the way I have said.

     Here are several further considerations that may cut against relativity.
     (1) The simplicity-complication tradeoff between contextualism and invariantism is a bit one-sided.  Though Unger is right to say that each view makes simple what the other makes complicated, the contextualist's complicated syntax/semantics is probably a good deal less complicated than is the invariantist's pragmatics.  Unger's idea is that the contextualist syntax/semantics must turn "The field is flat" into something like "According to contextually relevant standards, the field is sufficiently close to being such that nothing could ever be flatter" (p. 248), which would be a mighty operation.  But on no known semantic theory are actual references to context, standards, sufficient closeness etc. incorporated right into logical form.  "The field is flat" says nothing about such stuff.  Rather, its contextualist logical form would be just something like "Ffp," where "p" is a "purpose" parameter, all read as "The field is flat enough for purpose p."  It doesn't take a lot of syntactic/semantic work to get "Ffp" out of "The field is flat."
     By contrast, the invariantist ttheory must posit a chain of Gricean reasoning that begins with "He has literally said that the field is (absolutely) flat" and ends with "What he means is that according to contextually relevant standards, the field is sufficiently close to being (absolutely) flat."  It's easy to see the first step in that reasoning: "He can't mean what he literally said, because it's too obviously false."  But then the hearer has to work out using principles of conversational cooperation what the speaker positively did mean instead.  Quite complicated.
     Unger may fairly protest that I have swept something under the rug, and so I have.  Since on the contextualist view "The field is flat" has a hidden parameter, the parameter has to be fixed in context by semantic-pragmatics.  And the semantic-pragmatics will have to advert to standards and sufficient closeness.  So all that stuff will still have to be represented and computed over, even if it does not turn up in the logical form assigned to the target sentence.  Probably the latter assertion is right, but that computation still seems simpler to me than the Gricean reasoning posited on the invariantist view.
     (2) A distinctive feature of Gricean implicature is what Grice called "cancellability": An implicature can be pre-empted by the speaker.  "Martha shed a single tear, walked to the edge of the precipice, and jumped.  But don't get me wrong; I don't mean she jumped off the cliff.  She just jumped up and down near the edge."  If invariantism is true, the implicature should be cancellable in that way.  So can one say, "The field is flat," and immediately add "But don't get me wrong; I mean just what I said, that it's absolutely flat, not that according to contextually relevant standards it's sufficiently close to being absolutely flat"?  The matter is complicated by the fact that in this case, "The field is absolutely flat" entails that according to standards etc.; so one cannot deny the truth of the implicatum one is trying to cancel, but only deny that it is what one means.  But in any case the utterance would be a very odd one.  (Not that that proves much.)
     (3) Remember DeRose against Unger (p. 213, especially the "physician" example).  (Neat of DeRose as volume editor to have got his criticism of Unger into the book earlier than Unger's excerpt appears.)  One well-known desideratum for semantic theories is charity: other things being equal, we try to make the natives' sentences come out true.  And so, as DeRose says, "[t]hat... [the invariantist] theory involves us in systematic falsehood continues to constitute a strike against it," though theorists differ in how much weight they think should be given to charity.
     Moorean considerations may come in as well.  The invariantist theory, taken together with the fact that nothing is absolutely flat and the invariantist T-sentence "The field is flat" is true iff the field is absolutely flat, entails that nothing is flat at all.  But I am more certain that my kitchen counter top is flat than I am of the premises of any philosophical or semantic argument designed to establish the contrary.
     (4) Finally, we should ask why our language should enshrine in one of its precious four-letter words the invariantist concept of absolute flatness rather than the contextualist concept of flatness-enough-for-the-purpose-at-hand.  The latter concept is constantly useful.  The former, since it is never exemplified, is useful only in that it gives rise to implicatures that themselves contain the contextualist concept.  For a designer of our language to have assigned to "flat" the invariantist concept rather than the contextualist concept would have been pointless.  Which is some evidence (call me Panglossian if you will) that "flat" does not express the invariantist concept.