veridicality

["John E. Clifford" <pcliffje@CRL.COM>]

        While I generally share the feeling that, when a linguist says
"pragmatics," he means the as yet undeveloped part of his theory which
will solve all the problems that the developed parts of his theory cannot
cope with, I have to admit that there are now a number of theories that
have _developed_ segments called pragmatics, some of which do solve
problems left over from other parts (but leave new unsolved parts for some
unnamed further chunk).  What a theory means by "pragmatics,"  how it
differs from syntactics and semantics (and morphology and phonology for
that matter) differs from theory to theory (as do the meanings of the
other terms here as well) and are all pretty remote from the original
Peirce and Morris notions.  Before I would want to attack And too much for
his discussion, then, I would want to know more about the whole theory in
which he is working and how he draws the line.  The little I have seen
suggests Gricean antecedents for pragmatics but little about which of the
several routes from there are followed (but it does not look like the
formal - e.g Karttunen -- line).
        But the questions about veridicality do not really depend upon
sorting out theory parts.  The questions about whether it can be or will
succeed in being an active feature in Lojban are empirical questions and
will presumably be answered eventually by examining what happens in
Lojban.  The same may be said for other languages, except that the data
are more unclear for them -- as witness the diverse interpretations they
have received.
        First, veridicality, if it is meaningful at all, is a distinctive
feature in some modern version of the old Prague sense (and so will not
fit into every linguistic theory even to begin with).  Thus, it is a rule
governed part of the description of some linguistic category and within
that category there are items that take a positive value and others that
take a negative value (and perhaps some that are neutral).  Outside of
that category (or perhaps several categories) in which it is defined, it
has no meaning (cf. in logic, questions about true arguments and valid
claims).  Thus, to say that a _language_ is +veridical makes no sense.
Good Griceans that we are, we take the fact that someone has broken a rule
(loosely "Don't babble") to indicate that some meaningful message is being
hinted at.  My best guess about what that might be is that veridicality is
never in fact a distinctive feature in any language because all items in
all languages satisfy what would be the positive value for the feature.
Since this claim makes sense and might even be true, we could stop to
examine it in this context.  However, the general claim is almost
certainly false for natural langauges (and is reputed to be false for
Lojban -- but that is begging the question): every language I know of has
items which are not veridical, not literal, and which are marked as such
so as to constitute a distinctive kind of item.
        So, perhaps we should not take the babble as this claim but as a
more precise and relevant claim, that descriptive sumti (or what
corresponds to them in a natural way) in all languages are veridical,
i.e., that the referent of the sumti is required to have the property
mentioned in the description.  The explicature of a sentence involving
such a descriptive sumti will contain a segment involving the predication
in the description in such a way that it affects the truth of the sentence
as a whole and also affects what the description refers to.  This claim,
whatever may be its status for natural languages, is directly challenged
by Lojban.  For _le_broda_cu_brode_ is said to have as explicature simple
brode(B), where B is the direct reference to the thing that _le_broda_
refers to; _broda_ nowhere appears in this explicature in any role.  Of
course, somewhere in the theory we must have a device for hooking
_le_broda_ up with B as its reference.  But there is no gurantee in Lojban
that _broda_ plays any role there either.  It may, of course, and it is
often useful (for the hearer, at least) if it does, but the connection may
be totally casual -- an earlier scrap of dialog where the speaker says
"_le_broda_ there" waving toward B, who is not at all brodaish.
         Clearly, the same sort of thing happens in NLs as happened with
_le_broda_ in this tale.  Why,then, do would we insist that the
descriptive sumti in English, say, are always veridical?  The short answer
seems to be that "the" is taken as univocal (or not very equivocal anyhow)
in English.  so, since the vast majority of uses of "the" are veridical,
all of them must be.  Only when the veridical interpretation fails do we
fall back on some other method -- for understanding the message of the
utterance, not for interpreting the meaning of the sentence or the
referent of the description.  (Actually, of the half-dozen or so
well-formulated views about how the first part of this is done, nearly
half make reference to truth even in the process of fixing reference;
Frege's definite descriptor, for example, refers to a different thing
depending upon whether or not there is a veridical referent, but the
reference is fixed all in the same process and before the overall sentence
is evaluated).  We could equally well say that, since some cases of "the"
are nonveridical, none of them are inherently and then separate the
determining of what B is off from determining the truth of the claim (as
some logical systems do, as just noted).  The veridical reading might have
a favored status in this reference-finding procedure, but it would not be
the only one to be used nor could even it be used alone. There might, for
example, be several things satisfying the description and we need to pick
the one intended (this plays a role in using _le_ in Lojban, for example).
        More pressing, in some sense, is And's claim that
_lo_broda_cu_brode_ is going to be used to make statements understood
(eventually) as true when there is no broda which is also brode, but
rather (as we might say of the theoretical record) when _lo_broda_ is used
to refer to something totally unbrodaish.  And, again, the case of English
"the" can be taken to prove the point.  But, in Lojban, we can refuse to
write in -- or allow other to make use of unwritten -- the rules that
allow reinterpretation of false sentences with _lo_ descriptors.  So, we
come finally to another empirical question -- can we make this refusal to
allow reinterpretation stick.  My guess is that we can.  Since we have the
device that forces "reinterpretation," we are in a better position to say
that the person misspoke and force them to correct themselves.  Of course,
at the observational level, we may seem to allow them to get away with it
because we are (unobserved) rewriting what they say (always a part of the
courtesy of the cooperative convention) -- and maybe even hearing -- the
speaker's _lo_ as _le_.  But notice that this is theoretically a very
different move from the earlier one, since it functions at the
morphologoical or , at worst, the syntactic level.