RT list: another take on logical form

Thanks again to all of you who have contributed to this
ongoing discussion
on logical form. Let me offer one other set of
observations that may be
of use.

If we assume that logical form refers to the basic
semantic representation of an utterance
(without its propositional content necessarily being
determined), the question again
is whether people really ever compute such a thing during
online language understanding.

Looking at the question from a purely logical point of
view (as linguists and philosophers
sometimes do), we can speculate that having some sort of
preliminary semantic representation
is useful for building up a more complete understanding of
a speaker's pragmatic meanings,
or communicative intentions, along with other cognitive
effects that could possibly be inferred.

But the problem for me comes down to whether there really
are some set of linguistic or
nonlinguistic processes that operate by default, and are
autonomous from other sources
of information that may be active throughout the time
course of moment-by-moment interpreting
of ostensive communication.

My claim is that there are very few, if any, processes
that are truly default and always
work autonomously from other linguistic and nonlinguistic
processes. This position is
very anti-modular, and controversial to some, but I
maintain that the bulk of experimental
evidence in psycholinguistics supports this alternative,
highly interactive (often now
called "constraint satisfaction") perspective.

One implication of my position is that some of the classic
distinctions often discussed in
theories of pragmatics may not have psychological reality.

For instance, there is the idea that we typically do some
sort of "decoding" of language
to determine the language spoken, the phonemes produced,
the words stated, and so on,
before "inferential" processes come into play later on to
create fuller, pragmatically rich
interpretations of meaning.

But there is plenty of evidence showing that people's
processing of small bits of speech,
such as phonemes, syllables, and even smaller, can easily
be modified by other sources of
information including other words just spoken and context,
that work "top-down" to guide
speech processing. If any of this is true, and I think an
empirical case can be made for it, then
it is never really clear that "decoding" always starts and
ends at one place and "inferential"
processing always begins and ends at some later place.
Thus, one ends up questioning if there
really is a nice decoding vs. inference distinction with
each working in some independent,
default manner.

A second implication of this alternative vision is that if
many aspects of context and
pragmatics have their influence early on in language
processing, then this makes the
classic semantics vs. pragmatics distinction less useful.
Thus, if people do not compute
complete logical forms or semantic representations for
utterances, this raises the possibility
that there is no default semantics per se at work in
online language processing.

Others of you have mentioned psychological studies showing
that context-invariant word
meanings may be typically processed during language
processing. I suggest that the debate
over whether this is really the case has not been settled
(and has been raging in the
literature for over 40 years), and that plenty of evidence
suggests a more contextually
informed processes of lexical access.

But even if you adhere to the idea that some sort of
lexical meanings are always, invariably
accessed, there is still no evidence that these are
necessarily combined to create full
semantic representations for utterances.

My reason for raising these specific points is that one
way to further debate issues about
the reality of constructs like logical forms,
propositional contents, etc. is to consider
whether these must be constructed in a completely
automatic, default manner without influence
from high-order sources of information. If we can't find
evidence that such things really
are create by default (and this in ALL cases), then it
suggests, to me, that these theoretical
constructs may not play quite the role in realistic
utterance interpretation as once thought.

None of what I say implies that people do not engage in
compositional analysis of
language or that people never infer implicatures given
some analysis of what a speaker has said.
But I view these other pragmatic possibilities as matters
of different types of contextual knowledge
activation coming into play at different times, rather
than as a process of going from logical forms
(semantic representations) up to pragmatic processes that
ONLY come into play late in the process.

Finally, I confess that my love of RT comes from the
detailed ways it talks about various pragmatic
processes operating, given other cognitive and
communicative constraints. Much of this I strongly
believe is psychologically real. I am just not sure that
RT must necessarily assume other lower-level
default linguistic processes to be in place for the theory
to work overall.

Thanks for your attention

Cheers

Ray
Received on Fri Feb 18 17:19:52 2011