Alec Marantz: Lecture Notes II

[Martha McGinnis]

Dear List-ers,

Without waiting for response to the first set of lecture notes, I present
you with set two.  Please refer to the readings listed at the beginning of
the first set of notes for an indication of what I'm talking about when I
discuss "Fodor" and "Pustejovsky."

--Alec


Morphology II, 9/17/99
Roots and decomposition

1.	Fodor asks us to separate the linguistic from the conceptual, and
decompositional features from other properties.  Fodor agrees that it's odd
to say, "John began the rock" (out of context) because rocks aren't events
nor imply a characteristic event (as does "book").  However, he asks what
reason there is to believe our knowledge about rocks is either linguistic
or decompositional.

2.	Pustejovsky retorts that Fodor is simply avoiding the interesting
question:  how can we represent the knowledge that speakers have about
"rock" that leads to their judgements about sentences containing "rock"?
If the knowledge that rocks don't name/imply events is conceptual, it at
least must be represented in a way that interacts with the representation
of linguistic semantics to yield sentence meanings from the linguistic
system.

3.	Fodor points out that inferences based on Pustejovsky-style
representations are defeasible ("I baked a cake" sometimes implies I made
the cake, but, "I removed the twinkie from its cellophane wrapper and baked
it in the sun" is fine with no "creative" implications).  On the other
hand, if "I baked a cake," then it's true that I baked a cake.  (And if "I
saw cats" is true, then there is more than one cat such that I saw it.)
The question would be, where do Pustejovsky's observations belong, and is
his "theory" (simply) an elaborate method to encode some of these
generalizations.

4.	The decompositional nature of Pustejovsky's lexical entries imply
that it makes sense to consider, e.g., what a lion would be if it weren't
an animal.  Or what a cake would be if it weren't made by baking.  Or what
a hammer would be if it weren't used for hammering.

Cf., what's "raising" without agentive cause (=rising).  If a "singer" is
"one who sings," what is, "John sings"?

>>From the standpoint of morphology, one would ask why it is that the
categories Pustejovsky deals in aren't morphologically expressed in any
language.  Why is there no "telos" morpheme indicating the class of
something according to what it's for?

5.	Three dimensions of controversy over root meanings:
	A.	Linguistic nature of meanings
	B.	Criteria of identity between roots
	C.	Private nature of meanings/source of meanings in individual

6.	Linguistic nature of meanings:

	a.	Are there strict meaning/syntax correlations - for the type
of meaning categories Pustejovsky uses in his decompositions?

Fodor questions whether distributional facts distinguishing "eat" and
"devour" represent a true generalization about semantics/syntax mappings.

I ate it up/I devoured it up/I finished it up.
I ate it raw/I devoured it whole

Cf. Keyser & Roeper on "abstract clitic" hypothesis.

	b.	Are the word-internal decompositional meanings ever
expressed systematically in syntactic composition?

Compare eat/devour with 'walk around in circles'/'walk to the store'
bake/bake+applicative  in languages with overt applicative morphology and
double object applicative constructions

	c.	Are the semantic decompositional categories ever
systematically marked by overt morphology in any language?

Compare the animate/inanimate distinction with the 'inherently for doing
something'/'created in a particular way' distinction

Stage/individual level vs. state/event distinction

6.	Note that this discussion takes place against the bedrock
assumption of compositionality in the syntax.

Everyone agrees that the semantics of "kick the ball" is compositional.
One could ask whether monomorphemic "blick" could mean "kick the ball"

Cf. Jackendoff on "kick the bucket" = "die"

Or one could ask whether there isn't a different sort (different
primitives, different means of composition) of compositionality below the
word level, where the syntax and the below-the-word level composition mix
only at the word level; the syntax never does what the word internal
mechanisms do and vice-versa.

7.	Criteria of identity between roots

polysemy (same thing has multiple meanings) vs. homophony (different things
have same sound)

Pustejovsky claims that Fodor provides no account of productive polysemy.
Fodor claims that Pustejovsky's polysemy is sometimes homophony and
sometimes requires no linguistic account (and thus is technically not
polysemy or homophony).

Is "bake" polysemous or homophonous?

Is it the same "cat" in "The cat is on the mat" and "Don't let the cat out
of the bag"?
bank of a river vs. bank with the money

Clearly these questions can't be asked without a fairly articulated
linguistic theory.

Morphophonology will be crucial here - where is information about
allomorphy stored?  With the phonological form of a root?  With the
meaning/identify of the root as an interpreted object?  Somehow in a system
of interrelated stored words?  Consider the "flied out to center field"
type of example.  What does this tell us about the "fly" in "fly out"?  Is
this a question of structure or of identity?

Related to allomorphy: conjugational/declension classes of roots/stems.
What's the connection between the identity of a root and its (arbitrary)
class.  See Embick on deponent verbs in Latin.

8.	Private nature of meanings/source of meanings in individual

What makes your "cat" the same as my "cat"?

	a.	Nothing; they're not the same but they're close enough

The biology of cats and the biology/psychology of humans interact to make
my concept of "cat" close enough to your concept that communication is
possible and almost perfect.  If chimps could talk, their "cat" would be
more or less the same as our "cat" as well.  However, if snakes could talk,
their "cat" might be quite different.

	b.	The nature of cats

"Cat" is a natural class in nature.  A more or less "blank slate"
perceptual/learning system will discover the class through interaction with
the environment.  So "cat" is a class of nature and an emergent category of
the perceptual/conceptual system.  If snakes could talk, their cat category
would be the same as ours - or they would be getting things wrong.


	c.	The structure of the human linguistic and conceptual system

Jackendoff and Pustejovsky more or less adopt this stance.  "Cat" is like
"plural," only more complex.  The feature space of language (of natural
language semantics) leaves a "cell" in the paradigm for "cat," waiting to
be filled during language acquisition.

Here, chimps can't talk so they can't have our concept of "cat."  Why
chimps seem to be able to do complex things that would implicate complex
conceptual structures, yet don't/can't use language, is an interesting
question for Jackendoff (along with why most people have a very hard time
describing scenes they see and can operate within).

The issues here are tightly bound up with questions of acquisition.  Both
in a. and in c., the child will know a cat when s/he sees one and will be
able to say, oh "cat" means cat.  Both scenarios leave open the possibility
of suppletion for roots like "cat" (singular, "cat," plural, "blicks"),
depending crucially though on the exact structure of the grammar (where do
roots enter the grammar and in what form?).

9.	Pustejovsky can't win simply by pointing out what we know about the
meanings of words in sentential contexts.  Fodor is denying it makes sense
to separate the word from the concept, and he therefore refuses to accept
evidence about speakers' knowledge of concepts as evidence for anything
other than knowledge of concepts.  Fodor is thus committed to a modularity
between linguistic structure and conceptual meanings; what's "below" the
word level isn't strictly speaking linguistic, and isn't compositional.
Pustejovsky must show that there's a strong resemblance between the
compositionality below and above the word level.  Fodor accepts the
what-you-see evidence of syntactic composition as sufficient for supporting
syntactic decomposition; I imagine he could be convinced to see overt
morphological composition in the same way, arguing for word decomposition
when you can see the composition.  Since Pustejovsky is pushing
decomposition where there is no overt composition, he needs very strict,
reproducible evidence for syntax/semantics correlations (in putative cases
of polysemy and for supporting a claim of a semantic feature with syntactic
consequences).  Fodor says Pustejovsky fails empirically, and I tend to
agree.

marantz at mit.edu