September 15 & 17 class summary: 24.

902 could not be replaced by depend, because depend bears the distinct
feature [+ __ NP PP].
The week actually began with some unfinished business from last week, which is
OK, painful detail. What is interesting to us is a constraint proposed by Chomsky on the
contained in the previous summary. Then we moved on...
ability of a subcategorization rule to control the environment in which it applies.
1. Preliminaries Chomsky claimed that the environment of the subcategorization rule for X makes
A word often cares about its syntactic environment. A verb, for example, may reference to all and only the sisters of X. More simply still: the subcategorization property
require two, one or zero complements, and (seemingly -- see below!) may specify the of X cares only about the sisters of X.
syntactic category of its complement.
Terminological note: As mentioned in class, a slightly different way of speaking has
Examples: put requires NP PP, devour requiresNP, depend requires PP, eat takes an arisen in the field of syntax. We say that a verb like put "subcategorizes for" an NP and a
optional NP. PP. Likewise a verb like eat "subcategorizes for" an optional NP.

Chomsky's constraint has interesting implications for language acquisition. Granted

(1) a. Sue put the book under the table.
that a child must maintain some record of the syntactic environment in which lexical
b. *Sue put the book.
items occur (or else subcategorization information would not be acquired), Chomsky's
c. *Sue put under the table.
constraint suggests that this information is quite limited. The child does not attend to
the entire tree when keeping these mental records.
(2) a. John devoured the sandwich.
b. *John devoured. That said, it is not clear why subcategorization phenomena exist in the first place. But
c. *John devoured under the table. they do.
An interesting point that was came up (inadvertently) in class: the child must already
(3) a. Bill depended on his tablet PC. know a lot about syntactic structure to knowe what information to ignore when
b. *Bill depended from his tablet PC. creating lexical entries!
c. *Bill depended his tablet PC.
d. *Bill depended.
3. Chomsky's mistake
(4) a. Sue ate the sandwich.
b. Sue ate.
Chomsky's constraint is empirically on the right track: we do not find obligatory modifiers
or other adjuncts. Where a constituent is obligatory, it appears to be a complement. For
This information forms part of the lexical entry for a verb, and is partly arbitrary.
example:

(5) *Sue depended on her friends and Bill did so on his relatives.
2. Chomsky's subcategorization mechanism and his constraint
But it is also partly wrong. Let us focus on PP complements. Contrary to what
Chomsky's (1965) mechanism: Chomsky's (1965) proposal leads one to expect, many verbs actually ask for particular
How words get inserted into structure: prepositions in their PP complements -- not just any old PP. Thus, depend requires on in
a. Step 1: Nodes such as V, N, A, etc. undergo a subcategorization rule that English, not from:
assigns them to subcategories, whose name describes the environment in
which the rule applies. The feature on the right of the arrow, by the way. (6) a. Bill depended on his friends.
was called a "Complex Symbol" (CS): b. *Bill depended from his friends.

Example: V--> [+ __ NP PP] / __ NP PP Similarly, the noun preference (we didn't do this one in class), when it takes a PP
complement (which it doesn't have to do) requires that the preposition of the PP be for:
b. Step 2: A CS is replaced by a lexical item of the appropriate category that also
bears the same subcategorization feature. For example, "[+ __ NP PP]"
might be replaced by put, because put bears the feature [+ __ NP PP]. It
 September 16, 2002 summary/page 2

(7) a. Mary's preference for opera 4. Using the constraint to identify heads
b. *Mary's preference of opera / to opera / etc.

Likewise, the adjective fond requires a PP complement, whose preposition must be of: We can now use this generalization (assuming it is correct) to probe for heads of
phrases.
(8) a. John is fond of his children.
b. *John is fond for his children / at his children / etc, C and CP
c. *John is fond. (ok in older English, with different meaning)
Consider, for example:
Predictability: The semantic class of preposition required by a particular verb, noun or
adjective restricts the actual choice of preposition somewhat in these cases. It might be (12) a. Mary believes [that Sue can read the book].
surprising to find from selected by fond, for example. But the actual preposition, or b. Mary arranged [for Sue to read the book].
selection for a preposition in the first place, is unpredictable. The comparison between
English and Russian makes this clear. Note that from is actually the preposition used in (13) a. *Mary believes [for Sue to read the book].
Russian with 'depend': b. *Mary arranged [that Sue can read the book].

(9) English Russian The verb believe allows a clause as a complement when the clause is introduced by the
depend on [zaviset' ot 'from'] complementizer that. The verb arrange allows a clause introduced by the complementizer
speak to [govorit' s 'with'] for. But believe is incompatible with for and arrange is incompatible with that. Looks
looked at the wall [smotret' na 'on'] like subcategorization!
look for the book [iskat' knigu - no preposition]
This suggests that C (complementizer) is the head of an embedded clause. Following our
So we must modify Chomsky's proposal, using a term that picks out the P of a PP. The terminological conventions, we will call embedded clauses CP.
term we use is head. The P is the head of PP.

(10) Chomsky's Constraint, revised (informal version) An important note on complementizer for: It is crucial to distinguish complementizer
The subcategorization feature of a lexical item a dictates the heads of the for from preposition for. The complementizer for is followed by a sentence. An example
complements of a. of complementizer for is seen in (14). What is important here is the fact that the NP that
follows for -- the pronoun it -- is the special near-meaningless it that can be used as the
We will dispense with the extra mechanisms of Chomsky's approach, and just assume that subject of weather verbs like rain:1
nodes such as V bear the relevant subcategorization features, with lexical insertion below
such nodes as V limited to verbs that match V in subcategorization features.
CP
An additional observation (not discussed in class): It is worth noting that there is actually ??
no verb (or noun or adjective) that requires a PP complement and doesn't care about the
head of the PP. Verbs either name a particular head that a PP complement must have or
else require a particular semantic class of PPs. Put, for example, requires both an NP and NP VP
a PP complement, and demands that the head of the PP complement be a preposition that
names the goal (endpoint) of the motion of the NP: V C N' I
(14) The student arranged [for it to rain on Thursday.]
(11) a. Sue put the book on the table.
[endpoint: surface of the table] When for is a complementizer, the sequence for+NP is not a constuent. One sign of this is
b. Sue put the book inside the room. the fact that it cannot be moved to the front of the sentence when focused:
[endpoint: interior of the room]

c. *Sue put the book from the bookcase.

[from doesn't give the endpoint]

1 Called expletive, pleonastic or dummy it. All three terms are in use, and mean the same thing.
 September 16, 2002 summary/page 3

(15) *For it, the student arranged to rain on Thursday.2

CP
In contrast, when for is a preposition, the sequence for+NP may be moved: TP

(16) a. We sang the song for Bill. T'

b. For Bill, we sang this song. NP VP

And, of course, the preposition for may not take the special meaningless it as its object. If V C N' T
I come into the room say "it's raining", you won't ask me: "What's raining?". If I say (17) (19) The government arranged [for it to rain on Thursday.]
out of the blue, you will want to know what I was referring to with the word it!

(17) We sang the song for it. 5. Japanese phrase structure

It's not unusual to find a preposition also doing duty as a complementizer. It's quite Consider now some sentences in Japanese whose word order is perceived by native
common for meaningful words to be "demoted" over time to function words speakers as "normal" -- or, in linguists' jargon, unmarked. [You'll see next week why I
(complementizers, determiners, etc.). The process is called grammaticalization, and make this cautionary remark.] These sentences conform to a very neat generalization
there's a large (not always insightful, alas) literature on the topic. across all categories:

(20) Japanese word order

I and IP For any head H:
a. H follows its complements.
The same reasoning that led us to identify C as the head of the constituents that Sue can b. H' follows all adjuncts (modifiers).
read the book and for Sue to read the book tells us that to is the head of the sister to C. c. H' follows its specifier.
The syntactic category of to is variously labeled as T (for Tense) or I (for Inflection).
(Perhaps we'll take a vote in class on T vs. I. The reading by Carnie calls it T.) Expressing this using Phrase Structure rules, where complements stands in for whatever
the complements of a particular head may be, etc:
The key observation is the fact that the complementizer for requires to (also known as an
infinitival or non-finite marker), while the complementizer that requires anything but to. (21) Phrase structure rules for Japanese
H' --> complements H
(18) C cares about I H' --> adjuncts H'
a. ...for Sue to read the book. HP --> specifier H'
b. *...that Sue to read the book
c. *...for Sue can read the book / *...for Sue has read the book / etc. It is important (and reassuring) to note that the same elements singled out as heads when
d. ...that Sue can read the book / *...that Sue has read the book / etc. we examined subcategorization in English are identified as heads in Japanese according to
the generalization in (20) and (21).

Thus, for example, the fact that verbs in English whose complement is a a subordinate
This tells us, probably, that the head of the mysterious "??" constituent in (14) is I, since C clause "care about" the particular choice of C told us that C is the head of the
seems to subcategorize for it. The mysterious phrase itself The sister to C therefore, subordinate clause. Lo and behold, the C of a subordinate clause comes last in
is called IP: Japanese -- just where the head should be. That's why we call it CP!
Similarly, the fact that English C "cares about" its complement's I, told us that I is the
head of the complement of C. Lo and behold, the I of the complement to C comes last
in Japanese. So we call the phrase IP.

The Japanese handout has many relevant examples. Here's the example that shows C final
in CP and I final in IP (as well as V final in VP etc.):
2 Not actually starred if we allow a loony meaning in which we not only know something to which it refers,
but students can meaningfully be said to rain.
 September 16, 2002 summary/page 4

(22) IP

I'

VP
???
V'

CP

C'

IP

I'

VP

V'

NP NP NP V I C V I

Mary-ga John-ga hon-o yon da to omottei ru

Mary-SUBJ John-SUBJ book-DO read PAST that think PRESENT
'Mary thinks that John read the book'