Keir Moulton OUR GRAMMAR
A. TREE GENERATOR
Words and morphemes can be thrown together into phrase structures in any way you want. The labels of these phrases are random, the structures are random. The trees, though, must be well formed: no crossing branches; no multiple immediate dominance (two mothers), the tree must be “connected”, dominance is represented top-to-bottom.
B. LEXICAL ENTRIES
All morphemes/words have lexical entries that specify (at least) the category ad subcategorization.
The lexical entries of Vs, Ns, As and Ps include its category, θ-grids (specifying the kind and number of θ-roles it assigns).
C. i. WORD-TO-TREE RULE
a. A word/morpheme’s category label is the node that’s put in the tree. It is an X0 , a “head”. The head carries all the information that the lexical entry does (we’ll write its pronunciation or spelling below the X0
often). b. Every X0
must head a well-formed XP.
c. A well-formed XP conforms to the X-BAR SCHEMA.
ii. X-BAR SCHEMA
XP → (ZP) X’
X’ → QP X’ takes care of endocentricity, labels, etc. X’ → X’ QP removes phrases that have multiple heads,
X’ → X0
(YP) (UP) and trees that have heads dominating their phrases…etc…
D. THE THETA CRITERION
a. For every θ-role there is a unique position to which that θ-role is assigned.
b. For every θ-position, there is one unique phrase with an appropriate semantic value
that occupies that position (i.e., the argument).
E. PROJECTION PRINCIPLE
i. If X assigns a theta-role to Y, X and Y are sisters.
ii. If anything is sister to a head X0, then X0 assigns a theta-role to it.
F. MODIFICATION PRINCIPLE
X modifiers Y iff.
i. X is sister to Y’ and daughter to Y’
2. What we lose when we don’t have category-specific rules
When we move from having lists of specific phrases structure rules (even the X-bar versions) for each category and replace that with the X-bar schema, we lose a lot of information we need to account for syntactic patterns.
A. We lose information about the categories of non-head elements, such as: i. the category of the complement phrases a head allows
ii. the category of the modifier a X’ allows
Our solution to (Ai):
We put information about the category of complements into lexical entries for heads (like nouns and verbs and so on) by listing their subcategorization in the theta-grids.
Our solution to (Aii):
Here, the hope is that the meaning of modifier phrases and the meanings of the X-bars they attach to is sufficient to determine what category the modifier is. This is embodied by the MODIFICATION PRINCIPLE.
B. We lose information about the linear order of elements in a phrase, such as: i. that complements follow the head in English
ii. that specifiers precede it
iii. that modifiers can precede or follow, depending on the category of the modifier and the category of the X’ modified
Here we will follow an idea found in the syntactic literature on cross-linguistic differences in word order, one which is outlined nicely in your textbook (pp 174-175). The order of the elements after the “→” is now replaced by settings word order parameters. These settings apply to the X-bar schema.
The options that a language has are made available by Universal Grammar. The child fixes the parameter for her language.
Specifier rule: XP → (ZP) X’ or XP → X’ (ZP) Specifier: first/last Adjunct rule: X’ → QP X’ or X’ → X’ QP Adjunct: first/last Complement rule: X’ → X (YP) (UP) or X’ → (YP) (UP) X Head: first/last
(1) Hasan kitab-i oku-de
Hasan-SUBJECT book-OBJECT read-PAST
(2) [Babasnn arabasnn ] altnda father-PERS-GEN car-PERS-POSS-GEN under ‘under (his/her) father’s car’
How do the languages set their X-bar Schema Parameters?
The adjunct rule for English doesn’t set the parameter – adjuncts can come before or after:
(3) a. The [AP happy ] woman.
b. The woman [AP happy with syntax ]
(4) a. John has quietly crept into the hall. b. John has crept into the hall quietly.
Other combinations are possible, but only within the limits set by Universal Grammar.
Also note: the parameters are set for all instances of the rule: if complement precede their verbs, then complements of nouns will too.
(5) Bambara (p. 98)
a. A kasire c. N sonna a ma
He cried I agreed it to
‘He cried’ ‘I agreed to it’
b. Den ye ji min child PAST water drink
‘The child drank water’
(6) Hiaka (see page 132)
a. Tékil né-u ‘aáyu-k b. Hunáa’a yá’uraa hunáka’a hámutta nokriak work me-for is that chief that woman defend ‘There is work for me. “That chief defended that woman”
c. Taáwe tótoi’asó’olam káammomólim híba-tu-‘ure
Hawk chickens young like
‘(The) hawk likes young chickens’
C. Finally, we lose information about:
i. The obligatoriness of specifiers for certain phrases (like TP) ii. The number of complements a head allows
iii. The order of those complements
Recall the TP rule requires that either CP or DP be in the specifier. Our X-bar Schema needs to say that specifiers are optional – since some phrases don’t have any and in some cases they are optional.
Our solution to (Ci):
Expletives
(7) a. It rained/snowed/hailed b. *rained/*snowed/*hailed.
The it doesn’t get a θ-role, but it still must be there.
(8) a. [CP That Bill loves chocolate] is likely. b. It is likely that Bill loves chocolate.
Likely assigns one θ-role – to the CP (which can appear as a subject or object)
When the CP appears as an object, the subject position is filled by a meaningless DP – what is known as an expletive.
Expletives cannot get θ-roles: they do not sit in positions where θ-roles are assigned
(complement position). In the examples below, it gets a θ-role and hence does not act like an expletive.
(9) Non-expletive ‘it’
a. I love it.
b. John’s book of it.
What forces meaningless DPs, then, to appear in the subject position as in (6)?
(10) EXTENDED PROJECTION PRINCIPLE (the “EP P”)
All clauses must have subjects. (i.e. the specifier of TP must be filled by DP or CP)
Let’s think of expletives as a way for sentences to satisfy the EXTENDED PROJECTION PRINCIPLE
expletives anywhere you want!)
Our solution to (Cii)
Here the THETA CRITERION and the PROJECTION PRINCIPLE kick in.
Recall these two principles had the effect that the only complements to a head are phrases that get θ-roles.
(11) a. Mary put the book on the table. b. *Mary put the book.
c. *Mary devoured.
d. *Mary devoured the cake the soup.
Our θ-grids tell us the number (and type) of non-subject θ-roles the verbs put and devour have. In conjuction with the THETA CRITERION and the PROJECTION PRINCIPLE we can only put as
many complement phrases as there are non-subject θ-roles (no more, no less).
Our solution to (Ciii) – Ordering of Complements
Our problem here is that our old PSRs told us the order of complements; this allowed us to explain why PP arguments followed DP arguments.
(1) a. Sheila placed this on the table. b. *Sheila placed on the table this. V’ → V (DP) (PP)
Now that we have purged our grammar of particular phrase structure rules for each category (and purged information about the order of complements), we can’t explain this.
In his 1981 dissertation, Tim Stowell argued that this information too can be taken out of phrase structure rules.
Where do we put this information about the order of complements?
We don’t want to put this information in the θ-grids – every verb that has DP and PP arguments puts them in this order.