Agreement

Linguistics 322

Agreement is a syntactic phenomenon which requires certain features in two different constituents to match under terms of government. The best known kind of agreement is subject-verb agreement which occurs in various languages throughout the world. Subject-verb agreement occurs in the present tense in English:

 

(1)     A tree grows in Brooklyn.
 
(2)     Some trees grow in Brooklyn.

The verb must agree with the subject in person and number:

 

(3)     I am, therefore I think.

(4)     You are late.

(5)     He is a coward.

(6)     They are at the movies.

(7)     She was such a good student.

(8)     They were such ambitious go-getters.

How does agreement formally work in a grammar. Chomsky proposed the following percolation hypothesis. The features of a constituent modifying the head of the phrase are copied unto to head. This is down be percolation. A feature is copied upward to its maximal projection, and if that maximal projection modifies a head, the features are copied onto the maximal projection of the head . The features then percolate downwards. The dubious specifier also acts like a modifier in this respect. If specifiers were considered adjuncts (as I do (x-bar theory)) then the property holds of all adjuncts.

However, we are claiming, using Chomsky's notion of a link, that feature information is passed through links extablished by government [government], and not by percolation, because many fewer steps are required. We propose, in effect, that agreement is limited to terms of government. We will go so far as to claim that agreement is obligatory when one node governs another and both contain the same feature.

Let us start with a simple sentence like:

(9)     The bird flies.

Let start with the following structure depicting birds in the subject position following raising (raising, internal subject). FLY has not been spelled out. BIRD includes the feature [-Pers] (-Personal) (the third person). The first and second persons are usually considered personal and are represented here as [+Pers]. The follow structure is the occurs after the first pass through the lexicon and the raise of NP to subject position. Government links are extablished beween the NP and T, t and V, and V and NP. The linkes of this type of government are shown in red. The blue link is what we will call antecedent government:

(10)

The features of the head noun [-Pers] and [+Pl] percolate upwards to NP. Percolation apply to all heads and their adjuncts. First, let 'j' represent the features of T. 'j' percolates upwards to TP.Next, since NP is adjoined to TP (or a specifier of TP) the features percolate upwards to TP. We represent the features of the head first and those of adjunction following and the symbol '<' meaning adjoined to the right: 'j<i' ('x>y' means x is adjoined to the left). Then the percolate down to T, and they are copied to the lowered T. To keep the derivation legible, we will use the subindex 'i' to represent the relevant features of N which percolate to T.

After percolation and all other rules have applied, a second pass through the lexicon spells out the features into phonological form:

(11)

The features of number and person in the lowered T must match the same features in the subject NP. A checking mechanism is required to ensure that they do match. Otherwise the sentence will crash.

Finally, we go to the lexicon in order to insert the phonological form of each node. C receives no phonological form ([empty]), NP is spelled as the bird (remember, we haven't considered the internal structure of NP yet), the original T is empty since there is no host for the tense affix, V is spelled out as /flaj/ 'fli-' and T is spelled out as /z/ '-es':

(12)

Returning to the lexicon, the features adjoined to FLY must match the same in the lexical entry. The null suffix '0' matches the features [-Past], [-Pers], [+Pl]. The suffix '-s' ('-s' is spelled s '-es' after vowels) matches the features [-Past], [-Pers], [-Pl]. .FLY is spelled out as fli-':

(13)    fli+es.

In the plural variant of (9):

(14)     The birds fly,

The features of tense are copied to V. There is no inflectional ending, all the features are spelled out in one form ( a morpheme). :

(15)

C is spelled out as [empty], D as the, BIRD as bird-, Nu as -s, and FLY as fly:

(16)

 

Features may also percolate onto the complement of a head. There is no agreement between a complement and a head in English, except, perhaps in Case assignment. But some languages do show head complement agreement. Hence the universal need for head-complement percolation. This system could work in Case assignment. Suppose that the verb generates the feature [Acc] (the accusative Case) as a default if the verb is transitive (Case theory). The Case feature then percolates to VP, and then down to the NP complement. In this way, the NP is assigned the accusative Case. There could be other mechanism for Case assignment, but the fewer the processes, the better.

Let us go a little further than is necessary here. If we adopt the head-complement percolation hypothesis, a curious observations develops. VP is a complement of T. Suppose that the features of T percolated up to TP, then to its complement VP, and then down to V. Going to the lexicon we find that FLY and the features are spelled out as the fli- plus -es:

(17)   

When percolation reaches V (only the upper V exists at this point), the derivation must return to the lexicon. It must match the entry for FLY. In the present tense the default rules apply: FLY is spelled out as fli- (a spelling variant of fly) and T is spelled out as the suffix '0'.

What is appealing about this approach to lowering is that lowering can be viewed as the result of percolation rather than the movement of a bundle of features. This would leave us with one process, rather than with two processes. This is a non-standard approach to the problem and we won't formally adopt it here. But it is well worth thinking about.

Another example of agreement in English is demonstrative-noun agreement. Here a demonstrative determiner must agree in number with the noun it modifies:

(18)     this book

(19)     these books.

This follows whether determiners are viewed as adjoined to N , as specifiers of N, or even as the head of the construction. Actually, the agreement is with the numeral three:

(20)    these three books.

(21)    *this three book.

(22)     

The feature [+Pl] percolates from the noun to the determiner. The determiner features percolate vacuously to the noun. Vacuous features are never spelled out. For expansion of the noun see propositional structure of the.noun.

Note that it is not common practice in include features in tree structures. It not always necessary to do so. It is necessary when the percolation of features is being discussed or perhaps for some other reason. Otherwise, they do not have to be included.

Agreement through government

We propose that agreement is limited to terms of government. We will go so far as to claim that agreement is obligatory when one node governs another and both contain the same feature. Let us illustrate this with the following quantifier phrase, where the quantifier governs its complement:

(23)     Six rabbits.

(24)

Q six governs N rabbits. Note that [+Pl] agrees. If they do not agree, the result is ungrammatical:

(25)     *six rabbit

(26)     *one rabbits.

That government is a requirement for agreement, consider the following with two NPs, each with a quantifier:

(27)     Six children each ate a cookie.

(28)*    *Six child each ate a cookies.

Neither does six govern cookies nor does a govern child. (26) fails because the quantifiers to not govern the indicated governees. See Definitions (government). See also Logical Structure of Noun Agreement.

To return to course outline Click here.

This page last updated 31 DE 2000