6    Subject movement and case theory

Over the past several weeks, we have been investigating the syntactic structure of sentences and their constituent phrases, and our investigation has revealed evidence that subjects of sentences originate in the V projection and move to a structurally higher position in the I projection. But although we have argued that subjects move, but we have not asked why they do so. This chapter discusses the structural conditions under which case is assigned and derives subject movement from considerations of case assignment.

A first look at case

Consider the contrast between (1) and (2).

(1) a. ok They will help her.
b. ok She will help them.
(2) a. * Them will help she.
b. * Her will help they.

Why are the sentences in (2) ungrammatical? The answer is that the grammar of English requires subjects and objects to appear in a specific case form: subjects in the nominative and objects in the oblique. (3) shows the nominative and oblique forms of personal pronouns in English.

(3)     Nominative Oblique
1 sg   I   me
2 sg/pl you you
3 sg m he him
3 sg f she her
3 sg n it it
1 pl we us
3 pl they them

As is evident, both of the subjects in (2) are oblique forms, and both of the objects are nominative forms. Each of the sentences in (2) therefore contradicts the requirement just stated in two ways.

Morphological case

As the table in (3) indicates, not all pronouns exhibit a morphological distinction between a nominative and an oblique form. Full noun phrases in English also have no morphologically distinct forms for the nominative and the oblique case, as shown in (4).

(4) a. ok The neighbors will help my friend.
b. ok My friend will help the neighbors.

Of particular interest in this context are the case forms of the interrogative pronoun who.. In older forms of English, the nominative was who, and the oblique whom, and the schools more or less valiantly attempt to uphold the distinction. But in vernacular usage, whom is essentially obsolete, and many contemporary speakers appear to have reinterpreted it as a stylistically conditioned formal variant ("Say whom when you want to sound fancy"), rather than as a syntactically conditioned oblique case form.

Besides failing to express its case distinctions in a robust manner, English also exhibits a comparatively limited range of case forms: the nominative and the oblique just discussed, and the possessive. Other languages have more grammatical cases. For instance, instead of a single oblique case, German has a dative and an accusative, giving it four cases in all (nominative, genitive, dative, accusative). As in English, the nominative marks subjects of sentences, and the German genitive corresponds closely to the English possessive. With ditransitive verbs, the first object (the beneficiary or recipient) appears in the dative, and the second object in the accusative, as shown in (5). (In this and following examples, relevant inflectional material is underlined.)

(5) a. deinen Eltern einen Brief schicken
your parents.dat a letter.acc send
'to send your parents a letter'
b. * deine Eltern einem Brief schicken
your parents.acc a letter.dat send
c. * deinen Eltern einem Brief schicken
your parents.dat a letter.dat send
d. * deine Eltern einen Brief schicken
your parents.acc a letter.acc send

With simple transitive verbs, the situation is more complex. Such verbs are said to govern the case of their object. That is, depending on the verb, the object appears in the accusative, in the dative, or (rarely) in the genitive. For all practical purposes, the case that a particular verb governs is not predictable, since even verbs with similar meanings may govern different cases, as (6) and (7) illustrate.

(6) a.   dem Kind { helfen, nachlaufen }
the child.dat help run after
'to help, run after the child'
b. * das Kind { helfen, nachlaufen }
the child.acc help run after
(7) a. das Kind { unterstützen, verfolgen }
the child.acc support pursue
'to support, pursue the child'
b. * dem Kind { unterstützen, verfolgen }
the child.dat support pursue

Prepositions, too, idiosyncratically govern the dative, the accusative, or (rarely) the genitive. Several German prepositions can govern both the dative and the accusative, in which case the dative marks location, whereas the accusative marks direction.

(8) a. Location: in { der Bibliothek / * die Bibliothek } arbeiten
in the library.dat the library.acc work
'to work in the library'
b. Direction: in { die Bibliothek / * der Bibliothek } schicken
in(to) the library.acc the library.dat send
'to send (in)to the library'

Four-case systems similar to the one just described for German are found in Japanese, Korean, and many languages of the Indian subcontinent. Old English (700–1150) also had the same four cases as German, but the distinction between the dative and the accusative was lost during the Middle English period (1150–1500), leaving the modern language with only a single oblique case.

Some languages have even more cases than the ones just described. Finnish has been argued to have over a dozen productive cases, and Turkish has six: nominative, genitive, dative, accusative, locative (for locations), and ablative (used to mark the origin or source of movement). Proto-Indo-European, the reconstructed ancestor of the Indo-European language family, had eight cases: the ones just listed for Turkish (which does not itself belong to the Indo-European family), along with an instrumental (for instruments and means) and a vocative (used to address someone by name). The original Indo-European case system is best preserved in the Baltic languages (Latvian and Lithuanian) and in many Slavic languages, including Ukrainian and Czech. In these languages, the ablative merged with the genitive, leaving seven cases. Several other Slavic languages, including Russian, have almost completely lost the vocative, leaving only six cases (nominative, genitive, dative, accusative, locative, instrumental). In Latin, the Proto-Indo-European ablative, instrumental, and locative merged into a single ablative case that serves all three functions. In Ancient Greek, all three of these cases were lost, but the vocative survived, leaving five cases (the four cases discussed for German and the vocative).

There are also languages with even weaker case systems than English. For instance, the Romance languages distinguish nominative and oblique only on pronouns, but in contrast to English, they have no possessive. Other languages, like Chinese, lack morphological case entirely.

Case features

In traditional grammar, the concept of case is apt to be equated with morphological case. But generative grammar is less interested in the case forms themselves than in their structural licensing conditions (that is, the conditions under which it is grammatical for them to appear). Consider (1), repeated here as (9), and (10).

(9) a.   They will help her.
b. She will help them.
(10) a. You will help her.
b. She will help you.

As is evident, they and she exhibit distinct forms in the nominative and oblique, whereas you doesn't. But in generative grammar, the focus is not on this morphological contrast, but instead on the fact that the form you appears in the same structural position as they and she in (10a), but as them and her in (10b). Contrary to superficial appearances, the two instances of you in (10) are therefore treated not as tokens of a single type, but rather as two grammatically distinct forms that happen to be homonymous.

In order to disambiguate between such homonymous forms and also, more generally, in order to describe the distribution of case forms, it is useful to postulate case features, whose values range over the various cases that a language exhibits. The case features represent abstract case, which may, but need not, be overtly expressed by morphological case, as we just saw in (9) and (10). We will further assume that case features are assigned from heads to noun phrases. That is, they originate on heads, and they are transmitted from those heads to noun phrases under certain structural conditions. At the end of the derivation of a sentence, no case feature may remain on the head it originated on, and every noun phrase must be associated with a case feature. The latter requirement is often referred to as the case filter. Given these assumptions, the aims of case theory include specifying which heads in a language bear which case features, and under what structural conditions case features can pass from heads to noun phrases.

In some versions of case theory, including that in Minimalism, case features are said to be checked rather than transmitted. The general framework is the same as that just presented, but the perspective is declarative rather than procedural. The idea is that noun phrases have case features from the very start of a derivation, and that each case feature on a noun phrase must be checked off against a corresponding case feature on a head, subject to appropriate structural conditions. In this chapter, we adopt the case assignment perspective over the checking perspective for expository reasons.

Licensing configurations

Government

In early work in the Minimalist Program, it was argued that case features are invariably checked in a specifier-head configuration (see below). However, many of the additional theoretical assumptions necessary in order to implement this approach have since fallen by the wayside. As a result, we will adopt a more traditional approach, according to which oblique case is assigned under government, a concept that generative grammarians have adopted from traditional grammar, but redefined, as might be expected, in structural terms. A number of several closely related definitions of government have been proposed in the syntactic literature, of which (11) is representative.

(11)     A governs B iff
a. A is a head,
b. B is a maximal projection, and
c. A and B are sisters (= mutually c-command each other).

In English, oblique case is a feature associated with V and P, and oblique case can therefore only be assigned to sisters of V and P.

Specifier-head agreement

Given the association of oblique case with V, it might be supposed that nominative case is a feature associated with V as well. But this idea is untenable in view of the contrast between the finite and the nonfinite clauses in (12).
Note: We discuss the grammatical counterpart of (12b), He claims to be a rock star, in Chapter 7.

(12) a. Finite: He claims [ that he is a rock star ] .
b. Nonfinite: * He claims [ he to be a rock star ] .

Since the subordinate verb is the same in both examples, the grammaticality contrast in (12) would remain mysterious if it were the verb that bears a nominative case feature. The obvious alternative is that nominative case is instead a feature of I—specifically, of finite I.

What structural relationship must the nominative noun phrase bear to finite I? From the inability of an intermediate structure like (13) to surface, it is evident that the relationship between finite I and the nominative noun phrase cannot be government, since I does not take DP complements, and since any DP position within a VP complement of I is too deeply embedded to be governed by I.

(13)    

In English, it is therefore generally assumed that elements sharing a nominative case feature must be related as specifier to head. This relation is standardly referred to as spec(ifier)-head agreement. Subject movement is a straightforward consequence of this assumption. Since Spec(IP) is conveniently empty in intermediate structures like (13), nothing stops the subject (or part of it) from moving and being assigned nominative case there.

Possessive case in English is also assigned under spec-head agreement. A unitary analysis of possessive full noun phrases and possessive pronouns like those in (14) is possible if we assume a silent possessive morpheme and structures like those in (15) (recall Assignment 4, Exercise 6).

(14) a.   John's book, Mary's car, the men's coats
b. his book, her car, their coats

(15) a.       b.  

Case agreement within noun phrases

Our discussion so far has focused on case features on a noun phrase in relation to corresponding case features on a head that is external to the noun phrase. This section, by contrast, will focus on the distribution of case features within noun phrases.

Notice that in the German examples in (6a) and (7a), repeated here in (16), the distinction between dative and accusative is expressed on the determiner, the head of the noun phrase according to the DP analysis.

(16) a.   dem Kind { helfen, nachlaufen }
the child.dat help run after
'to help, run after the child'
b. das Kind { unterstützen, verfolgen }
the child.acc support pursue
'to support, pursue the child'

In many languages with morphologically overt case distinctions, case can also be expressed on the nouns themselves. Although this is not generally true in German, certain nouns allow or even require such redundant marking, as shown in (17). As before, relevant inflectional material is underlined; notice that the inflections on the determiner are not necessarily identical to those on the noun.

(17) a.   Dieser Schimpanse heißt Nim Chimpsky.
this.nom chimpanzee.nom is called
'This chimpanzee is called Nim Chimpksy.'
b. Wir kennen diesen Schimpansen.
we know this.acc chimpanzee.acc
'We know this chimpanzee.'
c. * { Dieser Schimpansen / diesen Schimpanse } heißt Nim Chimpsky.
d. * Wir kennen { dieser Schimpansen / diesen Schimpanse }.

In traditional grammar, the external relation between a noun phrase and a governor that assigns case to it is distinguished from the noun phrase–internal relation between a determiner and a noun. In the first relation, government, the head is thought of as imposing requirements on the noun phrase, whereas in the second relation, traditionally referred to as agreement, the determiner and the noun are said to agree with each other; that is, the relation is symmetrical.

Given the DP analysis, however, it is possible to express both relations in terms of the same structural notion: namely, government as defined in (11). In order to do so, we need to specify our assumptions concerning case assignment a bit more closely. We have been saying that case features are transmitted from a head to a noun phrase, and the simplest way to interpret this statement is that the case feature is transmitted to a DP node. Modifying this idea just slightly, let us say instead that the case feature is transmitted to the entire spine of the DP projection. Imagine clones of the case feature being smeared along the entire spine. Since the D head of a DP now bears a case feature, we can simply require the case features of a D and its NP complement to match, in much the same way that the case features of a V or P and its DP complement must match. If need be, we can still distinguish between government (in its traditional sense) and agreement, by noting that the case features of a governing head are not shared by its projections (that is, V' or PP bear no case features), whereas the case features of an agreeing head are. If the distinction between government and agreement is not relevant, we can use the more general term selection to refer to the relation between a governing or agreeing head and its complement.

Mediated case assignment

In certain languages, including English, case can be assigned in a way that combines the two simple forms of case assignment discussed earlier (government and spec-head agreement). Evidence for this comes from a special cass of verbs exemplified in what follows by expect. As (18a) shows, expect can take a DP object. As is clear from the structure in (18b), expecting assigns oblique case to her under government.

(18) a.   I was expecting her.
b.

Now consider (19).

(19)     I was expecting her to dislike him.

Here, expect doesn't have quite the same meaning as in (18a). In (18a), it is a person (= an entity) that is expected, whereas in (19), it is a state of affairs (= a proposition), just as it is in (20).

(20)     I was expecting (that) she would dislike him.

The semantic similarity between (19) and (20) strongly suggests that the complement of expect in (19) is the entire sequence her to dislike him. Assuming that to in (19) is the counterpart of would in (20) yields an underlying structure for (19) as in (21a), and subject movement of the matrix and complement subjects then yields the structure in (21b).

(21) a.       b.  

The IP status of the complement in (19) is further supported by a syntactic argument that is based on the distribution of so-called expletive there Expletive there differs from ordinary adverbial there in being unable to bear stress. The contrast in (22) suggests that expletive there is restricted to subject position and unable to function as the object of a verb.

(22) a. ok There is a fly in my soup.
b. * I dislike there in my soup.

Now it was noticed early on in generative grammar that there is a contrast in examples like (23).

(23) a. * I was expecting there.
b. ok I was expecting there to be trouble.

The contrast in (23) would be mysterious if there were a complement of expecting in both examples. But the mystery dissolves if there is a subject in (23b), just as it is in the paraphrase in (24).

(24)     I was expecting (that) there would be trouble.

The question of course remains of how the complement subject in (21b) can receive oblique case from expecting, given that it is not itself the verb's complement. In order to answer this question, let us be clear about the goal of case assignment, which is to transmit a case feature from a case assigner to a noun phrase. In the instances of case assignment discussed so far, this transmission is direct. But in a configuration like (21b), the best the verb can do is to transmit its case feature to the IP projection under government. But the case features are useless on the IP projection, since case can only be realized on noun phrases. Let us therefore let I transmit the oblique case feature of expecting that it has inherited to its specifier under spec-head agreement. We can think of this second step as a last resort measure allowing the verb's case feature to be transmitted to a noun phrase that otherwise would remain without a case feature.

The two-step process just described is often referred to as exceptional case marking (ECM). In order to highlight the fact that it is not completely exceptional, but rather involves both of the structural configurations introduced earlier— government and spec-head agreement—we will generally refer to it instead as mediated case assignment. In ordinary case assignment, a case feature is transmitted directly from a case assigner to that case assigner's complement or specifier. Mediated case assignment, on the other hand, requires an mediating head in addition to the original case-assigning head.