NP SHELLS[1]

Andrew Radford (radford@essex.ac.uk)

 

 

1.  Introduction

Much research in the past two decades (beginning with Abney’s seminal 1987 study) has concluded that there are significant symmetries between the constituent structure of nominal and verbal projections. In work within the Minimalist Program, Chomsky (1995, 1998, 1999) has argued (extending earlier work by Larson 1988, 1990 and Hale and Keyser 1991, 1993, 1994) that verb phrases have a complex internal structure comprising an outer vP shell headed by an abstract light-verb (denoted by lower-case v) and an inner VP core headed by a lexical verb. There have been occasional suggestions in the literature that noun phrases have a similar shell structure, comprising an outer nP shell headed by a light noun and an inner NP core headed by a lexical noun (see e.g. Carstens 2000, who builds on earlier work by Sportiche 1990 and Valois 1991): the nP shell in turn serves as the complement of higher-level nominal projections (See Cinque 1994 and Longobardi 1994, 1996 for some discussion of nominal superstructure). In this paper, I propose a specific variant of the NP-shell analysis, focusing mainly on the syntax of argumental (i.e. argument-selecting) nouns in English.

 

 

2.  Argument merger

If there is symmetry between the structure of noun phrases and verb phrases, we should expect to find a uniform set of UG principles determining how (in the unmarked case) specific types of argument are first-merged within split xP/XP projections. For present purposes, I shall assume a simplified argument projection hierarchy along the lines of (1) below:

 

(1)      [xP bare agent [XP experiencer > theme > oblique > agentive PP > clause]] 

This is intended to indicate that a bare agent (like John in John killed Fred) is projected as an external argument in spec-xP, and that other arguments are projected as internal arguments within XP, with > indicating  c-command; where a head has a single internal argument, this is projected as its complement; where a head has two internal arguments, the higher of these is projected as its specifier and the lower as its complement; where a head has three internal arguments, this requires the projection of a double XP structure with the highest argument serving as the specifier of the superordinate XP, the intermediate argument serving as the specifier of the subordinate XP, and the lowest argument as the complement of the subordinate XP. An agentive PP is a phrase like by John in Fred was killed by John; oblique arguments include PP arguments headed by a P other than genitive of or agentive by; a clausal argument is a non-nominal clause (i.e. one which is case-resistant in the sense of Stowell 1981). The relative ordering of arguments within the hierarchy (1) has correlates in other work: for example, the relatively ‘low’ position of agentive PPs correlates with their chômeur status in Relational Grammar (Perlmutter 1983); and the maximally low position of clausal arguments may be a consequence of the need to avoid the parsing problems posed by centre-embedded clauses (see e.g. Chomsky 1965, §1.2).

In order to establish that the hierarchy in (1) is not specific to nominal projections, I will briefly show how it applies to verbal projections. In relation to unaccusative sentences such as He has arrived, (1) determines that the theme argument he is first-merged as the highest internal argument of V; since it is the sole argument of V, it follows (from the binarity of merger) that it will be projected as the complement of V (an assumption which receives empirical support from the analysis of Italian ne cliticisation in Burzio 1986). In simple transitives like He ate it, (1) specifies that theme it is merged as the complement of V and that the bare agent he is merged as the specifier of v: the verb raises from V to adjoin to the null light-verb heading vP to derive the structure shown in simplified form in (2) below at the end of the vP phase[2]:

(2)      [vP he [v ate] [VP [V t] it]]

In sentences like He gave it to her, agent he is merged in spec-vP by (1), theme it as the highest internal argument (in spec-VP), and oblique goal to her as a lower internal argument in VP-complement position, so that after V-to-v raising we derive the structure (3) below at the end of the vP phase:

(3)     [vP he [v gave] [VP it  [V t] to her]]

In He explained to her that there were complications, agent he is merged in spec-vP, oblique to her in the highest internal argument position (spec-VP) and the that-clause in the lowest internal argument position (VP-complement position), so deriving (4) after V-to-v raising:

(4)      [vP he [v explained] [VP to her [V t] that there were complications]]

In He was criticised by her, the prepositional agent by her originates in the lowest internal argument position (VP-complement position) and the theme argument he in the highest internal argument position (spec-VP), subsequently raising to spec-TP by A-movement. By contrast, in It was agreed by the committee PRO to delay the decision, the complement clause PRO to delay the decision is merged as the lowest internal argument of V and therefore occupies VP-complement position and the by-phrase is merged in the highest internal argument position (spec-VP), so that the by-phrase precedes the infinitive complement and c-commands its PRO subject[3].

 

 

3. Other key assumptions

In addition to assuming the Argument Projection Hierarchy in (1) above, I shall make the following assumptions about genitive case-marking in English:

 

(5)(i)      an experiencer or theme argument of N can be assigned inherent

              genitive case (encoded by of[4])

    (ii)     a null f-complete definite determiner assigns structural genitive case

              (encoded by the affix `s or by a genitive pronoun like my) to the closest

              nominal which it agrees with[5]

 

There are several reasons for thinking that of genitives mark inherent case and `s genitives[6] mark structural case. For one thing, it is characteristic of inherent case in English that it is encoded via a preposition[7] (e.g. to encodes inherent dative case in structures like He lent his car to Mary), and characteristic of structural case that it is morphologically encoded (in the case of pronouns like I/me/my, via inflectional morphology; in the case of genitives like John’s, via an `s affix which has much the same range of overt allomorphs[8] as third person singular present tense –s). Moreover, (structural) `s genitives can encode a wide range of q-roles, whereas (inherent) of genitives can generally encode only theme/experiencer arguments: hence (e.g.) the former – but not the latter – can be assigned to an agent argument[9], as we see from the contrast in (6) below:

 

(6)(a)      the government’s ban on tobacco advertising

    (b)     *the ban of the government on tobacco advertising

 

In addition, a temporal expression like last week can be assigned structural but not inherent genitive case, as (7) below shows:

 

(7)(a)      last week’s announcement that a new peace treaty had been signed

    (b)     *the announcement of last week that a new peace treaty had been signed

 

The ungrammaticality of (7b) can be accounted for in a straightforward fashion if we posit that of encodes inherent genitive case, and hence can only be assigned to (internal) arguments of N, not to adjuncts. Moreover, (for some speakers, at least [10]) `s genitives can undergo raising, whereas of genitives cannot: cf.

 

(8)(a)      the president’s certainty to be re-elected

    (b)     *the certainty of the president to be re-elected

A final argument in support of treating of genitives as inherently case-marked comes from contrasts such as:

 

(9)(a)      extensive criticism over the past few weeks of President Mugabe by

               leading Commonwealth politicians

    (b)     *I have come to believe over the past few weeks President Mugabe to be

                intransigent

 

As the ungrammaticality of (9b) shows, there is a case-adjacency requirement on structural case-marking in ECM structures to the effect that case-assigner and case-assignee must be immediately adjacent. However, no such adjacency effect is found in structures like (9a)[11].

 

In addition to the assumptions made in (1) and (5) above, I shall make the further assumption in (10) below:

 

(10)      The light-noun heading nP is always strong in English, so that nouns

             always raise from the head N position in NP to the head n position of nP

 

If V always raises to v (as argued in Radford 1997), it may be that there is a greater generalisation here, to the effect that a null light head always triggers head-raising.

 

Having outlined the three key assumptions (1/5/10) made here, I now turn to look at how a range of nominals are derived within the framework outlined above.

 

 

4. Unaccusative nominals

Consider first of all unaccusative nominals such as:

 

(11)(a)      the return of the president from Cincinnati

      (b)      the president’s return from Cincinnati

 

By (1), the theme argument the president is projected higher than the oblique (source) argument from Cincinnati. If the theme argument is assigned inherent genitive case by (1), we will have the structure (12) below at the end of the NP cycle:

 

(12)      [NP of the president [N return] from Cincinnati]

 

On the next cycle, the NP in (12) is merged with a strong null light-noun, triggering raising of the noun return from N to n, and so deriving:

 

(13)      [nP [n return] [NP of the president [N t] from Cincinnati]]

 

Subsequent merger of the nP in (10) with the determiner the forms the DP:

 

(14)      [DP [D the] [nP [n return] [NP of the president [N t] from Cincinnati]]]

 

If attributive adjectives are contained in some projection intermediate between D and nP, an attributive adjective like unexpected can be positioned between the and return, so giving rise to structures such as the unexpected return of the president from Cincinnati[12].

 

In the case of (8b) the president’s return from Cincinnati, (1) will once again determine that the theme argument the president is projected higher than the oblique (source) argument from Cincinnati. This time, however, we assume that the president does not receive (optional) inherent case, so that at the end of the NP cycle we have the structure (12) below:

 

(15)      [NP the president [N return] from Cincinnati]

 

On the next cycle, the NP in (15) is merged with a strong light noun which triggers raising of the noun return from N to n, so deriving the nP (16) below:

 

(16)      [nP [n return] [NP the president [N t] from Cincinnati]]

 

On the final cycle, the nP in (16) is merged with a null f-complete determiner, so forming the DP (17) below:

 

(17)      [DP [D ø] [nP [n return] [NP the president [N t] from Cincinnati]]]

 

This null determiner agrees with (and assigns structural genitive case to) the DP the president; the null determiner also carries an EPP-feature which triggers raising of the president to spec-DP, so deriving:

 

(18)     [DP the president’s [D ø] [nP [n return] [NP t [N t] from Cincinnati]]]

 

If attributive adjectives are contained within a projection intermediate between D and nP, we correctly predict that the moved genitive DP the president can ‘cross’ intervening attributive adjectives, so giving rise to structures such as the president’s unexpected return from Cincinnati.

 

 

5. Experiential nominals

Now consider how we derive experiential nominals such as:

 

(19)(a)      the unwillingness of the chairman to admit responsibility

      (b)      the chairman’s unwillingness to admit responsibility

 

By (1), the experiencer argument the chairman will originate in spec-NP and the clausal argument to PRO admit responsibility in NP-complement position. If the experiencer argument the chairman is assigned inherent genitive case by (5i), we will have the structure (20) at the end of the NP cycle:

 

(20)      [NP of the chairman [N unwillingness] to admit responsibility]

 

Merger of the NP in (20) with a strong light-noun will trigger raising of the noun unwillingness from N to n, thereby forming:

 

(21)      [nP [n unwillingness] [NP of the chairman [N t] to admit responsibility]]

 

Subsequent merger of (21) with the determiner the will derive the DP (19a) the unwillingness of the chairman to admit responsibility.  In the case of (19b), the noun unwillingness is again merged with its complement to admit responsibility and with its specifier the chairman, but this time the chairman is not assigned (optional) inherent genitive case by (5i), so that at the end of the NP cycle we have:

 

(22)      [NP the chairman [N unwillingness] to admit responsibility]

 

Merger of the NP in (22) with a strong light-noun will trigger raising of the noun unwillingness from N to n, thereby forming:

 

(23)      [nP [n unwillingness] [NP the chairman [N t] to admit responsibility]]

 

The resulting nP will in turn be merged with a null determiner, deriving[13]:

 

(24)      [DP [D ø] [nP [n unwillingness] [NP the chairman [N t] to admit resp]]]

 

Since the null determiner is (by hypothesis) f-complete, it will agree with (and assign structural genitive case to) the DP the chairman. Since the determiner also has an EPP-feature, it triggers movement of the genitive DP the chairman’s to spec-DP, so deriving (25) below:

 

(25)      [DP the chair’s [D ø] [nP [n unwillingness] [NP t [N t] to admit resp]]]

 

As would be expected (given our earlier assumptions), the bold-printed genitive DP can move across an intervening attributive adjective, resulting in structures like the chairman’s apparent unwillingness to admit responsibility.

 

 

6. Agentive nominals

Now consider how we derive agentive by-phrase nominals such as:

 

(26)(a)      the destruction of the city by the enemy

      (b)      the city’s destruction by the enemy

 

By (1), the theme argument the city is projected as the higher (specifier) argument of N and the agentive PP by the enemy as its lower (complement) argument; by (5i) the city can be assigned inherent genitive case and realised as of the city. If so, at the end of the NP-cycle, we will have the following structure:

 

(27)      [NP of the city [N destruction] by the enemy]

 

Merger of the light-noun n with the NP in (27) will yield the structure (28) below after N raises to n:

 

(28)      [nP [n destruction] [NP of the city [N t] by the enemy]]

Merger of the resulting structure with the overt definite determiner the ultimately derives (26a) the destruction of the city by the enemy.

If the of-phrase is heavy, it can be extraposed, as in (29) below:

 

(29)      the destruction by the enemy of the city and the surrounding villages

 

The word-order in (29) is marked (in the sense that the canonical word-order illustrated in (26a) is for the of-phrase to precede the by-phrase), with the italicised of-phrase showing a characteristic heaviness effect. One possibility is that structures like (29) may be the result of a movement operation traditionally termed extraposition: if there is parallelism between the syntax of nominal and verbal structures and if (as Chomsky 1999, p.18 suggests) extraposition in the verbal domain involves adjunction to vP, it seems likely that extraposed nominal arguments are adjoined to nP. However (as Bob Borsley point out), an extraposition analysis would violate Kayne’s (1994) Linear Correspondence Axiom/LCA by virtue of involving rightward movement. An alternative in situ analysis which would not involve movement and hence not violate LCA would be to suppose that heavy constituents occupy the rightmost position on the hierarchy (1) and are merged in situ without undergoing movement[14].

 

Now consider the derivation of (26b) the city’s destruction by the enemy. At the end of the NP-cycle, we have the structure (30) below (with the theme argument the city not having been assigned optional inherent genitive case):

 

(30)      [NP the city [N destruction] by the enemy]]

 

Merger of n with the resulting NP and N-to-n raising derives:

 

(31)      [nP [n destruction] [NP the city [N t] by the enemy]]

 

Merger of a null f-complete definite determiner with this nP in turn derives:

 

(32)      [DP  [D ø]  [nP [n destruction]  [NP the city  [N t] by the enemy]]]

 

The head D of DP c-commands (and agrees with) the city, and values its case-feature as genitive (with structural genitive case being spelled out as the suffix `s on the city’s); D carries an EPP feature which triggers movement of the city’s from spec-NP to spec-DP, so deriving:

 

(33)     [DP the city’s  [D ø]  [nP  [n destruction]  [NP t  [N t] by the enemy]]]

 

A genitive expression moving to spec-DP can cross an intervening nP-modifying adjective, e.g. like recent in a structure such as the city’s recent destruction by the enemy.

 

If a referential DP is always a strong phase and nP is a strong phase only when it has an external argument, such a derivation will not lead to violation of Chomsky’s (1999, pp. 10-11) Phase Impenetrability Condition which specifies that:

 

(34)     Only the head H and edge a of a strong phase HP (not its domain YP) are

            accessible to the head Z of an immediately superordinate strong phase ZP in

            structures of the schematic form [ZP Z... [HP a [H YP]]

 

This is because nP (by virtue of lacking an external argument in spec-nP) is not a strong phase in (32), and so does not block agreement between D and the city.

 

Although movement of the noun destruction from N to n is not directly visible in (33), some evidence that N-to-n movement does indeed take place is provided by nominals like (35) below:

 

(35)      the city’s destruction last week by the enemy

 

If we assume that expressions like last week can serve as ‘adverbial’ NP-modifiers (first-merged on the edge of NP), a nominal like (35) will have the structure (36) below:

 

(36)     [DP the city’s  [D ø]  [nP  [n destr]  [NP last week t [N t] by the enemy]]]

 

and movement of the noun destruction from N to n thereby becomes directly visible.

 

 

7. Transitive nominals

Now consider the derivation of ‘transitive’ nominals like (37) below:

 

(37)      the enemy’s destruction of the city

 

Here, the theme argument the city is merged as the complement of N by (1) and assigned inherent genitive case by (5i); the external argument the enemy is merged as the specifier of the light noun n by (1), so that at the end of the nP cycle (after N-to-n raising), we have the structure:

 

(38)      [nP the enemy [n destruction] [NP [N t] of the city]]

 

Merger of a null f-complete definite determiner with the resulting nP yields:

 

(39)      [DP [D ø] [nP the enemy [n destruction] [NP  [N t] of the city]]]

 

The null f-complete head D of DP agrees with and assigns structural genitive case to the enemy; the EPP-feature of D drives movement of the enemy’s to spec-DP, so deriving:

 

(40)      [DP the enemy’s  [D ø]  [nP t  [n destruction]  [NP  [N t]  of the city]]]

 

As before, the moved genitive DP the enemy’s can cross intervening nP-modifying adjectives like recent in structures such as the enemy’s recent destruction of the city. Agreement between D and the enemy in (39) is not blocked by the Phase Impenetrability Condition (34) since the enemy is at the edge of nP and therefore accessible to D.

 

Next, consider a more complex (3-argument) transitive structure such as:

 

(41)      the withdrawal of troups from the occupied territories by the Israelis

 

By (1), the theme argument troups is merged as the highest argument of N (and assigned inherent genitive case by (5i)) and by the Israelis is merged as the lowest argument of N; the oblique argument from the occupied territories is an intermediate argument of N. Since merger is a binary operation, a double-NP structure has to be projected to accommodate the three arguments. At the end of the inner NP cycle, we will have the following structure:

 

(42)     [NP from the occupied territories [N withdrawal] by the Israelis]

 

At the end of the outer NP cycle, we derive the following structure (after N-to-N raising[15]):

 

(43)     [NP of troups  [N withdrawal] [NP from the occup territories [N t]  by the Israelis]]

 

Merger of n with the resulting NP (and N-to-n raising) yields:

 

(44)     [nP [N withdrawal] [NP of troups [N t] [NP from the occ terr [N t]  by the Israelis]]]

 

Merger of the determiner the with the structure in (44) yields (41) the withdrawal of troups from the occupied territories by the Israelis.

 

Now compare the derivation of (41) with that of (45) below:

 

(45)      the Israelis’ withdrawal of troups from the occupied territories

 

Here, of troups and from the occupied territories are the two internal arguments of withdrawal, so that at the end of the NP-cycle we have the structure (46) below:

 

(46)      [NP of troups [N withdrawal] from the occupied territories]

 

The DP the Israelis is the external argument of the light noun heading nP, so that at the end of the nP cycle (after N-to-n raising) we have the structure (47) below:

 

(47)      [nP the Israelis [n withdrawal] [NP of troups [N t] from the occ territories]]

 

Subsequent merger of the nP in (47) with a null determiner, and movement of the Israelis from spec-nP to spec-DP will derive:

 

(48)      [DP the Israelis’ [D ø] [nP t [n withdrawal] [NP of troups [N t] from the occ terr]]]

 

The analysis in (48) correctly predicts that the genitive DP the Israelis’ can cross an nP-modifier like recent in the Israelis’ recent withdrawal of troups from the occupied territories.

 

 

 

8. Nominals with clausal complements

Now consider nominals with clausal complements such as:

 

(49)(a)      the jury’s decision to acquit the defendant

      (b)      the decision by the jury to acquit the defendant

 

In (49a), the clausal argument to acquit the defendant is the only internal argument of decision, and so is merged as its complement; the external agent argument the jury is merged in spec-nP, so that at the end of the nP cycle we have the following structure (assuming N-to-n raising):

 

(50)      [nP the jury [n decision] [NP [N t] PRO to acquit the defendant]]   

 

PRO is c-commanded by its controller the jury in spec-nP. Subsequent merger of (50) with a null D with abstract agreement properties, assignment of structural genitive case to the jury and movement of the jury to spec-DP will derive (49a) the jury’s decision to acquit the defendant. 

 

However, the derivation of (49b) is somewhat different. Here, there are two internal arguments; by (1) the clausal argument must be the lower argument (and hence the complement of N) and the agentive by-phrase the higher argument (and hence the specifier of N), so that at the end of the NP cycle we have:

 

(51)      [NP by the jury [N decision] PRO to acquit the defendant]

 

As required, PRO is c-commanded by its controller, namely the agentive argument by the jury. On the nP-cycle, N will raise to n, deriving:

 

(52)      [nP  [n decision]  [NP by the jury  [N t] PRO to acquit the defendant]]

 

On the DP cycle, the resulting nP will be merged with the to derive the DP (49b) the decision by the jury to acquit the defendant.

 

 

9. Temporal genitives

An interesting variant of (49b) are structures like the following (noted by Fukui 1986) which contain a temporal genitive like yesterday’s:

 

(53)      yesterday’s decision by the jury to acquit the defendant

 

One analysis of temporal genitives (which we adopted earlier in (36) above) is to suppose that they are NP-modifiers which are first-merged on the edge of NP, so that at the point where D is merged with nP we have the following structure[16]:

 

(54)     [DP [[D ø] [nP [[n decision] [NP yest [by the jury [[N t] PRO to acquit the def]]]]]]]

 

D will then agree with (and assign structural genitive case to) yesterday, and the EPP-feature of D will drive movement of yesterday’s to spec-DP, deriving the following structure:

 

(55)     [DP yesterday’s [[D ø] [nP [[n decision] [NP t [by the jury [[N t] PRO to acquit the def]]]]]]]

 

Since nP has no external argument, it is not a strong phase and so will not block assignment of structural genitive case to yesterday by D[17].

 

 

10. Prenominal arguments

Now let’s turn to consider the syntax of the italicised prenominal expressions in structures such as the following:

 

(56)(a)      the government decision PRO to ban French beef

      (b)      the French decision PRO to ban British beef

 

The italicised expressions seem to occupy spec-nP in that they precede the noun decision (which, via N-to-n raising, comes to occupy the head n position of nP), serve as controllers for PRO and occur after nP-modifying adjectives like recent, as the examples below illustrate:

 

(57)(a)      the recent government decision PRO to ban French beef

      (b)     *the government recent decision PRO to ban French beef

 

 

(58)(a)      the recent French decision PRO to ban British beef

      (b)     *the French recent decision PRO to ban British beef

 

Since the status of French as an adjective in (56b) seems relatively clear, it could be that government in (56a) serves a similar (quasi-adjectival) function: indeed, it might be claimed that French serves as an argumental adjective in (56b), fulfilling the role of the external argument of decision. Alternatively, it may be that government is an N rather than a DP in (56a), and is projected into spec-nP by virtue of being the external argument of decision, escaping the case filter and consequent raising to spec-DP by virtue of having the categorial status of N rather than DP. (See Radford 1993, p.79 for a suggestion along these lines.)

 

One of the many puzzles surrounding structures like (56) concerns how we account for contrasts such as the following

 

(59)(a)      the defeat of the government by the opposition

      (b)      the opposition defeat of the government

 

(60)(a)      the opposition’s defeat of the government

      (b)     *the government’s opposition defeat

 

(59a) seems to have essentially the same derivation as (26a), the destruction of the city by the enemy, and (60a) as (37) the enemy’s destruction of the city. In (59b), the government is merged as the complement of defeat by (1) and assigned inherent genitive case by (5i). On the nP cycle, N raises to n and the external argument opposition is merged in spec-nP. The resulting nP is then merged with the determiner the to derive:

 

(61)      [DP  [D the]  [nP opposition  [n defeat]  [NP  [N t]  of the government]]]

 

But now consider why (60b) is ungrammatical. The derivation of (60b) will proceed essentially as for (60a) up to the stage represented in (61), except that the overall DP will be headed by a null f-complete D, and the government will not be assigned inherent case and so will still be active by virtue of having an unvalued case-feature. At relevant stage of derivation, we thus have:

 

(62)      [DP [D ø]  [nP opposition  [n defeat]  [NP  [N t]  the government]]]

 

We might therefore expect that the null determiner occupying the head D position of DP can agree with (and assign structural genitive case to) the government, and that the EPP feature of D will drive movement to spec-DP, so deriving:

 

(63)      [DP the government’s [D ø]  [nP opposition  [n defeat]  [NP  [N t]  t]]]

 

However, the resulting structure is ungrammatical. Why should this be? The answer is that the derivation will violate the Phase Impenetrability Condition/PIC (34) if a referential DP is always a strong phase, if nP is a strong phase when it has an external argument, and if opposition is the external argument of nP: this is because PIC will block agreement between D and the government across the intervening strong phase nP[18]. Phase impenetrability means that the case feature of the DP the opposition cannot be LF-erased and PF-valued (nor indeed can the EPP-feature of D be erased), causing the derivation to crash at both LF and PF. By contrast, the corresponding agentive nominal (64) below is grammatical:

 

(64)      the government’s defeat by the opposition

 

and this has a derivation parallel to that of (26b) the city’s destruction by the enemy.

 

 

11. Possessive nominals

Although my main concern in this paper is with the syntax of argumental nominals in English, I will briefly look at the syntax of possessive nominals[19] such as:

 

(65)(a)      Mary’s new hat

      (b)     *the new hat of Mary

 

Since nouns denoting concrete objects like hat are non-argumental, it is clear that Mary does not function as an argument of the noun hat. Rather, it seems more likely that Mary originates as the specifier of an abstract possessive morpheme poss which assigns the role of possessor to its specifier (Mary) and possessum to its complement (new hat). If (as Cinque 1994 argues) attributive adjectives are specifiers of a superordinate functional projection FP[20], (65) will have the simplified structure (66) below at the point where D is first-merged with its possp complement:

 

(66)      [DP [D ø] [POSSP Mary [POSS ø] [FP new hat]]]

 

Since Mary is not an argument of hat, Mary cannot be assigned inherent genitive case (marked by of) under (5i) – hence the ungrammaticality of (65b) *the new hat of Mary. Instead, the null f-complete D heading DP agrees with the nearest nominal which it c-commands (namely, the possessor Mary) and assigns it genitive case (so that the possessor surfaces as Mary’s); the EPP-feature of D forces movement of Mary’s to spec-DP, so deriving:

 

(67)      [DP Mary’s [D ø] [POSSP t [POSS ø] [FP new hat]]]

 

And (67) is the structure associated with (65a) Mary’s new hat.

 

 

12. Conclusion

The overall conclusion argued for here is that there is a significant parallelism between the structure of (argumental) nominal and verbal projections, and that both may comprise an inner XP core with a lexical head X contained within an outer xP shell with a light head x. The parallelism also extends to their superstructure in that both are in turn contained within a superordinate strong phase (CP for complete clauses, DP for complete nominals), and if both vP and nP are only strong phases if they project an external argument in spec-vP/spec-nP.

 

 

13. References

Abney, S P (1987) The English Noun Phrase in its Sentential Aspect, PhD diss,

   MIT

 

Boskovic, Z (1997) The Syntax of Nonfinite Complementation: An Economy

   Approach, MIT Press

 

Bowers, J. (1993) ‘The syntax of predication’, Linguistic Inquiry 24: 591-656.

 

Burzio, L (1986) Italian Syntax, Reidel, Dordrecht

 

Carstens, V (2000) ‘Concord in Minimalist Theory’, Linguistic Inquiry, vol. 31,

   no.2

 

Chomsky, N (1965) Aspects of the Theory of Syntax, MIT Press, Cambridge Mass

 

Chomsky, N (1995) The Minimalist Program, MIT Press, Cambridge Mass

 

Chomsky, N (1998) Minimalist Inquiries, MIT Working Papers in Linguistics no.

   15

 

Chomsky, N (1999) Derivation by Phase, MIT Working Papers in Linguistics

 

Cinque, G (1994) ‘Evidence for partial N-movement in the Romance DP’, in G

   Cinque, J Koster, J-Y Pollock, L Rizzi & R Zanuttini (eds) Towards Universal

   Grammar, Georgetown University Press, Washington DC, pp. 85-110.

 

de Wit, P. (1995) ‘Double genitives’, OTS Yearbook 1995, Utrecht University,

   pp.101-113

 

Fukui, N. (1986) A Theory of Category Projection and its applications, PhD diss,

   MIT

 

Hale, K & Keyser, S J (1991) On the Syntax of Argument Structure, Lexicon

   Project Working Papers, Center for Cognitive Science, MIT

 

Hale, K & Keyser, S J (1993) ‘On argument structure and the lexical expression of

   semantic relations’, in K Hale & S J Keyser (eds) The View from Building 20,

   MIT Press, Cambridge Mass, pp.53-109

 

Hale, K & Keyser, S J (1994) ‘Constraints on argument structure’, in B Lust, M

   Suñer & J Whitman (eds) Heads, Projections and Learnability, Erlbaum,

   Hillsdale NJ, vol.1, pp.53-71

 

Kayne, R (1994) The Antisymmetry of Syntax, MIT Press, Cambridge Mass.

 

 

Larson, R (1988) ‘On the double object construction’, Linguistic Inquiry 19: 335

   -91

 

Larson, R (1990) ‘Promise and the theory of control’, Linguistic Inquiry 22: 103-39

 

Lobeck, A (1990) ‘Functional heads as proper governors’, Proceedings of NELS

   20: 348-362

 

Longobardi, G. (1994) ‘Reference and proper names’, Linguistic Inquiry 25: 609

   -666

 

Longobardi, G. (1996) ‘The syntax of N-raising: a minimalist theory’, OTS

   Working Papers no 5, Research Institute for Language and Speech, Utrecht

 

Murusagi, K & Saito, M (1994) ‘Adjunction and cyclicity’, Proceedings of the 13th

   West Coast Conference on Formal Linguistics, pp. 302-317

 

Perlmutter, D (1983) Studies in Relational Grammar, University of Chicago Press

 

Radford, A. (1993) ‘Head-hunting: on the trail of the nominal Janus’, in G Corbett,

   N M Fraser & S McGlashan (eds) Heads in Grammatical Theory, Cambridge

   University Press, Cambridge, pp. 73-11

 

Sportiche, D (1990) Movement, Agreement and Case, ms. UCLA

 

Stowell, T (1981) Origins of Phrase Structure, PhD diss, MIT

 

Valois, D (1991) The Syntax of DP, PhD diss, UCLA

 

Zribi-Hertz, A (1997) ‘On the dual nature of the “possessive” marker in Modern

   English’, Journal of Linguistics 33: 511-537

 

__________________________________________________________________

 

 

 

 

 



[1] I am grateful to Bob Borsley, Claudia Felser and Mike Jones for comments on an earlier draft of this paper. This version appeared in Essex Research Reports in Linguistics (November 2000), vol. 33, pp 2-20.

[2] To simplify exposition, labelled bracketings show only heads and maximal projections, not intermediate projections: null light verbs and light nouns are omitted; traces and their antecedents are printed in the same bold type-face (italic or bold). I am not concerned with the syntax of superordinate (e.g. TP and CP) projections here, or with whether transitive objects move to a higher position.

[3] This will also be the case if PRO remains in situ within VP, as suggested in Bowers 1993 and Boskovic 1997.

[4] This should not be taken to imply that the sole function of the word of is to mark inherent genitive case. Arguably of has a different predicative function in of-phrases such as those italicised in ‘a man of great distinction’ and ‘a portrait of John’s’. (See Kayne 1994, de Wit 1995 and Carstens 2000 on the latter.) The force of or in (5i) is that when a noun has both an experiencer and a theme argument, only one of these two arguments can be assigned inherent genitive case (marked by of). The grammaticality (i) a mother’s love of her children and (ii) the love of a mother for her children and the ungrammaticality of *the love of a mother of her children suggests that when a noun (like love) has both an experiencer and a theme argument, only one of the arguments can be inherently case-marked by genitive of. This presumably follows from a case uniqueness requirement of some kind, requiring each argument of a given head to be assigned a unique case.

[5] The assumption that the relevant determiner is definite accounts for the fact that genitives like John’s car are interpreted as meaning ‘the car belonging to John’ rather than ‘a car belonging to John’. If Lobeck (1990) and Murusagi and Saito (1994) are correct in their observation that only a f-complete functional head allows ellipsis of its complement, the fact that `s possessives allow complement ellipsis (in structures such as ‘John’s car is red, and Mary’s is yellow’) suggests that the relevant determiner is f-complete. For concreteness, the genitive `s affix is here taken to a marker of the genitive case carried by the nominal to which it is affixed, rather than (as in Chomsky 1995, p.263) the head D of DP – though this does not affect the analysis offered here in any significant way. Not taken into account here is the so-called adjectival use of  `s discussed in Zribi-Hertz (1997).

[6] Throughout, the term `s genitives should be taken to include pronominal genitives like my. These are arguably the only ‘true’ genitives found in English, for the reasons given in the following footnote.

[7] It is far from clear that it makes sense to say that (e.g.) inherent dative/genitive case is encoded via the preposition to/of, for the obvious reason that the complement of  to/of surfaces with accusative (not dative/genitive) case. What seems to be meant by saying that a preposition encodes inherent dative/genitive case is that a given head indirectly

q-marks a given argument via the use of the transitive preposition to/of. From this perspective, it is misleading to talk of prepositions like to/of as encoding dative/genitive case. However, since this traditional terminology is so deeply embedded in the relevant literature on case-assignment, I shall continue to make use of it here.

[8] As is well known, the two differ in that possessive `s has a null allomorph e.g. in the soldiers’ behaviour.

[9] Bob Borsley points out to me that questions arise about structures in which a possible agent surfaces as an of-phrase, e.g. in nominals such as the shooting of the hunters (where the hunters has a subject interpretation). However, as Mike Jones observes, such nominals have a manner interpretation rather than an event interpretation – hence we say The shooting of the hunters was accurate (with hunters interpreted as a subject) though not (on the same subject interpretation) *The shooting of the hunters took place yesterday. Also unaccounted for within the present framework are contrasts like *The decision of the Faculty Board about the withdrawal of the degree scheme vs. The decision of the Faculty Board to withdraw the degree scheme.

[10] Not all speakers accept (5a), it seems. However, for speakers like me who do, there is a strong contrast between (5a) and (5b). Note also the ungrammaticality of *her belief of him to be innocent, showing that an infinitive subject cannot be realised as an of genitive: this follows if of marks inherent genitive case, since him is not an argument of belief.

[11] An additional argument in support of the analysis in (2) can be formulated in relation to structures such as Mary’s fear of spiders. If (contrary to what is suggested here) both `s and of genitives involved structural case assignment, `s genitives might be case-marked by a null f-complete D, and of genitives by a null f-complete light-noun. But on the nP cycle where we have the structure [nP [n ø] [NP Mary [N fear] spiders]], the light-noun [n ø] would agree with and assign (of) genitive case to the nearest nominal (Mary) which it c-commands. This not only wrongly predicts that Mary will surface as of Mary but also provides no means of case-marking spiders. By contrast, under the analysis in (2) in the main text, spiders is assigned inherent genitive case (so surfacing as of spiders), and Mary is assigned structural genitive case (surfacing as Mary’s and moving to spec-DP) on the DP cycle, via agreement with a null f-complete D with an EPP feature.

[12] As Bob Borsley points out, some potential empirical evidence in support of the analysis in (14) might come from co-ordinate structures such as The return of the President from Cincinnati and of the First Lady from Cleveland.

[13] Where lack of space makes this expedient, I abbreviate words in structural representations.

[14] It may be that binding data from marked-order structures like my presentation to Mary of herself/??my presentation to herself of Mary favour the in situ analysis, since the to-phrase would c-command the of-phrase on the in-situ analysis though not on the extraposition analysis. However, the relevant binding data are somewhat murky.

[15] The noun withdrawal here raises from the subordinate to the superordinate N position in order to q-mark troups, under the assumption that q-marking involves a local relation between a head and its complement or specifier.

[16] In (31/32), intermediate projections are shown (in unlabelled form) as well as heads and maximal projections. The fact that we can say both (i) the decision yesterday by the jury to acquit the defendant and (ii) the decision by the jury yesterday to acquit the defendant might suggest that a temporal expression like yesterday can be first-merged either immediately above or immediately below the argument in spec-NP (hence immediately before or after the agentive argument by the jury). See also the following note.

[17] As Bob Borsley points out, the analysis of  temporal modifiers like yesterday presented here would predict that temporal expressions are canonically positioned after head nouns. However, the word order in the return yesterday of the president from Cincinnati seems to me to be more marked than that in the return of the president yesterday from Cincinnati. It may be that yesterday is a temporal argument of the noun return and that the preferred position for temporal arguments is intermediate between theme and oblique arguments on the hierarchy (1).

[18] An alternative answer might be in terms of the minimal link requirement which in effect requires D to agree with the closest nominal which it c-commands: if (contrary to the assumption made earlier) opposition is a f-complete DP with an unvalued case feature, D will agree with (and assign structural genitive case to) opposition rather than the government by virtue of being closer to the former than the latter. However, if opposition is a f-complete DP, we should expect it to be modified by the, so deriving the opposition’s defeat of the government. The fact that the noun opposition seems to resist structural case-marking (cf. *the recent opposition’s defeat of the government) and is bare (i.e. lacks the determiner the) might support the suggestion made in the main text that it is a caseless N in structures like the recent opposition defeat of the government.

[19] I am concerned here only with what Zribi-Hertz (1997) terms transitive possessive structures.

[20] It is clear that new cannot be the specifier of hat within a binary-branching framework, and hence that new must be the specifier of a superordinate functional projection FP.