2.1. Distinct Analyses of Genitive Phrases and Possessive Adjectives

Genitive phrases and possessive adjectives in the West Germanic languages have certain characteristics in common. They have similar semantics, as they both indicate possession; they have similar positioning and both indicate definiteness in prenominal position, as illustrated in 6.

1. (6)

   

Because of these commonalities, one might be tempted to analyze genitives and possessive adjectives as belonging to the same syntactic category, and as having the same base position and landing site.

There are, however, several compelling arguments against such a uniform analysis. One argument is that genitives and possessive adjectives are not in the same case. In examples 7a,b from German, the noun Tochter ‘daughter’ is in the nominative case. Both components of the genitive phrase ihres Vaters ‘her father's’ in 7a are in the genitive case. However, the possessive adjective seine ‘his’ in 7b is in the nominative case because it must agree in case with the noun it modifies.

1. (7)

   

Genitive phrases and possessive adjectives also differ in Φ-feature agreement. The possessive adjective seine ‘his’ in 7b is feminine singular like the noun it modifies. The genitive phrase in 7a shows no such agreement.

Another major difference between possessive adjectives and genitive phrases is that they differ in positioning and definiteness. Possessive adjectives in German are definite and prenominal. Genitives need not be either. The genitive in 8c, for example, is postnominal and the phrase is indefinite.

1. (8)

   

Possessive adjectives are also not interchangeable with other possessive or genitive forms. This is illustrated in the contrast between 9a,b and 9c,d:

1. (9)

   

Unlike genitive phrases, which are typically DPs in the genitive case, possessive adjectives behave like heads in that they show Φ-feature agreement with all the other heads in the nominal phrase. In the following Italian example, the possessive adjective suoi ‘his/her’ shows the same inflection for masculine plural as the quantifier, the definite article, the adjective, and the noun:

1. (10)

   

Another difference between possessive adjectives and genitives is illustrated by the fact that in German, a genitive phrase triggers strong/primary morphology on an adjective regardless of whether the adjective is marked for singular or plural number, as shown in 11a,b. A possessive adjective, however, triggers strong/primary morphology if it is singular but weak/secondary morphology if it is plural, as in 11c,d.

1. (11)

   

I should point out here that Georgi & Salzmann (Reference Georgi and Salzmann2011) demonstrate that in some ways possessive adjectives in German behave more like determiners than like adjectives. Consider the following examples:

1. (12)

   

One can see here that the genitive ending on the possessive adjective in 12b is indeed the same as the ending on the definite article in 12a, while the genitive ending on the adjective in 12c is [n]. However, I would suggest that the inflection on the possessive adjective actually depends on its position rather than on its category. That is, if it moves to D, as in 12b, it takes on the inflection of a determiner. Yet there are instances in which a possessive adjective does not move to D. In those instances, it bears the inflection of an adjective. In 13, the possessive adjective meinen ‘mine’ bears the genitive ending of an adjective, as in 12c, rather than the ending of a determiner seen in 12a,b.

1. (13)

   

Following Sleeman Reference Sleeman1996 and Zribi-Hertz Reference Zribi-Hertz1997, I assume that in this example there is an empty category (an elided noun) following the possessive adjective meinen. The point is that if a possessive adjective moves to D, it behaves like a determiner. In contrast, if it does not move to D, it behaves like an adjective. It is thus positioning that makes the difference, and the term possessive adjective is justified.Footnote ²

Last but not least, a very important difference between German genitives and possessive adjectives lies in their interaction with universal quantifiers, as shown in 14. This interaction is the subject of this article.

1. (14)

   

Based on this evidence, I follow authors such as Delsing (Reference Delsing1993, Reference Delsing, Artemis and Wilder1998), den Besten (Reference Besten, Thornburg and Fuller2006), Þráinsson (Thráinsson) (Reference Thráinsson (Þráinsson)2007), and Norris (Reference Norris, Nicholas, Thompson and Tucker2011) in treating genitive phrases as maximal projections base-generated in the specifier position of NP or nP, and possessive adjectives as heads of a PossP located somewhere between nP and DP.

Before moving on to the next subsection, I should address the positioning of PossP. Assuming that PossP is somewhere above NP, in order to generate constructions with postnominal possessive adjectives, such as the ones found in the Scandinavian languages and Romanian, the head noun must move past the possessive adjective. Of course, the movement of a head across another head causes locality problems. Julien (Reference Julien2005) avoids this problem by treating possessive adjectives as complex heads base-generated in a phrase (here: PossP) that originates in Spec, NP. This enables the head noun to undergo head movement above the possessive adjective without contravening locality-based restrictions on movement. A similar approach is taken by Cinque (Reference Cinque2010) in his treatment of adjectives. Cinque places APs in the specifier position of an FP above NP, thereby allowing head nouns to move above adjectives and produce a postnominal adjectival construction. The similarity between Julien's handling of possessive adjectives and Cinque's handling of adjectives is to be expected given the adjectival nature of possessive adjectives. I do not elaborate further on the exact position of PossP; I simply point out that I am open to an approach such as Julien's and do not feel that it is fundamentally incompatible with the one presented here.

2.2. The Split NP Hypothesis and Genitive Case Assignment

Following Abney Reference Abney1987, Adger Reference Adger2003, Radford Reference Radford2004, and Longobardi & Silvestri Reference Longobardi, Silvestri, Silvia and Parodi2012, I assume a Split NP Analysis (nP/NP), patterned after the Split VP Analysis that began with Larson Reference Larson1988. The parallels between Split VP and Split NP are quite striking. For example, both involve raising (V to v and N to n), and both involve the assignment of thematic roles to one or more arguments. Under the Split VP Analysis, V assigns a thematic role to objects (or to the subjects of unaccusative and passive verbs), while v assigns a thematic role to agentive or experiencer subjects.

The nominal domain appears to work in a similar way. In the German example in 15, the noun Entdeckung ‘discovery’ has two arguments, an agent and a theme. The theme is base-generated in Spec, NP, just as the theme of a verb is base-generated in Spec, VP. N-to-n movement takes place in the same way that V-to-v movement takes place.Footnote ³ Furthermore, again based on example 15, it also seems that at least deverbal nouns lend themselves to an analysis in which agentive subjects are base-generated in Spec, nP, the way agentive subjects of verbs are base-generated in Spec, vP. Even in the case of nondeverbal nouns it is widely believed that nP parallels vP. Szabolcsi (Reference Szabolcsi, Kiefer and Kiss1994), for example, argues that a possessor in a phrase such as John's house is the subject argument (specifier) of the possessee and is assigned a possessor thematic role by the possessee.

1. (15)

   

A possible issue with placing possessors in Spec, nP is that possessors are located higher than adjectives, as seen in 16. This only means that the possessor has been raised to a position above APs, namely, Spec, DP. This issue is discussed in section 2.3. Regarding possessors in phrases such as John's house, below I provide evidence that they are base-generated in Spec, NP rather than Spec, nP (Georgi & Salzmann Reference Georgi and Salzmann2011 also place possessors in Spec, NP).

1. (16)

   

For the purposes of this article, an important question is whether the verbal and nominal domains are similar in the way case is assigned to subjects. In the verbal domain, objects can be assigned case by the verb, but v does not assign nominative case to subjects. It is T/I that is normally associated with the assigning or valuing of nominative case. The question is whether the nominal domain is similar. Consider again 15 and 16b. Since the object Amerikas follows the head noun Entdeckung, it is reasonable to conclude that it was assigned genitive case by that head noun. The position where case is assigned to the subject Erikson is less transparent. In 16b, Eriksons is in Spec, DP, but that does not mean that it was assigned genitive case there (or in Spec, PossP). It could very well have been assigned case in Spec, nP and moved to Spec, DP for other reasons, such as definiteness, which are discussed shortly. The evidence that an agentive subject can be assigned genitive case within the nominal domain comes from Georgi & Salzmann Reference Georgi and Salzmann2011:2080:

1. (17)

   

However, from examples such as this, one cannot conclude whether genitive case can be assigned in Spec, nP. In this example, an agentive subject follows the deverbal noun, which means that the deverbal noun has moved from N to n, past the subject. The subject is therefore clearly in Spec, NP, not Spec, nP. Example 17, therefore, confirms what has already been established, namely, that genitive case can be assigned in Spec, NP in German. It also shows that there is flexibility in the nominal domain with respect to where an agent can be base-generated: in Spec, nP, as in 15, or in Spec, NP, as in 17. Below, I provide evidence that possessors such as the one in John's friends are also base-generated in Spec, NP. This article is therefore ultimately about genitives base-generated in Spec, NP. The question of whether case can be assigned in Spec, nP, which would indicate asymmetry between the nominal and verbal domains, is not explored here, but it is not central to the arguments presented in this article.

2.3. Definiteness and the D-Position

Longobardi (Reference Longobardi1996, Reference Longobardi2001) argues that the Germanic and Romance languages differ with respect to the significance of an empty D-node. Following his arguments, Julien (Reference Julien, Dahl, Bentzen and Svenonius2003, Reference Julien2005) argues that in the Germanic languages, definiteness requires that the D-node—that is, D or Spec, DP—be overtly occupied. This of course means that the D-node in a definite DP has a strong feature that attracts an overt element. I offer the following illustrative examples from German:Footnote ⁴

1. (18)

   

In all three examples, the head noun Freunde ‘friends’ has moved from N to n. The DP in 18a is indefinite. Consequently, it is not necessary for the D-level to be overtly occupied, and no additional movement takes place. The DPs in 18b,c are definite, and there must be something overt at the D-level. Two strategies are available. In 18b, the genitive form Johanns has moved from Spec, NP to Spec, DP. In 18c, the nP has simply merged with the definite article die.Footnote ⁵

It is important to emphasize here that in German the genitive case can be assigned in Spec, NP. This is clear from the above examples. However, unlike German, English and Dutch require preposition insertion for case assignment in Spec, NP:

1. (19)

   

Note that it is truly the genitive case that is assigned in English, but not in Dutch:

1. (20)

   

English seems to still have the vestige of a true genitive case, while Dutch does not. This can be seen from the fact that Dutch does not have the equivalent of the following English genitive forms:

1. (21)

   

These genitives probably contain an empty/elided noun as discussed in Sleeman Reference Sleeman1996 and Zribi-Hertz Reference Zribi-Hertz1997. The point I am making here is that English still has this kind of genitive construction, while Dutch does not. I discuss this issue in more detail in section 5.

It is of course well-known that in addition to the genitive with of as in 19a English also has nongenitive PPs with of that have a possessive meaning, for example, the brother of John. Given the semantics of the preposition of, this phenomenon is not surprising. An of-PP without a genitive case is the norm in the Romance languages (Italian di, French de, etc.) and is also possible in Dutch and in German, as seen in 19b and 4c, respectively. Interestingly, there is a certain complementary distribution between of-PPs with and without the Saxon Genitive. The postnominal Saxon Genitive is clearly associated with indefinites, as shown in 22.

1. (22)

   

Delving too deeply into this matter would go beyond the scope of this paper. This issue is partially addressed in section 5, where I explain why a genitive form is usually prenominal in definite DPs. However, I do not present a complete analysis of the phenomenon illustrated in 22. I leave this highly relevant topic for future research.

To summarize this subsection, in the Germanic languages, definite DPs require an overt element in D or in Spec, DP. Indefinite DPs (presumably) have a null element in the D position. German allows genitive case assignment in Spec, NP. Dutch and English cannot assign case at the N-level without a preposition. Unlike of, the Dutch preposition van does not assign genitive case but objective or accusative case.

4.1. The Inflection on the Universal Quantifier

Before presenting my own analysis, I need to discuss the nature of D when a genitive phrase occurs in combination with a universal quantifier. German has a rich inflectional system that requires prenominal modifiers to show Φ-feature and case agreement with their head noun. However, unlike other modifiers, a universal quantifier need not be inflected for Φ-features and case as long as there is an overt element in D, such as a definite article, a demonstrative, or a possessive adjective:

1. (31)

   

In example 31a, there is no overt element in D, so the quantifier must bear the -e ending. This is confirmed by the ungrammaticality of 31b. In 31c–e, D is overtly occupied, and there is no inflection on the quantifier. If inflection is added to the quantifier when D is overtly occupied, downgrading or ungrammaticality is often the result. Observe the following examples, which could be either in the nominative or the accusative case and mean ‘all the children’:

1. (32)

   

Many speakers accept the example in 32d, but many consider it to be ungrammatical. Even speakers who accept it generally prefer 32a,b.

As already mentioned, the examples in 32 are in the nominative or accusative case. The data remain interesting if the dative and genitive cases are examined. The following examples mean ‘He has helped all the children’. The verb helfen ‘help’ takes the dative case:

1. (33)

   

Furthermore, if one examines the genitive case, one can also see that co-occurrence of the inflection on alle and the definite article produces marginal results. The following examples mean ‘The parents of all the children’, and the phrase meaning ‘all the children’ is in the genitive:

1. (34)

   

Incidentally, the examples in 32–34 would produce similar results if the definite article were replaced with a demonstrative or a possessive adjective, but it must be noted that grammaticality judgments among speakers vary greatly.

At least two issues emerge from these data. First of all, there are ramifications for the claim in Julien Reference Julien, Dahl, Bentzen and Svenonius2003, discussed in section 2.3 above, that definiteness requires overt occupation of D or Spec, DP. Since in the b-examples in 32–34 there is no overt element in D or Spec, DP, one might suggest that the universal quantifier compensates for the lack of an overt element in D or Spec, DP. This is not implausible. As argued in Giusti Reference Giusti1990, the universal quantifier is the head of a QP and selects a DP as its complement. The selected DP is always definite, which means that the universal quantifier bears a definite feature. Given this definite feature, and the fact that the quantifier is adjacent to D, it is quite conceivable that the presence of the quantifier satisfies the need for overt occupation of D or Spec, DP, if a DP is definite. Another possibility, as proposed below, is that the inflection on alle is itself an overt D-element that has attached to the quantifier.

The second issue is the question of why inflection on the quantifier is redundant and causes downgrading for many speakers if D is occupied. One might perhaps propose that it has something to do with the fact that the quantifier is not really in DP but in the higher nominal projection QP, and is therefore not subject to the same inflectional requirements as elements in DP. As long as the item occupying D is inflected, that is enough, and any inflection on the quantifier is superfluous and causes downgrading. The problem is that a claim like this is very difficult to defend because there is no independent evidence for it, especially since all heads in a German DP are obligatorily inflected despite redundancy.

Another possibility that comes to mind is that the inflection on the quantifier in the b-examples in 32–34 is simply a definite article that has been phonologically reduced and attached to the quantifier. In other words, the b-examples are the same as the a-examples except that the definite article has been reduced and moved to Q to attach to the quantifier. The definite articles in the c-examples in 32–34 would thus be the spell out of traces or lower copies, and this could be what causes the downgrading. The problem with this idea is that it does not explain where the inflection on a quantifier comes from if something other than the definite article occupies D. Consider the following phrases:

1. (35)

   

These examples are a bit downgraded for some speakers but are certainly possible. One would not want to claim that the -e on the quantifier in these examples is a reduced definite article, since definite articles do not co-occur with demonstratives or possessive adjectives. One would also not want to say that the -e is a reduced demonstrative in 35a and a reduced possessive adjective in 35b: These two categories do not lend themselves to reduction, given that they bear more semantic weight than a definite article and are bi-syllabic.

To quickly recapitulate the last few paragraphs, two issues need to be addressed. First, the data in 31a, 32b, 33b, and 34b seem to contradict the well-established principle in Julien Reference Julien, Dahl, Bentzen and Svenonius2003 that in a definite DP, an overt element must occupy D or Spec, DP. Second, the data in 32d, 33c, 34c, and 35a,b show that the co-occurrence of an overt element in D and inflection on the quantifier create marginal results, and the reason is not clear. What I propose is that both of these issues are resolved if the inflection on the quantifier is analyzed as an overt D-element that has moved up and attached to the quantifier. This would immediately explain why 31a, 32b, 33b, and 34b are grammatical. It would also offer an explanation for the marginality of 32d, 33c, 34c, and 35a,b because the -e ending and the other overt D-element (definite article, possessive adjective, and demonstrative) would be competing for the D-position, producing a doubling effect.

If the -e ending is a D-element, what is it exactly? As already discussed, one does not want to claim that it is a reduced definite article, possessive adjective, or demonstrative. The best idea that comes to mind is that the -e ending on the quantifier is the phonetic realization of the definiteness feature on D, and that it moves up and attaches to the quantifier. How such a derivation works is briefly discussed in section 4.4. First, however, several other issues must be discussed.

4.2. Two Elements in D

The claim that -e (-en or -er, depending on case) is a D-element implies that in 32d, 33c, 34c, and 35a,b there are two elements in D. Is this plausible? There is in fact very good evidence that D can be occupied by more than one element in German. Consider examples such as the following, which are well-known to Germanicists:

1. (36)

   

One way to analyze these sentences would be to say that the possessive adjective has not moved to D, which is occupied by a demonstrative. I reject this analysis for very good empirical reasons. In 36a, for example, if the possessive adjective meine had not moved to D and were still in its adjectival position, it would have to be inflected with a weak/secondary ending yielding the ungrammatical ∗diese meinen guten Freunde. Example 36b is even more interesting. If the demonstrative were the sole occupier of D, there would have to be a weak/secondary ending on both the possessive adjective and the adjective groß, producing the ungrammatical ∗dieses meine große Glück. If one posits a compound formed by movement of the possessive adjective to D and having it right-adjoin to the demonstrative, the inflection in both sentences in 36 is readily explained.

One might say that in the sentences in 36, the possessive adjective occupies D and the demonstrative is an independent DP that is simply in apposition to the DP headed by the possessive adjective. This explanation does not work. For example, if 36b consisted of two appositive DPs, these DPs would be separated by a slight pause—which they are not—and the demonstrative dieses would appear in the pronominal form dies. The best explanation for 35 and 36 is the double-occupation of D. In the next subsection, I provide more evidence that -e is a D-element.

4.3. Additional Evidence that -e Behaves Like a D-Element

I now present additional evidence that the -e on alle is a D-element in German and Dutch. I begin with German, in which the inflection on the quantifier has the same morphological effect on adjectives that follow it as the plural definite article has. They both trigger the weak -n ending:

1. (37)

   

If there is no overt element in D in a plural context, there is no weak/secondary morphology. This is seen in the following example, in which D is empty, and the adjective bears a strong/primary ending:

1. (38)

   

Another way of saying this is that the combination of definiteness and an occupied D position means weak/secondary inflection in a plural context:Footnote ⁸

1. (39)

   

These examples demonstrate the D-like characteristics of the inflection on the quantifier. Genitive phrases in German do not have this quality. In the case of a prenominal Saxon or non-Saxon/standard genitive, both of which signify definiteness, D is probably empty. Otherwise, there would be weak morphology on the adjectives in the following examples:

1. (40)

   

The example in 40a contains a Saxon genitive, the example in 40b a non-Saxon genitive. Based on the data just presented, I conclude that the -e ending on alle in German originates in D and moves to Q to attach to the quantifier. In contrast, prenominal genitive endings, Saxon and non-Saxon, are not in D but in Spec, DP, along with the noun that they are attached to.

I now demonstrate that in Dutch, the -e ending on alle also originates in D, but the Saxon Genitive ending is in D, not Spec, DP as it is in German. I begin with the Dutch examples in 41.

1. (41)

   

Dutch does not have as rich a nominal inflectional system as German, but it has certainly not lost its nominal inflectional system, and it is practically identical to German in the way it handles the inflection of the universal quantifier. Both languages require either inflection of the quantifier or an element in D. As seen in 41c, Dutch is even stricter than German on the co-occurrence of the definite article and the inflection on alle. That is, whereas 41c produces marginal results in German for many speakers, it is consistently ungrammatical in Dutch. The ungrammaticality of this “doubling” implies that the -e ending on alle is in D in Dutch, just as it is in German.

The following two examples provide evidence that the Saxon Genitive ending is in D in Dutch:

1. (42)

   

Example 41d shows that in a definite phrase with a universal quantifier, D needs an overt element. Example 42a is grammatical because the Saxon Genitive ending is in D, which means that inflection of the quantifier is unnecessary. I am proposing that 42b is ungrammatical because there are two elements competing for the D position: the -s and the element that has become the inflection on alle.

One question that arises here is why there is one instance in which English does not require an overt element in a definite DP, which is impossible in German and Dutch:

1. (43)

   

German and Dutch, unlike English, are both inflectional languages in the nominal domain, and some type of marking is necessary. This is one reason why the German and Dutch examples in 43 are not possible, while the English one is. The other reason why the German and Dutch examples are ungrammatical is, of course, that there is no overt element in D. However, this raises the question of why the English example is possible in the absence of an overt element in D. I treat this as an English-specific issue that requires future research, but conjecture that it has to do with the lack of any kind of nominal inflection in English.

Given that 43b is ungrammatical because Dutch requires inflection in the nominal domain, one may ask why the phrase al mijn vrienden ‘all my friends’, in which there is no inflection on the quantifier or the possessive adjective, is grammatical. There is an explanation. No inflection is needed on the quantifier because, as already mentioned, the possessive adjective mijn occupies D. No inflection is visible on the possessive adjective mijn for the simple reason that this possessive adjective, unlike other adjectives, is not inflected: mijn vriend ‘my friend’ and mijn vrienden ‘my friends’.

4.4. Mechanics of the Phonetic Realization of the Definiteness on D

When the definiteness feature on D is phonetically realized, the derivation proceeds as follows: The universal quantifier selects definite DPs, and the definiteness feature on D is phonetically realized as -e in German and Dutch but remains unrealized in English, which is totally non-inflectional. (The -e ending will of course vary as -en or -er, according to which case it is assigned.) After the universal quantifier has selected a definite DP and the definiteness feature has been phonetically realized as -e, one of the following scenarios is possible:

1. (i) If there is no other element in D (no determiner, no possessive adjective, no demonstrative), -e will attach to the quantifier:

   

2. (ii) If there is an overt element in D, (a determiner, possessive adjective, or demonstrative), the -e ending is optionally but preferably deleted in German and obligatorily deleted in Dutch:

   

3. (iii) If -e is not deleted (possible only in German), it is raised to the quantifier. This unusual procedure produces downgrading for many speakers:

   

Thus, analyzing –e as a D-element can explain the behavior of universal quantifiers in German and Dutch.

4.5. Summary

To briefly summarize this section, the inflection on the universal quantifier in both Dutch and German shows all signs of being a D-element. I have proposed that it is the phonetic realization of the definiteness feature on D. I have argued that the -e ending can co-occur with another D-element in German, such as a determiner, a possessive adjective, or a demonstrative, which produces, for many speakers, a doubling effect. As a result, phrases such as all e seine Freunde ‘all his friends’ are perceived as marginal. My claim that -e can co-occur with another D-element is supported by evidence that more than one element can occupy D in German. Finally, although the -e ending is the same in German and Dutch, the two languages differ with regard to the Saxon Genitive. The Saxon Genitive ending is in D in Dutch but not in German, where it is a true case ending located in Spec, DP, with the possessor it attaches to.