In Swedish predicative adjectives agree with the subject (or the object) in grammatical gender and number:
(1)
There are, however, contexts where the predicative adjective and the subject seem to disagree. As will be shown, there are two variants of this construction, which I will refer to as Construction NOM, which is short for the Nominal Construction, and Construction PROP, which is short for the Propositional Construction.Footnote 1 Consider (2) for two examples:Footnote 2
(2)
Both senap and älskare are lexically common gender nouns – nevertheless agreement is in the neuter on the predicative adjectives in (2a, b). Consequently, the agreement morphology, -t, is the same as on the predicative adjective in (1b). In (2a) the subject senap has a mass reading, whereas the subject in (2b), två älskare, has a propositional reading: ‘to have two lovers’. The predicative adjectives in (2) thus seem to display disagreement in gender and/or number. Traditionally the construction in (2) is referred to as Ärter är gott-konstruktionen (‘peas is good’ construction) in Swedish and Pannekaker er godt-konstruksjonen (‘pancakes is good’ construction) in Norwegian, hence the title of this paper. The construction is found also in Danish, and a Danish example of the construction is Sild er godt ‘Herring is good’. It should be stressed that it would be somewhat unintuitive to think of the predicative adjectives in (2a) and (2b) as displaying disagreement, since ‘canonical agreement’ is not really an option in these cases. For (2a) agreement on -Ø would simply be ungrammatical; for (2b) plural agreement, i.e. on -a, would not be ungrammatical as such, but a different reading would be triggered, ‘the two lovers are immoral’, implying that immorality is a property of each one of the lovers. In this article I will show that the ‘disagreement’ in cases like (2) is only apparent – in fact agreement holds. Counter to the traditional assumption, I will also show that the agreement pattern is not default, but motivated by morphosyntactic features on the subject.
Having argued that agreement holds in (2a) and (2b), I will focus on Construction PROP. First of all I will show that the subject of (2b) is not a simple noun phrase, but a clause-like constituent, where the DP två älskare ‘two lovers’ is the syntactic object. Evidence showing this is case properties, the possibility of adding VP-adverbials without inducing a V2 violation, and properties of anaphors. Secondly, I will argue that there is a null verbal element responsible for the propositional reading of (2b). The null verb in question is located in the head of a vP, a ‘small vP’. This element is the null equivalent of a light verb such as ha ‘have’ as well as the preposition med ‘with’, and it assigns accusative case to the DP.
The outline of the paper is as follows. In section 1, I demonstrate the properties of Construction NOM and Construction PROP. The purpose is to point out similarities and differences between the two constructions. A brief overview of earlier work on the constructions is also given. The focus of section 2 is Construction PROP (as in (2b) above). I will show that the subject in this type of sentences is clausal, and that there is a range of readings for the ‘missing’ predicate, corresponding to a set of verbs that are usually referred to as light verbs. I also propose licensing and identification mechanisms for this null element. In section 3, I discuss properties of the med-phrase ‘with-phrase’, which may replace the subject in (2b), if combined with an expletive det ‘it’ as subject:Footnote 3
(3)
Section 4 contains a concluding discussion.
1. THE TWO ‘DISAGREEMENT’ CONSTRUCTIONS: AN OVERVIEW
The agreement pattern illustrated in (2a) and (2b) is not exclusive to Swedish, but found also in e.g. Norwegian and Danish, as pointed out above.Footnote 4 It has been the subject of a vivid discussion in the literature, see Wellander Reference Wellander1949, Reference Wellander1973 [1985]; Heinertz Reference Heinertz1953; Teleman Reference Teleman1965, Reference Teleman1969; Widmark Reference Widmark1966, Reference Widmark and Molde1971; Faarlund Reference Faarlund1977; Malmgren Reference Malmgren1990 [1984]; Hellan Reference Hellan, van Riemsdijk and Muysken1986; Källström Reference Källström1993; Teleman, Hellberg & Andersson (Reference Teleman, Hellberg and Andersson1999, part 3:702ff.); and Faarlund, Lie & Vannebo (Reference Faarlund, Lie and Vannebo1997:776–779) (for a detailed overview, see Källström Reference Källström1993).
More recently Enger (Reference Enger2004) has discussed the construction in Norwegian, and Josefsson (Reference Josefsson2006) from the point of view of Swedish. Enger's main claim is that the agreement in question is semantic, i.e. that it reflects the semantic properties of the subject. The semantic interpretation that he claims to be the trigger of neuter predicative agreement is ‘low degree of individuation’ (p. 26); in other words, the subjects in question ‘rank low on the individuation scale’ (p. 26). Enger rejects – in my view on good grounds – the idea that the subject of ‘pancake’-sentences are pruned infinitival clauses, as has been suggested by Faarlund (Reference Faarlund1977) and Faarlund et al. (Reference Faarlund, Lie and Vannebo1997:767) for example. However – and this is crucial – Enger also rejects the possibility of analysing the agreement in question syntactically: His view is that the agreement in question is semantic, hence ‘agreement of a kind that cannot be described by standard syntactic features’ (Enger Reference Enger2004:21). It is this standpoint that I question in this paper. In my view, the introduction of semantic agreement in the sense ‘non-syntactic agreement’ is an undesirable solution, a device that could be applied only if standard analyses fail. This is not the case here.
As we shall see below, we have good reason to believe that there are two distinct construction types, which I will refer to as Construction NOM(inal) and Construction PROP(ositional). The subject of Construction PROP is clausal, whereas the subject of Construction NOM is a noun phrase. When the subject is propositional, i.e. clausal, agreement in neuter is what we expect. Construction NOM sentences, where the subject appears to be a noun phrase, will be given an analysis along the same lines.
It should be stressed that I do not reject the idea that agreement in neuter is semantic in nature per se. What is rejected is the idea that it is the semantic interpretation of the subject that triggers agreement. With the solution that I propose the semantics of the subject is a function of the feature content of the subject, and this content is mirrored by the feature makeup of the predicative adjective – in the way agreement generally works in the grammar. Agreement is thus ‘semantic’ in the proposed analysis too, but in a trivial sense, and it works in the same way as number, for example: a subject in the plural, such as bilarna ‘the cars’ and husen ‘the houses’ in (1c), has the semantic interpretation ‘more than one’, and this interpretation is reflected in the plural agreement of the predicative adjective, which, consequently, also carries the meaning ‘more than one’.
As pointed out above, one of the main points of Josefsson (Reference Josefsson2006) is that the ‘disagreement construction’ falls into two distinct categories, which I have called Construction NOM and Construction PROP. There are basically four ways in which Construction NOM (see (2a) above) and Construction PROP (see (2b)) differ: (i) the subject of Construction PROP can be paraphrased by an infinitival phrase, which is not possible for the subject of Construction NOM; (ii) Construction NOM cannot be paraphrased by expletive det ‘it’ + a med-phrase ‘with-phrase’, which is possible for Construction PROP (see (3) above); (iii) definite subjects are ungrammatical in Construction I, whereas they are allowed in Construction PROP (though marginally, a fact that will be discussed in detail below); and (iv) the subject of Construction NOM disallows attributive adjectives, whereas this is grammatical for subjects in Construction II.Footnote 5 An overview of these differences is given in Table 1.
Table 1. A survey of the differences between Construction NOM and Construction PROP.
In previous studies (Josefsson Reference Josefsson and Riemsdijk1999, Reference Josefsson2006) I have argued that the subject of Construction NOM sentences contains a null pronominal element, marked +neuter, in the topmost projection of the subject noun phrase. This analysis will be used as the point of departure in this paper. The pronoun in question is responsible for the neuter agreement on the predicative adjective, as well as the substance reading of the subject. The presence of a null pronominal element of this kind is motivated by a parallel construction in Swedish, with an overt pronoun. An example of the latter type is shown in (4a), in which the pronoun, hon ‘she’ occupies a position above the DP, i.e. it precedes the definite article.Footnote 6 Josefsson (Reference Josefsson and Riemsdijk1999, Reference Josefsson2006) refers to this pronoun hon in (4a) as a prenominal apposition.Footnote 7
(4)
The prenominal apposition, hon in (4a), is probably in many ways similar both to the proprial article in northern Swedish and to the obligatory or near-obligatory use of personal pronouns together with proper names in Icelandic; in argument positions hún Lísa (she Lisa) ‘she/Lísa’ is preferred over the simplex Lísa.Footnote 8 The important property of (4a) is that hon ‘she’ adds a semantic feature, +feminine, which is not a part of meaning of the noun professor. Josefsson (Reference Josefsson and Riemsdijk1999, Reference Josefsson2006) refers to the phrase hosting the pronoun hon in (4a) (and consequently also the null pronominal element in (4b)) as a Semantic phrase, a SemP, since the pronouns specify the semantic gender of the whole phrase.
It should be pointed out that a functional projection on top of the DP, hosting personal pronouns, is not unique to the proposed analysis. Similar solutions are suggested for instance by Platzack (Reference Platzack2004:85, Reference Platzack2008), who refers to the phrase in question as a Person phrase, a PersP.Footnote 9, Footnote 10
A difference between the leftmost pronominal element in (4a) and (4b) is not only the distinction overt vs. null, but also the size of the noun phrase, which is hon + a full DP in (4a), but presumably Ø + an NP in (4b).Footnote 11 The proposed analysis resembles in many ways the one in Platzack (Reference Platzack2004:85), where it is suggested that the PersP (which, as pointed out above, corresponds to the SemP in this paper) could host free pronouns or bound agreement elements, and that these pronominal elements can be both +phonological and null.
According to Josefsson (Reference Josefsson and Riemsdijk1999, Reference Josefsson2006) the neuter feature has a dual nature. First of all it is a morphosyntactic feature associated with nouns, in other words a ‘lexical gender feature’. As such the neuter gender does not carry any meaning; there is simply no element of meaning shared by all neuter nouns.Footnote 12 Information about the lexical gender is necessary for the spell-out of the definiteness and indefiniteness articles, as well as for agreement on adjectives (at least in their singular forms). This means that the lexical gender is operative within the boundaries of the DP. In the pronominal system, however, neuter vs. common gender may have a meaning. The pronouns den (it.common) ‘it’ and det (it.neuter) ‘it’ can be used as deictic pronouns, namely in cases when they make reference directly to entities in the world, without going by way of a noun or a noun phrase. As deictic pronouns, the difference between den and det is that den refers to a bounded entity, whereas det refers to an unbounded entity, more specifically a substance or a proposition. In Josefsson (Reference Josefsson and Riemsdijk1999, Reference Josefsson2006) han ‘he’, hon ‘she’, den ‘itbounded entity’ and det ‘itunbounded entity’ represent the four semantic genders in Swedish.Footnote 13 The details of this gender system are not of direct relevance here; what is important is that neuter is a meaningful category within the pronominal domain.Footnote 14
The null hypothesis is that a DP or a SemP subject of a Construction NOM sentence starts out as the argument of the predicative adjective, and that predicative agreement between the adjective and this phrase is established in the canonical way, for instance in a Spec-head configuration in the AP, as proposed in Chomsky (Reference Chomsky1995:354).Footnote 15 As previewed, this means that agreement in neuter in (2) is motivated by features of the subject, it is neither default nor triggered by the semantic interpretation of the noun. The structure of (2a) is shown in (5). (The dotted lines indicate omitted structure.)
(5)
What is important in (5) is that the neuter feature is not a property of the noun itself, but part of the extended projection containing the noun.Footnote 16 Sem° is the head of the subject phrase, which means that it projects and defines the properties of the phrase as a whole.
The proposed analysis explains why the subject in a Construction NOM sentence can take neither definiteness inflection nor an adjectival modifier (see Table 1 above). If a definiteness feature were present on the noun, this feature would also require the presence of a lexical gender feature maybe only for the sake of phonology, since a definiteness suffix in the non-plural is always marked for a grammatical gender. (The definiteness suffix -n/-en is used for common gender nouns and -t/-et for neuter nouns.) Consequently, a definite common gender noun as a subject of an (ungrammatical) Construction NOM sentence, e.g. *Senapen är gul-t (mustard.common.def is yellow.neut) would have to have a +common (lexical) gender feature on the noun and a +neuter feature in the SemP, its topmost projection, according to Josefsson (Reference Josefsson2006). This would presumably cause a gender conflict within the noun phrase – the upstairs projection, the SemP, would carry the neuter feature, whereas the downstairs projection, the DP/NP, would carry a common gender feature – causing the derivation to crash. (For a more detailed account, see Josefsson Reference Josefsson2006.Footnote 17) However, a +definite noun in Construction PROP sentences is acceptable, as shown in Table 1, though there are certain restrictions, which will be discussed in detail in section 2.3 below. Within the proposed framework this is expected, since the definite DP in such cases is not the subject, but an object, embedded inside a clausal structure, hence there will be no conflict between the neuter feature of the SemP and the common gender feature of the (object) DP.
Adjectival agreement, at least in the non-plural, is marked for grammatical gender, and a gender conflict similar to the one described for definite DPs will arise if an attributive adjective is added to the subject of a Construction NOM sentence. The background assumption is that an attributive adjective needs access to the lexical gender of its head noun in order for agreement to be spelled out. Hence, *Fransk senap är gul-t (French.common mustard is yellow-neut) is out, according to Josefsson (Reference Josefsson2006), because the common gender feature on the attributive adjective fransk clashes with the neuter feature hosted in the SemP.Footnote 18 With subjects such as fransk senap ‘French mustard’, the canonical Ø-agreement is fine, as expected: Fransk senap är gul-Ø. In such cases we have no reason to believe that a SemP is present on top of the DP fransk senap. Speaking in general terms, the possibility of having definiteness inflection and adjectival modifiers on the noun depends on how deeply embedded the nominal head is. If the noun is embedded in a clausal structure, definiteness inflection and adjectival modification is fine, otherwise it is not. As we shall see in section 2, we have good reasons to believe that the overt noun phrase in Construction PROP subjects is really more deeply embedded than the corresponding noun in Construction I.
So far I have discussed the fact that definite subjects and adjectival modifiers are disallowed in Construction NOM subjects. Table 1 lists two more differences between Construction NOM and Construction PROP: Construction PROP subjects may be paraphrased by an infinitival phrase, and by a med-phrase ‘with-phrase’ + expletive det, possibilities that are unavailable for Construction NOM sentences. These properties will be discussed in sections 2 and 3.
The traditional explanation for neuter agreement on Construction NOM and PROP sentences is that neuter is default. This is suggested for instance by Corbett (Reference Corbett1991:216), although he prefers to call it ‘neutral agreement’, and Källström (Reference Källström1993:244).Footnote 19 The Swedish Academy Grammar (part 2:226) refers to this type of agreement as ‘non-agreeing predicative’. (The Swedish wording is ‘icke kongruensböjd predikativ’.) In fact, we have good reasons to assume that default agreement (‘retreat to the general case’, in terms of Distributed Morphology, see Halle & Marantz Reference Halle, Marantz, Hale and Keyser1993) is not -t, but -Ø in Swedish. One piece of evidence is that agreement is generalised to -Ø, not -t, in the plural of the northern Swedish dialects. If the default form had been -t we would have expected this agreement form to have been generalised instead. When it comes to Norwegian, Trosterud (Reference Trosterud2001:34–35) has shown that masculine, not neuter is the default gender. Even though Norwegian is a different language, the argument can be carried over to Swedish; there is no indication that neuter is a default gender in any sense in Swedish.
2. CONSTRUCTION PROP
2.1 The subject is clausal
The DP två älskare ‘two lovers’ is the surface subject in (2b), repeated below. I will claim that the DP is not really a subject, but rather an object, embedded in a clausal structure, which, in turn, is used as a subject.Footnote 20
The first argument is that the subject can be paraphrased by an infinitival phrase, where the DP shows up as the syntactic object, as shown in (6).
(2)
(6)
The example in (6) shows that the subject in (2b) has a propositional reading. In earlier versions of generative theory, Construction PROP was analysed in terms of pruning or deletion, as indicated in (2b′) (see Faarlund Reference Faarlund1977; see also Faarlund et al. Reference Faarlund, Lie and Vannebo1997:767).
(2)
Free or unrestricted deletion/pruning of this kind is of course unappealing. Furthermore, as Enger (Reference Enger2004:7) points out, an important question remains with a deletion analysis, namely which verb is deleted. In many cases different lexical items could be supplied:
(2)
The second argument in favour of analysing the surface subject as an object is the possibility of having reflexives. The anaphoric pronouns sin/sina in (7) indicate the presence of a subject antecedent. Crucially, the antecedent is not the overt DP. (Example (7c) is inspired by Teleman & al. (Reference Teleman, Hellberg and Andersson1999, part 3:703))
(7)
The use of a reflexive within an ordinary noun phrase subject is straightforwardly ungrammatical:Footnote 21
(8)
I will refrain from discussing the nature of the subject within the subject phrase in the sentences in (7) in detail, but a plausible candidate for the binder of the reflexives is generic PRO, which we know can bind reflexives in infinitival clauses. This is illustrated in (9). For the sake of convenience – especially since the true nature of the subject in question is not of crucial relevance for me – I will assume that the subject is an instance of generic PRO.
(9)
The third argument indicating that the noun phrase in the subject position is an embedded object is that it has non-nominative case, which is evident if it is a pronoun:Footnote 22
(10)
It is possible to use the corresponding pronoun in nominative as subject, i.e. hon ‘she’, as shown in (11).
(11)
Notice, however, that (11) differs in meaning from (10). In (11) the PP med senap och ketchup is construed as an attributive to the noun: ‘the girl/woman who has/holds mustard and ketchup’. In this case, agreement between the pronoun hon, and the adjective läcker ‘delicious’ is the canonical one. Crucially the propositional meaning ‘to have’ or ‘to eat’ is absent in (11), which means the sentence is trivial, consisting of the subject hon med senap och ketchup ‘she/the female with mustard and ketchup’ + verb + agreeing predicative adjective.Footnote 23
Fourthly, more than one phrase can precede the finite verb in Construction PROP sentences without inducing a V2-violation. If the same constituent is used sentence initially with an ordinary predicate like bruka ge ‘usually give’ or ha ‘have’, the sentence turns ungrammatical:
(12)
The examples in (12) show that the subjects of Construction PROP sentences are not ordinary DPs, but larger phrases, containing an overt DP.Footnote 24
The main point so far is that the surface DP subject of Construction PROP sentences is not simply a noun phrase, but a larger structure, presumably a clausal constituent. It contains a non-overt subject, presumably a generic PRO, a predicate and an object. VP-adverbials can be present too. In 2.2 we shall take a closer look at the structure of this clausal subject.
2.2 The structure of the clausal subject
The examples in (7) show that there is a hidden subject within the subject phrase and (10) that the structure contains a case assigner, responsible for the accusative case to the pronoun. The propositional reading of the subject indicates that the case assigner is a verbal element. As pointed out above it was proposed in earlier days of generative theory that the subject in Construction PROP sentences is an infinitival phrase, where the leftmost part is deleted. Enger (Reference Enger2004:7–9) argues that such an analysis is untenable, since we in that case would be unable to identify the deleted element; different verbs can be supplied in this position.
(13)
In my view, we have no reason to assume that the verb is deleted; instead the verb is null, i.e. devoid of phonological features. The idea is that the null verb is, more specifically, a light verb, corresponding to a basic concept, which we may, at a first attempt, gloss as HAVE. In order to show that a null light verb would not be a unique possibility for this construction, we shall first take a look at a different construction where the presence of a null light verb has been suggested. The construction in question consists of subject + modal + directional PP or adverbial (see Josefsson Reference Josefsson1998:118–120):
(14)
The only overt verb in (14a) is the modal vill ‘want to’ and in (14b) the modal, ska ‘shall, will’, both of which are unambiguous modal auxiliaries in Swedish. Normally these modals combine with main verbs, and are unable to take complements headed by the infinitival marker att ‘to’, a fact that provides clear evidence of their status as modals. Thus, (14) shows that a main verb that includes the meaning GO can be left out, provided a modal is present, and provided there is a directional PP or adverbial. Different actual verbs could be supplied in (14): åka ‘go’, fara ‘go’, resa ‘travel’, etc. Josefsson (Reference Josefsson1998:118–120) argues that the null main verb in (14a, b) is licit if properly licensed and identified in the sense of Rizzi (Reference Rizzi1986). The licensing requirement is fulfilled by the auxiliary and the identification requirement by the directional PP/adverbial, which identifies the content of the null verb. The concept GO does not correspond straightforwardly to any particular lexical item in Swedish, but there is a cluster of verbs with this concept as a core part of their meaning, as pointed out above.
The verb GO is a good candidate for a light verb. It is neither a modal nor an auxiliary, and it corresponds to a basic concept, which we may illustrate by the source, goal, and path schema as follows:
(15)
The representation in (15) is unspecified for agentivity, which means that the entity that is transferred from a source to a goal could be a human, acting volitionally, or an entity, human or non-human, that is being transported.
Drawing on a light verb analysis of examples such as (14), I propose that Construction PROP has a similar structure. Thus, no deletion has taken place; the null head corresponds to the basic concept HAVE, which, just like the concept GO, can be phonologically realised by different language-specific lexemes, for example ha ‘have’, få ‘get’ and äta ‘eat’, yielding sentences like (6), repeated here as (16), which, accordingly, should be compared to (2b).
(16)
The function of infinitival marker in (16) will not be explored in this paper, but I assume that it is related to tense (which could have a plus or a minus value).
The important point so far is that what appears to be the subject of Construction PROP sentences is an object DP embedded in a clausal constituent. The verb is null, and in the cases discussed so far it corresponds to the basic concept HAVE:
(17)
The next question concerns the precise nature of the null verb.
The notion of light verb has been the focus of attention in recent literature. The concept was introduced already by Otto Jespersen, who pointed at examples consisting of have, take and give + NP, for example have a rest/a read/a cry, take a sneak/a drive/a walk/a plunge, give a sigh/a shout/a shiver/a pull/a ring (Jespersen Reference Jespersen1965 [1909–1949], vol. VI:117). The term light verb was coined by Grimshaw & Mester (Reference Grimshaw and Mester1988), who examined the Japanese verb suru ‘do’. Butt (Reference Butt1995) provides an extensive overview of ‘the light verb jungle’ in a variety of languages, and she shows that in a language like Urdu verbs like take, give, let, fall, go and hit seem to be light verbs. Other languages have similar sets of light verbs. Lundin (Reference Lundin2002:131ff.) suggests that låta ‘let’ and få ‘get’ are light verbs in Swedish, and Thurén (Reference Thurén2008:chapter 4) analyses Swedish komma ‘come’ in conjunction with participles, for instance in the construction komma gående (come walk.present.participle), as a light verb. Butt (Reference Butt2003:13) argues that light verbs are in a sense the same verbs as the corresponding main verbs. Diachronically, the two have coexisted, which means that light verbs do not simply constitute a stage in a grammaticalisation process, where a main verb gradually transforms into an auxiliary. Finally, according to Butt (Reference Butt2003:18), light verbs are passepartouts: ‘their lexical semantic specifications are so general that they can be used in a multitude of contexts, that is, they “fit” many constellations’. Butt & Lahiri (Reference Butt and Lahiri2004:36) posit ‘that a handful of verbs universally act as passepartouts. These verbs are always drawn from the lexically “simplest” part of the lexicon . . . i.e. they mainly encompass motion verbs and basic relations such as “give”, “take”, “put”, “make” and “do”’. Also have is mentioned by Butt & Lahiri as a light verb.Footnote 25
I will adopt the idea that light verbs are passepartout verbs, as suggested in Butt (Reference Butt2003) and Butt & Lahiri (Reference Butt and Lahiri2004), as well as Adger's assumption that light verbs are instances of little v (Adger Reference Adger2003:134). If this is correct, a subject in a Construction PROP sentence is at least as large as a vP. The diagram in (18) shows the first approximation of the structure of the subject in (2b).
(18)
This structure provides a position for the antecedent subject (assumed above to be an instance of arbitrary PRO) for reflexives. The propositional meaning of the subject and the absence of a violation of the V2-constraint when a VP-adverbial is added also follow straightforwardly from the proposed analysis (given the assumption that VP-adverbials are adjoined to the VP, which seems to be a reasonable assumption).
The difference between the subject in (2a), två älskare ‘two lovers’, and the corresponding infinitival paraphrase in (16) is not only the presence vs. the absence of the infinitival marker. The verb in (16) is in the infinitival form. Drawing on Chomsky (Reference Chomsky1999) I will assume that a verb in the infinitival carries tense information, more specifically it is a form marked -tense, which means that the tense is either defective or unspecified. This would imply that the subject att ha två älskare in (6)/(16) is at least as large as a TP. In Construction PROP sentences, such as (2b), we have no reason to assume that the subject phrase is a TP.Footnote 26 The main reason is that there is no place for lower adverbs, such as negation.Footnote 27
(19)
Assuming either that the NegP immediately dominates the vP or that the negation is adjoined to the vP, which should be uncontroversial, the conclusion is that the structure is in fact no larger than a vP. As pointed out in section 1, it is unsatisfactory to assume that the neuter agreement in Construction PROP sentences is default; hence we need to give a proper account for this type of agreement. Gender is a nominal feature, and the head of a vP is verbal. This makes it unappealing to assume that the gender feature will combine with the verb. To solve this problem we may tentatively assume the presence of a nominal head, F°, hosting the nominal gender feature that triggers predicative agreement in the neuter:
(20)
The structure in (20) is supported by a construction that we have reason to believe has a parallel structure – though with an overtly realised head. Assuming that the analysis of the example in (4a) above is correct, it is fully reasonable to assume that the FP in (20) is identical to the SemP in [[hon] den nya professorn], and that the feature content of F° is what triggers agreement in neuter in Construction PROP sentences. This would also allow us to account for the similarities and difference between Construction NOM and Construction PROP sentences in a more precise way: the subject of Construction NOM sentences is a SemP taking an NP complement, whereas the subject of a Construction PROP sentence is a SemP taking a vP complement. In both cases the neuter feature is hosted in Sem°. This feature triggers agreement on the predicative adjective. The meaning associated with this feature is that of the fourth gender in the four-way semantic gender system, proposed in Josefsson (Reference Josefsson2006), i.e. SUBSTANCE or UNBOUNDED ENTITY.
I will remain agnostic as to the more precise relation between the head F° in (20), the head C° in finite and non-finite embedded clauses. We may conclude, however, that the feature contents of F° and of C° in infinitival clauses are not identical, since F° cannot be lexicalised with att:
(21)
So far I have discussed Construction PROP sentences with the null light verb HAVE ‘have’, which alternatively may have the flavour ‘get’, which is a dynamic version of the stative HAVE. In addition, there seem to be cases with a null GIVE:Footnote 28
(22)
The reason why the null verb is identified as GIVE is the presence of the PP till svärmor ‘to mother-in-law’, which carries the theta role goal. Since this theta role in the unmarked case has a sibling role, theme, in ditransitives, the null verb can be identified. In a way similar to null HAVE, different verbal lexemes can be supplied, for example, ge ‘give’, överlämna ‘give’, and överräcka ‘hand over’.
The sentence in (23) shows that PERCEIVE may be used as a null verb.
(23)
Examples such as (22) and (23) raise the question of the identification of the null verb. It looks as though not only adverbials play a role. Our world knowledge (films are normally watched, not eaten, for instance) helps us to retrieve the meaning of the null light verb such as PERCEIVE or possibly even the more specific SEE. It might even be the case that the identity of the null light verb may remain underdetermined, thus ‘oscillating’ between different readings:
(24)
The sentence in (24) could mean that it is unpleasant to listen to angry custumers on the phone, to see them in the store, or to just have them around. One possibility is to assume that HAVE has a very broad meaning, including, ‘eat’, ‘see’,’ listen to’, ‘smell’, ‘hold in the hand’, ‘control’, etc. The other option is that there is a range of concepts with which the null element may be associated, and that a sentence can be undetermined as to exactly which one.
It is possible that there is as null light verb corresponding to TAKE too:Footnote 29
(25)
The example in (25) should be compared to that in (26):
(26)
The light verb discussed in Grimshaw & Mester (Reference Grimshaw and Mester1988) is the Japanese verb suru ‘do’. DO and MAKE seem to be the lightest of all light verbs, i.e. the prototypical light verbs. A null DO seems to be an option in Swedish as well:
(27)
All the initial DPs in (27), delbetalning, vattentvättning and avrättning, are typical event nouns. However, only for (27a) is a paraphrase with the verb göra ‘do’ impeccable:
(28)
However, as pointed out above, the null light verbs discussed in this paper do not necessarily correspond to actual language-specific verbs. They are, to borrow the term used by Butt & Lahiri, passepartouts: a group of verbs that encompass ‘motion verbs and basic relations such as “give”, “take”, “put”, “make” and “do”’ (Butt & Lahiri Reference Butt and Lahiri2004:36). The range of actual verbs in a language that fit into the light verb position is language-specific.
In addition to the null light verbs discussed so far at least two more options seem to be available: hålla ‘hold’ and sätta ‘put’:
(29)
The sentences in (29) should be compared to those in (30):
(30)
So far I have proposed a number of null light verbs in Swedish, HAVE, PERCEIVE, GIVE, TAKE, DO, HOLD and PUT. Following Josefsson's (Reference Josefsson1998) proposal, the verb GO should be added to this list, even though the term light verb is not mentioned in that article. I will leave the question open as to the precise membership of the list of Swedish null light verbs. The proposed analysis directly addresses Enger's (Reference Enger2004) objection to a deletion analysis of the apparent disagreement construction. The ‘missing’ verb is a light verb, which means that the number of verbs that could fit in is restricted. According to a strong version of a general theory of light verbs, this class of verbs would be the same cross-linguistically, probably because they encode basic human activities, acts, and experiences.Footnote 30
The light verb in constructions such as take a bath, take a rest, give a shout, etc. seems to have very little concrete meaning associated with take and give in examples such as take a pencil and give flowers to someone. However, Ekberg (Reference Ekberg1993) shows clearly that there is a very close link between the ‘concrete’ main verb ta ‘take’ and the more abstract ta, used as ‘a function verb’ (which I take to refer to the same category of verbs as light verbs). Thus, it should pose no problem that the null light verbs proposed in this paper retain a ‘concrete’ lexical meaning component. A similar objection could be raised for another class of verbs, namely those used in pseudocoordinations, exemplified in (31) below. Wiklund (Reference Wiklund2005:chapter 6) proposes that the first verb in pseudocoordinations, such as sitter ‘sits’ and körde ‘drove’ in (31), is a light verb.
(31)
Verb 1 in pseudocoordinations is often a motion or a posture verb, for instance sitta ‘sit’ or köra ‘drive’, as in (31). Josefsson (Reference Josefsson1991:136) showed that even though the pseudocoordination determines the aspect/aktionsart of the sentence, yielding something that looks like a progressive form in (31a) or an instigation of an event in (31b), the concrete meaning of the verb is retained. This means that Peter actually sits in (31a) and that Maria drives in (31b). Thus, in my view, light verbs can indeed have a light lexical meaning, that is, they encode basic human activities, acts, and experiences, such as doing, taking, getting, holding, perceiving, going, but they can also take the step up and become ultra-light, as in the case of Japanese suru ‘do’ (Grimshaw & Mester Reference Grimshaw and Mester1988).
2.3 Definite DPs in the subject of Construction PROP sentences
It has been noted in the literature that definite DPs are heavily constrained as subjects (or rather as DP objects within the clausal subject) in Construction PROP sentences (see e.g. Wellander Reference Wellander1949:109f., Faarlund Reference Faarlund1977, Källström Reference Källström1993:196). Faarlund (Reference Faarlund1977) even states that definite DPs are ungrammatical in Norwegian. An ungrammatical example of this type is shown in (32).
(32)
The restriction against definite DPs is not absolute, however, as witnessed by examples such as (10) and (22), repeated below, as well as (33), and (34)–(35) (the latter two are authentic, found on the Internet).
(10)
(22)
(33)
(34)
(35)
Regarding the ungrammatical example in (32), I will argue that it is not definiteness per se that is problematic, but specificity. Thus, (33)–(35) are grammatical because väskan, kvaliteten and den kommunala sponsringen are non-specific; those DPs do not refer to a specific bag, quality, or sponsorship, but to such elements in general. In the unmarked case definite DPs in Swedish have a specific interpretation; non-specific DPs are normally indefinite or bare. This is the reason why it might be difficult to construct examples such as the ones in (34) and (35). Evidence that it is specificity and not definiteness that restricts definite DPs is that an indefinite DP, such as en chokladbit ‘a piece of chocolate’ in the subject of Construction PROP sentences can only receive a non-specific interpretation; consider (36).
(36)
The restriction against specific DPs in the clausal subject of Construction PROP sentences remains to be explained, however. In my view, this restriction could be linked to the more general role of DP objects in the syntax. Arad (Reference Arad1996:219–220) shows that DP objects – more specifically specific DPs – typically play the role of delimiting an Event. They are, in other words, Event measurers. (See also Krifka Reference Krifka, Sag and Szabolcsi1992 and Tenny Reference Tenny, Sag and Szabolcsi1992.) For example, a sentence like Peter eats the apple is construed in such a way that the specific DP object, the apple, is successively consumed until it is all gone. In order for a specific DP object to be licit in a Construction PROP sentence, the predicate has to be dynamic. Since stative HAVE – the typical predicate in the construction under discussion – is not dynamic by definition, it cannot combine with a specific, definite DP.Footnote 31
Specific, definite DPs are found in (10) and (22) above. The predicates in those examples are dynamic, GIVE and HAVE/GET, and, consequently, specific (and definite) DPs that play the role of Event measurers, are grammatical. The conclusion seems to be that there are two cases where a definite DP is licit in Construction PROP sentences. The first case is stative HAVE + non-specific, but definite DP object. The second case is a dynamic predicate + a specific, definite DP.Footnote 32
The proposed analysis explains another property of Construction PROP sentences, namely that definite DPs are more likely to combine with irrealis vore ‘were’ or past tense var ‘was’, than present tense är ‘is’ in a sentence such as (10). Compare (10), repeated here again, (37) and (38):
(10)
(37)
(38)
The fact that (38) is odd has to do with the temporal interpretation of the null predicate within the clausal subject. This interpretation is in a way similar to that of a verb in infinitive. Infinitive is not a tense form per se but, drawing on Chomsky (Reference Chomsky1999), I assume that the infinitive has ‘defective’ tense. In my view, this implies that the temporal interpretation of a verb in infinitive is dependent on the temporal and modal interpretation of the matrix verb.Footnote 33 Consider (39), which illustrates this (the arrow should be read as ‘is interpreted as’).
(39)
Informally we may say that (39) shows that the temporal/modal interpretation of the matrix verb spreads to the embedded infinitival predicate. If this analysis is on the right track, the difference in acceptability between (10) and (37), on the one hand, and (38) on the other, is due to the temporal interpretation of the embedded null predicate.
Consider (40)–(42):
(40)
(41)
(42)
The reason why (42) is odd is that the present tense interpretation of the null dynamic predicate within the clausal subject + a DP with specific reference, ‘her’, makes it necessary to interpret the utterance as a comment on an ongoing event, uttered by one of the consumers (since only a consumer could evaluate the tastiness of what is being consumed). The oddity of the utterance is thus due to pragmatics. The event referred to by the clausal subject in (41) is interpreted as having taken place in the past, whereas the event described by the clausal subject in (40) is in the irrealis mood. From a pragmatic point of view, (40) and (41) are more likely to occur.Footnote 34
We may conclude that restrictions regarding the definiteness of the DP within the subject and tense/mode on the matrix verb are interrelated. A definite, non-specific DP is fine, if the null predicate is stative or, rather, if it can be construed as stative. Definite specific objects are licit only if the null predicate is dynamic or, rather, can be construed as dynamic. The tense/mood of the matrix clause is restricted by pragmatics, since matrix tense/mode interpretation ‘spreads’ to the null predicate in the clausal subject. It seems as though the possibility of construing a null predicate as stative or dynamic has to do with the identification of the predicate, which will be discussed in more detail in section 2.4.
2.4 The licensing and identification of null verbs in Construction PROP sentences
We now turn to the licensing and identification of the null verbal predicates assumed in Construction PROP sentences. Josefsson (Reference Josefsson1998:118–120) suggests that sentences like (14) above, repeated below, contain a null GO, and that this null verbal predicate has to be properly licensed and identified in the sense of Rizzi (Reference Rizzi1986).
(14)
According to Josefsson (Reference Josefsson1998), the licensing requirement in cases such as (14) is fulfilled by the auxiliary, and the identification requirement, by the PP/adverbial. The same kind of constraint seems to hold for the null verbal predicates in Construction PROP sentences. The intuition behind the proposed licensing requirement is that some element has to indicate that the structure is larger than shown by the phonological properties of the clause or phrase, i.e. that there is a slot in the structure for a null element. As pointed out above, the modal auxiliary is what indicates a position for a null main verb GO in (14a, b). In Construction PROP sentences, different types of licensers seem to be operating. First of all, the neuter agreement on the predicative adjective indicates the presence of a functional projection hosting the neuter feature. The idea is that gender is a nominal feature, which implies that the functional projection hosting this feature is nominal. In this case the neuter feature carries a meaning, namely the semantics related to the fourth semantic gender in the sense of Josefsson (Reference Josefsson2006), namely SUBSTANCE/UNBOUNDED ENTITY. The -t agreement on the predicative adjective in examples such as (2b) – and in fact also (2a) – thus both license and identify the null head of the SemP. The solid arrow in (43) indicates this relation. (Intermediate projections, as well as the CP level are omitted.)
(43)
I have suggested that the fourth semantic gender encompasses events and substances. Events are expressed by vPs and substances by NPs, hence the adjective omoralisk ‘immoral’ disambiguates the xP selected by the null head of the SemP as being a vP, not an NP. (An adjective like ‘immoral’ does not normally characterise a substance, but is more likely to describe a stative event.) We could thus assume that Sem° licenses the null verb (a relation indicated by the broken-line arrow in (43)), whereas the adjective omoralisk identifies it as a particular light verb, presumably in conjunction with the fact that the DP (två älskare ‘two lovers’ in (2b)) carries a theta role, assigned by this null light verb. The identification of the null light verb is indicated by the dotted arrow in (43).
The typical verb used in Construction PROP sentences is stative HAVE. In non-prototypical Construction PROP sentences, i.e. with other types of null predicate verbs, the identification requirements seem to be stricter. In (10), for instance, a PP like med senap och ketchup ‘with mustard and ketchup’ is obligatory, probably since the DP henne ‘her’ is not typical food, hence the null light verb is not normally identifiable as HAVE, with the special meaning ‘eat, consume’. The PP could be exchanged for a location adverbial, for instance i en sportbil ‘in a sports car’, as illustrated by (44).
(44)
The null light verb in the subject clause in (44) would presumably be identified as HAVE/PERCEIVE/SEE.
In (22) the Goal PP till svärmor ‘to mother-in-law’ identifies the null verb as GIVE, and in (33) the DP väskan ‘the bag’ in conjunction with the PP på ryggen ‘on the back’ identifies the null verb as HAVE. In this case the definite form requires a location PP in order for the non-specific reading of the definite DP väskan ‘the bag’ to be available.Footnote 35
3. CONSTRUCTION PROP SENTENCES AND THE MED-PHRASE PARAPHRASE
As noted above, Construction PROP sentences can be paraphrased by det ‘it’ + a med-phrase, i.e. a PP with the preposition med ‘with’. The example illustrating this point given in (3) above is repeated here:
(3)
As we shall see, we have reasons to believe that the med-phrase has a structure that is parallel to the vP assumed for Construction PROP sentences.
It is well known that the verb ha ‘have’ and the preposition med ‘with’ are closely related (see Benveniste Reference Benveniste1966, Kayne Reference Kayne1993). In traditional grammar a construction with med + DP + location adverbial/predicative is analysed as a non-finite clause (‘satsförkortning’ or ‘satsekvivalent’ in Swedish), see for example Teleman et al. (Reference Teleman, Hellberg and Andersson1999, part 3:697ff.). The verb substituting for the preposition med is ha ‘have’. (45a) gives an example where med takes two phrases in its complement, a DP, handen ‘the hand’, and a PP-adverbial, i bandage ‘in a bandage’. (45b) paraphrases (45a), but the med-phrase is exchanged for a full clause with the verb ha ‘have’:
(45)
From this we can gather that med + DP + PP/adverbial has some kind of clausal properties. The fact that två älskare ‘two lovers’ in (2b) and med två älskare ‘with two lovers’ in (3) have the same basic reading suggests that also simple med-phrases, i.e. med-phrases with only a single phrase as its complement, have or may have clausal properties. Another indication is that such phrases may contain a reflexive pronoun, as shown in (46).
(46)
The reflexive pronoun sin in (46) indicates that there is a subject inside the med-phrase, binding the reflexive. In view of this, it would seem quite natural to analyse Swedish ha ‘have’ as the spell-out of BE + preposition, as argued for English by Kayne (Reference Kayne1993); null HA would in a sense be BE + the preposition med ‘with’. (The reading would in that case be ‘be two lovers at/with DP’.) However, not only ha ‘have’ seems to alternate with the preposition med; this holds true for the other assumed null light verbs too, a fact that calls for a slightly different analysis than the kaynian one. Consider (47), which includes examples related to some cited earlier in the present paper.
(47)
As is shown in (48), all the examples in (47) can be paraphrased by sentences where med is replaced by att ‘to’ + a VP with the verb in the infinitival form.Footnote 36
(48)
It should be pointed out that not all the sentences in (48) are equally well formed, maybe they are not as natural as those in (47), which are all unproblematic. However, all of them are grammatical in Swedish.
Given the semantic and structural similarity between the assumed null light verbs and the preposition med ‘with’ we may hypothesise that med and the null light verbs are located in the ‘same’ position, i.e. in the head of a small lexico-functional projection that corresponds to the vP, with the important difference that the head is not v° but p° (‘little p’), in other words a pP.Footnote 37 If this is correct we arrive at the structure in (49).Footnote 38
(49)
I have proposed that a SemP can be added on top of a vP, hosting the neuter feature that triggers agreement in the neuter on the predicative adjective, as shown in (2b). This agreement is not default, but semantically motivated, since the neuter feature carries a meaning that corresponds to SUBSTANCE, or UNBOUNDED ENTITY. In a parallel fashion it is reasonable to assume that we should be able to add a SemP on top of the pP, yielding (50):
(50)
A question brought about by the proposed analysis is why an overt det ‘it’ cannot be spelled out in the SemP, yielding (51). As shown in (52), which should be compared to (4a), the pronoun hon can be spelled out in the corresponding position:
(51)
(52)
It should be pointed out that (51) is not ungrammatical as such, but det ‘it’ has a referential reading in this context, meaning ‘that’ or ‘that thing’, hence det is presumably not spelled out in Sem°. I do not have a full answer as to why det cannot be spelled out in Sem°, whereas hon ‘she’ can, but it is reasonable to assume that it is due to the spell-out convention for pronouns. Since det in (51) can be exchanged for demonstrative det där ‘that’, we may hypothesise that it is spelled out in the head of the DemP (or the functional projection hosting demonstratives). Generalising this idea, we may assume that all instances of overt det combined with a PP as a modifier (which presumably is generated in the complement of N°) are instances of demonstrative det + PP. This analysis is supported by the fact that the topicalisation of the med-phrase across an expletive det makes the weak pronoun referential:
(53)
The reading of (53) is ‘With two lovers that/that thing becomes immoral’. The example in (53) shows in fact that the PP med två älskare cannot be raised across an expletive subject det; det in (53) is not an expletive.
The idea that det in (53) is different from det in (3) is supported by the observation that det in (53) could be exchanged for the demonstrative det där ‘that’. This is the same reading that would be obtained for (3) as well, if we used det där as subject instead:
(54)
The natural way of accounting for the fact that med två älskare cannot be in sentence-initial position in (53) without the subject det receiving a referential reading is that the raising of Øneut med två älskare across subject det would induce a cross-over effect; a subject would be raised over a subject with the same referential index. In my view, the presented data suggest that med två älskare in (53) and (54) are bare adjunct PPs, i.e. PPs with no SemP on top. The t-agreement on omoraliskt in (53) and (54) is thus triggered by det/det där in the subject position in a canonical way.
A final question that needs to be addressed is why a SemP taking a vP complement with a null head is fine as a subject, but not a SemP + a pP, as witnessed by (55). (The background assumption is that a pP can indeed have a SemP on top, with a nominal head.) In other words, why is (55) ungrammatical?
(55)
It is a well-known fact that PPs cannot be subjects in Swedish.Footnote 39 The ungrammaticality of (55), as well as the ungrammaticality of PPs more generally in the subject position, is probably due to the nature of EPP on SpecIP/SpecTP. EPP is a visibility criterion, which means that the structure is sensitive to the category of the phonological head of the phrase in this position; the visible head must be nominal. Hence even though a pP may have a null nominal projection on its top, it cannot function as subject. The reason is that this neuter feature lacks overt realisation.
4. CONCLUSION AND DISCUSSION
I have proposed that the apparent disagreement between the two types of construction, called Construction NOM and Construction PROP, is not a case of disagreement – instead agreement holds. In Construction PROP sentences the subject is clausal and in Construction NOM sentences the neuter feature is a feature of the noun phrase, not a feature of the nominal head. In both Construction NOM and Construction PROP sentences the topmost projection within the subject is a SemP, hosting the neuter feature that triggers agreement in the neuter on the predicative adjective.
A grammatical gender feature may be generated low in the NP, on the lexical head, but it may also be merged above the DP, presumably in the SemP. When the neuter gender feature is merged low, for instance in the noun hus-et in Hus-et är grön-t (house.def.neut.sg is green.neut.sg) ‘The house is green’ (see (1b) above) it carries no meaning. When the neuter feature is generated high, as in Senap är gul-t ‘Mustard is yellow’ and Två älskare är omoralisk-t ‘To have two lovers is immoral’, as in (2), the neuter feature corresponds to the meaning of what in earlier work I have called the fourth semantic gender – SUBSTANCE, UNBOUNDED ENTITY (Josefsson Reference Josefsson2006). The proposed analysis thus suggests that one and the same feature may be located in different positions, and that this feature may have different meanings – or no meaning at all – depending on its location. This conclusion should come as no surprise; the same system is at work in the context of numbers notated through Arabic digits. There are ten different digits, but the value of a digit in a calculation depends on where it is located hierarchically and linearly. Thus, the digit 1 corresponds to the value ‘one’ in isolation, but to the value ‘ten’ if it appears in the second position to the left, etc.
The second conclusion concerns the nature of the subject in Construction PROP sentences. I have argued that the subject of this type of clauses is a SemP, taking a vP as its complement. The subject within the subject phrase is phonologically null, presumably an instance of generic PRO. The main reason for assuming that the subject is clausal in Construction PROP sentences, in turn containing an embedded null subject, is the fact that reflexives are fine. This means that the overt DP is in fact an embedded object. The head position, v, is filled by a null verb, which I have identified as a light verb. In the typical case this light verb is HAVE, but it could also be construed as GET, GIVE, PERCEIVE, TAKE, HOLD, and PUT. The verbs in question are almost exactly identical to the set of light verbs, from point of meaning analysed as passepartout verbs, listed in Butt & Lahiri (Reference Butt and Lahiri2004:36). My proposal is that light verbs can indeed be null in Swedish, provided they are properly licensed and identified. Whether this suggestion holds for other languages and other types of constructions remains to be investigated.
I have argued that the assumed restriction against definite DPs in Construction PROP sentences is in fact a restriction against specific DP objects. This restriction holds for cases where the predicate is a stative HAVE, which cannot combine with specific DP objects. Specific DP objects can combine only with dynamic null predicates. I have proposed a system where the null elements are licensed and identified in the sense of Rizzi (Reference Rizzi1986).
In the last section I proposed that the preposition med ‘with’ typically corresponds to the light verb HAVE, though devoid of the verbal features hosted in v. By being prepositional it cannot head the projection occupying the subject position. This is the reason why an expletive det will have to be present in order to satisfy the EPP feature on the subject position. The ‘expletive’ det is chain-related to the SemP, which explains why the ‘with-phrase’ cannot raise across it.
ACKNOWLEDGEMENTS
Previous versions of this paper have been presented at the Grammar Seminar at Centre for Languages and Literature at Lund University, and at Grammar in Focus, 2006, Lund University. I want to thank participants at those events for comments and suggestions for improvement. I also would like to thank the anonymous reviewers for their thorough work and helpful comments, all of which has helped to improve the paper. I am responsible for all remaining errors and inadequacies.
