1 Introduction

Since Russell (1919), it is commonplace to distinguish between a meaningful identity-copula (1a), and an essentially semantically vacuous copula of predication (1b).Footnote 1

figure a

In his seminal work, Higgins (1973/1979) distinguished between four types of copular sentences (2), all of which may be realised by the surface string ‘DP1 be DP2’.Footnote 2

figure b

This paper investigates the syntactic and semantic properties associated with the alternation between Nominative and Accusative case on the postcopular DP in Swedish copular clauses (3). The alternation appears to be a relatively new phenomenon in Swedish, which has traditionally allowed only Nominative case on postcopular DPs (Maling and Sprouse 1995; Sigurðsson 2006).Footnote 3

figure c

The investigation has an essentially comparative focus. In Sect. 2, I compare Swedish predicational and equative copular clauses with data from a range of languages (primarily Polish, but also Russian, Dutch, English and Danish). This comparison is motivated by the observation that, despite appearing to have quite different systems, the Swedish alternation not only shares a number of distributional properties with these other languages, but the distribution of the two variants is also shown to track the same interpretive dimension.Footnote 4

Based on this data, I argue that the case alternation tracks an interpretive distinction, mediated by two distinct LFs, and is not purely a PF-phenomenon, contra Sigurðsson (2006), who hypothesises that Swedish is undergoing a change at PF, towards a system like that in English and Danish, where all postcopular DPs are marked with Accusative case, regardless of their interpretation. More specifically, the data indicates that the Swedish alternation is conditioned by the same semantic factors that condition the Nominative/Instrumental alternation in Polish (and to some extent also Russian and Dutch): sentences with Nominative case on DP2 are interpreted as equatives or as involving individual level predication, whereas sentences with Accusative case (Instrumental in Polish) are interpreted as involving stage level predication. This contrast is well-established in the Slavic literature (e.g., Matushansky 2000, 2008, 2010; Citko 2008), but has not been previously discussed in the context of Swedish. Furthermore, I show that the Swedish alternation also displays the same syntactic properties as the Polish alternation: as is again well known in the context of Slavic, Nominative postcopular DPs cannot undergo A-bar extraction (e.g., Matushansky 2000; Citko 2008). However, what has not been previously noted, for neither Swedish nor Polish, is that Nominative postcopular DPs also show ‘non-local’ properties with respect to binding.

Having shown that predicational and equative copular sentences in Swedish behave semantically and syntactically like those in Polish, I turn to previous analytical claims made in the context of the Slavic alternation, and identify some empirical issues inherent to these. As an alternative, I propose an analysis of the alternation—based primarily on the Swedish and Polish data, whereby the morphological, syntactic, and semantic contrasts are all mediated by the same underlying structural difference, proposing a biclausal analysis for Nominative DPs and a monoclausal analysis for Accusative DPs. In particular, I argue that that the postcopular DP in these sentences receive Nominative case via Agree with \(\hbox {T}^{o}\) inside a silent free relative clause, whereas Accusative case is assigned by a semantically inert Pred\(^{o}\) head to its sister DP; a predicate of type <et>. In Polish, this predicational case is realized as Instrumental. The presence or absence of this tensed clausal layer accounts for both the morpho-syntactic and the semantic properties associated with the alternation. Section 3 provides an interim summary.

In Sect. 4, I turn to specificational copular clauses, showing that also in this case does Swedish pattern with Polish, along with Dutch and Russian, but unlike Danish and English. In the former group of languages, the postcopular DP (DP2) is the subject, whereas in the latter group the initial DP (DP1) is the subject.Footnote 5 I argue that a particular type of ‘predicate inversion’ analysis is needed to account for the Swedish type of specification, such that specification is derived from the same types of structures underlying predicational and equative sentences. However, I show that this cannot simply be topicalization of the lower DP. I end my discussion with a note on some of the empirical challenges facing accounts attempting to provide a unified account of predicational, equative and specificational copular sentences in languages of the Swedish type.

2 Predication and equation

2.1 Basic properties

The basic characteristic of predicational copular sentences, as in (2a), is that they attribute some property, provided by the postcopular element, to the subject DP. In this respect, they are like non-copular sentences, with the difference that the property is provided exclusively by the post-verbal element (Higgins’ 1973/1979 ‘predicate complement’), whereas the copula itself is essentially meaningless. The semantics of predicational copular sentences can thus be characterised in terms of Partee’s (1986) family of DP types: the subject (DP1) is referential, of type e, and the complement (DP2) is predicational or property-denoting, a function of type <et>. This is then different from equative copular sentences, as in (2b), where both DPs are referential, of type e. For this type of sentence, the basic intuition is that it equates the referents of the two arguments of the copula. Evidence for this type ambiguity of the postcopular DP comes from the fact that it can pronominalize either with the pronoun ‘who’ (4a), which must range over animate referents, or with the pronoun ‘what’ (4b), which must range over inanimate referents (Partee 1986).

figure d

The obligatorily referential status of DP1 can be demonstrated by the fact that it only pronominalises like an individual, as shown in (5).

figure e

In Sect. 2.2.1, I examine in detail the semantic properties of these types of copular clauses in Swedish, comparing them to their counterparts in Polish, Dutch, Russian, Danish, and English. I demonstrate that the Swedish case-alternation tracks an interpretive distinction, such that Accusative postcopular DPs are interpreted as stage level predicates, and Nominative DPs as individual level predicates, or as involving equation of the two DPs. In Sect. 2.2.2, I turn to the syntactic properties associated with the case-alternation.

2.2 Data

2.2.1 Interpretation

At a first glance, Swedish copular sentences look quite different from their Polish and Russian counterparts. First, unlike in these languages, Swedish has only one copular element, the verbal copula ‘vara’ (be), shown in (6).

figure f

Polish and Russian, on the other hand, have two types of copular elements,Footnote 6 the pronominal copula ‘to’ (Polish)/‘éto’ (Russian), which is always uninflected, and the verbal copula ‘być’ (Polish)/‘byt\('\)’ (Russian).

The choice of case on the postcopular element correlates with the type of copula. In Polish pronominal copular sentences (7a), only Nominative is possible on DP2 (note that here, the verbal copula jest is entirely optional, in the sense that its presence has no interpretive effect). When there is only the verbal copula (7b), however, only Instrumental is possible on DP2.

figure g

The distribution of the pronominal copula in Russian is more complex than in Polish. For the purpose of the comparison with Swedish, I focus on predicational sentences where the verbal copula is non-zero—i.e., in the past or future tense. In these contexts, both Nominative and Instrumental are possible on DP2, as shown in (8).Footnote 7

figure h

Swedish, on the other hand, is traditionally taken to allow only Nominative case on post-couplar DPs. However, as noted by Sigurðsson , a subtype of Swedish couplar sentences—where “the predicative DP takes on the role of the subject DP, rather than its identity” (Sigurðsson 2006, 34) allows both Nominative and Accusative case, as shown in (9).Footnote 8 As (9) shows, there is variation among speakers as to whether Accusative is accepted here.

figure i

As noted by Maling and Sprouse (1995), the presence of Accusative case on postcopular DPs appears to be a rather recent development in Swedish.Footnote 9\(^{,}\)Footnote 10 Sigurðsson (2006), who discusses this alternation in the context of Germanic, argues that it can be understood in terms of a split between those languages that assign Accusative to all ‘dependent arguments’ (Predicate Nominative languages), and those that assign Accusative to all ‘dependent DPs’ (Predicate Accusative languages). Swedish, he claims, is changing from a Predicate Nominative grammar (like Icelandic, which allows only Nominative on postcopular DPs), to a Predicate Accusative grammar (like English and Danish, which allow only Accusative on postcopular DPs). He posits a similar trajectory for Dutch, although notes that the two languages differ somewhat with respect to the contexts where Accusative is possible (see discussion in footnote 11).

However, if this was an across-the-board morpho-syntactic change, as Sigurðsson (2006) suggests, we would expect that all postcopular DPs in Swedish should be eligible for Accusative case, regardless of their interpretation. However, this is not what we find. In Swedish, only Nominative is possible in equative sentences.

figure j

In this respect then, Swedish is similar to Polish, Russian, and Dutch, which all allow Accusative case on the postcopular element in predicational contexts, but allow only Nominative on the postcopular element in equatives:

figure k
figure l

This is different from Danish and English, which do behave like Sigurðsson’s Predicate Accusative languages. Here, all copular sentences mark DP2 with Accusative, regardless of their interpretation: both the equatives in (14a) and (15a), and the predicational sentences in (14b) and (15b), require Accusative.

figure m

This is unexpected on Sigurðsson’s account, whereby the case-alternation is a pure PF-phenomenon, with no semantic import. Even more problematically, it turns out that for speakers who allow both Nominative and Accusative, within the contexts where either appears to be possible, it’s not a matter of free variation or semantically unconditioned optionality. Instead, we find that the choice of case tracks interpretation: if the DP is interpreted as temporally bounded, contingent, or eventive, Accusative is preferred (16a), whereas if the DP is interpreted as invariant, intrinsic, or inherent, Nominative is preferred (16b).

figure n

This contrast recalls that between ‘stage level’ predication and ‘individual level’ predication (Carlson 1977), which has in fact been argued to be what drives the choice of Instrumental vs. Nominative case on predicate DPs in Polish and Russian.Footnote 11 As shown in (17) from Polish, Nominative case on DP2 is infelicitous with intrinsically stage level predicates. Here, Instrumental case on DP2 is the only option. (See also Matushansky 2000, 2008, 2010; Babyonyshev and Matushansky 2006, and Geist 2007, among others on Russian.)

figure o

Further evidence that this is the relevant distinction comes from Citko (2008), who shows that the use of (typically) stage level predicates with the pronominal copula and Nominative case on DP2 improves, if coerced into an individual level predicate (18a). She also points out that Nominative gives rise to—so called—‘lifetime effects’, which are associated with individual level predication (18b).

figure p

Of course, since in Swedish only pronouns have morphological case, testing this hypothesis with a wider range of constructions and contexts is difficult. However, it does predict that the use of Accusative case should be limited, given that pronouns, like proper names, are typically non-eventive, and thus generally excluded from functioning as stage level predicates. In Polish, these elements typically receive Nominative case. However, as shown in (19a), proper names can be coerced to receive eventive readings, given appropriate context. An example of such a reading of a proper name is given in (19).Footnote 12 Here, Instrumental case is the preferred choice.

figure q

Similarly, the only contexts that I am aware of, in which pronouns can be felicitously interpreted with respect to some eventuality, are precisely the type of ‘role’ or ‘assumed identity’ contexts discussed in the context of example (9) above, which are ambiguous between contingent, or temporally bounded readings, as in (16a), and invariant, inherent, or intrinsic readings, as in (16b). And indeed, when presented with the types of disambiguating scenarios illustrated in (16), those of my consultants who accept both Nominative and Accusative in the sentences in (9) now show a clear preference for Accusative case in the former context, and for Nominative in the latter.Footnote 13

To summarise our discussion of the interpretive properties associated with the Nominative-Accusative alternation in Swedish, we have observed that (contra Sigurðsson 2006) it does not pattern with English and Danish, but rather, seems to be tracking a semantic distinction, whereby Accusative is preferred when the predication is interpreted as temporally bounded, contingent, or eventive (stage level predication), whereas Nominative is preferred if the predication is interpreted as invariant, intrinsic, or inherent (equation or individual level predication). What appears to be free variation in ‘role’ or ‘assumed identity’ sentences is argued to be due to their ambiguity in this regard. Next, we turn to the syntactic properties associated with the alternation.

2.2.2 Structure

In Polish, extraction of Instrumental, but not Nominative, postcopular DP is possible, as shown in (20), adapted from Citko (2008).

figure s

The same also holds of Russian. As Matushansky (2000) shows, only Instrumental-marked predicates can undergo A-bar extraction.Footnote 14\(^{,}\)Footnote 15

figure t

Interestingly, we find that also in Swedish is it possible to topicalize Accusative, but not Nominative postcopular DPs, as illustrated in (22).

figure u

Citko (2008) in fact claims that in copular clauses of the form ‘DP1.NOM be DP2.NOM’, neither DP can move. She uses this as an argument to motivate an analysis of the Polish alternation as involving different functional heads, which she labels \(\pi \).Footnote 16 In verbal copular clauses, she assumes that \(\pi \) assigns Instrumental case to its complement, which behaves in all respects like the direct object of a transitive verb. For pronominal copular sentences, on the other hand, she assumes that these are headed by a ‘defective’ \(\pi \)-head, similar to the head of a coordination structure, with “no independent syntactic features of their own, [such that, KD] their only role is to mediate a relationship between a specifier and a complement” (Citko 2008, 288). Along with Matushansky (2008, 2010) and Bondaruk (2014), Citko (2008) assumes that the way in which DP2 gets Nominative case in these structures, is through some form of Multiple Agree with \(\hbox {T}^{o}\) (as in Hiraiwa 2002). Similarly to the head of a coordinate phrase, she takes this defective \(\pi \)-head to block extraction of both of its arguments. The example in (23) is intended to illustrate this point.

figure v

The sentences in (23) are problematic, however, as they fail to control for two confounds. First, it is well-attested that long-distance extraction from embedded indicative clauses, such as (23b), is generally degraded (e.g., Progovac 1993; Giorgi and Pianesi 1997; Antonenko 2010). However, Citko’s (2008) examples of extraction from a clause with an Instrumental DP2 all use subjunctive complements, as in (23a), whereas her examples of extraction from a clause with a Nominative DP2 all use indicative complements, as in (23b). Second, (23b) sounds odd also in English, which has no ban on DP1 movement in copular clauses. This is likely due to discourse factors (compare with ‘Which of these subjects do you think is physics?’).Footnote 17 Controlling for these two factors, however, we find that extraction of Nominative, but not Instrumental DP2, leads to ungrammaticality (24a,b)—but that extraction of DP1 in sentences with a Nominative DP2 is indeed grammatical (24c).

figure w

Thus, Citko’s analysis places too strong of a ban on movement, in that it predicts that neither DP should be able to move. An even bigger problem for this analysis is that it also predicts that A-movement of DP1 to Spec,TP should be impossible in these sentences, given that extraction from coordinate phrases is bad both with respect to A and A-bar movement, as shown in (25) (and importantly, does not improve with d-linking).Footnote 18

figure x

However, a ban on A-movement is clearly incompatible with the structure of Polish pronominal copular sentences, given that DP1 appears above all sentence adverbs, modals and negation, as well as above the pronominal copula to (which Citko 2008 takes to be merged in \(\hbox {T}^{o}\)).

figure y

The same is clearly true for Swedish as well, as shown in (27).

figure z

Hence, we may conclude that in both languages, it is specifically Nominative postcopular DPs that are prohibited from A-bar extraction. Next, I present a new observation, namely that Nominative DP2 also differ from Accusative/Instrumental DP2, in both Swedish and Polish, with respect to binding.

As shown in (28), Polish has two third person singular possessive pronouns: reflexive ‘swoja’ (fem)/‘swoim’ (masc), used with local antecedents (28a), and ‘jej’ (fem)/‘jego’ (masc), used elsewhere (28b).

figure aa

In verbal copular sentences with an Instrumental DP2, the expected reflexive possessive (‘swoim’) is used when DP1 is the antecedent, as in (29a). In pronominal copular sentences with a Nominative DP2, however, only the non-local pronoun (‘jego’) is grammatical—even though DP1 is the antecedent, as shown in (29b).

figure ab

Similarly, we find that in Swedish, regular reflexives, which also require a local antecedent, are obligatorily marked with Accusative case, even in those cases where Nominative case would otherwise be available, as shown in (30).Footnote 19

figure af

To summarise this section, we observe for Swedish, as in Polish, that Accusative/Instrumental postcopular DPs behave as though there were ordinary direct objects of the copula, both with respect to movement and binding. Nominative postcopular DP, however, behave with respect to both movement and binding, like we expect a DP to behave when it is embedded inside certain types of embedded clauses, such as factive or relative clauses that are islands for movement. This insight will become central in our analysis, which we turn to next.

2.3 Towards an analysis (part I)

Above, I discussed Citko’s coordination account, and saw that it places too strong of a ban on movement. Before moving on to the analytical alternative proposed here, let us first consider a different option, proposed by Matushansky (2010) to account for the Nominative-Instrumental alternation in Russian.

Like Citko (2008), Matushansky also argues that Nominative case is assigned through Multiple Agree with matrix \(\hbox {T}^{o}\), in the absence of a case-assigning head, such as Bowers’ (1993) \(\hbox {Pred}^{o}\), or Citko’s (2008) non-defective \(\pi ^{o}\). She follows Bailyn and Rubin (1991); Bailyn and Citko (1999); Pereltsvaig (2001), and Bailyn (2001, 2002), in analysing Instrumental case as being assigned by \(\hbox {Pred}^{o}\). However, she points out the following issue with analyses based on the absence or presence of PredP: if the semantic function of \(\hbox {Pred}^{o}\) is to convert its complement—a semantically saturated entity (of type e), into a predicate or property of type et, and the syntactic function of \(\hbox {Pred}^{o}\) is to assign predicate case (Instrumental in Russian and Polish, Accusative in Swedish), then it is not clear how to account for the presence of Nominative case on the postcopular DP in clearly predicational contexts, such as (31) (a point also raised in footnote 11 above).Footnote 20

figure ag

Matushansky’s solution is to abandon a standard Agree-based analysis of case-assignment, and to adopt a view (argued for in more detail and on a more global level in Matushansky 2008), whereby case-assignment involves the copying of all the formal features of a head onto its sister domain—only some of which get spelled out at PF. Given the semantic function that she assumes \(\hbox {Pred}^{o}\) to fill, she takes it to be required also with Nominative DP2. In sentences with a Nominative DP2, however, she takes PredP to be selected directly by \(\hbox {T}^{o}\), whereas in copular sentences with Instrumental DP2, she takes PredP to be selected by an eventive vP (in turn, selected by \(\hbox {T}^{o}\)), which introduces the verb’s eventuality argument, and which assigns ‘Eventive’ case to its complement. For Matushansky (2010) then, it is the combination of the case features from \(\hbox {T}^{o}\), \(\textit{v}^{o}_{evt}\) and \(\hbox {Pred}^{o}\), that get spelled out as Instrumental, whereas the case features from \(\hbox {T}^{o}\) and \(\hbox {Pred}^{o}\) alone get spelled out as Nominative. An appealing aspect of this analysis is that it straightforwardly captures the correlation between the case on DP2 and its interpretation (eventive vs. non-eventive). However, as noted by Sigurðsson (2006) for a number of languages, including Swedish, Nominative case on DP2 does not depend on the presence of finite \(\hbox {T}^{o}\), as shown by the ECM-sentences in (32).

figure ah

Second, it is not clear how Matushansky’s analysis, if applied to Swedish and Polish, would capture without further stipulation the syntactic differences between Nominative and Instrumental/Accusative copular sentences discussed in Sect. 2.2.2 above, repeated in Table 1. That is, why would the fact that PredP is selected directly by \(\hbox {T}^{o}\) impose these types of restrictions on the movement and binding possibilities for the postcopular DP?

Table 1 Summary of properties of Nominative and Instrumental/Accusative DP2

A final point to make in this section is that it is not simply the case that Swedish is currently undergoing a change in default case from Nominative to Accusative (as pointed out also in Maling and Sprouse 1995). The default case is stably Nominative, as illustrated in (33) (compare this with the translation showing that in English the default is Accusative).

figure ai

Hence, appealing to an across-the-board change in default case will not work as an explanation for the availability of Accusative case on postcopular DPs. And again, it is not clear how such an explanation would capture any of the semantic and syntactic facts associated with the alternation.

Here, I want to consider a different analytical route, according to which the syntactic and semantic contrasts associated with the morphological case-alternation (Type I and Type II, in Table 1) are all mediated by the same structural difference. In particular, I suggest that the source of Nominative case on DP2 is a \(\hbox {T}^{o}\) head embedded inside a silent free relative clause, whereas ‘predicate case’ (Accusative or Instrumental, depending on the languageFootnote 21) is a non-structural (inherent) case, assigned by a predicational head to license its sister DP.Footnote 22 This idea, illustrated in (34) (to be revised), has the advantage of straightforwardly capturing not just the morphological, but also the syntactic facts associated with the case-alternation—something which we have seen remains a challenge for previous accounts.

figure aj

More precisely, the idea is that DP2 in a Type I copular clause is the subject of a free relative clause (such as ‘I can help who’s next in line’), as illustrated in (35), and as such, gets assigned Nominative case by this lower \(\hbox {T}^{o}\) head.

figure ak

This has the syntactic consequence that a Type I DP2 is not local to its DP1, as it is merged inside the embedded relative clause. This is unlike DP2 in a Type II copular clause, which is local to its DP1 (the sister of \(\hbox {Pred}^{o}\), essentially a direct object).Footnote 23 This contrast in locality then accounts for the binding facts discussed above. Regarding the movement facts, the hypothesis advanced here is that by being merged inside a free relative, DP2 in a Type I sentence is no longer available for extraction to the matrix clause, given that relative clauses are islands for extraction. This is unlike DP2 in a Type II clause, which is not subject to this kind of restriction.

Let us consider now the semantic consequences of this move. First, consider the role of \(\hbox {Pred}^{o}\). One option would be to take it to be present in both structures—as in (34). We then have to determine both what syntactic and semantic function this head fills. In the literature following Jacobson (1995), Caponigro (2003), and others, DP-like free relatives are typically analysed as CPs denoting maximal (atomic or plural) individuals of type e, such that maximalization applies directly to the set of individuals denoted by the TP, as shown schematically in (36).Footnote 24\(^{,}\)Footnote 25 In (36), I follow Caponigro (2003), who uses the \(\delta \)-operator (originally from Link’s 1983 work on definite DPs) to apply to the set of entities denoted by the CP, returning its maximal entity.

figure am

Hence, if the same type of predicational head is indeed present in both types of copular sentences, it should be of type <e,et>, as in (34). One possible denotation for this head would then be Partee’s (1986) type-shifter ident, which takes a type e argument and shifts it into an identity predicate, which is thus able to combine directly with the type e subject.

figure an

Note, however, that this move would force us to also analyse DPs like ‘a teacher’, ‘the best dancer’, etc. as individuals, rather than predicates, and to analyse sentences of the form ‘DP1 be DP2.ACC’ (e.g., (7b) and (8)) as identity predicates. This, however, would seem like an undesirable move, given that these DPs are clearly interpreted as stage level or eventive predicates.

Another, more promising, option would be to say that ident is not the meaning of \(\hbox {Pred}^{o}\), but rather an operator that applies directly to the free relative itself, similarly to Partee’s (1986) original proposal. This then would no longer require us to analyse Accusative/Instrumental DPs as individuals—a welcome consequence. We can simply analyse these as predicates of type <et>, and assume that \(\hbox {Pred}^{o}\) itself is semantically vacuous (of type <et,et>), such that its only role is to license its predicate DP complement in the syntax. CPs, on the other hand, do not need case to be licensed, so \(\hbox {Pred}^{o}\) would never enter the derivation in a sentence involving a Nominative DP2 (SC stands for Small Clause). This option is illustrated in (38) (final version).

figure ao

This then is similar, at least in spirit, to the proposals in Williams (1983) and Partee (1985, 1986), according to which there is only one type of copula or small clause head—a predicational head of type <et,et> (as in (1b)), and equation is accomplished using Partee’s type-shifter ident.

Analysing DPs like ‘teacher’, ‘the best dancer’, or ‘a nice guy’ as predicates is not unprecedented (e.g., Heggie 1988; Williams 1994; Moro 1997; cf. Higginbotham 1985, 1987). However, this analysis raises the question of how to handle cases where an Accusative or Instrumental DP2 is a proper name or a pronoun, as these elements are not typically analysed as predicates. Here, I note that there is independent evidence indicating that there must be an accessible property (type <et>) denotation at a sub-DP level, which does not depend on \(\hbox {Pred}^{o}\). This is illustrated by the availability of adjectival modification inside the DP, as in (19) above (‘He’s a real Einstein today.’), where there must be an available property meaning for the adjective to combine with.Footnote 26 Further support for this claim comes from the fact that these kinds of uses of names are possible also in non-predicational contexts, such as ‘This looks like it was written by a real Einstein’, or ‘Every young Mozart dreams of his first solo performance’. Here, I don’t take a strong position on how these readings are derived; one alternative is to posit a pred type-shifter, that applies directly to the name at sub-DP level, along the lines of (39) (though see also Saab and Lo Guercio 2020 and references therein for alternatives):

figure ap

This assumption, which is central to the current analysis, has the consequence that there is nothing special about predicational uses of DPs in copular constructions.Footnote 27 The important point is that this observation offers independent support that Accusative/Instrumental marked pronouns and proper names have available predicate readings themselves, and are therefore able to combine with \(\hbox {Pred}^{o}\) in the same fashion as other DP predicates.

On the morpho-syntactic side, we saw the T\(^{o}\)-head inside the silent free relative clause gives us Nominative case. However, as shown in (36), it also gives us a contrast in interpretation. (40)-(41) represent the consequences of combining the analysis in (38) with that in (36) for Nominative DP2, and with that in (39) for Accusative DP2.Footnote 28

figure aq

That is, in Type II copular sentences, both DPs are bound by the matrix tense, as shown in (41), whereas in Type I sentences, the lower DP is bound by a separate, embedded, tense. This has the effect of fixing the referent of the Nominative DP2 at a particular, invariant, time; and then stating that the referent of DP1 is identical to that specific individual. For the Accusative DP2, on the other hand, the sentence simply ends up stating that whatever (relevant) property holds of that individual at the matrix time, that property also holds of the referent of DP1. I propose that this is what gives us the contrast in temporal or eventive (in)flexibility associated with the case-alternation, as discussed in Sect. 2.2.1 above.

Note, however, that on the current analysis as it presently stands, there is no way to distinguish between individual level predication and equation in structural or lexical terms. While this may seem like a potentially undesirable consequence, I submit that this is in fact not so. First, what I want to suggest is that this ambiguity of Nominative postcopular DPs may in fact fall out as a natural consequence from the familiar contrast between ‘singular definite’ and ‘universal’ readings of free relatives—a difference in intensionality (Jacobson 1995). The nature of this contrast is illustrated with the regular free relative in (42).

figure ar

For the denotation posited for Nominative DP2 (36), this would result in the following type of ambiguity, which seems to me appropriate for this contrast, though I leave it for future work to evaluate this hypothesis.

figure as

Clearly a typology which takes individual and stage level predication to instantiate one grammatical type, distinct from equation, would fail to capture the fact that equation and individual level predicates pattern together morphologically and syntactically, at the exclusion of stage level predicates (see the summary in Table 1). An alternative would of course be to posit three distinct grammatical types: individual level predication, stage level predication, and equation. On this kind of analysis, however, the morphological and syntactic facts about equation and individual level predication would have to be accounted for independently for each sentence type. The current analysis captures not just these different facts, but also their coincidence. It also has two further advantages. Apart from delivering a simpler taxonomy and case correspondence,Footnote 29 it also predicts that the case-alternation in a given language should correlate with the other grammatical and interpretive properties enumerated in Table 1, as they follow from the same underlying structures.

Note, finally, that this analysis also correctly predicts the absence of a case-alternation along the stage vs. individual level dimension for AP and PP predicates. Given that these too are of type <et>, and therefore compatible with \(\hbox {Pred}^{o}\), we might at first sight expect that they should. However, given that these (like CPs) do not need to be externally licensed, \(\hbox {Pred}^{o}\) never enters the derivation with these predicates (and of course, there is no source of Nominative in this case).

3 Interim summary

The first point of Sect. 2 was to show that the Swedish case-alternation, given the associated interpretive and distributional facts, cannot be purely a PF phenomenon (contra Sigurðsson 2006). We then looked at comparative data, primarily from Polish, but also from English, Danish, Dutch and Russian, showing that Swedish predicational and equative copular clauses pattern with their counterparts in Polish (and to some extent apparently also Russian and Dutch). This data pointed to a typological generalisation regarding the way in which syntax mediates interpretation, whereby Swedish copular sentences follow the same strategy as in Polish, Russian and Dutch, unlike English and Danish. Finally, I proposed a theoretical account of the former type of strategy, whereby the alternation, and its syntactic and semantic correlates, are mediated by a structural difference.

In Sect. 4, I show that the typological split identified here runs deeper. Adding to the work of Mikkelsen (2002, 2004a, 2004b, 2005, 2011); Geist (2007); Partee (2010); Heycock (2012), and others, I show that also specificational copular sentences in Swedish pattern with the former set of languages, and not with Danish and English—thus making the typological generalisation a more broad-reaching one about how syntax, semantics, and information structure mediate interpretation.

4 Specification

In the first part of this section (Sect. 4.1), I review the basic linguistic properties of specificational copular sentences. Next, in Sect. 4.2, I give comparative data showing that, as we saw for predication and equation, Swedish specificational sentences behave like specificational sentences in Polish, Dutch and Russian, but unlike their counterparts in Danish and English, in terms of their morpho-syntactic properties. In Sect. 4.3, I argue that a particular type of predicate inversion analysis is needed to account for the Swedish type of pattern, whereby specification is derived from the same types of structures underlying predication and equation. However, this cannot simply be topicalization of the lower DP. I finish this section with a proposal for the derivation of specificational sentences in Swedish.

4.1 Basic properties

Specificational sentences are characterised in part by their particular semantic properties, such that DP1 (typically a definite descriptionFootnote 30) denotes a property containing a variable slot, which gets valued by the individual denoted by the postcopular DP (Partee 2010; Mikkelsen 2011). Evidence for the non-referential status of DP1 comes for instance from the fact that it pronominalises like a predicate, as shown in (44).Footnote 31

figure at

In this respect then, specificational copular sentences can be viewed as the inverse of predicational sentences, in that DP1 is of type <et>, and DP2 of type e (see Heggie 1988; Moro 1997; Partee 1998; Mikkelsen 2005; den Dikken 2006, and Geist 2007).

However, this characterisation is not uncontroversial. As discussed originally by Heycock and Kroch (1999), the predicate must be able to be interpreted as being equated with the subject—that is, the two DPs have to be of the same semantic type, as shown by the contrast between (45a), involving two predicates, and (45a), involving a predicate and an argument.

figure au

This parallelism requirement appears to be a fairly robust generalisation, and is precisely the kind of observation that has motivated the analysis—originally due to Heycock and Kroch (1999), and pursued more recently by Heycock (2012), that specificational sentences are in fact a subtype of equatives (contra Heycock 1991; Moro 1997; Mikkelsen 2002, 2005; den Dikken 2006). See also Rothstein (2001); Heycock and Kroch (2002); Mikkelsen (2004b); Romero (2005); Comorovski (2007); and Heycock (2012) for arguments in favour of this characterisation. As shown in (46), this semantic parallelism requirement holds in Swedish as well.Footnote 32

figure aw

Another property characteristic of specificational sentences is their particular information structure. As noted by several authors (e.g., Mikkelsen 2004a, b; Partee 2010; Heycock 2012), DP1 in these sentences must be a topic, whereas the postcopular DP is obligatorily focused, as shown by the sharp contrast between (47a) and (47b) (small caps are used to indicate focus).

figure ax

Again, this is true of specification in Swedish as well.Footnote 33

figure ay

4.2 Two types of languages

As discussed by Heycock (2012), some languages that allow Accusative case on postcopular DPs in predicational sentences, such as Dutch, nevertheless allow only Nominative case on DP2 in specificational sentences, as illustrated with the contrast between the specificational sentence in (50a), and the predicational sentence in (50b), repeated from (11b) above.

figure az

The same is also true of Swedish, Polish and Russian: first, contrast the Swedish specificational sentences in (51) with the predicational ones in (52) (repeated from (9)), where both Nominative and Accusative are available.

figure ba

Unlike the predicational sentences we looked at in the previous section, specificational sentences in Russian and Polish also allow only Nominative case on DP2, as shown in (53) and (54).Footnote 34 The DP1 predicate however, can be either Nominative or Accusative (I return to this point shortly).

figure bb

Furthermore, verbal agreement in Dutch, Polish and Russian is with the Nominative DP2 in specificational sentences, and with the Nominative DP1 in predicational sentences. This is important as agreement provides an independent test for which of the two DPs is the underlying subject, in the sense of standing in an Agree relation with the \(\hbox {T}^{o}\) head of the clause (e.g., Chomsky 2000, 2001). I will use this property to point to a typological split between languages where the Nominative DP1 in specificational sentences is the subject, e.g., Danish and English (see also Mikkelsen 2004b and Partee 2010) and languages where the Nominative DP2 in specificational sentences is the subject (and DP1 is an inverted predicate), e.g., Polish, Russian, Swedish, and Dutch (see also Moro 1997, 28 for evidence that Italian is of the latter type).Footnote 35 In predicational and equative sentences on the other hand, DP1 is the subject in both types of languages.

This is demonstrated for Dutch, with the contrast between the specificational sentence in (55a) and the predicational one in (55b). The sentences in (56)–(57) further show that the same is true in Russian and Polish: agreement in specificational sentences is with the obligatorily Nominative DP2.

figure bc

Footnote 36

Taken together with the agreement facts, the fact that DP1 in these sentences allow both Instrumental and Nominative case provides strong indication that DP1 is a predicate, whereas DP2 is the subject.Footnote 37 Moreover, in (56a), the natural choice of case on DP1 is Instrumental, whereas in (56b), the natural choice is Nominative. This is what we would expect if DP1 in specification is underlyingly the same as DP2 in predicational sentences at some stage in the derivation, given the semantic conditions on the case-alternation (discussed in Sect. 2).

Predicational sentences, on the other hand, which allow both Instrumental and Nominative case on DP2, but only Nominative on DP1, obligatorily show DP1 agreement, as shown in (58)–(59) (contra claims by Bondaruk 2014 and Geist 2007).

figure bd
figure be

Note that Citko (2008) argues, from the example in (60) that—unlike in Russian (see Geist 2007, Partee 2010)—specificational sentences in Polish can only be formed using pronominal copular sentences with Nominative case on DP1, given that (60) can only be interpreted as ‘my best friend is impersonating, or trying to be Jan’, which is not a specificational interpretation.Footnote 38

figure bf

Of course, this is precisely the kind of ‘assumed identity’ reading discussed in Sect. 2.2.1. That is, given that Instrumental case on DP2 is associated with a property reading, the only interpretation consistent with Instrumental case on proper names and pronouns, are precisely assumed identity readings.Footnote 39 However, the sentences in (54) have the particular semantic and pragmatic properties associated with specificational sentences: (i) DP1 provides a description with a variable slot, which gets valued by the referent of the postcopular DP, and (ii) DP2 is focused, as indicated by the preceding question. In fact, it turns out that a specificational (focused DP2) reading is the only reading available for this kind of sentence, as shown by the fact that in a context which makes DP2 topical, the sentences in (54a,b) are both infelicitous, as shown in (61).

figure bg

Furthermore, recall, that DP1 in specificational sentences must be either a definite or a strong indefinite. As Citko (2008) points out, indefinites are typically bad in Polish sentences of this form (as in English, as shown by the translation in (62)).

figure bh

Altogether, this pattern is strikingly different from what we find in English and Danish, which as we saw above only allow Accusative case on the postcopular DP in equative and predicational sentences. These languages also permit only Accusative case on the postcopular DP in specificational sentences, as shown in the Danish example in (63) from Mikkelsen (2004b), and its English translation.

figure bi

Moreover, in English, agreement is always with DP1, in all types of copular sentences, as shown in (64).

figure bj

Neither Swedish nor Danish show verb agreement. However, given the assumption that case is tied to agreement (Accusative assigned via Agree with \(\textit{v}^{o}\) and Nominative assigned via Agree with finite \(\hbox {T}^{o}\), a standard assumption following Chomsky 2000, 2001Footnote 40), we may conclude that in Swedish specificational sentences, as in Dutch, Russian and Polish, DP2 is the grammatical subject. For Danish on the other hand, as in English, DP1 is the subject of specificational sentences—a conclusion defended also by Mikkelsen (2004b) and Partee (2010) (see also Geist 2007). This is illustrated in (65)–(66) below. In the following, I focus on the Swedish type (65), referring the reader to Mikkelsen (2004a, b, 2005) for a discussion of Danish, and Heggie (1988), Heycock (1991, 2012), Heycock and Kroch (1999, 2002), Rothstein (2001), den Dikken (2006), and Partee (2010), among several others, on English.

4.3 Towards an analysis (part II)

Above, I argued that in specificational sentences in Swedish, Polish, Russian, and Dutch, DP2 is the subject, whereas in Danish and English, DP1 is the subject. Thus, I take it that in order to account for specification in the former set of languages, some kind of inversion analysis is called for, unlike in the latter set (see Moro 1997; Geist 2007; Partee 2010; Heycock 2012; and references above). This is illustrated in (65)–(66); see Sect. 2.3 on SC and PredP.

figure bk
figure bl

However, focusing on Swedish, I will argue that it cannot be the case that specification is derived simply by topicalising the predicate DP, as suggested by Heggie (1988) for English (see also Partee 1998 and Geist 2007 for versions of this analysis applied to Russian). On Heggie’s account, DP2 is underlyingly the subject, in Spec,TP. The predicate DP then raises to Spec,CP, accompanied by V-to-C movement of the copula. As demonstrated by Heycock (1991, 186–189), Rothstein (2001, 259), and Mikkelsen (2004b, 2005), this analysis cannot be right for English, since it would predict the grammaticality of sentences like (67)—clearly not a well-formed sentence of English.

figure bm

Mikkelsen (2004b) also argues convincingly against this kind of analysis for Danish—which, unlike English, does have the requisite kind of predicate topicalization structures (main clause Verb Second [V2]). As we saw in (63) above, DP2 in Danish specificational sentences is marked with Accusative case. In topic-initial V2 clauses on the other hand, as shown in (68), DP2 is obligatorily marked with Nominative case.

figure bn

Indeed, the same is true also for predicate topicalizaton in Swedish.

figure bo

However, in Swedish (as in Dutch, Polish and Russian, but unlike in Danish and English), DP2 is obligatorily Nominative also in specificational clauses. This then leaves open the possibility that predicate topicalization might indeed be the correct analysis of specification in these languages.

However, there are a number of reasons for rejecting this analysis. As Mikkelsen (2004b) demonstrates for Danish, predicate topicalization and specificational sentences are structurally distinct. This is true also for Swedish. Here, I will focus on one of the several diagnostics used in Mikkelsen (2004b), namely the placement of negation and medial adverbs, such as ‘inte’ (not), which in Mainland Scandinavian are standardly assumed to occur at the left edge of vP.Footnote 41Mikkelsen (2004b) shows that in specificational clauses, these elements obligatorily occur to the left of DP2 (70a), whereas in predicate topicalization sentences, they occur to the right of DP2 (70b).Footnote 42

figure bp

Mikkelsen (2004b) shows that this syntactic contrast is associated with a difference in meaning. As shown in (71), from Swedish (Mikkelsen 2004b gives similar examples from Danish), NEG-DP2 order is only felicitous when DP2 is focused, i.e., when the sentence is assigned a specificational interpretation.Footnote 43

figure bq

Further, as illustrated by the contrast between the predicate topicalization sentence in (72) and the specificational sentence in (51a) above, in (51a), reference to DP1 is with the non-gendered pronoun ‘det’ (it), whereas in (72), reference to DP1 is with the gendered pronoun ‘hon’ (she).

figure br

It is also worth noting that predicate topicalization sentences in both Danish and Swedish differ from specificational sentences in terms of their word order in modal sentences, as shown with the Swedish examples in (73) (see Mikkelsen 2004b, 48 for Danish examples).

figure bs

Hence, as shown by Mikkelsen (2004b) for Danish, we find that Swedish has both predicate topicalization and specification, and that these are different from each other, both in terms of their syntactic, semantic, and information structural properties.

To summarise this section, Swedish specificational copular sentences are characterised by the following facts:

figure bt

Additionally, in relation to the first observation, we also saw evidence from Dutch, Polish and Russian (which behave otherwise like Swedish), that agreement in these languages is with DP2. Moreover, in relation to the second observation, we saw evidence from Polish and Russian that DP1 allows both Nominative and Instrumental case, depending on whether it is interpreted as an individual or stage level predicate.

For Danish, Mikkelsen (2004b) concludes that DP1 in specification is a predicative (non-referential) subject. For Swedish, as for Polish, Russian, and Dutch, however, this analysis cannot be right, given that in these languages, unlike in Danish and English, DP2 is the subject (and in Polish and Russian, DP1 shows the Nominative-Instrumental case alternation).

The basic structure proposed for Swedish specification is given in (65) above. However, implementing this kind of analysis is not uncomplicated. I have shown that a predicate topicalization analysis (à la Heggie 1988) is not viable. A different theoretical alternative comes from Heycock (2012), building on Romero (2005), who argues that specification is a type of inverted equation—in both types of languages discussed here (her focus is on the contrast between DP1 and DP2 agreement). She posits the structure in (75) as underlying both specificational and uninverted equatives, where the two arguments differ in terms of intentionality.

figure bu

She argues that since the functional head \(\hbox {F}^{o}\) raises to merge with the copula, the two DPs end up in a small clause configuration, such that both DPs become eligible for movement to the specifier of the copula. In specificational clauses, this will be the type <se> DP (NP2), whereas in regular equatives, it is the type e DP (NP1) that moves. (For further details, I refer the reader to Heycock 2012.)

Although an appealing option, it is not clear how this account would handle the fact that, in languages of the Swedish type, the case on DP1 (NP2 in (75)) in specificational sentences alternates along the same semantic dimension as DP2 in predicational and equative sentences. While this does not amount to a flat-out refutation of Heycock’s account, it seems like her account would at least need to be modified along the lines sketched in (38)/(65) to accommodate this fact. Of course, a problem for the movement analysis of Nominative DP2 in specificational sentences is the observation from Sect. 2, that such DPs cannot undergo A-bar movement. This (along with the binding facts) was one of the major empirical points motivating the bi-clausal analysis for equation and individual level predication (38). This raises the question of why this DP should be eligible for inversion in specificational sentences with predicate Nominative case, as in (53)–(57).

I propose that specificational sentences involve a low FocP. DP2 (the grammatical subject and an individual of type e) then enters into an Agree relation with both T\(^{o}\) and Foc\(^{o}\) in its base-position, thereby receiving Nominative case and becoming marked as an obligatorily focused constituent. It then moves to Spec,FocP where it becomes ‘criterially frozen’ (having valued all of its features), and is no longer eligible for further movement (see Rizzi 2006, 2016). The (Nominative or Accusative/Instrumental) predicate (see (65)) then moves to Spec,TP, in order to satisfy the EPP-feature on \(\hbox {T}^{o}\). (76) illustrates:

figure bv

More can, and should, no doubt be said about specification in these types of languages, in particular regarding the semantic composition. A corollary of the current proposal is that DP1 must move as a predicate and leave a higher type trace, in order for the composition to work out. I take this to be a desirable consequence, given that DP1 in these sentences is interpreted as a property with a variable slot, to get valued by the individual denoted by DP2. While the details of the syntax-semantics mapping remain to be worked out in more detail, the analysis proposed here has the benefit of capturing both the word order and the information structural properties of specificational sentences, as well as the interpretation of specification as involving property containing a variable slot, which gets valued by the individual denoted by the postcopular DP.

5 Conclusion

The first goal of this paper was to show that the Swedish case-alternation, given the associated interpretive and distributional facts, must be an LF phenomenon (contra Sigurðsson 2006). I then presented comparative data, primarily from Polish, but also from English, Danish, Dutch and Russian, showing (i) that the Swedish alternation is sensitive to the same types of semantic factors that condition the choice of predicate case in Polish, Russian and Dutch, and (ii) that the Swedish alternation also shares syntactic properties with copular sentences in these languages, regarding movement and binding. I suggested that the morphological and semantic contrasts are mediated by the same structural difference, proposing a biclausal analysis for Nominative case, and a monoclausal analysis for Accusative/Instrumental case. I also pointed to a typological generalisation regarding the way that the syntax mediates interpretation, whereby Swedish copular sentences follow the same strategies as in Polish, Russian and Dutch, which are different from those in English and Danish. I further showed that this typological split runs deeper. Adding to the work by Mikkelsen (2002, 2004a, 2004b, 2005, 2011); Geist (2007); Partee (2010); Heycock (2012), and others, I demonstrated that also specifiational copular sentences in Swedish pattern with those in the former group of languages, and unlike those in the latter—thus making the generalization one about how syntax and information structure mediate interpretation. This then echoes the view articulated in Partee (2010, 44) whereby “one language may have an information-structure-motivated diathesis choice where another language uses information-structure alone.” Regarding specification, I argued that the data supports a particular type of predicate inversion analysis, whereby specificational copular sentences are derived from the same kinds of structures underlying predicational and equative sentences.