Passive with control and raising in mainland Scandinavian

Abstract

This article gives an overview of the use of non-local passives in mainland Scandinavian, i.e. passives where the subject of the first verb is a thematic argument of a second verb. Three factors are important: whether V1 is a control verb or a raising passive, whether V2 is a passive participle or an infinitive and whether the passive is morphological or periphrastic. Danish and Norwegian allow passive control verbs such as forsøge ‘try’ with passive participles whereas this pattern is only found with semi-control verbs like begära ‘request’ in Swedish. In Swedish there is an alternative strategy for strict control verbs, viz. active control verb plus passive infinitive. All three languages allow both passive infinitival complements and passive participles with raising passives such as påstås ‘is claimed’. These passive constructions need to be distinguished from so called long passives and double passives where a passive feature on either V1 or V2 can spread to the adjacent verb.

1. Introduction

Passive normally involves the reordering of the arguments of a single predicate but in some languages there are non-local passive constructions which involve more than one predicate. By a non-local passive I understand constructions where the subject of a passive verb, V1, is an argument of another verb, V2, in an embedded VP, as shown in (1).

Different kinds of non-local passives have been studied in various languages. In Danish and Norwegian, there is a construction, known in the literature as complex passive, where V2 is a passive participle (see e.g. Engh, 1984; Hellan, 1984; Christensen, 1986; Ørsnes, 2006). This construction, was previously thought not to be found in Swedish but a closer look reveals that Swedish instead uses a construction where V2 is a passive infinitive. In all three languages we find a productive construction where V1 is a verb of communication. Ørsnes (2013) refers to this as reportive passive; I will use the term raising passive. German, Italian and French use a construction known as long passive where V2 is an infinitive, either active or passive. (see Wurmbrand, 2001; Cinque, 2006; Hobæk Haff & Lødrup, 2016). Lødrup (2014) argues that long passive is also found in Norwegian. The primary aim of this article is to investigate how these non-local passive constructions are used in Danish, Norwegian and Swedish in the light of recent studies by Julien & Lødrup (2013) and Lødrup (2014).

Three factors turn out to be important: whether V1 is a control verb or a raising passive, whether V2 is a passive participle or an infinitive and whether the passive V1 and/or V2 is morphological, formed by adding -s to the verb, or a periphrastic construction with copula and pariciple. ¹ Although a number of analyses of non-local passives have appeared, the interaction between these factors has not been sufficiently studied. In this article data from previous articles are complemented with searches in morphologically annotated corpora. The searches give some indications of relative frequency which turns out to be especially relevant for the development of certain constructions.

Section 2 is devoted to passives with control verbs. It is shown that we need to distinguish strict control verbs like försöka ‘try’ from semi-control verbs like önska ‘wish’. Danish and Norwegian allow complex passives with passive participles with both strict and semi-control verbs whereas Swedish only allows the latter type. In Section 3 I discuss cases where a passive V1 behaves as a subject-to-subject raising verb. This construction is common in all three languages. In Section 4 I show that Swedish has developed an alternative to the complex passive with strict control verbs: active control verbs with passive infinitives. In this Section I also trace the historical development to the middle of the 19th century. Section 5 looks briefly at so called double passives (Julien & Lødrup 2013) where the passive feature on V2 spreads onto V1. In Section 6 I look closer at the voice agreeing long passives with infinitives in Norwegian (Lødrup 2014), which are similar to long passives in German, and try to explain why their Swedish counterparts differ. Section 7 brings together the interaction between verb type, the form of the complement and the choice of passive form. Although there is considerable variation in the data, both within and across the languages, it is possible to discern certain patterns which I summarize in a table.

2. Control passive with participle

2.1 Strict control

In the 1980’s several researchers started to investigate the so called complex passive in Norwegian (see e.g. Engh, 1984; Hellan, 1984; Christensen, 1986).

In (2) the object bilen ‘the car’ of the embedded V2 passive participle reparert ‘repaired’ has been promoted to subject of the passive V1 ble forsøkt ‘was tried’ and is realized in the subject position (Spec,IP). ² There is no agent and the interpretation can be conveyed by a there-construction in English.

It was noted at the time that the Swedish version of (2) is clearly ungrammatical, both with the morphological s-passive and with the periphrastic bli-passive.

It was concluded that Swedish lacks complex passives, despite the fact that Engh (1984), 18 mentions the example in (4) which looks very much like a complex passive; the subject målet ‘the case’ corresponds to the object of the embedded participle avgjort ‘decided’.

There have since been a number of studies on complex passive in Norwegian (see e.g. Christensen 1991; Engh 1994; Hellan 2001), and at least one on Danish (Ørsnes 2006). Since the term complex passive is also used for other constructions, I will refer to this type as control verb with passive participle, abbreviated CoPP. A schematic structure is shown in (5).

CoPP is, if not common, at least not unusual in present-day Danish and Norwegian. ³ Both the morphological passive and the periphrastic passive can be used as shown in the following examples.

The participle has the form that is used in the periphrastic passive in the language. I refer to this form as the passive participle and distinguish it from the invariant perfect participle or supine that is used in the perfect tense (Larsson 2009). ⁴ In Danish and bokmål Norwegian ⁵ there is no agreement and the neuter singular form is used with all subjects. In the nynorsk variety, abbreviated NNo., the participle, in principle, agrees in number and gender with the subject (as in Swedish). ⁶ See the second participle in (8).

The number of first verbs that are used in CoPP is rather small. The most common ones are listed in (9). ⁷

The verbs are typical subject or object control verbs; in the active they take an infinitival complement and the matrix subject or object and the understood subject of the infinitival clause are coreferential. They do not take finite complements and the predicate in the infinitival clause must be controllable (Zaenen 1993). These properties are illustrated in (10).

In the passive, these verbs may combine with passive participles, see (6) and (7), and the matrix subject is understood as the internal argument of the participle. Supine forms (perfect participles of active verbs) are impossible and infinitival complements are dispreferred.

CoPP is typically found with passive particples of transitive verbs; the participle can be predicated of the internal argument.

CoPP is not found with unaccusative V2 although such verbs can be predicated of the subject. This presumably follows from the fact that they are not controllable, see (10c).

CoPP is also not found with unergative verbs like cykle ‘to bike’. It seems plausible that this follows from the argument structure since participles of unergative verbs cannot be predicated of their argument, see (14). ⁸

Examples of object control verbs used with CoPP are shown in (15) and (16). Note that neither of the contollers is expressed. In the nynorsk variety, the passive participles may show agreement with the matrix subject although this is no longer common, as noted above.

Already Engh (1984) noted that CoPP is blocked by prepositions.

Data like this have been used by e.g. Holmberg (2002) as evidence that the passive and the participle are reanalyzed as a complex predicate. ⁹ Ørsnes (2006) argues against this analysis and points out that a complex predicate typically has a single external argument whereas the suppressed agents of the control verb and the participle need not be coreferential. Example (2) can be understood to convey that someone tries to arrange for the car to be repaired by someone else. ¹⁰ Furthermore complex predicate formation is not possible when the second verb has independent tense (Wurmbrand 2001, 79ff.) but the two verbs in CoPP may have different reference times.

Another indication that CoPP does not involve complex predicate formation is the existence of examples like (21) where two participles are conjoined. ¹¹

In the English translations I have followed Ørsnes (2006) and tried to convey the meaning by using a paraphrase which introduces an event of trying, recommending, forbidding etc. Overt agent phrases are possible but are seldom present. In many ways, sentences with CoPP resemble impersonal passives, i.e. passives where no argument is promoted to subject, and it is not surprising that CoPP is also found in impersonal passives.

CoPP is also found in presentational sentences in Danish and Norwegian, provided that the pivot (the subject associate) is indefinite (Ørsnes 2006, 394).

Note that unergative verbs may be used in impersonal passives, see (22), but not in presentational sentences.

The difference possibly has to do with whether the argument of the passive participle is expressed or not. CoPP with participles of unergative verb are unavailable when the argument is expressed, as in (14) and (24), but possible when the argument is suppressed, as in (22). ¹²

As mentioned above, corresponding examples of CoPP in Swedish are blatantly ungrammatical with both passive types, see (3). The same holds for object control verbs as shown in (25). ¹³

2.2 Semi-control

However with certain verbs that can be used as control verbs, the participle version is possible in Swedish as well. We have already seen an example with the verb önska ‘wish’ in (4), repeated here in (26). Another verb that is commonly used in the passive with a passive participle is begära ‘request’. ¹⁴

Unlike the strict control verbs discussed in Section 2.1, these verbs can, in the active, be construed with a finite complement.

When used with an infinitive, the understood subject must be controllable. Regular transitive verbs are infelicitous unless they are preceded by a verb like få ‘get’, see (29). This induces a control shift similar to examples with promise. ¹⁵ The infinitival complement can also be expressed as a bli-passive, as in (30).

When these verbs are used in the passive, only participles of agentive transitive verbs can be used, as in (27). Unaccusative or unergative pariciples are not possible.

In formal genres the active verb begära can take an overt object followed by a participle. It thus resembles an ECM-verb (Postal 1974), but note that it cannot combine with an infinitive.

The corresponding verbs in Danish and Norwegian also allow both a CoPP construction and an object plus participle, at least in formal genres, whereas infinitival small clauses are not used. ¹⁶

Given that the examples with overt object plus participle are grammatical, the CoPP in (27), (33a) and (34a) can be analyzed as passivized ECM-constructions. For Swedish this seems to be a plausible analysis which brings out the fact that only verbs that can be construed with a participial small clause allow CoPP. Note that strict control verbs need to be distinguished from semi-control verbs in Danish and Norwegian as well as the former don’t allow overt objects.

This presumably follows from the fact that strict control verbs only take infinitival complements with a controlled understood subject (PRO). Consequently there is no overt argument that can be raised to object.

Previous studies of CoPP in Danish and Norwegian have not observed the need to distinguish between strict control and semi-control verbs, but this turns out to be important when we bring Swedish into the picture. The fact that semi-control verbs with overt objects cannot be used with infinitival complements in any of the languages becomes relevant when it comes to distinguishing CoPP from non-local passives with raising passives.

3. Raising passives

At first glance the example in (37) looks like a CoPP but closer inspection reveals that it is not.

The verb påstå ‘claim’ is not a control verb. In the active it takes a finite complement, not an infinitive; the understood subject in (38b) cannot be identified since it is not controlled.

In the passive this verb takes either a passive participle, as in (37), or an infinitive, active or passive, as in (39).

The subject of the passive verb påstås is interpreted as the understood subject of the infinitive ha ‘have’. The construction thus resembles examples with a lexical subject-to-subject raising verb such as verka ‘seem’ in (40).

Because of this similarity I refer to them as raising passives, short for verbs that in the passive resemble subject-to-subject raising verbs. ¹⁷ In the Latin grammar tradition, the term used for this construction was Nominativus cum infinitivo (NCI) (Noël & Colleman 2009), to be distinguished from Accusativus cum infinitivo (ACI), see below.

Raising passives are formed with verbs of communication and cognition. Ørsnes (2013) dubs them reportive passive since they typically “attribute a proposition to a (generally) unknown external source” (2013, 321). Raising passives are quite common in the Scandinavian languages and are found with around 50 verbs in Swedish (Lyngfelt 2011) and Danish (Ørsnes 2013). Faarlund et al. (1997, 1028) list eleven verbs that can be used as raising passives, mainly in written bokmål. ¹⁸ Examples of verbs that often used as raising passives are given in (41), in their Swedish version.

I first discuss raising passives with infinitives since they are better studied and then, in Section 3.2, raising passives with participles, as in (37).

3.1 Raising passive with infinitive

Schematically raising passives with infinitive have the following structure.

As shown above, these verbs take a finite subordinate clause in the active, see (38a). The entire subordinate clause can be realized as the subject of the passive verb, as in (43a), although this is uncommon. The extraposed version in (43b) is somewhat more common, but the version where the subject has been raised out of an infinitival clause is most common, as in (39b), repeated here in (43c) (Lyngfelt 2011). ¹⁹

Like other raising verbs, raising passives do not assign a thematic role to the subject. If the embedded verb takes an expletive subject, so does the raising passive, and idiom chunks can appear as subjects.

These two tests can serve as diagnostics for distinguishing raising passives from passives with control verbs which are clearly ungrammatical with expletives and idiom chunks (Lødrup 2014, 372f.).

Swedish examples of raising passives are typically construed with a bare infinitive without infinitive marker, as shown in the examples above. ²⁰

In Danish raising passives, the overt infinitive marker at is used, see (46).

According to Faarlund et al. (1997), the infinitive marker å is used with Norwegian raising passives, see (47a), but there is actually variation and examples without å are also found, see (47b) and (47c). ²¹

In all three languages, only s-passives are used productively; the bli-passive is seldom used with raising passives.

Most of the raising passives in (41) can only be used as subject-to-subject raising verbs and cannot be construed with overt objects in the active.

In this respect the raising passives behave like say in contemporary English which can be construed with an infinitival clause in the passive but does not admit an overt object, see Los (2009), 119f. and Noël & Colleman (2009).

However, a few raising passives can be used in the active with an object plus an infinitival complement, i.e. the construction known as accusative with infinitive (ACI) or exceptional case marking (ECM), see e.g. Postal (1974) and Lødrup (2008).

The corresponding Danish verb is construed with an object and a prepositional complement.

Given that the verb anse can be used in the active with an overt object, the passive versions can be analyzed as subject to object raising followed by regular passive. However, the majority of raising passives are not ECM-verbs and do not take overt objects in the active. ²² For this reason I follow Lyngfelt (2011) and Ørsnes (2013) and analyze most of the raising passives as instances of subject-to-subject raising, not as passivized ECM constructions. Note furthermore that when the exceptional verb anse is used with an overt object, the infinitival V2 is practically always a copula verb. The other raising passive verbs are used with a variety of V2 verbs as shown in the examples.

3.2 Raising passive with participle

What has seldom been made explicit in the literature is that passive raising verbs also can be used with passive participles, abbreviated RaPP, with the following structure.

This structure is identical to the one given for CoPP in (5) and this may have led some researchers to include such examples among CoPP (complex passives), (see e.g. Engh 1994; Faarlund et al. 1997; Hellan 2001). Others distinguish them from CoPP, see e.g. Christensen (1991, 47) and Ørsnes (2006, 389).

All the verbs listed in (41) can be used with passive participles, with varying frequency. The Swedish example in (37) is repeated here as (58), followed by Danish and Norwegian examples. ²³

Recall that in the active, these verbs take a finite complement without imposing any constraints on the type of subject or verb. In the raising passive construction only transitive verbs and unaccusative intransitive verbs are possible, as noted by Engh (1994) who gives the pair of examples in (61).

Just as with control verbs, this restriction presumably follows from the argument structure of the participle; participles of unergative verbs cannot be predicated of the single argument, see (14), (Hellan 2001; Platzack 2010).

Most verbs that can be used in the RaPP construction do not take overt objects plus participle in the active. ²⁴

The exceptional verb anse, which can be used with an overt object plus infinitive in all three languages as discussed in Section 3.1, can also be used with object plus participle. In the majority of the cases, the object is a reflexive pronoun, as in (66) (Lyngfelt 2011).

A few examples with non-reflexive objects were found in the corpus searches. In (67) det is an anticipating expletive for the extraposed infinitival clause.

3.3 Summary

Raising passives are used in all three languages. They differ from both strict and semi-control verbs in that they can take expletive subjects and raised idiom chunks. The main criterion is whether V1 can be used as a control verb or not. None of the verbs in (41) can take an infinitival complement in the active whereas they often do in the passive.

Semantically raising and control passives differ in perspective; raising passives typically report on a state of affairs, a proposition, whereas control and semi-control verbs express a relation to a future state of affairs. ²⁵ There are a few verbs which can do both and which take both sentential and infinitival complements. One such verb is frygte, frykte, frukta ‘fear’. In all three languages this verb can in the active be used with both complement types, with sentential complements being more frequent. See the Norwegian examples in (68).

This verb can also be used as a raising passive in all three languages and in Danish and Norwegian it is found with passive participles.

Given that the verb is frequently used in the s-passive as a raising passive, it seems plausible that the example in (69b) should be analyzed as a RaPP construction and not as a CoPP with an unaccusative verb, as has sometimes been assumed (Engh 1994). The fact that the bli-passive is not possible, (69c), further supports the RaPP analysis, compare (49). ²⁶ We can thus maintain our generalization in Section 2 that CoPP is only found with controllable transitive verbs. Raising passives do not impose any selection on the type of verb in the complement.

4. The Swedish strategy

Let us return to the strict control passives in Danish and Norwegian and consider what message they convey. The Danish example in (7b) is repeated here as (70).

The message is that there are plans to try to resume the match at a later time. We have seen that this cannot be expressed with a CoPP construction in Swedish, see (3)and (25), but the same message can be conveyed by an active strict control verb followed by a passive infinitive. I refer to this construction as CoPI.

I start by looking at this construction in modern Swedish and show in Section 4.2 that it is not a raising construction. I end by tracing its historical origin in Section 4.3.

4.1 Present-day Swedish

Examples like (72a) are occasionally found in Swedish texts. ²⁷

Note that neither the version with bli-passive in (72b) nor the bare passive participle in (72c) is grammatical. Many of the hits are from unedited informal blog texts.

But the construction is also found in edited newspaper texts from both Sweden and Finland.

The construction can also be found in legal texts which presumably have been carefully worded. This suggests that it has become part of standard language, although rather infrequently used.

The construction is most common with försöka but occasional examples are found with the verbs söka ‘seek’ and våga ‘dare’. ²⁸

CoPI in Swedish seems to be restricted to a few strict control verbs and there are no indications that the construction is spreading to semantically similar verbs. It has not been attested with the verbs orka ‘have enough energy to’ or slippa ‘avoid, not have to’, and the constructed examples in (78) do not seem acceptable.

The construction has not been found with object control verbs, like rekommendera, or with semi-control verbs like begära ‘request’.

As shown by the paraphrases, the interpretations of the examples in (73)–(77) resemble the interpretations given for the CoPP examples in Danish and Norwegian. A further indication that the CoPI construction is functionally equivalent to the Danish and Norwegian CoPP is that it can be used in presentational sentences, see (80), and impersonal passives, see (81) (slightly adapted from a blog example).

4.1.1 Animacy

All the examples of the CoPI construction in Swedish given so far have inanimate subjects or the expletive det. This presumably blocks the ordinary control interpretation as inanimate referents cannot be agents of försöka. ²⁹ But animate subjects are possible if the context makes it clear that someone else is responsible for making the attempt.

There is an interesting interaction with the choice of passive form. In a regular control construction with a passive infinitival complement, the bli-passive is strongly preferred in Swedish, see Engdahl (2006, 32) from where the examples in (83) are taken (see also (30)).

As mentioned above in Section 2, control verbs presuppose that the subject has some control over the situation. In (83a), the subject is understood to have acted in a way so that s/he would be reelected, whereas this interpretation is not available for (83b). The use of s-passives in the CoPI construction thus contributes to the interpretation that someone else is responsible for bringing about the event.

In Danish and Norwegian as well, most of the attested examples of CoPP have inanimate subjects, but animate subjects are possible. ³⁰

One further observation can be made here. In the CoPI examples, the control verbs are often modified by temporal or modal auxiliaries, see e.g. (82). With an unmodified verb, the example is somewhat less acceptable.

The auxiliaries ska ‘shall’ and måste ‘must’ are common in CoPI examples. It may be that these auxiliaries add an indirectness that helps block the regular control assignment.

4.2 Has ‘försöka’ become a raising verb in Swedish?

The syntactic frame in (71) resembles that of a lexical raising verb like verka ‘seem’.

One hypothesis that comes to mind is that försöka, söka and våga have developed into auxiliary-like raising verbs in present-day Swedish, similar to aspectual auxiliaries like börja ‘begin’, sluta ‘end’ and fortsätta ‘continue’. In present-day Swedish these verbs allow raised subjects.

Bylin (2013) has studied what she calls the auxiliation process of these verbs in the history of Swedish. ³¹ The oldest examples with börja ‘begin’ have animate subjects, (89a), but quite early the verb starts to be used with inanimate subjects, as in (89b). Expletive subjects, (89c), turn up around 1750 and around the same time Bylin finds the first examples with a passive infinitive, (89d). ³²

However, försöka does not undergo this development, as Bylin shows. It does not accept an inanimate subject with an active infinitive, nor can it be used with a weather expletive.

Holmberg (2002), 122 cites the presentational example in (91) as evidence that försöka has undergone restructuring to a raising verb which can take an expletive subject.

However försöka still requires the pivot (subject associate) to be animate which means it is still a control verb, as noted by Wiklund (2005, 45) who contrasts (91) with (92). ³³

Wiklund also points out that försöka cannot be used with an idiom, unlike true raising verbs.

I conclude that the control verbs we find in the CoPI construction in Swedish are indeed control verbs and have not turned into raising verbs. In Norwegian and Danish, the CoPI construction is rare and most examples are judged to be ungrammatical. As expected the CoPP version is preferred. ³⁴

4.3 Historical development

Engh (1994) includes a survey of the emergence of non-local passive constructions in Norwegian. ³⁵ He traces the first appearances of strict control verbs with passive participles to the end of the 18th century and the beginning of the 19th century. The first verbs to be used are søke ‘seek’, attested in 1761, and forsøke ‘try’ from 1837 (Engh 1994, 302 fn 15). ³⁶

During the 19th century the CoPP construction became noticeably more frequent. Engh (1994, 283) attributes this to the growth of newspapers and in particular to the increased use of advertisements.

Corpus searches for corresponding s-passive forms of söka and försöka followed by a passive participle in the Swedish historical corpora in Korp (1,35 G) yielded no examples but there were a few examples with bli-passive, both from the newspaper Aftonbladet. ³⁷

There were no examples from later periods, nor from other genres. I take this as an indication that CoPP did not become established in Swedish at the time when it increased rapidly in Norwegian. However, in the historical corpora I found around 30 examples of the CoPI construction with active söka, försöka or våga, followed by a passive infinitive. ³⁸

The following example is one of only a few with an overt agent phrase:

We noticed above that most of the modern examples of CoPI have an inanimate subject, i.e. a subject that cannot be interpreted as the agent of the control verb. The same is true for most of the examples from the 19th century with a few exceptions. In (105) the subject is alla ‘everybody’.

However, the meaning is clear: someone else must not try to force everybody to get married in a rush. Similarly in (106), it is someone else who tries to win over the military on their side.

It is maybe not surprising that the CoPI construction started to spread during the 19th century when several other control verbs began to be used as raising verbs, allowing an underlying object of a passive infinitive to raise to become subject. We have seen this development with aspectual verbs like börja ‘begin’ in (89d). Even more common is this pattern with hinna ‘have time to’ (about 12 000 hits in the historical corpus, 1,3 G).

But recall that the CoPI verbs have not developed into raising verbs, see Section 4.2.

4.4 Summary

The CoPI construction in Swedish seems to correspond quite closely to the CoPP construction in Danish and Norwegian. Nevertheless the CoPI construction has not become as established in Swedish as the CoPP pattern in Danish and Norwegian. CoPP is used rather frequently, in many genres and with both subject and object control verbs and can, at least in principle, be iterated.

CoPI in Swedish has only been found with three subject control verbs, försöka, söka and våga, and the low incidence (less than 0.3 per million words in the investigated corpora) indicates that it is not very productive. The small number of possible verbs makes it hard to test whether CoPI can be iterated. The following constructed example sounds marginally acceptable to me.

5. Double passives

The non-local passive constructions discussed in the previous sections need to be distinguished from two other constructions which are discussed in this and the following section. In Danish, Norwegian and Swedish a modal or temporal auxiliary verb may show up in the passive when the main verb is passive, as shown in (110) from Språkriktighetsboken (Svenska språknämnden 2007, 280).

I follow Julien & Lødrup (2013) and refer to this type as double passive. (110) resembles the raising passive with infinitive discussed in Section 3.1, but note that V1 does not have to be passive, (110b), whereas this is necessary for raising passives, see (111).

The active verb behöva ‘need’ has developed into a raising verb with modal meaning. ³⁹ The passive form tends to be found when the infinitival complement is in the passive, as in (110). If V2 is not passive, then only active forms of behöva are possible. ⁴⁰

Unlike the raising passive in (111), double passives only involve one passive predicate, as Julien & Lødrup (2013) point out. Only the main verb, V2, is passivized and the passive feature has spread from V2 to V1. ⁴¹

Julien & Lødrup (2013) also note that double passives are only found with inanimate subjects and constructed examples with animate subjects are judged to be strange.

They suggest that this may be a consequence of the preference for inanimate subjects with s-passive, see Engdahl (2006) and Laanemets (2012).

Double passive is also found with temporal and aspectual auxiliaries in all three languages, see Julien & Lødrup (2013) for frequencies. In Swedish V2 is practically always an s-passive, but in Norwegian V2 can also be a bli-passive.

Double passives are mainly found in informal, unedited texts. The example in (115) was part of a recorded message. The future auxiliary komma has a doubled s-form which is otherwise not used.

Normative grammars frown on double passives but they seem to have been in use for a long time, as Julien & Lødrup (2013) note. The earliest examples are from the 15th and 16th centuries and the construction becomes rather common during the 19th century. Given that the passive form of V1 is dependent on V2 being passive, it seems likely that the passive feature spreads from V2 to V1. ⁴²

6. Long passives

In Section 2.1 we looked at passive control verbs like forsøkes with passive participles, CoPP. But passive control verbs can also be used with a passive infinitive, as in the Norwegian examples in (116) from Lødrup (2014, 368).

Lødrup (2014) takes the examples in (116) to be instances of long passive in which two verbs are reanalyzed as one complex predicate and the internal argument of V2 is promoted to subject of V1, see Wurmbrand (2001). Other terms for this phenomenon is long-distance passives (Bader & Schmid 2009) and long object movement (Wurmbrand & Lohninger 2022). Long passives with infinitival V2 are found in several European languages, including German and French (Wurmbrand 2001; Wurmbrand & Shimamura 2017; Hob æ k Haff & Lødrup 2016), as well as in many Austronesian languages (Wurmbrand & Lohninger 2022). In the European languages, the matrix verb, V1, is always passive and the embedded verb, V2, is an infinitive. In German this infinitive is active, see (117), in French, it can be either active or passive, see (118). In all cases the subject corresponds to the internal argument of the embedded infinitive.

Long passives differ from double passives in that it is V1 that must be passive. In double passives it is the other way round; V2 must be passive. Double passives in Scandinavian are mainly found with temporal and modal auxiliaries whereas long passives can be used with both aspectual and control verbs, see the overview for German in Wurmbrand (2001, 327).

Using corpora and native speaker intuitions, Lødrup shows that in Norwegian, long passive is found with several different control verbs, including verbs that cannot be used in CoPP. For example the verb huske ‘remember’ can be used in long passives but not in CoPP, compare (116b) and (119).

As shown by the German and French examples above, the passive feature can be realized on both V1 and V2 or just on V1 in long passives. Lødrup shows that this applies to Norwegian long passives as well. In addition to examples like (116), there are parallel examples where V2 is active, as in (120).

Lødrup (2014) takes this as evidence that the examples in (116) arise through voice agreement, optional spreading of the passive feature from V1 to V2. Part of his evidence for this analysis comes from the existence of similar feature spreading with active verbs, as in (121a). Here V1 is a supine, an active perfect participle, and this feature has spread to the embedded V2; the expected form would be an infinitive.

In the around 100 examples of long passives that Lødrup (2014) found, V1 is most often an s-passive, but bli-passives are possible. V2 is most often also an s-passive. He notes that examples with long passive vary in accptability, both within and across speakers. Nevertheless they have been used at least since the 19th century, see (122) from the Norwegian newspaper Morgenbladet in 1843, supplied by Helge Lødrup.

Examples that resemble the Norwegian examples in (116) can be found in Danish and Swedish as well. In Danish the construction is found with huskes ‘be remembered’ and ønskes ‘be desired’.

Especially the type in (123c) is common on the internet. No examples with forsøges plus a passive infinitive were found in KorpusDK and Ørsnes (2006) finds the constructed example in (11), repeated here as (125), with an infinitival bli-passive unacceptable.

Searches in the Swedish Korp (2.1 gigabyte) yielded around 70 examples, mainly from blog texts and Twitter.

Just as in Norwegian, such examples have been used in Swedish for a long time. In the historical corpus, there are examples from newspapers starting from the 19th century.

But there are also some differences compared to the Norwegian data. First, V2 in the infinitival clause is always an s-passive; no examples with periphrastic passives have been found. Second, there are hardly any examples with active V2, apart from examples like (127) which are standard in advertisements.

The Swedish version of the Norwegian example in (120) is clearly degraded and searches in Korp give few hits. ⁴³

The pattern with passive försöka followed by an active infinitival clause is only found with extraposed infinitival complements in Swedish, see (129a). Note the overt infinitive marker which is obligatory in extraposition. ⁴⁴ The non-extraposed version in (129b) is possible but hardly used.

In Norwegian, an extraposed infinitival clause can be passive, see (130) from Lødrup (2014, 369).

This is not possible in Swedish, see (131). Only the version with active V2 is posssible, as in (129a). ⁴⁵

Both these differences follow if we assume that the spreading goes from V1 to V2 in Norwegian, as Lødrup (2014) assumes, but from V2 to V1 in Swedish. ⁴⁶ Since passive forsøkes can be used in the CoPP construction in Norwegian, it is conceivable that it can also be used with an infinitive with spreading of the passive feature to the infinitival V2. This raises the question whether the Swedish examples in (125) should be analyzed as long passives, as Lødrup (2014) does. An alternative presents itself which makes use of the fact that Swedish has the CoPI construction where an active control verb is followed by a passive infinitive. Since this resembles the precondition for double passives, see (110), the Swedish examples could arise through the same kind of feature spreading from V2 to V1 that we find in double passives. Given that both the CoPI construction in (132a), and the double passive in (132b) are used, then (132c) could arise through an extension of the feature spreading in double passive to CoPI verbs.

An additional indication that there is a kind of double passive spreading in CoPI constructions comes from the fact that we find this also in impersonal passives, cf. (81).

However, it is not clear that all apparent long passives in Swedish arise as double passives of CoPI constructions. There are several long passive examples with the verb glömma ‘forget’, see (125b), but the CoPI version sounds unacceptable, both with and without the particle bort.

There was one CoPI version without particle in Korp which actually sounds better. This illustrates that there is a lot of subtle variation, partly due to choice of tense, as well as considerable context sensitivity.

Since Swedish differs from Norwegian in that V2 is always passive in these examples, it seems that Lødrup’s analysis, based on optional voice agreement, is not applicable. In the Swedish data that I have looked at so far, it seems that the passive feature can spread from V2 to V1, as in double passives, also in CoPI constructions. The data come mainly from informal written media and there is considerable variation. Informants that I have consulted are often uncertain whether they could use a particular sentence themselves. More research is clearly needed. ⁴⁷

7. Concluding remarks

In this article I have described several non-local passive constructions involving more than one verb which are used, with varying frequency, in Danish, Norwegian and Swedish and which can be difficult to tease apart. At least the following distinctions need to be made. First, passive with control verbs needs to be distinguished from sentences where the first verb is a raising passive, i.e. a passive of a reporting verb. The latter type is found not only in the Scandinavian languages but is well known from e.g. Latin and English. Unlike these languages, however, most raising passives in Scandinavian cannot be derived via subject-to-object raising. Second, we need to distinguish strict control verbs like försöka ‘try’, which don’t alternate with finite complements and which never allow overt objects, from semi-control verbs, like begära ‘request’, which also take finite complements and which, to some extent, can be used with overt objects. Whereas Danish and Norwegian use complex passives with participles with both kinds of control verbs, this is only possible with semi-control verbs in Swedish. Third, the form of the complement matters. Complex passive is only found with passive participles (CoPP) whereas raising passives can be used with both passive participles (RaPP) and infinitives. Fourth, regarding the choice of passive form in V1, CoPP can be used with both morphological and periphrastic passive in Danish and Norwegian. In raising passives, double passives and long passives, the morphological s-passive is clearly dominant although examples with the periphastic bli-passive can be found in Norwegian. In Table 1 I summarize how the different verb types can be used with participial and infinitival complements.

Table 1. Overview of verb types and constructions in non-local passives in Danish and Norwegian and Swedish

Starting from the leftmost column, we see that in Danish and Norwegian, strict control verbs, which in the active are only used with infinitival complements, can in the passive be used with passive participles (the CoPP construction) whereas this is not possible in Swedish, see Section 2.1. Instead Swedish uses the CoPI construction with an active V1 followed by a passive infinitival V2, see Section 4. In all three languages, semi-control verbs, which are used with both finite and infinitival complements, can be used with passive participles, but not with passive infinitives. This is discussed in Section 2.2. The last row shows that raising passives, i.e. passive forms of reporting verbs such as påstå ‘claim’ which take sentential complements, can be used with both passive participles and passive infinitives, as shown in Section 3. The column ECM shows that strict control verbs cannot be used with overt objects. Semi-control verbs can take objects but only with passive participles, see 2.2. As for raising passives, most of the verbs do not allow overt objects but there are some verbs that do, see (54b).

In addition to the CoPP, CoPI and RaPP constructions, passive forms of auxiliaries and control verbs can also be used with passive infinitives, see Sections 5 and 6. The two rightmost columns, Double passive and Long passive, indicate two ways of analyzing such examples, see (110) and (116), repeated here as (136) and (137).

In a double passive, as in (136), the passive feature spreads from V2 to V1 (Julien & Lødrup 2013). In a long passive, as in (137), the passive feature spreads from V1 to V2 by optional voice agreement, (Lødrup 2014). The differences between Norwegian and Swedish discussed in Section 6 suggest that the Swedish examples with strict control verbs arise through an extension of the double passive spreading from V2 to V1, indicated by a plus in the Double passive column and a minus in the Long passive column. For Danish and Norwegian, it is the other way around according to Lødrup (2014); Lødrup (in prep.). For semi-control verbs there is considerable variation in the data and more research on these types is needed, as indicated by the question marks in these columns. It is also not clear if raising passives of reporting verbs should be subsumed under long passives. Lødrup (2014, 373) mentions that some Norwegian examples are ambiguous between a raising passive analysis and a voice agreeing long passive analysis. In Swedish these examples must be analyzed as raising passives. Finally, it is not clear that Danish behaves exactly like Norwegian with respect to double passives and long passives, as suggested in the table. There is clearly a need for investigating more Danish data.

In this article I have concentrated on describing the variation found among the mainland Scandinavian languages. It would be desirable to look at the emerging patterns in a wider cross-linguistic perspective and to relate my verb classification to other proposals in the literature as well as to the discussion of complex predicate formation. Hopefully my attempt at systematizing the Scandinavian data will prove to be useful to other researchers.