Complex Predicates in Norwegian: New Evidence from Passive and Impersonal Sentences

1. Reanalysis (or restructuring) and complex predicates have been an important research topic for decades. Cases of two verbs being reanalysed as one predicate in a monoclausal structure have been reported in many languages (see for example Butt 1995, Alsina 1996, Wurmbrand 2001, Cinque 2006 and references there). It is striking that the Scandinavian languages have been largely absent in the discussion of reanalysis.

Wiklund 2007:87 claims that Swedish does not have the transparency phenomena that have been used to argue for reanalysis in other languages, such as long passives. (She mentions, however, that the so-called complex passive in Norwegian has been analyzed as a long passive, see Holmberg 2002.) Wiklund 2007 still finds evidence for reanalysis in Scandinavian in her account of some cases of verbal feature agreement.

I want to show that Norwegian has monoclausal reanalysis with a number of verbs, motivated by the same kind of transparency phenomena as those reported for other languages - including long passives (when complex passives are put aside). Verbal feature agreement will be shown to have a role to play.

2. A type of passive in Norwegian that has hardly been mentioned in the literature is exemplified in (1)-(3). In these sentences, a passive verb takes a passive infinitive. The subject of the clause is usually the patient of the infinitive; another option is to use an expletive subject, as in (3). Note that these verbs are not raising verbs, as can be shown using the traditional criteria.

Passives with passive infinitives are productive and reasonably acceptable with a number of verbs, including anbefale 'recommend', beslutte 'decide', foreslå 'suggest', foretrekke 'prefer', forsøke 'try', glemme 'forget', huske 'remember', klare 'manage', mistenke 'suspect', planlegge 'plan', prøve 'try', tillate 'allow', unngå 'avoid', våge 'dare', ønske 'wish'

An important fact is that Norwegian has sentences that are identical to the ones mentioned, except that the infinitive has the active form. These sentences can be somewhat less acceptable than the type (1)-(3), but examples are easily found, cf. (4)-(5).

Examples (4)-(5) are clearly long passives. The passives with passive infinitives could then be seen as long passives in which the voice feature of the second verb agrees with the one in the first verb. This can be seen as a case of multiple exponence. When long passives are monoclausal, the values of the voice features will unify at f-structure (Niño 1997, Bresnan 2001:134-41, Sells 2004).

Reanalysis has been implemented in different ways in the literature (se e.g. Alsina 1993, 1996, Butt 1995, Andrews and Manning 1999, Wurmbrand 2001, Sells 2004, Cinque 2006). It is understood here, as is standard in LFG, as a process in which two verbs combine their argument structures, and constitute one predicate in a monoclausal functional structure.

Norwegian has two ways of realizing the passive, the morphological passive with the suffix -s, and the syntactic passive with an auxiliary and a participle. In passives with passive infinitives the morphological passive is the frequent form on the first verb, and the dominating form on the second verb. The latter fact could be explained on the standard assumption that the verbs in a complex predicate do not have independent time reference, taken together with the proposal in Lundquist 2012 that the morphological passive is not time referring.

3. Evidence for complex predicates can also be found in active presentational sentences. Very many active one-place verbs -- including many unergatives -- can take an expletive subject and realize their thematic argument as an object in Norwegian, an example is (6). This option also applies to complex predicates to some extent (mentioned in Holmberg 2002). In (7), the thematic argument is in the complement position of the second verb

4. Pseudocoordinations also give evidence for complex predicates. This construction looks like a coordination of two verbs; the first belongs to a small group of verbs including posture verbs (e.g. Han ligger og sover 'he lies and sleeps'). I have argued elsewhere that these pseudocoordinations are biclausal subordination constructions. The first verb takes a verbal complement whose tense feature agrees with the first verb; this does not in itself make them monoclausal in my view. However, pseudocoordinations sometimes show a different behavior than they usually do, and there is some evidence that they have the option of being reanalysed as monoclausal.

In (8), both verbs are passive. The subject of the sentence is the patient of the second verb; it has no thematic relation to the first verb. This must be considered a long passive; the parallel to (1)-(2) above is clear.

Sentences such as (8) are admittedly not frequent, and they are a bit marginal in my intuition. However, there is also another indication that reanalysis must be an option in pseudocoordinations. Presentational pseudocoordinations, such as (9), usually have the characteristics of a biclausal construction: the argument of the first verb is in the complement position of the first verb, controlling the subject position of the second verb. However, the thematic argument is sometimes in the complement position of the second verb, as in (10), which indicates reanalysis. The parallel to (6)-(7) above is clear.

5. It was shown that Norwegian has evidence for complex predicates in monoclausal structures. Data similar to the above seems to exist in Danish and Swedish, but these languages have not been investigated.