Benjamin Lyngfelt, Göteborg University

Title: Optimal control in Swedish relative infinitives

Abstract:

Most studies on control focus on verb complements and/or adverbial adjuncts. In these constructions, the implicit subject of a non-finite clause - PRO - is usually interpreted in accordance with c-command relations and lexical properties of the verbal heads. Adnominals, however, behave somewhat differently in this respect, since the control relation also depends on the nominal head. This seems to be the case regardless of whether the adnominal infinitive functions as an argument (1a), an apposition (1b) or a relative (1c):

(1) a. kunde vara tillräcklig orsak att PRO utebli från arbetet
could be sufficient cause to PRO stay-away from work
'could be a good enough excuse not to PRO appear at work'

b. för att PRO undgå det skandalösa ödet att PRO bli ogift mor
'to avoid the scandalous fate of PRO becoming an unmarried mother'

c. Han har ju inget arbete att PRO stiga upp till
'He has (you-know) no work to PRO get up to'

As shown in (1a-c), PRO in adnominal infinitives are typically controlled by the possessor (in a broad sense) of their head nouns - or, put differently, the referent that would correspond to a genitival determiner. That is not always the case, though. On closer inspection it shows that the three types of adnominals exemplified above display slightly different control patterns.

The argument type fits nicely into a regular analysis of complement control. Indeed, in many such cases the heads are deverbal nouns and behave exactly like their corresponding verbs, as in (2). Appositions, on their hand, behave essentially like predicative complements (compare (1b) with (3) below).

(2) a. Jag var belåten med mej själv för min halva vägran att PRO göra mig vacker
'I was quite happy with myself, due to my partial refusal to PRO make myself pretty'

b. ...för att jag hade vägrat att PRO göra mig vacker
' ... for having refused to PRO make myself pretty'

(3) Hennes öde var att PRO bli ogift mor
'Her fate was to PRO become an unmarried mother'

Because of these parallells, any analysis that makes the right predictions about verb complements should also be able to cover these two types of adnominals. The tricky bit is how to include the relative infinitives. They seem to require something extra to account for why the controller occasionally is identical with the nominal head, as in (4a) - but usually not, as in (4b):

(4) a. Jag är inte rätt man att PRO svara på detta
'I'm not the right man to PRO answer this'

b. Jag är inte rätt man att PRO fråga om detta
'I'm not the right man to PRO ask about this'

During the talk, I will propose an interpretive OT analysis of control, which can be applied to these cases as well as to more well-known types of control constructions.