Subject and object extraction from embedded clauses

Many languages display a restriction on extraction of subjects from embedded clauses, in particular when the embedded clause is introduced by an overt complementizer. This phenomenon is called the “that-trace effect,” and it is illustrated with a Swedish example in (1a). However, the literature reports on variation among the Scandinavian languages on this point (cf. among many others Platzack 1986, Holmberg 1986, Hellan and Christensen 1986, and Westergaard et al. 2012 for a recent overview). According to the Swedish Reference Grammar (SAG, Teleman et al. 1999, cf. also Platzack 1986, Holmberg 1986), Swedish indeed displays a that-trace effect, in that the subject cannot be extracted when the embedded clause contains an overt complementizer as in (1a), while such extraction is possible when the complementizer is deleted, as in (1b) (Teleman et al. 1999 vol 4:426). The same pattern holds true for Danish (cf. Vikner 1995) and Faroese (cf. Thráinsson et al. 2004):


Introduction
Many languages display a restriction on extraction of subjects from embedded clauses, in particular when the embedded clause is introduced by an overt complementizer.This phenomenon is called the "that-trace effect," and it is illustrated with a Swedish example in (1a).However, the literature reports on variation among the Scandinavian languages on this point (cf.among many others Platzack 1986, Holmberg 1986, Hellan and Christensen 1986, and Westergaard et al. 2012 for a recent overview).According to the Swedish Reference Grammar (SAG, Teleman et al. 1999, cf. also Platzack 1986, Holmberg 1986), Swedish indeed displays a that-trace effect, in that the subject cannot be extracted when the embedded clause contains an overt complementizer as in (1a), while such extraction is possible when the complementizer is deleted, as in (1b) (Teleman et al. 1999 vol 4:426).The same pattern holds true for Danish (cf.Vikner 1995) and Faroese (cf.Thráinsson et al. 2004): (1) a. * Kalle har dom sagt att inte kommer.(Swedish) Kalle have they said that not comes b.Kalle har dom sagt inte kommer.

Kalle have they said not comes
'They have said that Kalle isn't coming.' In Fenno-Swedish, on the other hand, the that-trace effect is claimed to be absent, and constructions like (2a) are acceptable (Teleman et al. 1999 vol 4: 426).Moreover, for Norwegian, Faarlund et al. (1999:986) claim that, although it is preferred to delete the complementizer when a subject is extracted, leaving it expressed overtly is also an option, as in (2b) (cf.also Hellan and Christensen 1986, Taraldsen 1978, Fiva 1985, 1991, and Westergaard et al. 2012 for a recent overview): (2) a.We know that there is quite a lot of variation across the Scandinavian languages with respect to issues related to subject extraction, and therefore these constructions were tested for in the Nordic Syntax Database.
Extraction of objects from embedded clauses is not known to be restricted by the presence or absence of the complementizer to the same extent, but for completeness, constructions parallel to those with subject extraction were included in the questionnaires.

Nordic Syntax Database (NSD)
The  6b) is also quite generally accepted.In Finland and Sweden, this example receives a high score everywhere.In Norway, however, there are several medium scores throughout the country, and there is also a cluster of low scores in the southern part of the country, see map 2. This is somewhat unexpected and calls for further investigation.Third, it has been noted (Nordgård 1985, 1988, Westergaard et al. 2012) that several Norwegian dialects use relative pronoun som rather than the complementizer at 'that' in constructions when a wh-subject has been extracted.Thus, the example in (9) was included in the questionnaire, and tested in Norwegian and Swedish:

Discussion
Object extraction appears to be possible from that-clauses both with and without an overt complementizer at 'that'.However, certain places, particularly in the south of Norway, display some restrictions on extraction from clauses with the overt complementizer.Moreover, use of the complementizer som 'that' is completely out in these constructions.
Turning to extraction of a wh-subject, we have seen that such extraction is generally accepted from that-clauses without an overt complementizer.In this respect, subject and object wh-extraction pattern alike.Moreover, while in Finland and Sweden, wh-subject extraction out of a clause with an overt complementizer is rejected, we find a complementary distribution with respect to such extraction in Norway.In the Eastern part, such extraction is allowed out of complements with the overt complementizer at 'that' but disallowed out of a complement with the overt complementizer som 'that', while in the rest of Norway, we see the opposite pattern.These results have been reported on by Westergaard et al. (2012) based on the data from the Nordic Syntax Database, and readers are referred to this paper for further reading.
Finally, notice that although resumptive pronouns in constructions with extraction of a wh-subject from a that-clause were generally rejected, resumptive pronouns were accepted with extraction of a nonwh-subject from an if-clause in Finland, Sweden, and large parts of Norway.

'
following constructions with object extraction were tested in Norway, Sweden, and Finland: (6a) extraction from a that-complement without an overt complementizer; (6b) extraction from a thatcomplement with an overt complementizer; (6c) extraction from a that-complement with the complementizer som.This is illustrated here with Norwegian examples: Who do you think I am going to meet tomorrow?' (6a) receives high scores everywhere (except in Närpes in Finland), and appears to be the typically accepted variant of the sentences in (6), see map 1. (

''
Map 1: wh-object extraction from that-clause without a complementizer.(#325: Hvem tror du jeg skal treffe imorgen Map 2: wh-object extraction from that-clause with complementizer at 'that'.(#329: Hvem tror du at jeg skal treffe i morgen?) ('Who do you think I am going to meet tomorrow?')(White = high score; grey = medium score; black = low score) Bentzen Subject and object extraction NALS Journal 438 In contrast, (6c) was basically rejected everywhere: Map 3: wh-object extraction from that-clause with complementizer som 'that'.(#333: Hvem tror du som jeg skal treffe imorgen?'Who do you think I am going to meet tomorrow?')(White = high score; grey = medium score; black = low score) With respect to subject extraction from that-clauses, various types of constructions were tested in the NSD.First of all, the construction that has been reported as generally accepted, with complementizer deletion, was tested in Faroese (7a), Norwegian and Swedish (7b): Who do you think has done it?'As expected, this construction is typically accepted all over Norway, Sweden, Finland, and the Faroe wh-subject extraction from that-clause with complementizer deletion.(#324: Hvem tror du har gjort det? 'Who do you think has done it?/Hvørvaentar tú fer við? 'Who do you think will come along?')(White = high score; grey = medium score; black = low score) Secondly, constructions with an overt complementizer at 'that' were also tested in Faroese, Norwegian and Swedish (here illustrated with Norwegian examples): Who did you say hadn't come?'The results show that a that-trace effect is missing in Finland, as well as in various places in Eastern Norway; speakers in these areas accept examples like those in (8).Moreover, in Trøndelag and a few places in central Sweden, these examples receive a medium score.However, in other locations in Norway and Sweden, as well as on the Faroe Islands, the examples in (8) are rejected: wh-subject extraction from that-clause with overt complementizer at 'that'.(#328: Hvem tror du at har gjort det? 'Who do you think have done it?')(White = high score; grey = medium score; black = low score)

( 9 )'
Hvem tror du som har gjort det? #332 (Norwegian) who think you SOM has done it 'Who do you think has done it?'This sentence received a high score in Northern Norway, Trøndelag, and the North West coast of Norway (Sogn og Fjordane, Møre og Romsdal).However, it was rejected in South and East Norway, wh-subject extraction from that-clause with overt relative pronoun som 'that'.(#332: Hvem tror du som har gjort det? 'Who do you think has done it?')(White = high score; grey = medium score; black = low score) Finally, constructions with a resumptive pronoun in the embedded clause were tested in Norwegian and Swedish; with an overt complementizer, (10a-b), and with complementizer deletion, (10c): the think you that they have robbed bank.the'Which boys do you think have robbed the bank?Who do you think has done it?'All of the sentences in (10) were generally rejected everywhere in Norway, Sweden and Finland: Hvem tror du at han har gjort det? 'Who do you think has done it?')Map 10: wh-subject extraction from that-clause with a resumptive pronoun without an overt complementizer at 'that'.(#1204: Hvem trur du han har gjort det? 'Who do you think has done it?')(White = high score; grey = medium score; black = low score) The availability of a resumptive pronoun was also tested in one type of complement clause with object extraction in Finland and Sweden, (11a), and with subject extraction Norway, Finland and Sweden, (11b-'t know whether Per Erik is coming.'Sentence (11a) with a resumptive pronoun with object extraction was in general not accepted in Finland and Sweden: Map 11: wh-object extraction from complement if-clauses with a resumptive pronoun.(#1399: Den boken kommer jag inte ihåg om jag läste den.'That book I can't remember whether I read').(White = high score; grey = medium score; black = low score) As Map 12 shows, sentence (11b) with a resumptive pronoun with subject extraction was rejected in Southern Norway, a few places in Trøndelag (Selbu, Namdalen, Lierne) and in Finnmark (Kvaenangen, Hammerfest, Lakselv, and Vardø), in Leksand and Sproge in Sweden, and in Brändö in Finland.However, it was accepted in Northwestern Norway (between Bergen and Trondheim) as well as in a few places in Nordland county (Northern Norway).It was also accepted in various places in Sweden, especially in the South as well as in Munsala and Närpes, Dragsfjärd, and Borgå in Finland.It received medium scores in the Troms, Nordland and Trøndelag counties, and also here and there in Sweden and ) without a resumptive pronoun with subject extraction was tested in Sweden and Finland, and as illustrated in Map 13, it was rejected in most places in Sweden except for Särna and Dalby, two locations close to the Norwegian border.In Finland, however, this sentence was typically accepted in the southern part of Finland, as well as in Larsmo, with Munsala, Snappertuna, and Kyrkslätt giving medium scores.This type is interesting since the complentizer om ('if') cannot be deleted.Additional investigation is required on this construction in Norway.Map 12: wh-subject extraction from complement if-clauses with a resumptive pronoun.(#997: Ole Gunnar veit jeg ikke om han kommer.'I don't know whether Ole Gunnar will come.')Map 13: wh-subject extraction from complement if-clauses without a resumptive pronoun.(#1398: Per Erik vet jag inte om kommer.'I don't know whether Per Erik is coming.')(White = high score; grey = medium score; black = low score)