Licensing conditions for SF: comparative data from Scandinavian 4.1 “State of the art”
4.9. The complementizer system of Modern Scandinavian
4.9.2. Embedded Wh clauses
Norwegian and Swedish
The hypothesis that Mainland Scandinavian som has a double merger is confirmed by the structure of embedded Wh-clauses, like indirect questions and headless relatives, also exhibiting a subject/non-subject asymmetry as illustrated in the examples below: (325)a. Han spurte hvilken buss *(som) __ gikk til sentrum (Norwegian) He asked which bus that goes to centre
“He asked which bus goes to the centre” b. Han spurte hvilken buss (*som) vi ville ta __ He asked which bus that we would take “He asked which bus we would take”
c. Det spelar ingen roll vem *(som) __ väckte uppmärktsamhet (Swedish) It plays no role who that aroused attention
“It doesn’t matter who caught people’s attention” d. Det spelar ingen roll vem (?*som) du väckte __ It plays no role who that you woke-up
“It doesn’t matter who you woke up” [Taraldsen 1986; Thráinsson 2007] The idea that som in (325)a. and c. occurs in a lower position than its relative counterpart is confirmed by the fact that it linearly follows the Wh- element in embedded Wh- clauses. In a cartographic structure, the Wh- element checks its [Wh-] feature on WhP by moving to the Spec of a WhP in the Focus field, and reaches the Spec of SubP for clause-typing139 as shown in (326) below:
(326) [SubP vem…WhP <vem> FinP <vem> Fin som [IP…
Danish
The situation is more complicated in Modern Spoken Danish. In Standard Danish, embedded subject Wh-clauses license subject extractions by means of der, instead of som. According to Mikkelsen (2002) der is the very same form of the expletive subject, with the same syntactic properties:
(327) a. Jeg ved ikke, hvem *(der) __ har boet i det hus. (Standard Danish) I know not who that has lived in that house
“I don’t know who has lived in that house”
139 Movement from Spec,WhP to Spec,SubP does not violate criterial freezing (cf. Rizzi 2004), since the
b. Jeg ved ikke, hvem (*der) Jens har truffet __ I know not, who that Jens has met
“I don’t know who Jens has met”
The fact that an expletive (der) is commonly adopted instead of a complementizer in subject extractions does support the hypothesis that the Danish as well as the Norwegian/Swedish strategy consists of a mechanism checking the subject features on Fin. The Danish strategy uses merger of an expletive140, which Mikkelsen claims to be in
Spec,IP. A tentative structure for Danish subject Wh- clauses is sketched in (328) below: (328) [SubP hvem…WhP <hvem> FinP <hvem> Fin [SubjP der...
However, der is possible also in preverbal position in main clauses, thus, assuming that main V2 clauses are derived by V-to-Fin, the preverbal subject would arguably be in a subject position in CP. Independent evidence has been provided in the literature for the existence of such position (cf. Poletto 2000; Platzack 2009, a.o.), where preverbal expletives and other subjects are merged in main clauses of Germanic languages. From a theoretical standpoint the presence of a subject position in CP is not uncontroversial. Given that subject extractions are constrained by ECP effects, which Rizzi (2004) reformulates in terms of “Subject Criterion”, how is it possible that a subject moves to an A-position in CP –being it Subj2P or FinP, cf. (317) above- without being frozen in Spec,SubjP; i.e. the subject criterial position?
Intuitively, a possibility is that V-to-Fin licenses subject movement to a CP position. Specifically, the verb would satisfy the subject requirement by moving to Fin and checking the uninterpretable phi-features thereby expressed. This mechanism would allow the subject to move directly to a position in CP, by “inactivation” of the freezing properties of SubjP. The extent to which this mechanism enables subject extractions must vary crosslinguically, perhaps depending on the feature specification on the verbal head.
Modern Spoken Danish has other options for introducing relative and embedded Wh- clauses: according to Mikkelsen (2002), more complementizers can cooccur provided that at least som or der are present141. The different possibilities are given in (329)
below.
140 Recall that the Icelandic expletive það is ungrammatical in subject extractions. The contrast between
Danish and Icelandic embedded subject Wh-clauses can thus be attributed to a different feature specification on the two expletive elements. Danish der is a true expletive, in contrast to the quasi- argumental det. Since Icelandic has only the syncretic form það which, according to Falk (1993), is a quasi-argument, overt expletives are not found in extraction contexts.
141 However som is ruled out in some cases where der is possibile, as, for instance, in appositional subject
(329) Jeg ved ikke... (Spoken Danish) I know not
a. hvem som har gjort det. who SOM has done it b. hvem der har gjort det. who THERE has done it c. hvem som der har gjort det. who SOM THERE has done it d. hvem som at der har gjort det. who SOM that THERE has done it e. hvem som at har gjort det.
who SOM that has done it f. hvem at der har gjort det. who that THERE has done it “I don’t know who has done it”
Assuming that the expletive der is in Spec, IP, as argued by Mikkelsen (2002), and that at is the declarative head of Force142, it follows that som, in Danish, ends up in a high
position. Given the diverse options of complementation for subordinate subject Wh- clauses in the spoken language, it is not easy to identify the exact position of som in (328)a., c., d., e. above. One possibility is that som merges in Fin and then moves to the head of SubP, in a similar fashion to the relative som undergoing Fin-to-Sub. In this view, Spoken Danish som would be specific for subject constructions, as well as som found in Swedish and Norwegian embedded Wh-clauses (cf. (325) above). However, this hypothesis is problematic for the following reasons. First of all, the locality of head movement would be violated unless we assume that som incorporates into the head of Force, at, on its way to SubP. Secondly, it is not clear why such low occurrence of som must move to Sub and cannot remain in Fin, instead, given that the clause is typed by the Wh- element. In fact, this linear order would be ungrammatical:
(330) *hvem at som der har gjort det. who SOM that THERE has done it “Who has done it”
The puzzling facts in (328)-(329) above could be reinterpreted in terms of multiple available strategies of complementation. Basically, Spoken Danish would have the following possibilities for extracting subjects in embedded Wh- clauses:
1. Der-insertion
2. Som-insertion (like Swedish and Norwegian, cf. (325) above)
The facts are complicated by the possibility of spelling out som in a high position, in addition to the low occurrence indicated in strategy 2. above. Basically, a relative-type of som is possible not only as a complementizer to relative clauses (cf. (319) and figure 10. above) but also in subject embedded Wh- clauses. Thus strategies 1. and 2. above are not mutually exclusive in Spoken Danish. According to Mikkelsen (2002), Danish
142 The complementizer forms at/att/að are assumed to merge in Force, not in Fin (contrary to German
som is an invariant operator, given its unspecialized distribution both in subject and non- subject extractions. However, the fact that som can substitute der in subject extractions does not exclude the possibility that also a specialized use of som is enforced in Danish, as well as in Swedish and Norwegian. These two conditions provide support to the hypothesis that the same complementizer form may be spelled out in different positions. The degree of acceptability of the options presented in (329) is subject to variations depending on “EPP effects”, according to Mikkelsen (2002) (cf. (331) below). In other words, it seems that the productivity of different complementizer forms and combinations is subject to variations related to the syntactic context. Consider, for instance, the different options enabling subject relativizaton:
(331) a. Vi kender de lingvister der vil læse denne bog b. Vi kender de lingvister som vil læse denne bog c. ?Vi kender de lingvister som der vil læse denne bog d. ?Vi kender de lingvister at der vil læse denne bog e. ? Vi kender de lingvister som at der vil læse denne bog f. ??Vi kender de lingvister som at vil læse denne bog We know the linguists that will read this book
“We know the linguists that will read this book” [Mikkelsen 2002, 80, 30] Notice that the most degraded option is (331)f.: interestingly, som at is worse than som at der. Following the present proposal, the lower acceptability of som at is expected because in such a construction the head at intervenes in Fin-to-Sub movement of som. In (331)e., instead, unspecialized som could directly merge in Sub, and subject relativization be enabled by der, locally c-commanded by Fin. By contrast, the only possibility for checking Fin- and clause-type features in (331)f. is to move a subject- specific lower som (analogous to the Swedish/Norwegian one) to Sub. Notice that this option is fine if no intervener is present, as in (331)b., but bad if at is merged in Force, in (331)f. Further support to this analysis comes from crosslinguistic evidence. According to my Swedish and Norwegian informants143 the presence of at/att linearly
following som to introduce a relative clause is ungrammatical144. The structural
143Johan Brandtler; Björn Lundquist; Johanna Prytz for Swedish; Pål Kristian Eriksen; Odd Egil Johnsen
for Norwegian.
144One case when Swedish att can follow som is in comparatives and other kind of subordinate clauses,
whose complementation structure is different from that of a subject relative. The comparative/subordinating som is of the higher type, given its unspecificity, i.e. given that it doesn´t need to “fill in” a subject gap. See for instance an “as if” clause:
(i) Det är som att han inte vill veta! (Swedish)
it is as that he not want know
"It's as if he does not want to know!" [Johan Brandtler, p.c.] An analogous possibility is found in Norwegian, where som at introduces complements to the impersonal predicate det virker (it seems, lit. it works). Notice that in this case the complement clause can also be introduced by som om (lit. as if), with a different interpretation: som at complements are more certain; whereas som om complements are more uncertain to the speaker´s knowledge. Given the interpretive properties of om/at we can assume that they are related to the illocutionary force of the sentence (ForceP), which entails that som, in those cases, is directly merged in the high, SubP position:
(ii)a. Det virker som at hun skal ikke komme (Norwegian) It works as that she shall not come
representation of the sentences in (329) is schematized in (332) below, and can be adopted for subject relatives as well, with the appropriate modifications:
(332) [SubP hvem Sub (som) Force (at) WhP <hvem> FinP <hvem> Fin (<som>) [SubjP der...
It is worth observing that the subject/non-subject asymmetry must depend on merger of a specific element licensing subject extractions, i.e. lower som for Norwegian and Swedish, or der, for Danish. This function is not carried out by a generic declarative complementizer like at.
Faroese and Icelandic
In Faroese, embedded Wh- clauses may be introduced by multiple complementizers, but no subject/non-subject asymmetry is attested:
(333) a. Eg veit ikki hvør (ið) ____ kemur (Faroese) I know not who.NOM that comes
“I don’t know who will come”
b. Eg veit ikki hvønn (ið) hann hevur sæð ____ I know not who.ACC that he has seen
“I don’t know who he has seen” [Thráinsson 2007, 449, 8.134]
The complementizer ið is declarative and analogous to Mainland Scandinavian at/att and Icelandic að. The declarative complementizer seems to have a different function from Mainland Scandinavian som. Indeed Thráinsson et al. (2004) observe that
“the relative complementizer sum is not used after an interrogative pronoun in Faroese” (p. 449, fn. 32),
confirming the intuition that the complementation system of Icelandic and Faroese differs from the Mainland Scandinavian one.
It has been argued that Old Icelandic has only an unspecific relative complementizer, spelled out in a high position, namely SubP. The same situation is reflected in Modern Icelandic, which lacks a lower, subject-specific copy of sem spelled out in Fin in cases of subject extractions. This is visible in the embedded Wh- clauses in (334) below.
(334)a. Ég veit ekki hver (*sem) __ kemur (Icelandic) I know not who.NOM that comes
“I don’t know who will come”
b. Ég veit ekki hvern (*sem) han hefur séð __ I know not who.ACC that he has seen “I don’t know who he has seen”
b. Det virker som om hun skal ikke komme It works as if she shall not come
Contrary to Swedish and Norwegian som, Icelandic sem cannot be present in embedded subject-Wh- clauses, as shown in (334)a. Moreover, (334)b. together with the analogous Swedish and Norwegian cases in (325)b.and c. above, repeated below for convenience, contrasts with the Faroese facts in (333)b.
(335)a. Han spurte hvilken buss (*som) vi ville ta __ (Norwegian) He asked which bus that we would take
“He asked which bus we would take”
b. Det spelar ingen roll vem (?*som) du väckte __ (Swedish) It plays no role who that you woke-up
“It doesn’t matter who you woke up”
The subject/non-subject asymmetry characterizing the distribution of som indicates its subject-specific nature, in contrast to the non-specific character of declarative complementizers, such as Faroese ið.
Further support to this hypothesis comes from the fact that in Icelandic many types of subordinate clauses are optionally introduced by a complementizer followed by the declarative head að. This is the case for indirect yes/no questions; conditionals; adverbial and also relative clauses, suggesting that sem is spelled out in a position higher than að, i.e. than Force:
(336) þetta er bókin sem (að) ég keypti (Icelandic)
this is book.the that that I bought
“This is the book that I bought” [Thráinsson (2007), 450, 8.135] Recall that relative clauses cannot be introduced by som at/att either in Norwegian or in Swedish (cf. p. 184, and fn. 144).