Verbs selecting lexically Case-marked objects
Icelandic, like Russian and German, also allows for lexically Case-marked objects as well as structurally Case-marked ones. Consider the following examples.
(1)
In (1a) the object is assigned accusative Case according to its configuration, whereas in (1b) the verb hjálpa selects an object in the dative Case.
This distinction between lexical and configurational Case is illustrated in the passive counterparts to (1a–b), given in (2a–b), respectively, where the corresponding accusative object becomes a nominative subject and the corresponding dative object becomes a syntactic subject but remains in the dative.
(2)
(2b) constitutes a different pattern from the corresponding Russian and German examples. The fact that the S-Structure subject barninu occurs in the lexically determined Case dative, rather than nominative, follows from the PLS. The phenomenon of lexical Case in such constructions strongly indicates the presence of a trace in the object position of passive constructions. Otherwise, the selection of lexical Case in these constructions would be simply unexplained since passive predicates do not select lexical properties of subjects at all. Presumably the reason that lexically Case-marked objects may occur in the subject position of corresponding passive constructions is that this position can be selected for lexical Case in regular active sentences. It appears that once lexical Case is selected in a position, lexically Case-marked NPs can be moved to such positions even when there is no direct selection of lexical Case in the position.
As Zaenen, Maling, and Thráinsson 1985 showed, the evidence that the lexically Case-marked NP in (2b) is in fact a structural subject is virtually identical to the evidence for lexically Case-marked subjects in active sentences reviewed above. The lexically Case-marked NP may be the antecedent of a reflexive in subject-oriented binding dialects.
(3)
Also, in yes/no questions, the lexically Case-marked NP behaves like a subject in that it immediately follows the finite verb.
(4)
And, finally, it can occur as the surface subject of a “seem”-type verb while being interpreted as the subject argument of the embedded infinitival.
(5)
That is, (5) is a standard case of subject-to-subject raising.
In Icelandic, as in German, it is possible for a lexically Case-marked object to remain in its D-Structure position at S-Structure in the presence of passive morphology. This is illustrated clearly in (6).
(6)
In (6) barni (DAT) appears in object position at S-Structure, as in the corresponding active.16 (6) demonstrates that passive morphology in Icelandic blocks neither the assignment nor the licensing of lexical Case with respect to object position. Thus, the movement of a lexically Case-marked object to subject position, as in (2b), is not forced by any Case Filter effect—again in contrast to its configurational Case counterpart.
Further evidence that passive predicates may license the occurrence of lexical Case in object positions comes from the analysis of ditransitive verbs, where the Case marking on one or both objects may be lexically selected. With respect to passivization possibilities there are two distinct classes of ditransitive verbs in Icelandic: the gefa-class, where the unmarked order of objects in active sentences is DAT-ACC, and the skila-class, representing all other occurring Case patterns. In both classes the±recipient NP precedes the theme NP. The classes may be summarized schematically as in Tables 1 and 2.
With verbs of class I either object may move to subject position in the passive. This is illustrated in (7a–c).
(7)
In (7b–c) either NP object may occur in the subject position, while the other remains in postverbal object position. (7b), like (6), shows that passive morphology blocks neither the assignment nor the licensing of lexical Case on objects. (7c) illustrates how the lexically selected Case marking on the object must be preserved at S-Structure when the object is moved to subject position. The nominative Case marking on the object NP in (7c) cannot be an instance of lexical Case, since nominative alternates with accusative in active sentences. A different pattern obtains for class II ditransitives. In this class, only the ± recipient NP may move to subject position. Compare paradigms (7) and (8).
(8)
Given that passive morphology does not block Case assignment via selection of lexical Case and that an NP that is lexically Case-marked by a passive form may occur in syntactic subject position (as illustrated in (8b)), the ungrammatically of (8c) is surprising. The most salient structural difference between the two passive constructions is that in the S-Structure representation of (8b) the lexically Case-marked trace is adjacent to the V that governs it, whereas in the S-Structure representation of (8c) it is not. We will suppose therefore that there is a general prohibition (9) against a lexically Case marked trace that is not strictly adjacent to a governing lexical head.
(9)
(9) has the flavor of a locality condition; though why this particular condition holds remains to be explained. In any event, the explanation cannot be that the lexical antecedent of the lexically Case-marked Trace is in some sense “too far away,” because it is possible to raise such antecedents out of their clauses, as illustrated in (10a).
(10)
(10) demonstrates that either underlying object NP of a class I ditransitive may be raised
to a higher subject position with the familiar pattern of Case retention by the lexically
Case-marked NP and Case alternation with the structurally Case-marked NP. Note that
(9) would also hold for the intermediate trace in (10a) given that the matrix verb taldir
governs the adjacent trace—which is plausible given that the familiar “exceptional Case
marking” (ECM) constructions exist in Icelandic, as in (11).
(11)
As noted above, we assume that because Icelandic allows lexical Case selection for subjects, it is possible to move lexically Case-marked objects into subject positions under the proper conditions—one of which is the adjacency constraint (9), as (12)–(13) demonstrate.
(12)
(13)
Both (12b) and (13b) violate (9).
Underlying subject NPs that involve lexical Case selection also occur in both ECM and raising constructions. (14) provides three standard examples of lexical Case selection for subjects, one in each of the nonnominative Cases of Icelandic.
(14)
The corresponding ECM constructions are given in (15).
(15)
Raising of the lexical Case selected subject is possible both in “seem”-type constructions and in the passive of ECM constructions. This is illustrated in (16) and (17), respectively.
(16)
(17)
These paradigms demonstrate that lexical Case selected subjects in Icelandic have the same syntactic distribution and behavior as configurationally Casemarked subjects at S-Structure.
