The standard argument for a movement analysis of nominal argument shift is that placement of the nominal argument in front of the modal clause adverbial leads to a freezing effect. We illustrate this in (37) with the so-called wat-voor split. Example (37a) first shows that the string wat voor een boek can be fronted as a whole and should therefore be considered a phrase; the full string functions as a direct object. This in turn strongly suggests that the split in (37b) is derived by wh-extraction of wat from the wat-voor phrase. The acceptability contrast between the two (b)-examples shows that the wat-voor split requires the remnant of the direct object to follow the modal adverb waarschijnlijkprobably; cf. Den Besten (1985). If the word-order difference between the (b)-examples is indeed due to leftward movement of the direct object across the clause adverbial, the unacceptability of (37b') can be accounted for by appealing to freezing: the wh-element wat has been extracted from a moved phrase.

Den Besten (1985) also claims that the wat-voor split is categorically excluded for subjects of transitive verbs, but Reuland (1985), Broekhuis (1987/1992), De Hoop (1992), Neeleman (1994a), and De Hoop & Kosmeijer (1995) have shown that the split is possible when the subject is not in the regular subject position, but occupies a position further to the right; this can be seen from the fact that the split is possible when the regular subject position in (38b) is occupied by the expletive er, but not when the expletive is absent.

This suggests that the subject is moved into the regular subject position from a more deeply embedded (more rightward) base position in the clause. This is consistent with the fact that in current generative grammar it is generally assumed that this base position is the specifier of the light verb v, as indicated in (39); cf. the introduction to this chapter.

Example (33b) has already shown that subject shift can cross the modal clause adverbial waarschijnlijkprobably, which was an important reason to assume that such adverbials demarcate the left boundary of the lexical domain of the clause (which is now taken to be the vP). If so, the movement of the object, as in (33a), also targets a position in the functional domain of the clause. A currently more or less standard assumption is that both types of nominal argument shift are motivated by case assignment. This means that the subject and the object (optionally) move into the specifier of some functional head responsible for structural case assignment: T for nominative case and some lower functional head X for accusative case. The many different proposals concerning the nature of X need not concern us here; we will therefore not elaborate on what X is, and refer the reader to Broekhuis (2008: §3.1) for an overview (including proposals that dispense with the category X altogether).

40

Flexible movement approach

For completeness, we conclude this discussion of the wat-voor split by noting that there may be reason to doubt that freezing should really be held responsible for the unacceptability of example (37b') and example (38b) without the expletive er. A viable alternative account might appeal to the fact that interrogative wat-voor phrases are non-D-linked, and that this may simply block object/subject shift; that nominal argument shift of wat-voor phrases may indeed be impossible is strongly suggested by the sharp contrast in acceptability between the two multiple wh-questions in (41).

Nevertheless, the absence of a freezing effect in (37b) and (38b) with the expletive er still supports the claim that remnants of wat-voor phrases should be located within the lexical domain of the clause, and thus also the claim that the subject and object are base-generated within vP.

Another classic argument for a movement analysis of nominal argument shift concerns VP-topicalization; cf. De Haan (1979) and Webelhuth & Den Besten (1987/1990). Since nominal argument shift is optional, the analysis in (40) correctly predicts that VP-topicalization can either pied-pipe or strand the direct object.

The analysis in (40) also accounts for the fact, illustrated in (43), that VP-topicalization cannot strand the object in a position following the clause adverbial, because there is simply no landing site for the object there; note that Section 13.3.2 will return to the fact that (43) is acceptable if the object is contrastively accented.

Note that the acceptability contrast between of (42b') and (43) is a problem for the flexible modification approach in (35), repeated here as (44), according to which the object is obligatorily moved into its case position. This analysis overgenerates by allowing us to derive both (42b') and (43) by VP-topicalization: the former can be derived from (44b) and the latter from (44a).

The contrast in acceptability between (42b') and (43) also poses a serious problem for the flexible base-generation approach in (45) because it is usually claimed that topicalization is only possible with maximal projections; if so, (42b') and (43) are both predicted to be ungrammatical, because they can only be derived by moving the verbal head in isolation. If we were to allow topicalization of the verbal head, there would still be a problem, because this would incorrectly predict both (42b') and (43) to be acceptable, since there would then be no obvious reason to assume that (43) cannot be derived from (45a) by V-topicalization.

The flexible movement approach in (40) can also easily account for the fact that it is not possible to pied-pipe clause adverbials by pointing to the fact that these are not included in the lexical projection of the verb (i.e. vP); cf. Section 8.4.

The flexible modification approach cannot account for the unacceptability of the examples in (46). The reason is that this approach can only account for the acceptability of the examples in (42b&b') by assuming that VP-topicalization moves XP in (44a) and VP in (44v). Consequently, it should be possible to derive example (46a) from (44a) by topicalizing the higher segment of XP, example (46b) from (44b) by topicalizing XP, and (46c) from (44b) by topicalizing the higher segment of VP. Even if we assume that only the lower segments of XP and VP can be topicalized, the unacceptability of example (46b) would still remain a problem, since it can be derived by topicalizing the XP in (44b). Similar problems arise for the flexible base-generation approach, since it should be possible to derive examples (46a) and (46b) from (45a) and (45b), respectively, by topicalizing the higher segments of V', and (46c) from (45b) by topicalizing the lower segment of V'. Even if we assume that only the lower segments of V' can be topicalized, an option that should be allowed in order to derive example (42b) from (45a), the unacceptability of (46c) would remain a problem. We conclude that the flexible modification and flexible base-generation approaches can account for the unacceptability of the examples in (46) only by appealing to ad hoc restrictions on what can and cannot be topicalized; the VP-topicalization data thus favors the flexible movement approach.

Example (57a) again shows that definite subjects and objects can be placed to the right of modal clause adverbials such as waarschijnlijk if they are part of the focus of the clause, i.e. the discourse-new information. The examples in (57b&c) show that both the subject and the object can move to the left of the clause adverbial if they are part of the presupposition of the clause. The effect of the word-order preservation constraint on nominal argument shift in (56) is illustrated by (57d); a presuppositional object cannot shift across the subject if the latter belongs to the focus of the clause and thus has to follow the modal adverb. This means that example (57a) is information-structurally ambiguous in that the direct object can either be part of the focus or part of the presupposition of the clause; in the second case, the discourse-old object would occupy a position within the lexical domain of the clause, in violation of constraint (54b).

The results are different if we replace the direct object dit boekthis book with the pronoun hetit. Example (58a) first shows that the object pronoun differs from non-pronominal objects in that it cannot remain within the lexical domain of the clause if the subject is part of the focus of the clause. Example (58b) shows that it behaves like non-pronominal objects in that it cannot cross the subject, but (58c) shows that it differs from non-pronominal objects in that it can push the subject up into the regular subject position of the clause. This means that the subject in (58c) can be interpreted not only as part of the presupposition of the clause, but also as referring to discourse-new information, in violation of constraint (54a); the latter is clear from the fact that this example can be used as an answer to the question Wie hebben het boek gelezen?Who have read the book?.

Note in passing that examples such as (58a) become perfectly acceptable when the subject is given contrastive stress; this shows that in such cases the subject can block object shift of the pronominal object in violation of the information-structural constraint (54b); we refer the reader to Section 13.3 for a discussion of the placement of contrastively focused phrases.

We find a similar pattern with indefinite subjects. However, the situation is complicated by the fact that the subject can receive a non-specific interpretation (i.e. unknown to the speaker and the hearer) or a specific interpretation (i.e. known to the speaker but unknown to the hearer). For example, the actual interpretation of the indefinite subject twee jongens in (60) depends on its placement relative to the modal clause adverbial: if it follows waarschijnlijkprobably, it is preferably interpreted as non-specific, whereas it must be interpreted as specific if it precedes it (cf. also Section N21.1.2, sub I).

The result changes again if we replace the direct object het boekthe book with the pronoun hetit. Placing the indefinite subject after the clause adverbial, as in (61a), again requires that the subject be assigned a contrastive accent; in the case of a neutral intonation pattern, the pronoun pushes the subject up into the regular subject position right-adjacent to the complementizer, as in (61b). This is again in violation of the information-structural constraints in (54): that the subject in (61b) can provide discourse-new information violates (54a), and that the contrastively accented subject in (61a) blocks object shift of the pronoun violates (54b).

The data above shows that referential object pronouns such as hetit can force subjects expressing discourse-new information into the regular subject position adjacent to the complementizer, in violation of the information-structural constraint (54a). Object shift of the referential pronoun can also be blocked, in violation of the information-structural constraint (54b), if the subject is assigned contrastive focus accent. This shows that the constraints in (54) are not absolute, but can be overridden to satisfy the “stronger” word-order preservation constraint on nominal argument shift in (56). This suffices to show that there is a complex set of factors that interact (in the technical sense of optimality theory developed in Prince & Smolensky 2004) in determining the surface position of the nominal arguments of the clause.

The discussion of the interaction of object and subject shift in Subsection 1 has shown that the information-structural constraints in (54) can be overridden by the word-order preservation constraint on nominal argument shift in (56). The same can be demonstrated by the interaction of indirect object and direct object shift. Since this is also discussed in detail in Section N21.1.4, sub IE, we will confine ourselves here to a brief review of the relevant data. The examples in (62) show more or less the same as the examples in (57); although both the direct and the indirect object can shift across the modal adverb, the direct object cannot cross the indirect object in its base position.

The examples in (63) show more or less the same as the examples in (58). Example (63a) first shows that the object pronoun differs from non-pronominal direct objects in that it cannot remain within the lexical domain of the clause when the indirect object is part of the focus of the clause. Example (63b) shows that the object pronoun behaves like non-pronominal direct objects in that it cannot cross the indirect object, while (63c) shows that it differs from them in that it is able to push the indirect object up into the functional domain of the clause. Note that, as in the cases discussed in Subsection 1, the judgments only hold under a non-contrastive intonation pattern; the orders in (63a&b) improve if the indirect object is assigned a contrastive focus accent.

The fact that example (63c) is still somewhat marked may be somewhat surprising at first sight; this may be related to the fact that the pronoun can precede the indirect object in (63c), dat hij het zijn moeder waarschijnlijk heeft gegeven, but we postpone discussion of this issue to Section 13.4. The markedness of (63c) may also be related to the fact that it competes with the periphrastic construction dat hij het waarschijnlijk aan zijn moeder heeft gegeventhat he has probably given it to his mother, which does not run afoul of the information-structural constraint (54a). The markedness of (64a) with a contrastively stressed indirect object blocking object shift of the pronoun hetit may have a similar reason: the periphrastic construction in (64b) does not induce a violation of the information-structural constraint (54b) that we see in (64a).

For completeness’ sake, note that Dutch is quite different from German, which allows the direct object to cross the indirect object. This is illustrated in (65) by examples taken from Vikner (1994).

The previous two subsections have argued that the word-order preservation constraint on nominal argument shift in (56) cannot be violated in Dutch, unlike in German. This subsection discusses a putative counterexample to this claim. The problem is illustrated by example (66), which shows that passive ditransitive and dyadic unaccusative constructions do not obey the word-order preservation constraint. Assuming that the orders in the primeless examples (goal/experiencer > theme) are unmarked, we would expect the primed examples to be unacceptable under a neutral, non-contrastive intonation pattern, but both orders seem perfectly acceptable, although some speakers may prefer a periphrastic indirect object to the nominal indirect object in (66a'), an option not available for nom-dat verbs.

However, the primed examples seem to impose specific constraints on the placement of modal clause adverbials such as waarschijnlijkprobably under a neutral intonation pattern: the number signs are used to indicate that the indirect objects may follow the adverbial only if they are assigned a contrastive accent. Since Section 13.3 will show that contrastively accented phrases are at least sometimes external to the lexical domain of the clause, it seems reasonable to conclude that the word-order preservation constraint is valid only insofar as it prohibits nominal argument shift across another nominal argument that remains within the lexical domain of the clause; for independent evidence in support of this claim, see the discussion of the interaction between nominal argument shift and wh-movement in Section N21.1.4, sub V.

If we can indeed draw the conclusion that the word-order preservation constraint in (56) should be reinterpreted as a prohibition of nominal argument shift across another nominal argument that remains within the lexical domain of the clause, the “reversed” orders can be derived as shown for examples (67b) in the derivation sketched in (68). First, (68a) gives the base order, which can actually be found when both arguments are part of the new-information focus of the clause. Second, (68b) is the order derived by shifting both nominal arguments to the left, and again this order can actually be found when the nominal arguments are both part of the presupposition of the clause; note that this structure satisfies the word-order preservation constraint because the order between the experiencer and the theme is the same as in (68a). Finally, the “reversed” order of (67b) is derived as in in (68c) by moving the theme argument de tocht into the regular subject position; this shift crosses the indirect experiencer object de meisjes, but this is allowed since it is no longer in the lexical domain.