That PP-complements can occupy different surface positions in the clause is illustrated by the examples in (81), taken from Neeleman (1994a).

That the difference in placement is the result of movement is supported by the fact, illustrated in (82), that R-extraction from the PP is possible only when the stranded preposition follows the clause adverbial (in this case nauwelijkshardly); if the (b)-examples in (81) and (82) are derived from the (a)-examples by leftward movement of the PP, the impossibility of R-extraction can be explained by appealing to the freezing effect. Note that we have added the time adverb toenthen in (82) in order to make the split of the pronominal PP daaropto that visible.

An important reason for assuming that the movement deriving the order in (81b) is different from nominal argument shift has to do with the distribution of PPs containing a definite pronoun. Subsection II has already mentioned that definite subject/object pronouns must normally undergo nominal argument shift: example (83a) would only be acceptable if the pronoun hem is given a contrastive accent: Jan nodigt waarschijnlijk hem uit (niet haar) Jan will probably invite him (not her).

The examples in (84) show that this is not the case for PP-complements: if the nominal part of the PP is a definite pronoun, leftward movement of the PP is optional, and even excluded if the pronoun is phonetically reduced. This shows that the division between discourse-old and discourse-new information has no (or a very different) effect on the leftward movement of PP-complements.

The unacceptability of the reduced pronouns in the primed examples is especially remarkable in view of the fact that nominal argument shift typically has the effect of deaccenting the moved element; cf. also Ruys (2008). In fact, some speakers report that they accept examples such as (81b) only when the nominal complement of the PP is contrastively stressed. That the moved PPs must be stressed can be seen from the fact that the pronouns in the primed examples of (84) differ from the shifted pronoun in (83b) in that they cannot be phonetically reduced; cf . All in all, this suggests that we are dealing with focus movement, which will be the topic of Subsection C.

Another reason for assuming that the movement in (81b) is different from nominal argument shift is that leftward movement of a complement PP under a neutral (i.e. non-contrastive) intonation pattern is only possible with a limited set of adverbial phrases. If we replace the negative adverbial phrase nauwelijkshardly in (81b) with the adverbial phrase gisterenyesterday, leftward movement of the PP leads to a degraded result (which can only be improved by giving the PP an emphatic or contrastive stress). This is illustrated in (85) for three different PP-complements.

The primed examples in (85) with the adverb gisteren contrast sharply with similar examples with object shift, which can easily cross adverbs such as gisteren: Ik heb <dat boek> gisteren <dat boek> gelezenI read that book yesterday. For completeness, note that some speakers report that the acceptability of the primed examples in (85) improves when gisteren is given an emphatic accent.

Finally, the (a)-examples in (86) show that leftward movement of a PP-complement across an adverbial PP is always blocked, whereas object shift across such an adverbial PP is easily possible. Note that the unacceptability of leftward movement in (86a) cannot be explained by assuming a constraint that prohibits movement of a complement of a certain categorial type across an adverbial phrase of the same categorial type, since such a constraint would incorrectly exclude object shift across the adverbially used noun phrase deze middagthis afternoon in (86b); cf. Verhagen (1986:78).

The above discussion has shown (contrary to Neeleman 1994a and Haeberli 2002) that the leftward movement of PP-complements exhibits a different behavior from nominal argument shift, which in turn suggests that it is a different type of movement. The following subsections will show that there are indeed other types of leftward movement, not restricted to nominal arguments, which can affect the word order in the middle field of the clause.

Haegeman (1995) has argued for West Flemish that negative phrases expressing sentence negation undergo an obligatory leftward movement into the specifier of a functional head Neg; she further claims that this functional head can optionally be expressed morphologically by the negative clitic en: da Valère niemand (en-)kent that Valère does not know anyone. Although standard Dutch does not have this negative clitic, it can be shown that it has the same kind of leftward movement of negative phrases; cf. Klooster (1994). At first sight, the claim that standard Dutch has negation movement may seem surprising, since negative direct objects as well as PP-complements with a negative nominal part are normally left-adjacent to the clause-final verbs.

That standard Dutch has obligatory negation movement becomes clear, however, when we consider somewhat more complex examples. First, consider the examples in (88) with the adjectival complementive tevredencontent/pleased, which takes a PP-complement headed by the preposition overabout. Although example (88a) shows that the PP-complement can either precede or follow the adjective, example (88b) strongly suggests that the A-PP order is the base order: leftward movement of the PP across the adjectival head produces a freezing effect.

Example (89) further shows that the PP-complement moves obligatorily to the left when its nominal part expresses sentence negation. The number sign in (89a) is used to indicate that examples in this order are acceptable, but only with a constituent negation reading: e.g. Jan is tevreden met nietsJan is content with anything does not mean that Jan is not satisfied with anything, but on the contrary that he is satisfied with very little (Haegeman 1995:130-1).

The reason why negation movement is usually not visible in standard Dutch is that the landing site of this movement is a relatively low position in the middle field of the clause and therefore often applies string-vacuously. This will be clear from the fact, illustrated in (90a), that the negative phrase from (89) preferably follows the clause adverbial waarschijnlijkprobably under neutral intonation (the unacceptable order improves somewhat when the negative noun phrase is given a contrastive accent). We have added example (90b) to show that it is no coincidence that the PP-complement of the adjective is moved to this position following waarschijnlijk: the negative adverb nietnot seems to be base-generated in this position.

That we are dealing with an obligatory leftward movement is also supported by the examples in (91); example (91a) shows again that PP-complements can normally either precede or follow the clause-final verb. However, when the nominal part of the PP-complement expresses sentence negation, the PP-complement must precede the verb; this immediately follows if it must undergo leftward negation movement.

This subsection has shown that phrases expressing sentence negation obligatorily move into a designated position to the right of the modal adverbial waarschijnlijkprobably. This shows that there are movement operations that affect the word order of the constituents in the middle field of the clause that are different from nominal argument shift, since the latter movement typically crosses the modal adverb.

The notion of focus used here refers to elements in the clause that are phonetically emphasized by means of accent, i.e. emphatic or contrastive focus. Emphatic focus highlights one of the constituents in the clause, as in (92a). Contrastive focus is normally used to express that a certain predicate applies exclusively to a certain entity or to deny a certain presupposition on the part of the hearer, as in (92b).

Although example (93a) strongly suggests that focused phrases can remain in their base position, while example (93b) shows that they can also occur in clause-initial position.

That focus phrases can occur in clause-initial position is not surprising, since they behave cross-linguistically very much like wh-phrases. In the Gbe languages (e.g. Kwa), both types of phrases must occupy the clause-initial position and are obligatorily marked with the focus particle wε∃, as shown in the examples in (94) taken from Aboh (2004: §7). The same is shown by Hungarian, where interrogative and focused phrases are placed in the same position left-adjacent to the finite verb; cf. É. Kiss (2002: §4) for examples.

Given that focus phrases have a fixed position in languages like Kwa and Hungarian, it may be somewhat puzzling that in standard Dutch focus phrases can also occupy different positions in the middle field of the clause. The examples in (95) illustrate this with the PP-complement of the adjective trotsproud in (93).

The fact that focused phrases can occupy a variety of surface positions in the clause has challenged the standard assumption that there is a unique position to which such phrases can move, and has led to proposals for a more flexible approach; cf. Neeleman & Van de Koot (2008). We will not take a position on this issue here, but simply conclude that the examples in this subsection show that focused phrases can optionally undergo leftward movement within the middle field of the clause.

The term topic is used here quite broadly as aboutness topic; it refers to the entity that the sentence is about. Typical examples are given in (96); they show that aboutness topics are accented and can precede the subject if the latter is focused (which we have forced in (96) by combining the subject with the focus particle alleenonly).

The fact that leftward movement of aboutness topics can change the underlying order of the arguments in the middle field (a property that, according to some, also holds for focus movement) shows that we are again dealing with a different kind of movement than the nominal argument shift discussed in Subsection II, which typically leaves this order intact. This is also shown by the fact, illustrated in (97), that aboutness topics need not be nominal in nature, but can also be PPs or (complementive) APs.