According to clause (i) of the definition of c-command in (36c), direct objects c-command prepositional objects and adjuncts, but not subjects. We therefore expect direct objects to be able to bind an anaphor if it is a prepositional object or an adjunct, but not if it is a subject. Section 23.1, sub I, has already shown that this prediction is correct in the examples repeated here as (43).

The use of a reflexive pronoun in (43b&c) may be somewhat marked; this is not due to a violation of any grammatical principle, but is related to our knowledge of the world: for instance, example (43c) with zichzelf simply refers to a highly unlikely situation. That the markedness of such examples is non-grammatical can be easily demonstrated by making clear that we are intentionally referring to such an unlikely situation: compare the odd example in (43b) with the structurally parallel but perfectly acceptable embedded clause in (44).

Clause (ii) of the definition of c-command makes the additional prediction that acceptability judgments are not affected when the anaphor is embedded in the relevant parts of speech. This is illustrated by (45) for the reciprocal elkaar used as the possessor of a noun phrase.

Since everything that is c-commanded by the direct object is also c-commanded by the subject, we would expect ambiguities to arise when both the subject and the object satisfy the agreement requirements imposed by the anaphor. The examples in (46) show the correctness of this expectation, since they allow either the subject wij or the object de meisjes to act as the antecedent of the anaphor.

The examples in (43) and (45) show that the predictions that follow from binding condition A with the c-command constraint on binding are correct. Note, however, that there are possible counterexamples to the claim that a direct object cannot bind an anaphor embedded in a subject; we will discuss these examples in Subsection E.

The predictions made for indirect object antecedents are similar to those made for direct object antecedents: indirect objects should be able to bind an anaphor when it is (part of) a prepositional object or an adjunct, but not when it is (part of) a subject. These predictions seem to be correct on the whole, although plausible acceptable examples for zichzelf are somewhat harder to construct; in fact, there are no examples with a PP-complement at all, for the independent reason that indirect objects and PP-complements are mutually exclusive. This leaves us with the cases in (47).

Again, the c-command constraint on binding predicts that ambiguity may arise if the subject (or the direct object, for that matter) also satisfies the requirements of the anaphor. Consider the following examples in which either the subject or the indirect object can act as the antecedent of the anaphor: the option of selecting the subject as the antecedent of the reciprocal triggers an interpretation in which the referent of the indirect object de mannen functions as an intermediary in the transfer of the toys; when the indirect object is selected as the antecedent, its referent is simply interpreted as the recipient.

The examples in (47) show that the predictions that follow from binding condition A in tandem with the c-command constraint on binding are on the right track. There are, however, counterexamples to the claim that indirect objects cannot bind an anaphor embedded in a subject; cf. Subsection E.

Direct and indirect objects are not ranked in the GF-hierarchy in (36b). The reason for not doing so is that it is not immediately clear how they should be ordered with respect to each other. Consider the examples in (49). Example (49a) again illustrates that prepositional indirect objects can be bound by direct objects: the direct object de gasten c-commands the prepositional indirect object and is therefore able to bind it. The corresponding double object construction in (49b) is less felicitous, but acceptable; inverting the anaphor and its antecedent, as in (49b'), leads to a highly degraded result.

The neutral order of nominal objects is that the indirect object precedes the direct object: it therefore seems likely that the first object in the (b)-examples in (49) is an indirect object. The fact that the first object can bind the second object, but not vice versa, seems to show that indirect objects c-command direct objects; cf. Daalder and Blom (1976). This conclusion is supported by similar data in English; cf. Barrs and Lasnik (1986).

However, the inversion of the two nominal objects is sometimes possible as a marked option in Dutch, so their relative order is not a foolproof argument for concluding that the second object in (49b&c) is a direct object. Now consider the following German data, taken from Webelhuth (1989:§5.6) and Haider (2010:§6.4).

In German, indirect and direct objects are morphologically marked with dative and accusative case, respectively. The examples in (51) therefore clearly show that in German the direct object can act as an antecedent of the indirect object, but not vice versa, i.e. that it is the direct object that c-commands the indirect object. The close relationship between Dutch and German suggests that also in Dutch only the direct object can bind the indirect object in (49b), i.e. that the direct object c-commands the indirect object and not vice versa. If this is the case, it might even provide us with an explanation for the marked status of example (49b), since it would involve the marked word order of the two nominal objects.

Admittedly, the cross-linguistic evidence is not robust, but there is also language-internal evidence for the claim that the direct object c-commands the indirect object in Dutch, based on attributive modification in noun phrases. Note that an attributively used past participle usually modifies a noun corresponding to the theme argument of the main verb from which it is derived; cf. Section A32.2 for a detailed discussion. This implies that the modified noun should correspond to the direct object of a ditransitive verb such as overhandigento hand (over); that this is the only possibility is illustrated by the examples in (52).

If the past participle of the ditransitive verb voorstellento introduce in (49) is used as an attributive modifier, as in (53), the modified noun should also correspond to the direct object of the verbal construction. Since the reciprocal pronoun elkaar can easily be used as a bare (i.e. preposition-less) object of the attributive participle voorgesteld, we have to conclude that it is the indirect and not the direct object in (49b), just like einander in the German examples in (51).

This seems to be pretty conclusive evidence for the claim that Dutch is like German in that it is the direct object that c-commands the indirect object, and not vice versa. It strongly suggests that our initial intuition about the grammatical functions of the nominal arguments in (49) is wrong.

If direct objects precede indirect objects in the GF-hierarchy, we expect direct objects to be able to bind anaphors embedded in an indirect object. The examples in (54a&b) show that this expectation is borne out. However, this also predicts that an indirect object cannot bind an anaphor embedded in a direct object. The German example in (54c) shows that this prediction is clearly wrong, since it is possible alongside (54b); cf. Lee & Santorini (1994), and Den Dikken (1995:§4.6) for an early analysis of this seemingly paradoxical behavior of German.

Since Dutch has no overt case marking, we cannot show that (54a) corresponds to the two German forms in (54b&c), but there is a semantic argument showing that structures comparable to (54c) do occur in Dutch: the animate object de meisjes in (55) is the goal argument and should therefore be analyzed as an indirect object. This would suggest that the indirect objects in (54c) and (55) can c-command the direct object, which contradicts our earlier conclusion.

Similar facts can be found with prepositional indirect objects, although it should be added that the acceptability judgments are much less clear here. Consider the examples in (56), which show that the direct object can bind the prepositional indirect object regardless of their order, even though the orders in the primed examples are marked and require an emphatic focus accent (indicated by small caps) on the prepositional indirect object.

The primeless examples in (57) further show that the prepositional indirect object cannot bind the direct object in the unmarked order. Things change, however, when the order is reversed: if the direct object itself is an anaphor, as in (57a'), the example remains unacceptable, but if the anaphor is embedded in the direct object, as in (57b'), the result improves considerably.

Apart from the fact that focus accent is needed, the pattern in (56) is strikingly similar to what we found in constructions with a bare indirect object: neither the bare nor the prepositional indirect object can bind the direct object, but both can bind an anaphor embedded in the direct object, provided that they precede the direct object. This suggests that cases in which an indirect object binds an anaphor embedded in a direct object are the special cases; we can describe them by the generalization in (58).

This generalization is more widely valid in that it applies to other prepositional objects as well. The (a)-examples in (59) show that the direct object can bind the prepositional object regardless of the order of the antecedent and the anaphor, provided that in (59a') the correct intonation contour is chosen. Example (59b) shows that the prepositional object cannot act as an antecedent of the direct object if it follows it. However, things change again when the order is reversed: (59b') shows that the example remains unacceptable when the direct object is itself an anaphor, but that the result is considerably better (with the correct intonation contour) when the anaphor is embedded in the direct object. This is consistent with the predictions of generalization (58).

To our knowledge, the descriptive generalization in (58) has not yet been proposed in the domain of binding, but a similar generalization is known under by the names of weak and strong crossover, which accounts for the contrast between the two interrogative clauses in the primed examples in (60).

The (a)-examples show that wh-movement across the pronoun zij does not affect the interpretation: the pronoun zij and the noun phrase dat/welk meisjethat/which girl must be disjoint in reference. The (b)-examples, on the other hand, show that wh-movement across the noun phrase haar oudersher parents can affect the interpretation; while haar and the noun phrase dat meisje in (60b) must be disjoint in reference, the pronoun haar can be referentially dependent on the wh-phrase welk meisje in (60b'). This suggests that generalization (58) may be a more general side effect of A'-movement, since it includes both wh-movement and focus-movement, which are used to derive the marked orders in (56), (57), and (59).

This subsection has shown that the GF-hierarchy in (36b) correctly predicts that nominal arguments can be the antecedent of anaphors in a PP-complement or an adjunct. We have not indicated the relative positions of the two types of nominal object on the hierarchy. The reason for this is that although the initial evidence suggests that nominal indirect objects are higher in the hierarchy than direct objects, there is also evidence that supports the opposite view. We have further argued that the initial evidence does not bear on the c-command relation, but on an interpretive effect related to A'-movement. If this is correct, the data discussed suggests that direct objects are higher in the GF-hierarchy than indirect objects, although it seems fair to say that the issue is far from settled.

For the purpose of the present discussion we will use the notion of PP-complement as a cover term for prepositional indirect objects and other PPs selected by the verb. Prepositional indirect objects illustrate the case most easily: the primeless examples in (61) show that they cannot be antecedents of anaphors functioning as subject or direct object (we saw earlier that the opposite is possible). The primed examples show that they cannot be antecedents of anaphors embedded in a subject or direct object either.

The second case can only be illustrated for subjects; the reason is that nominal and prepositional objects rarely co-occur, so that plausible examples are hard to find. Example (62a) shows that PP-complements selected by the verb cannot be the antecedent of an anaphor functioning as subject (while we saw earlier that the opposite is possible). Example (62b) shows that they cannot be the antecedent of anaphors embedded in a subject either.

The above examples support the claim that subjects and nominal objects are higher in the GF-hierarchy than PP-complements. Recall from subsection B, however, that examples such as (62b) improve when the PP-complement precedes the direct object in accordance with the descriptive generalization in (58).

Subsections B and C have already shown that subjects and nominal objects are able to bind anaphors functioning as adjuncts. Swapping places of the antecedent and the adjunct containing the anaphor leads to unacceptability, as predicted by the GF-hierarchy.

That A-movement feeds certain kinds of binding relations is often illustrated by passivization. On the traditional assumption that indirect objects c-command direct objects, the acceptability of example (65b) would show that passivization reverses the binding relation between the two arguments: while the experiencer binds the accusative theme in (65a), the nominative theme binds the experiencer in (65b).

Similarly, examples such as those in (66) would then show that passivization can bleed binding when the anaphor is embedded in a larger noun phrase: while the indirect object can bind the reciprocal in the active construction in (66a), where elkaars adres functions as the direct object, it cannot bind the reciprocal in the passive construction in (66b), where it functions as the subject.

However, Subsection C has shown that the claim that the indirect object is higher on the GF-hierarchy than the direct object (IO > DO) is probably incorrect: rather, it is the direct object that outranks the indirect object. This means that the arguments for the claim that A-movement affects anaphor binding collapse, since passivization does not affect the relative order of the experiencer and the theme argument on the GF-hierarchy: the examples in (65) are therefore irrelevant for this claim. Moreover, the bleeding effect in (66) cannot be attributed to passivization either, but involves the word-order effect described by generalization (58). This is also shown by the fact that (67), in which the derived subject follows the indirect object, is considerably better than (66b).

The claim that passivization has no impact on the relative ranking of clausal constituents in the GF-hierarchy does not mean that it cannot affect anaphor binding at all. Consider the examples in (68). The active sentence in (68a) first shows again that a subject (agent) can bind a direct object (theme). The two (b)-examples show that passivization reverses the binding relation between the agent and the theme. The theme de jongens has become the subject of the clause and therefore c-commands the agent elkaar, which is now part of an adjunct-PP headed by doorby; this correctly predicts that the theme can bind the agent in the passive door-phrase in (68b), while the agent cannot bind the theme in (68b'), regardless of the word order of the clause.

This finding is important because it confirms our earlier claim that the c-command constraint on binding can largely be captured by appealing to the notion of grammatical function; appealing to the notion of thematic role, for example, would not have yielded the desired results.

A-movement normally takes an NP that has not yet been assigned case and moves it into a case position; this movement is optional in Dutch, as already shown by the fact that derived subjects can follow indirect objects in passive constructions such as (67). Passivization is only one specific case of A-movement, which takes a theme argument of the verb and places it in the canonical subject position of the clause. This section discusses another instantiation of A-movement, which places the subject of certain infinitival clauses into the canonical subject position of a tense-bearing clause; cf. Section V5.2.2.2 for a detailed discussion. An illustration of subject raising in English is given in the examples in (69) with the verb seem: example (69a) shows that when seem takes a finite object clause, its subject position is filled by the anticipatory pronoun it; example (69b) shows that when seem takes an infinitival object clause, the subject of the embedded infinitival clause appears as the subject of the tense-bearing main clause, which is assumed to be the result of A-movement (hence the trace). Finally, the primed examples show that subject raising can feed anaphor binding: the noun phrase the girls can only function as the antecedent of the reciprocal experiencer when it is raised to the subject position of the main clause.

We can find similar Dutch examples, which differ only in that the counterpart of the English to-phrase appears as a dative phrase. This is illustrated in (70a&b) for the Dutch renderings of the primed examples in (69); these examples thus show that A-movement can also feed anaphor binding in Dutch. Note that example (70b) is slightly marked, but alternates with (70b'), in which the A-moved noun phrase does not act as an argument of an infinitival clause, but of a complementive adjective, which is assumed to be the head of a so-called small clause. For completeness’ sake, it should be noted that Everaert (1986:92) claims that complex reflexives cannot be used as experiencers in subject-raising constructions, and it seems that substituting zichzelf for elkaar does indeed lead to a somewhat marked result; it is not a priori clear whether this acceptability contrast should be considered syntactic in nature or whether some other kind of explanation should be invoked.

The crucial point for now is that the noun phrase de meisjes does not c-command the reciprocal in the main clause from its base position within the infinitival or small clause, but that, as in English, it is nevertheless able to bind the reciprocal when it is moved into the subject position of the main clause by subject raising. This shows that A-movement can feed anaphor binding. Finally, note that A-movement is also optional in the Dutch subject-raising construction; the derived subject of the main clause in (71a), de meisjesthe girls, can either be moved into the canonical subject position of the main clause preceding the object pronoun mij or remain in its original position following the pronoun. The contrast between the (b)-examples on a neutral intonation pattern shows that A-movement must apply to make anaphor binding possible; although assigning contrastive accent to elkaar seems to improve example (71b') slightly, it is still quite bad.

Note that (71) shows that the formulation of the c-command constraint on binding in terms of the GF-hierarchy is less adequate than a formulation in structural terms related to the hierarchical organization of the clause; while the former cannot distinguish the two (b)-examples because they do not differ in terms of grammatical function, the latter can because they differ in the structural position of the subject de meisjesthe girls: its base position or its derived position where it is assigned nominative case.

The previous two subsections have shown that (in Dutch) there are at least two A-positions for the subject: its base-generated position and a position where it is assigned nominative case, which we referred to earlier as the canonical subject position. With the advent of the minimalist version of generative grammar in the early 1990s a similar position is distinguished for direct objects: the theme argument is base-generated within the immediate projection of the verb and is moved into a higher accusative case position at a later stage of the derivation; cf. Vanden Wyngaerd (1988), Chomsky (1991/1995a), Zwart (1993/1997), and especially the papers collected in Lasnik (1999). In the literature on Dutch, the availability of these two positions has been exploited to account for object scrambling by assuming that it involves optional A-movement of the object into its case position; cf. Vanden Wyngaerd (1989), De Hoop (1992), Schaeffer (2000), and Broekhuis (2008). The hypothesis that object scrambling is an A-movement operation would predict that it feeds anaphor binding.

Consider the two examples in (72), which differ in that scrambling of the object de meisjesthe girls in (72b) is only possible if its referent is known from the context or the preceding discourse; cf. Section 22.1.3 for further discussion.

The hypothesis that object scrambling is an A-movement operator predicts that binding of an anaphor embedded in the temporal adjunct is possible only if the object precedes it. The examples in (73) show that this prediction is indeed correct: (73a) is unacceptable under a neutral intonation pattern comparable to that in (72a), while its scrambling counterpart in (73b) is perfectly acceptable.

Note that the examples in (73) show once again that formulating the c-command constraint on binding in terms of the GF-hierarchy is less adequate than formulating it in structural terms related to the hierarchical organization of the clause: the former approach cannot distinguish the two cases in (73) because they do not differ in terms of grammatical function, while the latter can because the two cases differ in the structural position of the object de meisjesthe girls: its base position or its derived position in which it is assigned accusative case.