The examples in (170), repeated here as (173), have shown that although the simplex reflexive zich and the weak referential pronoun ʼrher cannot be bound by their co-argument, the logical subject of the prepositional small clause, they can be bound by the subject of the sentence.

From this we can conclude that both must be free in their anaphoric domain, as defined in (174), taken from 23.3, sub II.

According to (174a), the prepositional small clause is the CFC of the preposition bij in (173), since it contains all its arguments; since bij assigns a thematic role and case to its nominal complement (here: the pronouns), according to (174b) the small clause is also the anaphoric domain of its nominal complement. Since (173a) shows that the simplex reflexive pronoun zich cannot be bound by the logical subject of the small clause, it follows that it must be free in its anaphoric domain. This correctly predicts that zich behaves like a referential personal pronoun in being in complementary distribution with complex reflexives and reciprocals. This is illustrated again by the examples in (175).

However, as already shown in (171), zich still patterns with the reflexives/reciprocal pronouns in that it must have an antecedent within its minimal clause; it differs from referential personal pronouns in this respect: these can remain free in their minimal clause and refer to an antecedent in some higher matrix clause, as in (176a), or to a contextually determined antecedent, as in (176b).

The above discussion indicates that an appeal to the notion of anaphoric domain of α as defined in (174) is insufficient to fully describe the differences in the binding behavior of anaphors, referential pronouns, and simplex reflexives in constructions with a small clause. We can remedy this by provisionally assuming that the terms minimal clause of α and sentence containing α also constitute domains relevant for binding theory; this allows us to describe the observations made so far by means of Table (177). Here, bound indicates that α must be bound in the given domain, free that α cannot be bound in the given domain, and optional that α may or may not be bound in the given domain. Note that the values of the shaded cells are predictable from the values to their left: to determine the binding conditions for the various elements, it is therefore sufficient to inspect the cells without shading.

	anaphoric domain	minimal clause	sentence
anaphor (zichzelf/elkaar)	bound	bound	bound
simplex reflexive (zich)	free	bound	bound
referential pronouns (hij/hem)	free	optional	optional

In (169), repeated here as (178), the three anaphoric domains of the pronouns more or less coincide, and it is therefore correctly predicted that only the complex reflexive pronouns can be bound by the subject of the clause, since the referential pronoun haar and the simplex reflexive zich are incorrectly bound within their anaphoric domain.

In (175) above, the anaphoric domain of the pronouns is constituted by the prepositional small clause, while their minimal clause and the sentence coincide: consequently, we predict that the complex reflexive and reciprocal personal pronouns must find an antecedent within the small clause, while the subject of the sentence can act as the antecedent for the simplex reflexive (obligatory) and the referential pronouns (optional). In (176a), the three domains are all different, and we correctly predict that only referential personal pronouns can be bound by an antecedent external to their minimal clause. The pattern in Table (177), based on three binding domains, is still fairly straightforward, but this changes as we consider more cases.

A serious complication is that we would expect the pattern in (175) to arise also in cases with adjectival and nominal small clauses, but the examples in (179) show that this expectation is not fully borne out. The primeless examples show that the logical subject of the small clause can bind the complex reflexive zichzelf, but not the simplex reflexive zich, as expected on the basis of (175a). However, the primed examples show that zich cannot be bound by the subjects of their minimal clauses either, which is surprising given the acceptability of (175b). The fact that the referential weak pronoun ʼmhim can be bound by the subject of the clauses is again in accordance with the predictions.

The fact that the wrong predictions are made for the simplex reflexive pronoun zich shows that relying on the distinction between anaphoric domain and minimal clause is not sufficient to account for its distribution. The easiest way to solve this problem is to assume that zich must not only be free but also bound within adjectival and nominal small clauses, as in Table (180): since these requirements cannot be satisfied simultaneously, we predict that zich cannot appear as the complement of such small clauses at all, while still allowing pronouns to be bound by the subject of the clause.

	anaphoric domain I (a) SC-AP (b) SC-NP (c) SC-PP	anaphoric domain II (a) Minimal clause (b) SC-AP (c) SC-NP	sentence
anaphor (zichzelf/elkaar)	bound	bound	bound
simplex reflexive (zich)	free	bound	bound
referential pronouns (hij/hem)	free	optional	optional

The pattern in Table (180) seems less natural and somewhat stipulative in nature: it at least raises the question of why prepositional small clauses differ from adjectival and nominal small clauses in allowing a simplex reflexive to be bound by an antecedent external to it. There have been attempts to give a more or less principled account of this restriction on the distribution of zich (see Broekhuis 1992 for a discussion), but we will not discuss these here, because we will see that it follows more naturally if we adopt the alternative approach to simplex reflexives outlined in Section 23.4.2, sub III.

Section 23.3, sub IIA, has established that bare infinitival clauses in AcI-constructions are CFCs in the sense that they form a verbal small clause. We therefore expect that by extending Table (180), binding in AcI-constructions can be accounted for in the same way as in Subsection I. We will see that this is the case in one prototypical instance, but that there are also cases that exhibit more ambivalent behavior.

Consider the examples in (181): since the complex infinitival verb schieten op assigns a thematic role and case to its complement α, the infinitival clause functions as the anaphoric domain of α according to the definitions in (174). Example (181a) shows that this correctly predicts that the logical subject of the bare infinitival clause can bind α if it is a complex reflexive, but not if it is a simplex reflexive or a referential personal pronoun; example (181b) shows that this also correctly predicts that the subject of the minimal clause can bind a simplex reflexive (obligatorily) or a referential pronoun (optionally).

The examples in (181) therefore only require the addition of SC-VP (infinitival clause of an AcI-construction) in the column anaphoric domain. This is to be expected and thus does not really add to the complexity of the system.

	anaphoric domain I (a) SC-AP (b) SC-NP (c) SC-PP (d) SC-VP	anaphoric domain II (a) Minimal clause (b) SC-AP (c) SC-NP	sentence
anaphor (zichzelf/elkaar)	bound	bound	bound
simplex reflexive (zich)	free	bound	bound
referential pronouns (hij/hem)	free	optional	optional

However, various problems arise in cases where noun phrase α functions as the direct object of the infinitival verb. We expect that zich can also be bound by the subject of the clause in such cases but the examples in (183) show that, contrary to our expectation, the only possibility is the use of a referential personal pronoun. This casts serious doubt on the feasibility of analyzing the binding behavior of simplex reflexives in terms of binding domains.

The sentences in (184), in which the logical subject of the small clause is left implicit, are fully acceptable with zich as the object of the infinitival verb. This would be consistent with Table (182) if the agent argument of the infinitival verb were syntactically realized as the phonetically empty PRO-subject; the infinitival clause would then be the minimal CFC of zich, which is then correctly free in its anaphoric domain I. If there were no logical subject (indicated by Ø in the examples), the infinitival clause would not be a CFC, so that the minimal CFC of the pronoun would be the finite clause, and zich would be incorrectly bound within its anaphoric domain. The judgments in (184) therefore follow only if PRO is obligatory.

That the phonetically empty subject PRO is obligatory in the examples in (184) is, however, unlikely for two reasons. First, consider the examples in (185) with the referential personal pronoun ʼrher: if PRO were present, we would expect the indicated binding relation to be possible, while in fact it is degraded: the pronoun cannot be coreferential with Marie but refers to some other female person. This follows only if the agent of the infinitival verb is not syntactically realized (which is indicated by Ø); the minimal CFC of the pronoun would then be the finite clause, so that ʼrher cannot be bound by Marie. This leads to the conclusion that PRO cannot be present.

The second problem with assuming the obligatory presence of a PRO-subject is illustrated in (186), which shows that the implicit agent can be expressed overtly by an agentive door-phrase; cf. De Geest (1972:§4.3.2) and Broekhuis (1988:§3.4). The inclusion of PRO in the representation in (186) is therefore ruled out by the same restriction that blocks the redundant use of a passive door-phrase in active sentences such as Jan wekt Marie (*door NPagent) Jan wakes Marie (*by NP). The subjectless infinitival clause in representation (186) is not a CFC, and this correctly predicts that the referential pronoun ʼr cannot be bound by the subject of the sentence, but also incorrectly predicts the same for the simplex reflexive zich.

The discussion above shows that the examples in (184) and (185) lead to a paradox: PRO must be obligatorily present to account for the examples in (184), but cannot be present to account for the examples in (185). Furthermore, the presence of an agentive door-phrase militates against a PRO-subject, which leads to the incorrect prediction that zich cannot be used in this example. Further discussion of these examples is deferred to Section 23.4.2, sub III.

The notion of anaphoric domain is relational in the sense that it is defined only for a noun phrase α in a specific structural position in the sentence: the phrases listed under the headings of table (182) function as local domains for noun phrases in the complement position of a small-clause predicate, but not necessarily for other noun phrases. This will become clear when we consider the binding behavior of logical subjects of small-clause predicates. The examples in (187) show that in this function simplex reflexives are not in complementary distribution with complex reflexive and reciprocal personal pronouns.

The acceptability of both the simplex and complex reflexive forms in (187) does not follow from the definition of anaphoric domain in (174b). The matrix clause functions as the anaphoric domain of the logical subject of the small clause because it is assigned case by the main verb of the clause; this correctly accounts for our earlier finding in Section 23.3, sub IIB, that personal pronouns in this function cannot be bound by the subject of the clause but incorrectly predicts that the same holds for zich.

One could solve this problem by claiming that logical subjects of small clauses have no anaphoric domain, i.e. by eliminating clause (174b,ii) on case assignment from the definition of anaphoric domain. The prediction would then be that it is sufficient for simplex reflexive subjects of small clauses to be bound within their minimal clause. Unfortunately, this claim wrongly predicts that zichzelf is impossible in the examples in (187), but this may be balanced by the fact that it correctly accounts for the fact that the complex reflexive cannot be used in examples like (188).

De Vries (1999:§5) has indeed claimed that in all cases in (187) and (188) only zich is possible; the form zichzelf in (187) is not the complex reflexive, but the simplex reflexive zich modified by the emphatic modifier zelf (discussed in Section 19.2.3.2, sub V). He acknowledges that the complex reflexive zichzelf is possible in more special semantic contexts; for example, the contrast in acceptability between the examples in (189) is attributed to the fact that the referents of the subject of the main clause experience the sensation expressed by the infinitival clause in (189a) in an immediate way, while they observe the eventuality expressed by the infinitival clause in (189b) indirectly, namely through the eye of a camera lens. The acceptability of zichzelf in (189b) is thus not syntactically but semantically motivated, viz. by the fact that we are dealing with a “second-world” context.

However, we question the validity of De Vries’ arguments: it is not clear, for example, whether the claim that zichzelf in (187) is actually zich modified by the emphatic element zelf is supported by phonetic evidence. Another reason for our reluctance to accept the claim that only zich is possible as the logical subject of a small clause is that it says nothing about the fact that reciprocals can also be used as a SC-subject in the examples in (187) and (189b), which means that our problem concerning anaphor binding from a subject position is still unresolved. Furthermore, there seems to be a good semantic generalization concerning the impossibility of the complex reflexive in examples like (188): the complex reflexive zichzelf is only excluded as logical subject of a small clause if this position cannot be filled by a referential phrase, as will be clear from a comparison of the examples in (188) and (190).

The dollar signs in (190) indicate that the unacceptability of the use of the referential phrases haarher and Marie is related to our knowledge of the world and thus is not (only) a matter of binding theory. It is therefore questionable whether we should account for the examples in (187) and (189b) in terms of the binding domains we have distinguished. We will therefore stop the discussion here, but we will return to examples of this kind in Section 23.4.2, sub II/III, where we will approach the lack of complementary distribution between complex and simplex reflexives from a different perspective on simplex reflexives.

Subsections I to III have shown that attempting to account for the distribution of simplex reflexives in terms of binding domains leads to a less natural and somewhat stipulative formulation of binding theory, which also leaves several issues unresolved. This subsection briefly considers a number of additional cases that may further complicate the description in terms of binding domains.