The fact that infinitival clauses of the type in (512/514a) arguably contain an empty operator that is moved into the clause-initial position suggests that we are dealing with regular relative clauses in which the relative pronoun simply has no phonetic realization. However, there are several problems with this proposal. The first is only circumstantial and has to do with the fact that German actually cannot easily use infinitival clauses in this modifying function (Kester 1994:§3.3.4.4), which is clear from the fact that a Dutch example such as (515a) cannot be directly translated into German, as shown by the unacceptability of (515b). Given that Dutch and German are so closely related, it would be very surprising if the former could make use of infinitival relative clauses, but the latter could not.

The second problem is more serious. As can be seen in (516a'&b'), infinitival clauses containing an empty operator can also occur in predicative position, an option that, as shown in (517), does not exist for regular finite clauses. This suggests that the infinitival clause is comparable to the set-denoting adjectives in that it can be used both predicatively and attributively.

The third problem is that infinitival clauses are often used in the presence of adjectives that can enter the so-called easy-to-please construction in the primed examples in (518), the analysis of which also involves an empty operator; cf. Section A29.5, sub IVA, for further discussion. Given the semantic similarities between the primeless and primed examples, it seems reasonable that the former are the attributive counterparts of the predicative constructions in the latter.

At this point, German comes in again. It should be noted that German does not have the easy-to-please construction of the type in (518a'&b'): *Das Buch ist einfach um zu lesen. German does have constructions such as Das Buch ist einfach zu lesen without the infinitival complementizer um, but these must clearly be analyzed as modal infinitives, which is also clear from the fact that the infinitive can be used in prenominal attributive position das \`1leicht\`1 zu lesende Buch, in which it is realized in the form of an attributively inflected present participle; cf. the discussion in Subsection IVB. It therefore seems plausible to relate the unacceptability of (515b) to this fact; infinitival clauses in the easy-to-please construction are actually more like infinitival clauses than like regular relative clauses.

Leaving the precise analysis of the restrictive infinitival clauses in this subsection to future research, we can safely conclude from the discussion above that infinitival clauses containing an empty operator are either set-denoting phrases that can be used either predicatively or attributively, or belong to the adjectival part of an easy-to-please construction that can be used in these functions. Despite appearances, there is no clear evidence that infinitival clauses are relative clauses.

Given the conclusion that the infinitival clauses under discussion are not relative clauses, and that the empty operator is therefore not a relative pronoun, it is not surprising that there are no restrictions on the antecedent in terms of number, gender, or animacy. This is illustrated in Table 6.

There seems to be no further restriction: the antecedent can be definite, indefinite, or quantified. Thus, while in Table 6 all antecedents are indefinite, the antecedent in (519a) is a definite DP, and in (519b&c) the antecedent is quantified.

For completeness’ sake, it should be noted that noun phrases modified by a restrictive infinitival clause often have as their counterpart a nominal compound in which the verb of the infinitival clause is used as the first morpheme. The relationship between the two morphemes of these compounds is typically one of purpose. Some examples are given in (520).

There seem to be some restrictions on the types of verbs that can appear in the modifying infinitival clause, and on the syntactic function of the empty operator. For obvious reasons, infinitival clauses with an empty operator do not normally occur when the infinitival verb is intransitive or unaccusative, as in (521a&b); of course, both examples are acceptable, but they must be analyzed without an empty operator: in these examples, it is the implied subject PRO that is interpreted as the element coreferential with the antecedent. Note, however, that this restriction is not absolute; example (524) will show that there are ways to circumvent it.

An infinitival clause gives rise to a perfect result if the infinitive is a transitive verb, as in (522a): in this case the empty operator corresponds to the direct object. If the verb is ditransitive, the result depends on the form of the indirect object; constructions with a nominal indirect object are much less acceptable than constructions with a prepositional indirect object. The (b)-examples show this for constructions in which the empty operator acts as the direct object, and the (c)-examples for cases in which it corresponds to the indirect object.

Given that the empty operator can act as the complement of the preposition aan in (522c'), it will come as no surprise that the result is also fine when the empty operator acts as the complement of a PP-complement of the verb, as in (523a). Example (523b) shows that the empty operator can act as the complement of a locational predicate.

Finally, the empty operator can correspond to the complement of a PP-adjunct, provided that this PP allows R-extraction. Examples of this kind were already given in (520), and some more examples are given in (524). Note that when such adjuncts are present, the construction can also occur with intransitive and unaccusative verbs.

The examples in (525) and (526) show that the construction is excluded if the adjunct PP does not allow R-extraction. First, consider (525b) which shows that it is not possible to strand the preposition in in the relative construction; relativization is possible, but it requires that the entire PP be replaced by the locational pro-form waar.

Example (526a) shows that it is not possible to have the infinitival construction with the stranded preposition, which seems to provide additional evidence for our earlier conclusion that the empty operator must be moved into the clause-initial position. More surprisingly, (526b) also seems unacceptable: the most prominent but incoherent reading of this example seems to be that een café functions as the direct object of drinkento drink. We leave open whether the construction should be considered ungrammatical under the intended reading, given that we have found a small number of such examples on the internet.

Subsection I argued that infinitival clauses with an empty operator are not relative clauses on the basis of the fact that they can be used predicatively, which is never possible in the case of relative clauses. However, this argument does not hold for infinitival clauses without an empty operator, which is clear from the fact that the primed examples in (531) are not interpretable.

The question whether non-restrictive infinitival clauses without an empty operator are relative clauses cannot therefore be decided in this way. To get closer to an answer, we can ask the basic question whether the modifying function of the infinitival clause is brought about by the coreference relation between PRO and the modified noun phrase, or whether this relation is epiphenomenal due to the fact that PRO often has to have an antecedent in order to be interpretable? If the infinitival clause is a regular relative clause, we should conclude that the former is the case. However, there are examples of modifying infinitival clauses that seem to refute this hypothesis. In (532), for example, it seems clear that the infinitival clauses are used as restrictive modifiers of the noun phrase (de) tijd(the) time. Nevertheless, the modified noun phrase is not coreferential with PRO (nor with an empty operator, since the noun phrase does not seem to play any semantic role within the infinitival clause). These examples therefore suggest that modification by the infinitival clause is not related to the coreference relation between the modified noun phrase and PRO, which in turn suggests that modifying infinitival clauses are not relative clauses. Other possible examples of this kind are given in (533).

Whatever the correct analysis of the non-restrictive infinitival clauses under discussion, it seems clear that they do not contain an overt relative element. It is therefore not surprising that there are no restrictions on the antecedent in terms of number or gender. This is illustrated in Table 7.

In many cases it is difficult to find examples in which the antecedent of PRO is inanimate, but this is due to the fact that PRO functions as the subject of an infinitival clause and is therefore typically agentive, hence animate. However, if we are dealing with an unaccusative verb, as in (534a), or an infinitival clause in the passive voice, as in (534b), the result of having an inanimate antecedent is perfectly acceptable. For completeness, note that the corresponding active construction of (534b) in (534b') involves an infinitival clause with an empty operator.

Restrictive infinitival clauses without an operator are not easily used with definite noun phrases, unless they are used as the nominal predicate of, for instance, a copular construction. This is clear from the contrast between the two (a)-examples in (535). The (b)-examples illustrate that a similar contrast does not arise in the case of indefinite noun phrases.

There seem to be few restrictions on the infinitive, which is not surprising given that the antecedent is generally coreferential with the implied PRO subject of the infinitival clause. In (536a) we find an infinitival clause with the intransitive verb werkento work and an implied subject PRO coreferential with the antecedent een typea type. Similar constructions can be found in (536b&c) with a copular and unaccusative verb. In (536d) the antecedent een machinea machine is interpreted as the subject of the transitive verb schurento sand, while in (536e), the antecedent een bedrijfa company is interpreted as the subject of the ditransitive verb gevento give.

For unclear reasons, however, a dyadic unaccusative verb cannot be used as an infinitive in these restrictive infinitival clauses, as shown by the unacceptability of the examples in (537).

Sentence-final infinitival clauses can easily be confused with adverbial purpose clauses, which can also take the form of an infinitival om-clause. A first difference between modifying infinitival clauses and purpose clauses is that the latter cannot contain an empty operator. Thus, while example (538b) can be interpreted as a purpose clause, this reading is absolutely impossible in (538a).

In (538b), however, the om-clause can still be interpreted as modifying the direct object of the main clause, nieuwe softwarenew software. In this case the implied subject PRO is interpreted as coreferential with the direct object, as indicated by the coindexing in (539a); the construction can be paraphrased as “software designed to solve the problem”. In the (more prominent) purpose clause interpretation, on the other hand, the implied subject PRO is coreferential with the subject of the main clause, as indicated by the coindexing in (540a). The two structures in (539a) and (540a) differ not only in meaning, but also in syntactic behavior. First, (539a&b) show that the modifying infinitival clause can appear either in extraposed position or immediately right-adjacent to the object. The purpose clause, on the other hand, clearly prefers the clause-final position in (540a). Second, the (c)-examples show that while the purpose clause can be placed in the clause-initial position, the modifying infinitival clause cannot be topicalized in isolation. Finally, the (d)-examples show that only the modifying infinitival clause can accompany the direct object in the clause-initial position, which of course follows from the fact that only in this case the infinitival clause is part of the direct object (the constituency test); in these examples we have added the adverbial phrase of time netjust to facilitate topicalization.

The examples in (541) show that the infinitival clause with an empty operator in (538a) behaves exactly like the infinitival clause without an empty operator in (539): like the latter, the former can appear both in extraposed position and right-adjacent to its antecedent, and can be pied-piped by topicalization of the direct object, but cannot be moved into the clause-initial position on its own.

That it is indeed the presence of an empty operator that blocks a purpose reading of example (538a) can be made clear by replacing this operator with the R-pronoun daar, which would result in the structures in (542). The examples in (542) show that in this form the sentence behaves like (540). The infinitival clause clearly prefers the clause-final position, it can be topicalized in isolation (provided the infinitival clause is assigned emphatic accent), but it cannot be pied-piped by topicalization of the direct object.

To conclude this discussion, note that in copular constructions like (543a&b), the infinitival clause cannot be construed as a purpose clause, but must be construed as a restrictive infinitival clause, which is clear from the fact that the infinitival clauses cannot be topicalized in isolation.

Modal infinitives, which are discussed in more detail in Chapter A32, can also occur in postnominal position. However, they are easy to distinguish from infinitival clauses, since they are never introduced by the infinitival complementizer om and can be used in either pre or postnominal position. Furthermore, they should not be considered as infinitival clauses but as (adjectival) phrases, just like their attributively used prenominal counterparts. Some examples of modal infinitives are given in (544), where the given English translations are intended to express the modal meaning of these examples: the root modality of obligation or ability in (544a), ability in (544b), and permission in (544c). Note the word order differences between the pre and postnominal occurrences of the modal infinitive: in accordance with the head-final filter on attributive adjectives, the prenominal modal infinitive must be immediately adjacent to the modified noun, whereas the postnominal modal infinitive may be separated from the head noun by all kinds of material.

Like infinitival clauses with, but unlike infinitival clauses without an empty operator, modal infinitives can often be used in predicative position. However, the examples in (545) seem to show that this is more or less restricted to cases where the modal infinitive expresses ability.

As the use of commas in (544) indicates, postnominal (as opposed to prenominal) modal infinitives are most easily interpreted non-restrictively. The examples in (546) serve to illustrate the difference between postnominal modal infinitives and non-restrictive infinitival clauses by means of a minimal pair: the infinitival clause in (546a) conveys the additional information that the bookcases are not yet assembled, but that the customer should do this himself; the modal infinitive in (546b), on the other hand, conveys the information that the bookcases are such that the customer is able to assemble them himself.

The examples in (547a&b) show that postnominal modal infinitives can sometimes (marginally) be used restrictively, provided that the phrase headed by the modal infinitive is complex; if not, the modal infinitive must be prenominal. In this respect, modal infinitives behave like non-restrictive adjectival phrases; cf. Section 17.3.5.