We conclude the discussion of the classification of verbs based on the number and type of their nominal complements by pointing out a possible problem with one of the unaccusativity tests used in the previous discussion: the claim that the selection of the auxiliary zijn is a sufficient condition for assuming unaccusative status for a verb. If this test is indeed valid, we predict that there are no transitive verbs that select zijn. This seems to be true in general, but there are a small number of possible counterexamples, which we will discuss in this section.

We start with a series of ostensible counterexamples: verbs like bijspringento help out, ontkomento escape, ontlopento escape, ontvluchtento flee, tegemoet gaan/komento meet, tegenkomento encounter, volgento follow all take an object, even though they form their perfect tense with zijn. This is not really surprising, since these verbs all take a dative object in German, which is allowed in unaccusative constructions. However, the verb volgen can still be a possible problem, since it can be passivized, which was considered a sufficient test for assuming ergativity (which implies transitivity in this particular case).

The seeming contradiction is resolved when we realize that the verb volgen is ambiguous with respect to the auxiliary test; it combines not only with zijn but also with hebben: De politieagent is/heeft de verdachte gevolgdThe policeman has followed the suspect. This suggests that volgen is undergoing a process of reanalysis; it is developing from a verb with a dative object into a verb with an accusative object. Such a reanalysis has also been found in other cases; for example, the German verb hilfento help takes a dative argument and cannot be passivized, whereas its standard Dutch counterpart helpen exhibits prototypical transitive behavior in that it can undergo regular passivization: Het slachtoffer werd door een voorbijganger geholpenthe victim was helped by a passer-by.

Even if we ignore those cases that are susceptible to a dative object analysis, we have to deal with at least the following two (notorious) problems: the transitive verbs vergetento forget and verliezento lose, which can take either hebben or zijn in the perfect.

Perhaps we can set the case of verliezen aside as part of the formal register, since Haeseryn et al. (1997:79) claims that the use of zijn is not generally accepted and is more common in written language than in speech. The case of vergeten is harder to explain. We might understand the acceptability of zijn in (169b) better if we relate this example to examples such as (170), in which the noun phrase mijn paraplu does not function as a complement of the verb vergeten but as an argument of the embedded infinitival predicate meenemento take along. This would suggest that (169b) has a phonetically empty embedded predicate.

But even if this were the case, it would leave us with cases such as (171), in which vergeten is interpreted more precisely as “to not remember”: although Haeseryn et al. (1997) claims that zijn is much preferred in this case, the stipulation of a phonetically empty embedded predicate seems less tenable. We will not speculate any further on this (long-standing) issue.

An incidental case is beginnento start in (172a); example (172b) shows that passivization leads to a reasonably acceptable result. Perhaps, such cases are relatively recent innovations in the language, given that beginnen can also be combined with a PP-complement with aan; cf. Zij is aan een nieuw boek begonnen She has started a new book. This line of thought may find support in the fact that its counterpart stoppen can only be combined with a PP-complement; Zij is *(met) haar nieuwe boek gestopt She has stopped with her new book.

We leave further investigation of such cases to future research.