Verb-frame alternations involve changes in the number and type of complements selected by the verb. The cases discussed in Section 3.2 are clear cases of verb-frame alternations in the intended sense, as they involve the demotion, suppression, or addition of an external argument by passivization, middle formation, or causativization, respectively. This also holds for the NP/PP alternations discussed in Section 3.3, provided that we assume that the PPs in question are selected by the verb. However, Levin (1993) includes a number of cases in her inventory of verb-frame alternations of which it is not so clear whether they should actually be characterized as such (at least in Dutch). Consider the two examples in (542). Pairs like these clearly do not involve verb-frame alternations in the sense defined above, since in either case the verb selects two arguments (an agent and a theme). Note that coreference is indicated in italics.

Things may be different in the (a)-examples in (543), which Levin calls the understood reciprocal alternation and which seems to involve the (optional) suppression of the theme argument. However, it seems implausible that exhibiting this alternation is a general property of verbs with an agent and a theme, since example (543b') seems to be infelicitous without the reciprocal.

This means that there must be some other difference between the verbs kussento kiss and slaanto hit. The relevant difference seems to be that kussen can be combined with a comitative met-PP, while this is impossible with slaan.

That this may well be the correct conclusion is strongly suggested by the fact, illustrated in (545), that the understood reciprocal alternation is more generally found with verbs that allow a comitative met-PP; Levin calls this case reciprocal alternation.

The question as to whether we are dealing with a verb-frame alternation now hinges on whether the comitative met-PP is a complement of the verb: we are dealing with a verb-frame alternation only if the answer is positive. An argument in favor of a positive answer is that the possibility of having a comitative met-PP clearly depends on the meaning of the transitive verb, but there are also reasons for assuming that the comitative met-PP is an adjunct, just like the instrumental PP met de balwith the ball in (546), which does not allow the alternation because it does not have the semantic function of co-agent.

A first reason for assuming that comitative and instrumental met-PPs are both adjuncts is that they can easily be omitted without being semantically understood: the sentence Jan speelde in de tuin leaves completely open whether Jan is playing with another person or with a specific object. The second reason is that they both behave like VP-adjuncts, which can be seen from the fact that example (546a) can be paraphrased by the ... en pronoun doet dat met-PP clause in (547a), regardless of the nature of the met-PP. Another reason may be that these met-PPs can both precede the adverbial place adverbs in (547b) in neutral (non-contrastive) contexts, whereas PP-complements usually follow adverbial phrases in such cases; note that we have used an embedded clause to illustrate this, to eliminate the intervention of extraposition. See Sections 2.3.1, sub VII, and 2.3.4, sub I, for further relevant discussion.

If we conclude from these facts that the comitative met-PP is simply an adjunct, we should also conclude that Levin’s understood reciprocal alternation is not a verb-frame alternation: we are simply dealing with (pseudo-)intransitive verbs. An additional argument against the assumption of an understood reciprocal verb-frame alternation is that the constructions with and without a reciprocal are not semantically equivalent. This is clear from the fact that there is no implication relation between the primeless and primed examples in (548): the primeless examples simply state that Jan and Peter like to kiss/play in general, without implying that they like to do so together; the primed examples, on the other hand, do express that Jan and Peter like to kiss/play together, but they do not imply that they like to do so in general, i.e. with other people.

The conclusion that there is no (understood) reciprocal verb-frame alternation holds not only for the above cases with a comitative met-PP, but also for other syntactic configurations in which a noun phrase can bind a reciprocal. This is especially true for resultative constructions such as (549), in which the logical subject of the predicative PP can act as an antecedent of a reciprocal embedded in the PP.

The examples in (550) show that here too the reciprocal construction is semantically different from the non-reciprocal construction. The primed examples are only possible if the primeless examples are symmetrical in the sense that they allow the two noun phrases to change places: cf. Marie legde de enveloppen bij de brievenMarie placed the envelopes with the letters versus $Marie legde de voordeur bij de brievenMarie put the front door with the letters. This clearly shows that the alternation is determined by the nature of the noun phrases and no by the nature of the verb.

Similar objections can be raised to other cases that Levin collects under the general denominator of reciprocal alternation, such as the samen-alternation in (551). Since the particle samentogether in (551a'&b') can precede the adverbial phrase and the PP-complement, it is clearly not a verbal particle selected by the verb, and consequently we can safely conclude that we are not dealing with a verb-frame alternation. Similarly, in the (c)-examples, the PP and samen seem to have the same syntactic function, that of complementive, so once again it is not justified to consider this a case of verb-frame alternation.

This section has discussed a number of systematic alternations and considered the question as to whether we are dealing with verb-frame alternations in the restricted sense defined earlier, i.e. as changes in the number and types of complements selected by the verb. We concluded that this is not the case for the alternations discussed in this subsection, which implies that such alternations are not interesting from a syntactic point of view (which of course leaves open the possibility that they might be interesting from e.g. a semantic point of view).