The discussion of conjunction reduction and gapping in the previous sections started from the generally accepted view that these reduction processes occur only in coordinate structures. However, Van der Heijden & Klein (1995) has argued that these processes can also occur in some subordinate contexts; this is illustrated for gapping in (183).

The claim that we are dealing with gapping in examples like those in (183) may be controversial, since it has certain properties not found in coordinate structures: this is immediately clear from the two examples in (183), where (presumed) gapping apparently applies in both a forward and a backward way. The claim that we are dealing with subordination in (183) is also controversial, as can be seen from the fact that behalve has been analyzed as a coordinator in Paardekooper (1966) and as a preposition in Landman & Moerdijk (1980): it has also been analyzed as a hybrid category in Van der Heijden (1999).

This section therefore continues the discussion of conjunction reduction and gapping by reviewing the arguments for the different positions in the literature. We discuss a small number of words/phrases that have been claimed to have the coordinator-like property of being compatible with gapping: the comparative linkers dan/alsthan, the element behalveexcept/besides in (183) above, the phrasal preposition in plaats vaninstead of, and the verbal expression Laat staanlet alone. These elements at least have in common that they can be followed by a nominative noun phrase or a finite clause; this is illustrated in (184) for the element danthan, which is found e.g. in comparative constructions. These examples provide negative arguments in favor of the assumption that danthan is a conjunction-like element; (i) the dan-phrase in (184a) cannot be analyzed as a prepositional phrase with a nominal complement, since prepositions assign objective (not nominative) case; (ii) the element dan in (184b) cannot be analyzed as a subordinator, since it is followed by an embedded clause introduced by the subordinator datthat; cf. Den Besten (1978).

Besides these two negative arguments for the assumption that danthan is a coordinator-like element, the gapping-like construction in (185b) has been given as a positive argument, because it is generally assumed that gapping occurs only in coordinate structures. We have added (185a) to show that it is at least possible to analyze example (184a) as a gapping-like construction.

The conclusion drawn from the data in (184) and (185) differs from study to study: Corver (1990:89ff/1993) and Hendriks (1991/1995: §2) state that dan can function as a coordinator; Haeseryn et al. (1997: §27.5.4) states that dan dat is a complex linker (with optional dat) with coordinator-like properties, and Van der Heijden (1999: §1) claims that dan is a “mixed” category between coordinator and subordinator. We will not consider these proposals in detail, but follow Van der Heijden & Klein (1995) in arguing that there are reasons to reject the idea that we are dealing with coordination-like structures altogether. This raises a number of novel questions about gapping, which are discussed in Section 39.4.2. Section 39.4.3 concludes with a possible problem with the assumption that gapping is involved in the derivation of the constructions discussed in this section.