Section 39.4.1 has shown that elements like comparative dan/alsthan/as, behalveexcept/besides, in plaats vaninstead of and laat staanlet alone are not coordinator-like elements, but subordinators that (can) introduce a gapped clause. This contradicts the traditional claim that gapping (and backward conjunction reduction) occur only in coordinate structures, and thus raises a number of new questions. One of them is what licenses gapping in subordinate clauses. In fact, the same question should have been asked in the context of gapping in coordinate structures, because not all coordinators license gapping: gapping is not possible in coordinate structures with the coordinator wantbecause and less common with the coordinator dusso. cf. Section 39.2, sub IA.

So, if gapping can also apply in subordinate clauses introduced by the elements listed above, it should not surprise us that there are other cases of subordination in which gapping is excluded. Some examples are given in (231).

A straightforward way to rule out gapping in examples like those in (231) would be to point to the A'-movement hypothesis proposed in Section 39.2, sub IID, according to which the remnants of gapping must occupy a designated A'-position. Gapping in (231) would then be unacceptable, because complementizers do not occupy such a position, but cannot be elided either, because they express information not recoverable from the linguistic context; toenthen, omdatbecause, and hoewelalthough cannot be elided, because they express temporal, causal, and concessive information. Elements like dan/alsthan/as, behalveexcept/besides, in plaats vaninstead of and laat staanlet alone, on the other hand, are not part of the gapped clause and are therefore not expected to interfere with gapping. Unfortunately, this approach leaves us in the lurch regarding the unacceptability of the examples in (230a): the coordinators wantbecause and dusso are also external to the gapped clause, and so are not expected to interfere with gapping.

Another proposal aimed at making the correct distinction between gapping and non-gapping contexts is offered in Van der Heijden & Klein (1995), which focuses on the logical and arithmetical properties of gapping constructions. The general idea is that the contrastively accented parts of the antecedent and the target clause stand in a simple logical or arithmetical relation specified by the linker (coordinator or subordinator). That something like this holds for the conjunctions in (232) is uncontroversial; cf. Section 38.1, sub IV, for further discussion. Note that we did not include the conjunction maarbut in (232), which is like en in that it can express conjunction and intersection, but adds an adversative meaning aspect.

The fact that wantbecause and dusso cannot be used in gapping contexts can now be ascribed to the fact that they do not specify a simple logical/arithmetical relation between subparts of the coordinands, but a complex relation between the clausal coordinands as a whole, as is clear from the meanings ascribed to them in Section 38.4.1, sub V, repeated here in (233).

Example (234) shows that the subordinators als and dan found in comparatives again specify a simple relation between the subparts of the clauses, which can be expressed by logical symbols like “=”, “>” and “<”, which can be applied to quantities and degrees.

Van der Heijden & Klein (1995) further claims that the subordinating elements in (235) can also be seen as simple logical/arithmetical operators, although some of them have no clear counterparts in standard logic/arithmetic. The general property of these linkers is that they perform some operation (reduction, expansion, or substitution) on a possibly singleton set given in the antecedent clause.

The question is now whether the semantic operations/relations should be used in the formulation of a restriction imposed on the syntactic operation of gapping, or whether they should simply be seen as an inventory of semantic relations that can hold between the contrastively accented remnants of gapping and their correlates in the antecedent clause. Since the answer to this question depends to a large extent on the model of grammar one adopts, we will not digress on this issue here.