An important difference between the comparative-deletion construction and the wh-movement constructions discussed in Sections 11.3.1 to 11.3.4 is that the former has no phonetically visible wh-moved antecedent for the interpretive gap: if such an antecedent is present, we have to assume that it is deleted or at least left phonetically unexpressed under “identity” with (a phrase containing) the equative or comparative form that selects the dan/als-phrase. This makes it difficult to firmly establish (or refute) the claim that wh-movement is involved in the derivation of comparative-deletion constructions. To do so, we need to show that the construction has at least the three characteristic properties of wh-movement listed in (489).

That there is an interpretive gap was already shown in the introduction to this section on the basis of the meaning of the constructions in (487): for example, we have seen that (490a) expresses that the number of books Jan owns exceeds the number of books he can read. That the gap is obligatory can also be shown by the fact, illustrated in (490b), that its position cannot be taken by an overt noun phrase (except for bare noun phrases with a more deeply embedded interpretive gap in comparative-subdeletion constructions, which we will discuss later).

This suggests that the comparative dan-phrase in (490a) has a phonetically empty direct object that is associated with the overt direct object in the matrix clause containing the comparative, meer boekenmore books. We will assume that wh-movement establishes this association by moving the (phonetically empty) phrase sufficiently close to its antecedent in the matrix clause. Since wh-movement usually results in the formation of some sort of operator-variable chain, we will henceforth refer to the moved phrase as empty operator (leaving aside the question of whether the construction involves deletion of the wh-phrase).

There is some dispute about the exact landing site of the empty operator, which is related to the fact that the categorial status of the element danthan (as well as alsas) is also unclear. Although it is sometimes claimed that dan is a complementizer (i.e. a subordinating conjunction), we will provisionally assume that it is a preposition-like element capable of selecting a clausal complement. This seems consistent with the fact that in colloquial speech the element dan can be optionally followed by dat in examples such as (490a); because dat should clearly be analyzed as the complementizer of the embedded clause, it seems unlikely that dan has the same function. If the above is correct, we can assign to example (490a) the structure in (491a). The claim that dan is preposition-like can be further supported by the fact, illustrated in (491b), that it can also be followed by a noun phrase. See Section A26.1.3.3 for a more detailed discussion.

If wh-movement is indeed involved in the derivation of comparative-deletion constructions, we expect that the interpretive gap can be embedded in complement clauses selected by bridge verbs like denkento think and zeggento say. The examples in (492) show that this expectation is indeed borne out.

Furthermore, we expect comparative deletion to be excluded when the interpretive gap is embedded in an island for wh-movement. This is again confirmed, as shown in (493) for an interrogative and an adverbial clause, respectively. While the intended meanings in the primed examples are intelligible (though perhaps a bit difficult to grasp at first), the corresponding sentences in the primeless examples are gibberish.

The island sensitivity of comparative deletion can also be illustrated by the contrast between the two constructions in (494). Den Besten (1978) claims that these examples differ in that the element dan takes a clausal complement in (494a), but a nominal complement in the form of a free relative in (494b). If this proposal is on the right track, the contrast between the two examples can be attributed to the independently established fact that wh-movement can only strand a preposition if the PP is pronominalized, i.e. if it has the form waar + P; cf. Section P36.3.

The discussion above has shown that comparative deletion does indeed exhibit the three characteristic properties of wh-movement in (489): (i) the interpretive gap in the dan/als-phrase is obligatory and cannot be filled by some overt phrase (leaving aside the comparative-subdeletion constructions to which we turn next); (ii) assuming that the empty operator must be located in the clause-initial position of the clause selected by dan, its relation to its trace is (apparently) non-local in bridge contexts; (iii) the movement of the empty operator is island-sensitive.

In determining whether the derivation of comparative-subdeletion constructions involves wh-movement, we should take a similar approach as in the case of comparative-deletion constructions, and show that the construction has at least the three properties of wh-movement in (489). That there is an interpretive gap has already been shown in the introduction to this section on the basis of the meaning of the constructions in (488): for instance, example (495a) expresses that the number of books Jan owns exceeds the number of CDs Els owns. That the empty quantifier is obligatory is shown by the fact, illustrated in (495b), that its position cannot be filled by an overt numeral or quantifier. This suggests that the direct object of the comparative dan-phrase in (495a) must contain a phonetically empty quantifier, which is associated with the quantifier meermore of the direct object in the matrix clause meer boekenmore books.

The examples in (496) show that, as predicted by the wh-movement hypothesis, the interpretive gap can also be more deeply embedded in bridge contexts; the question marks in brackets indicate that some speakers consider these examples to be slightly marked.

The examples in (497) further show that comparative subdeletion is sensitive to interrogative and adjunct islands. The intended interpretations in the primed examples are intelligible (though perhaps difficult to grasp at first), while the sentences in the primeless examples are utter gibberish.

The data discussed so far are consistent with the wh-movement hypothesis, but there are also problems with this hypothesis. The first is that the empty operator in (495a) is a quantifier modifying a noun phrase; the examples in (498) show that noun phrases are usually islands for wh-movement of such modifiers; movement of the quantifier hoeveel obligatorily pied-pipes the containing noun phrase.

The hypothesis that comparative-subdeletion constructions are derived by wh-movement therefore requires some special stipulation. One possible analysis could be constructed in analogy to the constructions in (499), which show that quantified noun phrases such as (499a) alternate with the construction in (499b) with so-called quantitative er, which replaces the lexical part of the noun phrase (here: cd’s). We follow Coppen (1991) and Barbiers (2017) by assuming that er is extracted from the noun phrase by leftward movement (although this analysis is not without problems, as shown in Section N19.3).

Quantitative er can optionally occur in the clausal complement of the comparative dan-phrases in (500); cf. Bennis (1977). We can easily account for this by assuming that the interpretive gaps in the two constructions in (500) are different: the gap e in (500a) receives the interpretation [NP [Δ many] books], while in (500b) it receives the interpretation [NP [Δ many] t], with t acting as the trace of quantitative er.

This means that the two constructions in (500) are both derived by comparative deletion, as the interpretive gap [e] corresponds to the full direct object in both cases. Note that this analysis requires that the noun phrase [[Δ many] t] operator can be wh-moved across quantitative er, but this is unproblematic, as can be seen from the fact that it is also possible to wh-move the remnant noun phrase in (501) across er.

The acceptability of extracting quantitative er from the noun phrase may lead to the conclusion that, more generally, it should be possible to extract the lexical part of a noun phrase while stranding the functional part of it. We have reason to believe that this is possible in principle, as can be seen from the fact that some varieties of (especially southern) Dutch and German exhibit this property in so-called split-topicalization constructions such as (502b); cf. Van Hoof (2006/2017) and Hendriks (2023). Again, we assume a movement analysis (although Van Hoof shows that this analysis may not be without problems). The number sign is used to indicate that the construction is impossible in standard Dutch.

If lexical projections can indeed be extracted from their noun phrase while stranding their quantifier, this opens up the possibility to reanalyze the comparative-subdeletion construction in (495a) as in (503); the underlying structure would then be approximately as in (503a), while the intermediate structure in (503b) is derived by scrambling the lexical part (i.e. the plural noun cd’s) from the noun phrase to a clause-medial position; the surface structure in (503c) is derived by moving the phonetically empty remnant of the noun phrase ([[Δ many] ti]) into clause-initial position.

This derivation unifies comparative deletion and comparative subdeletion (for cases involving quantified noun phrases), but the cost is high; for example, we have to make the ad hoc provision for standard Dutch that the leftward movement of the lexical part of the noun phrase is impossible when the numeral/quantifier is overt (cf. *dat Els cd’si [NP drie/veel ti heeft]), although such a provision is of course needed anyway to block the split-topicalization construction in (502b)) in standard Dutch. Another, even more compelling reason not to follow this line of inquiry is that a wh-movement analysis of comparative subdeletion would violate other well-known restrictions on wh-movement. For example, while (504a) shows that the interpretive gap can be part of a nominal complement of a PP, the (b)-examples show that the corresponding wh-extraction is usually impossible: wh-movement of wat does not lead to a wat-voor split, but obligatorily triggers pied piping of the whole PP.

This section has considered the role of wh-movement in comparative-deletion and comparative-subdeletion constructions. We have shown that there is good reason to assume that comparative deletion is derived by wh-movement of an empty operator into the initial position of the clause selected by the prepositional-like element als/dan; this movement may be motivated by the need to place the empty operator in a sufficiently local relation to its associate, i.e. (the phrase containing) the equative/comparative adjective phrase in the matrix clause. The proper analysis of comparative subdeletion is much less clear, as a wh-movement analysis seems to require several ad hoc stipulations. It is not surprising, therefore, that this construction is still the subject of ongoing debate. For a detailed discussion of the current state of affairs, we refer the reader to Corver & Lechner (2017).