Consider again the examples in (53), repeated here as (57). We will see at the end of this subsection that the comparative van-phrase van de klas (57c) functions as a regular PP headed by van, which takes the noun phrase de klas as its complement. However, there are reasons for thinking that the als/dan-phrases in (57a&b) cannot be analyzed as regular PPs with nominal complements.

If we were to assume that als and dan in (57a&b) are prepositions taking the noun phrase Jan as their complement, we would expect them to assign objective case to it. As the cases in (58) show, this expectation is not borne out; instead, the case of the noun phrase depends on the noun phrase to which it is compared. The noun phrase in the als/dan-phrase receives nominative case when compared to the subject of the main clause, while it receives accusative case when compared to the accusative argument in the main clause.

Let us first take a closer look at the (a)-examples in (58). They show that standard Dutch als and dan usually do not assign objective case to the noun phrase following them. The percentage signs in (59) are used because there are varieties of Dutch that allow object pronouns, but such cases are usually considered to be substandard; cf. taaladvies.net/taal/advies/vraag/355/ and the references given there.

Since nominative case is usually restricted to subjects of finite clauses, the fact that the pronouns in the (a)-examples of (58) have the nominative form strongly suggests that the complement of als/dan is clausal in nature. That the complement can be clausal in nature is also clear from the examples in (60), which have a finite verb in the complement of als/dan; some speakers also allow the complementizer dat to be present (cf. Van Zonneveld 2005). Note in passing that the subject pronouns in (60) cannot be replaced by object pronouns in any variety of Dutch.

That the subject pronoun is also used in the (a)-examples in (58) can now be accounted for by assuming that these examples are derived from those in (60) by deletion of the finite verb under identity with the finite verb of the main clause. Note that deletion is usually the preferred option.

Let us now turn to the (b)-examples in (58). The substitution of a subject pronoun for the object pronoun in these examples, as in (61), is never possible in the reading intended here, viz. that the speaker compares the two persons referred to by the pronouns haar and hij.

This would follow if we assume that, just as in the (a)-examples, the complements of the als/dan-phrases in the (b)-examples in (58) are clausal in nature; they can then be analyzed as in (62), i.e. by deleting the finite verb and the subject under identity with the verb and the subject of the main clause; again deletion is optional but usually preferred.

The number signs are used in (61) to indicate that the examples are fully acceptable when the speaker and the referent of the nominative pronoun hij evaluate the person referred to by haar; although this is not immediately relevant to the present discussion of the case-assigning potential of als/dan, we have added (63) to show that this reading is expected because the referent of hij is then compared with the subject of the main clause (i.e. the speaker). This reading of the examples in (61) can be analyzed as involving deletion of the finite verb and the object under identity with the verb and the object of the main clause; again, deletion is optional but usually preferred.

That identity is required for deletion is clear from the difference in acceptability between the examples in (61) in their intended reading and the examples in (64); the unacceptable examples in (61) cannot be derived from the acceptable examples in (64) by deletion because the copular differs from the verb of the main clause.

The discussion above has shown that the complement of als/dan in the comparative deletion construction is clausal in nature. This is clearly not true for the complement of the comparative van-phrase in superlative constructions, because the complement of van is always assigned objective case and never contains a finite verb. This shows that the van-phrase is just a regular PP; the preposition van simply takes a noun phrase as its complement.

There seem to be few restrictions on the reduction of the clausal complement of als/dan apart from the one already established in Subsection A, viz. that the omitted content must be recoverable under identity with material from the matrix clause containing the comparative. For instance, in the examples in (58) everything but the logical subject of the AP is deleted from the complement clause. The examples in (66) show, however, that the remaining part can also have other functions. In (66a), for example, the comparative form meermore functions as a clause adverbial of degree. and in the clausal complement of the dan-phrase everything is omitted except the noun phrase corresponding to the object of the main clause. In (66b) meer functions as the direct object and in the clausal complement of the dan-phrase everything is omitted except the noun phrase corresponding to the indirect object. In (66c) everything is deleted except the PP-complement of the adjective; in (66d), finally, an entire object clause has been omitted.

Despite the fact that there are few restrictions on the reduction, it is clear that one element can never be expressed overtly in the als/dan-phrases discussed so far, viz. the adjective corresponding to the adjective in the equative/comparative form in the matrix clause. The examples in (67) show this for the counterparts of the examples in (58), where the element in the comparison set corresponds to the logical subject of the AP: the examples in (68) show this for the counterparts of the more diverse cases in (66). Subsection C discusses the nature of this obligatorily suppressed element.

The nature of the interpretive gap has been the topic of a long and ongoing debate; cf. Corver (2006) and Corver & Lechner (2017) for comprehensive reviews of the literature. Probably the most influential proposal is that of Chomsky (1977), according to which the interpretive gap arises as a result of wh-movement and subsequent deletion of the moved phrase under identity with the adjective in the matrix clause.

A first reason for assuming this is that comparative deletion seems to be unbounded in the same sense as wh-movement. We have already seen an example of this in (68d), repeated here in a slightly different form as (69a), where the interpretive gap is found in a more deeply embedded clause. For completeness’ sake, (69b) provides the corresponding example with wh-movement for comparison.

If comparative deletion does indeed involve wh-movement, we predict that examples such as (69a) are possible when the embedded clause is the complement of a so-called bridge verb such as denkento think, but not when it is the complement of a non-bridge verb such as betwistento contest. Example (70a) shows that this prediction is indeed correct; (70b) again provides the corresponding case with wh-movement.

A second reason for assuming that comparative deletion involves wh-movement is that it cannot occur in so-called islands for extraction. We illustrate this with the (b)-examples in (71), which show that comparative deletion cannot target the nominal complement of a PP, just as wh-movement cannot apply to the nominal complement of a PP. Example (71a) serves only to show that examples of comparable complexity in which the interpretive gap serves as a direct object are perfectly acceptable.

The fact that the wh-movement approach can account for the unacceptability of (70a) and (71b) by independently motivated constraints is generally seen as strong support for Chomsky’s (1977) proposal; cf. Section V11.3.5 for further discussion.

Bresnan’s proposal can be supported by appealing to the fact that quantitative er can be used in comparative subdeletion contexts. Quantitative er, which is discussed in more detail in Section N19.3, is found in contexts such as (75a), where it licenses a phonetically empty nominal projection [e]; the content of [e] is determined by the nominal phrase (mooie) boeken in the first conjunct. Furthermore, [e] must be preceded by a weak quantifier or a cardinal numeral; (75a) becomes completely unacceptable if the quantifier veel is omitted. This means that if the occurrence of er in (75b) is also quantitative, we have independent evidence for the empty degree phrase Δ; cf. Bennis (1977).

Actually, there is little doubt that we are dealing with quantitative er in (75b). Example (76a) shows that the nominal projection [e] cannot be a non-count noun. Consequently, the unacceptability of (76b) also shows that we are dealing with quantitative er.

Moreover, the primeless examples in (77) show that the nominal projection [e] can be combined with postnominal modifiers but not with prenominal attributive adjectives; the contrast between the two primed examples thus again supports the claim that we are dealing with quantitative er.

Section 26.1.3, sub IC, has shown that there are reasons for assuming that the interpretive gap in the comparative deletion construction is the result of wh-movement and subsequent deletion of the moved phrase. We might now expect that the comparative subdeletion construction also involves wh-movement, but this does not seem to be borne out: the distribution of the interpretive gap Δ differs considerably from that of the trace left by wh-movement. One way in which the distributions of Δs and wh-traces differ is illustrated in (78) and (79). The (a)-examples in (78) show that wh-movement of interrogative quantifiers like hoeveelhow many necessarily pied-pipes the rest of the modified noun phrase; extracting the quantifier from the noun phrase leads to an unacceptable result. The fact that Δ in (78b) is in the same position within the noun phrase as the wh-trace in (78a') suggests that wh-movement is not involved in this example.

The examples in (79) show something similar for degree modifiers of APs. The (a)-examples show that wh-extraction of the interrogative degree modifier hoehow from the AP is excluded: wh-movement must pied-pipe the whole AP. The fact that Δ in (79b) occupies the same position within the AP as the wh-trace in (79a') again suggests that wh-movement is not involved in the comparative subdeletion construction. We refer the reader to Section V11.3.5 for a more detailed discussion.

Additional evidence for the hypothesis that comparative subdeletion differs from comparative deletion in that it does not involve wh-movement can be found in (80) and (81). First, the examples in (80) show that PPs are absolute islands for wh-movement: (80b) is just as unacceptable as example (80c) with subextraction of the wh-word hoeveel.

The comparative deletion construction in (81a) shows that having the interpretive gap in the same position as the wh-trace in (80b) yields an unacceptable result, supporting the earlier conclusion that comparative deletion involves wh-movement; cf. also the discussion of the (b)-examples in (71). The comparative subdeletion construction in (81b), on the other hand, shows that it is possible to have the interpretive gap Δ in the position of the wh-trace in (80c), supporting the hypothesis that wh-movement is not involved in comparative subdeletion.

Another difference between the two constructions that points in the same direction is that the comparative deletion construction can contain at most one interpretive gap, while the comparative subdeletion construction can contain multiple interpretive gaps. Consider the examples in (82).

Example (82a) seems perfectly acceptable, despite the fact that the meaning expressed is quite complex, since it asserts two things at once: (i) the number of women to whom neckties were sold exceeds the number of men to whom lipsticks were sold, and (ii) the number of neckties sold to women exceeds the number of lipsticks sold to men. On the other hand, examples such as (82b) are unacceptable because it seems impossible to give an intelligible interpretation to the two gaps in the structure at the same time. Since it is not possible in Dutch to place more than one wh-phrase in clause-initial position, the indicated contrast between (82a) and (82b) would follow from the proposal presented so far: comparative deletion involves wh-movement and consequently there can be at most one interpretive gap, whereas comparative subdeletion does not involve wh-movement and consequently there can be multiple gaps; cf. Corver (1990/2006) for more discussion.

The discussion above strongly suggests that comparative subdeletion is less constrained than comparative deletion. However, this does not mean that comparative subdeletion is completely free; for example, while comparative subdeletion is acceptable with the predicatively used APs in (79b) or (83a), it is excluded with the attributive APs in (83b).

Subsection B has shown that there are reasons for assuming that comparative deletion and comparative subdeletion cannot be assigned to the same analysis: while the former arguably involves wh-movement, the latter most likely does not. This in turn may affect the analysis of the complement of the als/dan-phrase. If comparative deletion involves wh-movement, it automatically follows, as argued in Subsection IA, that the complement of als/dan is clausal, since wh-movement targets a clause-initial position. If comparative subdeletion does not involve wh-movement, then the complement of the als/dan-phrase may or may not be clausal. It needs little elaboration that the complement in the comparative subdeletion construction may be clausal, since we have already seen several clear instances of this in the earlier discussion. Consequently, as in the case of comparative deletion, the primeless examples in (84) can easily be derived from the primed examples by deleting those parts that can be recovered from the matrix clause.

It is less clear whether the complement of als/dan can be non-clausal. However, we need to keep this possibility open: Subsection III will show that the complement of the als/dan-phrase need not be clausal.

The examples in (98) show that in clauses with a verb in clause-final position, the dan/als/van-phrase can optionally occur postverbally, i.e. in extraposed position.

The examples in (99) show that the dan/als/van-phrase cannot occur adjacent to an equative, comparative, or superlative phrase in attributive position; it must occur postnominally.

That comparative dan/als and superlative van-phrases cannot occur prenominally follows from the head-final filter on attributive adjectives in (100), which requires that adjectives with the attributive -e/-∅ ending should be adjacent to the noun they modify; cf. Section 27.3, sub IB, for a more thorough discussion of this filter.

Although the adjective and the dan/als/van-phrase need not be strictly adjacent, it seems impossible for the latter to precede the former. This can be easily established for the dan/als-phrases in (98a&b) by considering the examples in (101a&b); the fact that they cannot follow the clause adverbial zekercertainly suggests that their base position is to the right of the adjective, and the fact that they cannot precede zeker shows that they cannot be moved to the left into an AP-external position. The question as to whether leftward movement of the van-phrase is possible is more difficult to answer, since (101c) seems at least marginally acceptable if the van-phrase precedes the clause adverbial zekercertainly. However, there seems to be a subtle difference in interpretation between the van-phrases in (98c) and (101c); while the van-phrase in (98c) clearly establishes the comparison set, the van-phrase in (101c) seems to delimit the domain of discourse, suggesting that it does not function as a modifier of the superlative but as an independent restrictive adverbial phrase. For this reason, this order is marked with a number sign.

The (a) and (b)-examples in (102) show that topicalization of the adjective must pied-pipe the dan/als-phrase and vice versa: topicalization of the adjective or the dan/als/van-phrase in isolation leads to unacceptability. The (c)-examples are again somewhat more complex: example (102c) shows that topicalization of the adjective must pied-pipe the van-PP, while (102c') shows that it is possible to have a van-PP in clause-initial position when the AP occupies its base position. The unexpected acceptability of (102c') would again follow if we were dealing not with a modifier of the superlative but with a restrictive adverbial phrase; example (102c') indeed seems to get the special meaning (expressed by the number sign) ascribed to the acceptable order in (101c).

The placement of the dan/als/van-phrase in periphrastic constructions is more or less the same as in the morphologically derived cases discussed above; this can be seen by comparing the examples in (103) with the (b) and (c)-examples in (98) and (101). Again, the placement of the van-phrase in the acceptable version before the superlative gives rise to the special meaning that we have also ascribed to the acceptable examples in (101c) and (102c').

Just like dan/als/van-phrases, PP-complements of adjectives are either right-adjacent to the adjective or in extraposed position: cf. dat Jan dol <op vlees> is <op vlees>that Jan is fond of meat. This raises the question as to what happens in case a PP-complement co-occurs with a comparative dan/als/van-phrase. We start with extraposition. As shown in (104), the PP-complement must precede the dan/als/van-phrase when the two phrases are in extraposed position.

The primeless examples in (105) show that extraposition can also be restricted to the dan/als/van-phrase. On the other hand, extraposition of the PP-complement is blocked when the dan/als/van-phrase is in preverbal position, as shown by the primed examples in (105). Note, however, that the primed examples improve if the noun phrase following als/dan is heavy, e.g. when we replace vleesmeat with andijvie met een flink stuk vleesendive with a good piece of meat.

Extraposition of the dan/als/van-phrase seems to be preferred, but to the extent that it is possible to place both the PP-complement and the dan/als/van-phrase between the adjective and the verb in clause-final position, the complement must precede the dan/als/van-phrase, as in the extraposition constructions in (104).

As shown in (107), it is not possible to place the dan/als/van-phrase immediately to the right of the periphrastic elements of comparison even, minder or het minst. The number sign indicates that (107a) is acceptable if we interpret the sequence even + als as English “just like”, but the intended interpretation here is “as fond of meat as ...”.

Finally, note that the van-phrase can precede the PP-complement when it occurs more to the left of the superlative, but it seems plausible that in these cases we are again dealing with an independent restrictive adverbial phrase.

Section 24.3.1, sub III, has shown that the PP-complements of some pseudo-participles and deverbal adjectives can be base-generated either on their right or on their left. The examples in (109) show for the pseudo-participle gesteld opkeen on that this is also possible in periphrastic comparative and superlative constructions (cf. Section 26.3.1), but not in equative constructions. The examples in (109) also show that the PP-complement can be in extraposed position. Although cases with the PP-complement immediately after the adjective are certainly acceptable, some speakers may consider them somewhat marked compared to the other two orders.

Example (110) shows that the same results are obtained if we replace gesteld op vlees by the deverbal AP afhankelijk van hulpdependent on help (e.g. home care).

As in the case of the simplex adjective dolfond, extraposition of the dan/als/van-phrase is strongly preferred when a PP-complement is present. Since this is the case regardless of whether the PP-complement precedes or follows the pseudo-participle, we should conclude that the marginality of the examples in (111) is not due to some linear restriction that blocks the presence of both a PP-complement and the dan/als/van-phrase between the adjective and the clause-final verb position; we leave this issue to future research.

The examples in (112b&c) show that pseudo-participles like gesteld op and deverbal adjectives like afhankelijk van differ from the simplex adjectives discussed in the previous two subsections in that they allow the dan/als/van-phrase of comparison to be placed immediately after meer/minder or het meest/minst in the periphrastic comparative/superlative construction, regardless of the position of the PP-complement. However, this is not possible in the equative construction; the (a)-examples in (112) are acceptable, but only if we interpret the sequence even + als as English “just like”; the intended interpretation here is “as fond of meat as ...” and “as dependent on help as ...”.