Wh-exclamatives often express an extremely high degree. This is illustrated in example (423a); the exclamative wh-element watwhat expresses that the addressee has worked to a degree that exceeds a certain contextually given norm. This extremely-high-degree reading arises only when the wh-element does not function as an independent clausal constituent; cf. Bennis (1995/1998). In (423b), for example, the wh-element hoehow functions as a manner adverbial and this leads to an interrogative interpretation. Similarly, in (423c) the wh-element wat functions as a direct object and the construction must again be interpreted as a question. Ignore the elements Δi and ti in (423), which will be discussed shortly.

Nevertheless, Corver (1990) and Zanutinni & Portner (2003) hypothesize that wh-phrases in questions and exclamations perform a comparable function; they are operators that bind some variable in the clause. This means that questions and exclamations are similar in that they both denote open propositions or, in other words, sets of alternative propositions. For example, the manner adverb hoehow in question (423b) gives rise to an open proposition that denotes a set of alternative propositions that differ in manner: the addressee may have worked well, badly, hard, with pleasure, with reluctance, and so on. The non-wh part of the exclamation in (423a) can also be seen as an open proposition, but in this case the alternative propositions differ only in degree (here: intensity), which is why we have represented the variable by the Greek capital Δ. Of course, the representation in (423a) does not yet answer the question why this example is normally used to express an extremely high degree, i.e. that the addressee has worked exceptionally hard. We will return to this question in Subsection D.

We have claimed above that the exclamative wat in (423a) does not function as a clausal constituent. To substantiate this, we should show that wat differs from hoe in (423b) in that it cannot be used as a manner adverb. A first reason to assume that they are different is that (423a) does not allow an interrogative interpretation: if the wh-phrase wat were a manner adverbial, this would of course be quite surprising. Another reason is that exclamative wat is also possible when a manner adverbial is overtly expressed; this is shown in (424a), where wat can be assumed to bind a degree variable Δ of the manner adverbial hard. Note in passing that it is unlikely that Δ stands for a wh-trace of exclamative wat, since degree adverbs usually cannot be wh-moved from the preadjectival position. The (b)-examples illustrate this for the degree adverb hoe by showing that this wh-element obligatorily pied-pipes the full AP.

That wh-movement is not involved in the derivation of the type of wh-exclamatives under discussion is also clear from the fact, illustrated in (425a), that wat can bind a degree variable embedded in an attributive modifier of a noun phrase. The (b)-examples show that wh-movement of the degree modifier hoe again leads to an unacceptable result in questions, as does, in fact, wh-movement of the full attributively used AP; the only option is movement of the full noun phrase, in (425a'').

Subsection II will provide more evidence for the assumption that the derivation of examples like (423a), (424a), and (425a) does not involve wh-movement, but for now we will simply assume that exclamative wat is base-generated in the clause-initial position in suchlike examples. Furthermore, we assume that exclamative wat requires the presence of a degree variable in order to be licit. This requirement can be made to follow from a generally accepted economy constraint on natural language, namely that an operator is only licit if it actually binds a variable: if an operator does not bind a variable, it is superfluous and should be omitted. This ban on vacuous quantification is also empirically motivated, since it provides a simple explanation for the acceptability contrast between the two examples in (426), taken from Krijgsman (1983). Under the plausible assumption that the degree variable Δ can only occur with gradable adjectives, exclamative wat can be licensed by a gradable adjective such as grootbig but not by a non-gradable adjective such as houtenwooden. Note that the number sign indicates that (426b) is marginally acceptable when wat is associated with a contextually determined gradable property that is left implicit, e.g. with the meaning “impressive”, a possibility found in the perfectly acceptable sentence Wat is dat een huis!What an impressive house this is!.

The ban on vacuous quantification may also account for the acceptability contrast between (427a) and (427b); the fact that (427a) is perfectly acceptable is due to the fact that the degree modifier erg itself is gradable, as shown by [[heel erg] mooi], while the degraded status of (427b) is due to the fact that zeer is not gradable, as shown by *[[heel zeer] mooi]; cf. Krijgsman (1983). The same can perhaps be said for comparative forms such as mooiermore beautiful in (427c), since these cannot be modified by degree adverbs such as heel (cf. *heel mooier), although this raises the problem that comparatives do allow modification by quantifiers such as veelmuch (cf. veel mooiermuch more beautiful); we leave this problem for future research.

The acceptability contrast between (428a) and (428b) also follows from the ban on vacuous quantification: example (428a) is acceptable because exclamative wat properly binds a degree variable associated with the gradable quantifier veel in (428a), while (428b) is unacceptable because cardinal numbers are not gradable and thus cannot introduce a degree variable. Example (428c) is unacceptable for the same reason: definite noun phrases such as het antwoordthe answer do not contain a degree variable.

The account of the acceptability judgments in (426) to (428) based on the ban on vacuous quantification provides strong support for the hypothesis that wh-elements in wh-exclamatives are operators that must bind a phonetically empty degree variable.

The extremely-high-degree reading discussed in Subsection A is not the only reading found with wh-exclamatives: when the wh-element in clause-initial position is associated with a certain type of noun phrase, an extremely-large-quantity reading can also occur, as shown in (429a). The examples in (429b&c) show that the noun phrase must meet certain criteria in order for the extremely-large-quantity reading to be possible: it must be headed by a plural count noun such as boekenbooks, which must be preceded by the spurious indefinite article een. The term “spurious” is used here because the indefinite article een cannot normally be used in plural noun phrases; cf. Section N16.1.

The number sign in (429b) indicates that this example is at least marginally acceptable with an extremely-high-degree reading, in which case wat is associated with some contextually determined gradable property that is left implicit, such as “impressive”. In fact, this is also true of (429a), which is therefore ambiguous; cf. Subsection C for more examples of such ambiguities.

A non-count noun such as waterwater is also compatible with an extremely-large-quantity reading: it occurs in the singular (since non-count nouns have no plural), but again must be preceded by the spurious indefinite article een, as is clear from the fact that example (430b) is unacceptable.

If Zanutinni & Portner (2003) is correct in assuming that exclamative wh-phrases are operators that must bind some variable, then the acceptability contrasts in (429) and (430) strongly suggest that the spurious article een is capable of introducing a variable that ranges over quantities; cf. Bennis (1998) for a similar conclusion.

Plural noun phrases such as (431a), which contain both a gradable attributively used adjective and the spurious article een, are ambiguous between an extremely-high-degree and an extremely-large-quantity reading. If we omit the spurious article, as in (431b), the extremely-large-quantity reading becomes unavailable. If we omit the gradable adjective, as in (431c), the extremely-large-quantity reading becomes the most salient one (although an extremely-high-degree reading remains at least marginally possible with respect to some contextually determined gradable property that is left implicit). If we omit both the spurious article and the gradable adjective, the result is unacceptable.

The above interpretations and judgments are all to be expected if the spurious indefinite article een and the gradable adjectives are capable of introducing a degree variable that can be bound by the exclamative operator wat. However, if the spurious article een and the gradable adjective mooi in (431) are indeed both capable of introducing a degree variable, then we expect example (431a) to express the extremely-high-degree and the extremely-large-quantity reading at the same time, since Subsection IIB will show that the exclamative wat is can bind more than one variable. It seems that example (431a) is indeed capable of expressing these two readings simultaneously, but it is not clear whether this is obligatory, given that the extremely-high-degree reading is the most prominent and, for some speakers, even the only possible one. If the extremely-large-quantity reading is optional, we may have to conclude that spurious een has some additional function besides introducing a quantity variable; we leave this issue for future research.

Since Elliott (1974) and Grimshaw (1979) it is generally accepted that exclamatives are factive in the sense of Kiparsky & Kiparsky (1970); the speaker presupposes the truth of the proposition expressed by the utterance. Thus, a speaker who utters the exclamation in (423a), repeated here as the first part of (432), presupposes that the addressee has worked today. This can be seen from the fact that this utterance cannot be felicitously followed by the question in the second part of (432), since it contradicts the presupposition by questioning its truth. We have indicated this by using the dollar sign.

Exclamations and questions differ crucially in this respect, as can be seen from the examples in (433). While the exclamation in (433a) cannot be followed felicitously by the question Of heb je geen boeken gekocht? because it questions the truth of the presupposed proposition, the question in (433b) can be followed by it; this shows that the speaker does not presuppose that the addressee has bought books by uttering the question Welke boeken heb je gekocht? (although we can of course argue that the speaker does expect that the addressee has bought books).

Elliott and Grimshaw further support the claim that exclamatives are factive by showing that they cannot be selected by non-factive verbs; while we find exclamative clauses as complements of the factive verb wetento know, such clauses do not occur as complements of the non-factive verb bewerento contend.

That the speaker presupposes the truth of the proposition expressed by the embedded exclamative clause is also clear from the acceptability contrast indicated in (435): cf. Grimshaw (1979:283). Because the speaker presupposes the truth of the proposition expressed by the exclamative clause, the use of the first-person pronoun in (435b) leads to an incoherent result, since the speaker cannot deny having knowledge of the truth of a proposition that he presupposes to be true. Example (435a), on the other hand, is coherent; the speaker can easily deny that Marie has knowledge of the truth of a proposition that he presupposes to be true.

Subsections A through C have shown that wh-exclamatives prototypically express an extremely-high-degree or extremely-large-quantity reading. Other terms often used to describe the interpretation of exclamations include “surprise”, “unexpectedness”, “emotional reaction” and “noteworthiness”. Now consider the wh-exclamatives in (436), which are used to express that the book in question is very expensive and thus seem to imply the truth of the propositions expressed by the declarative clauses in the primed examples.

However, it would be wrong to conclude that the primeless and primed sentences are equivalent, since there are many cases in which speakers could easily use the primed examples without necessarily being able to use the primeless examples. For example, a 300-page hardcover book that costs 100 Euros (or dollars) would be considered very expensive by today's standards, so that any speaker could easily use the primed examples in (436) to discuss such a book. A speaker who opened the book and found out that it is written by a popular bestselling author would probably also be able to use exclamatives like (436a&b). On the other hand, a linguist who knows that the book is about linguistics would probably not use these exclamatives, knowing that many academic publishers charge twice as much for similar publications. This shows that it is the speaker’s expectation that determines the appropriateness of using wh-exclamatives.

Zanutinni & Portner (2003) claims that the notions mentioned above are not basic but rather pragmatic implicatures derived from the two core properties of wh-exclamatives discussed in the previous subsections: (i) such exclamations are constructions in which an operator binds a degree/quantity variable and thus denote a set of alternative propositions that differ in degree or quantity, and (ii) they are factive in that the speaker presupposes the truth of the proposition expressed by the non-wh part of the exclamation.

The above claim is based on a particular view on discourse semantics. In any conversation there is a set of propositions that the speaker and the addressee hold to be equally true, which are part of the so-called common ground. A proposition can only be successfully asserted, if it can be added to the common ground. Because the truth of the proposition expressed by the non-wh part of a wh-exclamative is already presupposed, such exclamatives are less useful for assertion. Since every utterance must have some function, wh-exclamatives must have a function (other than assertion) that is compatible with their factivity; Zanutinni & Portner proposes that this function is affecting, or more specifically, widening the common ground.

We will explain the notion of widening using the examples in (437). Suppose that the common ground includes a height scale applicable to adult humans, which ranges from 1.70 to 1.90 meters. The assertion in (437a) would establish that Jan occupies a high position on that scale. The wh-exclamative in (437b), on the other hand, widens that scale and places Jan on the extended part of it; this gives rise to the extremely-high-degree reading discussed in Subsection A. Note in passing that we might also expect an extremely-low-degree reading of (437b) to arise, but this can be ruled out by Grice’s (1975) maxim of quantity; the use of groottall to express this reading is blocked by its more informative antonym kleinshort.

Zanutinni & Portner does not discuss this, but it seems that widening does not necessarily lead to an extremely-high-degree reading of wh-exclamatives; what is predicted is simply a higher-than-expected degree reading, and it seems that this is correct. Suppose Jan has a garden that requires intensive watering. To conserve potable water, he has installed a 2000-liter water tank that is fed by rainwater. After a modest shower, he checks the contents of the tank and finds that it is already half full. Since this is much more than he expected, he can easily express his surprise by using the exclamation in (438a); the crucial point is that the adverb alalready makes it clear that we are not (necessarily) dealing with an extremely high degree, but simply a higher-than-expected degree. After the water tank is completely filled, there is a drought. Jan starts watering the garden and after two weeks he peeks into the water tank and to his surprise the tank is still half full. Since this is much more than he expected, he can easily express his surprise by using the exclamation in (438b); now the crucial point is that nog ’still’ makes clear that again we are not (necessarily) dealing with an extremely high degree, but with a higher-than-expected degree. For further discussion of the meaning of alalready and nogstill in these examples, we refer the reader to Sections A25.2.2, A25.2.3, and A25.3.1.

The examples in (438), which were inspired by a similar example in Castroviejo (2006), cited in Villalba (2008), clearly show that the extremely-high-degree reading prototypically found in wh-exclamatives is not an inherent part of the meaning of wh-exclamatives. This reading is pragmatically derived from the more basic semantic properties of exclamatives, as evidenced by the fact that it arises only under certain contextual circumstances.

Non-split exclamative wh-phrases can perform several syntactic functions. The examples in (441) show that they can easily be used as arguments and predicates; the wh-phrases are related to an interpretive gap within the clause with the function of subject, direct object, and complementive, respectively. Since this shows that non-split wh-exclamative constructions have the characteristic property of wh-movement in (440a), we indicate the interpretive gap by a trace. This subsection will show that this is fully justified, since the non-split pattern also exhibits the other characteristic properties of wh-movement in (440b&c).

The wh-movements indicated in (441) are obligatory; the unacceptability of the examples in (442) shows that leaving the wh-phrase in the position indicated by the trace leads to ungrammaticality. The number sign in (442c) indicates that this example is acceptable without an exclamative intonation contour if wat is interpreted as an degree modifier with the meaning “quite”; we can ignore this reading here. It should also be noted that, for unknown reasons, example (442c) improves considerably when the particle maar is added: cf. Die boeken zijn maar wat mooi! We leave this issue for further research.

The obligatoriness of wh-movement follows from the assumption that exclamative wat must be moved into the clause-initial position in order to create an exclamative operator-variable configuration; cf. the discussion in Subsection I. Since the clause-initial position can only be occupied by a single constituent, we should also conclude that exclamative wat can be part of a larger phrase and can pied-pipe that larger phrase under wh-movement. That pied piping is common in non-split wh-exclamatives is also illustrated by the two examples in (443), where the exclamative wat is more deeply embedded in a prepositional object and complementive, respectively: wh-movement of wat triggers movement of the whole PP.

Pied piping also occurs when exclamative wat is part of an adverbial phrase. This is illustrated in (444) by an adjectival and a prepositional adverbial phrase of manner.

The examples in (443) and (444) again illustrate that non-split wh-exclamatives exhibit the characteristic property of wh-movement in (440a): the wh-phrase in clause-initial position is the antecedent of an interpretive gap within the clause with different functions: argument, complementive, and adverbial.

Let us now turn to property (440b), according to which the antecedent-trace relation can be non-local in bridge-verb contexts. Extracting an exclamative wh-phrase from an embedded clause always produces a somewhat marked result, but there seems to be a consensus that it is possible when the matrix clause is headed by a bridge verb such as zeggento say; cf. Krijgsman (1983:132), Corver (1990: §4) and Bennis (1998).

That the examples in (445) are indeed relatively good becomes especially clear when we compare them with the examples in (446), in which the matrix clause is headed by the factive, non-bridge verb betreurento regret. In order to make the interpretation of these examples more plausible, we have replaced the adjective mooibeautiful with the adjective saaiboring, but the results are still infelicitous. From the contrast between the two sets of examples in (445) and (446) we conclude that non-split wh-exclamatives exhibit property (440b): the antecedent-trace relation can be non-local in bridge-verb contexts.

Finally, we show that non-split wh-exclamatives are sensitive to islands. First, the examples in (447) show that exclamative wh-phrases cannot be extracted from interrogative clauses.

Krijgsman (1983) shows that non-split wh-exclamatives are also sensitive to complex noun phrase configurations: example (448b) illustrates that it is impossible to extract an exclamative wh-phrase from a relative clause.

The examples in (447) and (448) thus show that non-split wh-exclamatives also exhibit the third and final characteristic property of wh-movement in (440c): the island sensitivity of the antecedent-trace relation. It is therefore safe to conclude that wh-movement is involved in the derivation of non-split wh-exclamatives.

Now that we have established that the non-split pattern is derived by wh-movement, our next task is to show that the pseudo-split pattern does not involve wh-movement. Consider example (449), which provides the pseudo-split counterparts of the non-split wh-exclamatives in (441).

The previous subsection has argued that the wh-phrases in clause-initial position in the non-split exclamatives in (441) are constituents, and a conceivable analysis of the split exclamatives in (449) would therefore be that wh-movement of exclamative wat does not have to pied-pipe the non-wh part of the larger constituent, but can also strand it. This would lead to the representations in (450), where t represents the wh-trace of the exclamative wat.

However, the following discussion will show that this analysis is not tenable: wh-movement cannot be involved in the derivation of the pseudo-split pattern. Instead, we will be led to assume that exclamative wat is base-generated in clause-initial position and that it enters into a syntactic dependency relation with a degree variable introduced by the gradable adjective mooibeautiful (in the present examples, which receive a higher-than-expected degree reading). The correct analysis of the examples in (449) is therefore the one sketched in (451), where Δ stands for the degree variable and the indices indicate the syntactic dependency relation between the exclamative wat and the variable.

The two analyses cannot easily be evaluated on the basis of the examples in (449), since they lead to more or less similar predictions, since wh-extraction is possible from interrogative wat-voor phrases functioning as subject or object. However, an evaluation can be based on the pseudo-split counterparts of the non-split wh-exclamatives in (443), where the exclamative wat is embedded in a PP. The primeless examples in (452) show that the pseudo-split counterparts of these examples are perfectly acceptable. The wh-movement analysis would assign to these examples the structures in the singly-primed examples, while the alternative hypothesis, according to which exclamative wat is base-generated in the clause-initial position, is given in the doubly-primed examples.

We will now show that the syntactic representations in (452a'&b') are syntactically ill-formed, which leaves us with the structures in the doubly-primed examples. We will do this with the help of a brief discussion of so-called wat-voor phrases, which are used to form questions and arguably involve wh-movement. The examples in (453) show that at first glance such wat-voor phrases behave exactly like wh-exclamatives in that they optionally split in certain cases; the reader can verify this by comparing the examples in (453) with the wh-exclamatives in (441a&b) and (449a&b).

However, wat-voor constructions exhibit a crucially different behavior from wh-exclamative phrases in that the split pattern is impossible when the wat-voor phrase is the complement of a preposition. It is important to note that the impossibility of the split pattern is consistent with the fact that wh-movement is normally excluded from prepositional phrases, the only exception being wh-movement from pronominalized PPs of the form waar + P P + what.

In short, the fact that the primed examples are unacceptable shows that Dutch prepositional phrases are islands not only for wh-extraction of their nominal complements, but also for wh-extraction of subparts of their nominal complement. Now that we have established this, we can return to the pseudo-split wh-exclamatives in (452). The fact that the two singly-primed structures derived by wh-movement are identical in all relevant respects to the unacceptable primed examples in (454) strongly suggests that they are syntactically ill-formed. We should therefore conclude that the wh-movement analysis in the singly-primed examples is incorrect; the alternative analysis in the doubly-primed examples in (452), according to which the exclamative wat binds the degree variable introduced by the gradable adjective, is therefore more promising.

A similar argument can be made on the basis of the pseudo-split counterparts of the adverbial wh-phrases in (444) repeated in (455); the wh-movement hypothesis assigns to these constructions the representations in the singly-primed examples, while the alternative hypothesis provides the structures in the doubly-primed examples. Because the earlier discussion of the examples in (452) to (454) has already shown that representation (455b') is arguably syntactically ill-formed, we will focus on the exclamative in (455a).

The reason to exclude analysis (455a') has to do with the fact that it violates another independently motivated restriction on wh-movement, namely that wh-movement of degree modifiers such as hoehow triggers pied piping of the full adverbial phrase of manner; stranding of the non-wh part of the phrase gives rise to a severely degraded result. The fact that the unacceptable structure in (456b) is similar to (455a') in the relevant respects again disfavors the wh-movement analysis.

The discussion above has shown that pseudo-split exclamative constructions are not sensitive to certain well-established islands for wh-movement, and thus do not exhibit the characteristic property of wh-movement in (440c). Pseudo-split wh-exclamatives do not exhibit the property in (440b) either; the examples in (457) show that the relation between the exclamative element wat and its associate (i.e. the trace/degree variable in the adjectival phrases headed by mooi) cannot be established in a non-local fashion in bridge-verb contexts; cf. Krijgsman (1983:150), Corver (1990: §4) and Bennis (1998). Note that the asterisk indicates that all three continuations are unacceptable

The unacceptability of the pseudo-split wh-exclamatives in (457) is difficult to explain on the basis of a wh-movement approach. One possibility would be to assume that the unacceptability is related to the fact that wat is subextracted from a nominal/adjectival phrase. However, the acceptability of the examples in (458) shows that such subextraction is possible in the case of uncontroversial wh-constructions involving wat-voor phrases; cf. Corver (1990) and Bennis (1998). The wh-movement approach would thus predict that at least the examples in (457a&b) are acceptable.

A more promising account of the unacceptability of the examples (457) can be built on the earlier proposal that exclamative wat must enter into a syntactic dependency relation with the degree variable Δ introduced by the adjective mooibeautiful. Since such syntactic dependencies are usually clause-bound, exclamative wat and Δ are simply not sufficiently local, the former being in the main clause and the latter in the embedded clause. Note that the clause-bound nature of the syntactic dependency makes it impossible to test whether the pseudo-split pattern is sensitive to islands evoked by e.g. embedded interrogative or relative clauses, since these inherently require the syntactic dependency to cross a clause boundary: all pseudo-split counterparts of the non-split examples in (446) to (448) are therefore (correctly) predicted to be impossible anyway.

The conclusion that non-split and pseudo-split wh-exclamatives have different underlying structures is consistent with the fact that they show different syntactic behavior in other respects as well. For instance, the examples in (459) show that they differ with respect to the presence of spurious een: while een is obligatory in non-split wh-exclamatives such as (459a) with a higher-than-expected degree reading, it can easily be absent in the corresponding pseudo-split counterpart.

We already mentioned in Subsection IC that it is somewhat surprising that spurious een can be present in pseudo-split wh-exclamatives such as (459b) with a higher-than-expected degree reading; we expect it to introduce a quantity variable that is not reflected in the meaning. However, it has been argued that the obligatory presence of spurious een in non-split wh-exclamatives is not only related to meaning, but may also have a syntactic motivation; cf. Bennis et al. (1998) for discussion. Such an approach to the obligatoriness of spurious een in non-split wh-exclamatives would be incompatible with the claim that the pseudo-split constructions are derived from the same base structures as their non-split counterparts, since we would then predict that spurious een is also obligatory in the pseudo-split pattern. The contrast between the two examples in (459) with respect to spurious een thus indirectly supports the hypothesis that non-split and pseudo-split wh-exclamatives have different underlying structures.

The conclusion that pseudo-split wh-exclamatives do not involve wh-movement may also explain why a single exclamative wh-element is able to bind more than one variable; cf. Corver (1990:110). This is illustrated in (460): the presence of the spurious article een in the examples in (460a&b) first shows that exclamative wat can be associated with either the subject or the object; the acceptability of (460c) further shows that wat can also be associated with the subject and the object at the same time.

This would be very difficult to explain in a wh-movement approach, because the one-to-many relation in (460c) is not found in the case of uncontroversial wh-extraction. We illustrate this with questions involving the wat-voor split: although the examples in (461a&b) show that both subjects and objects allow the wat-voor split, example (461c) shows that wat cannot be associated with two wh-traces at the same time; cf. Section 11.3.8, sub I, for further discussion of this restriction.

Finally, note that while example (462b) is significantly better than (461c), example (462b) is much worse than (460c). The contrast between the two examples in (462) would again be surprising if pseudo-split wh-exclamatives were derived by wh-movement of exclamative wat.

Consider again the wh-exclamatives in (474). We have seen that these constructions have several characteristic properties. First, they involve the exclamative wh-element wat, which acts as an exclamative operator. Second, the exclamative operator is licensed by binding a variable introduced by a gradable adjective and/or the spurious article een. Third, the operator and the variable enter into a syntactic dependency relation, which means that they must be part of the same clause. Example (474) also illustrates that there are two types of wh-exclamatives: non-split wh-exclamatives, like those in the primeless examples, which are derived by wh-movement, and pseudo-split wh-exclamatives, like those in the primed examples, which involve base-generation of the exclamative operator wat in sentence-initial position.

The introduction to this subsection has shown that the use of an exclamative intonation pattern is not sufficient to conclude that we are dealing with exclamatives, i.e. exclamations that are syntactically marked as such. We must therefore appeal to other properties to establish this. Subsection I has shown that wh-exclamatives are characterized by a higher-than-expected degree or a larger-than-expected quantity reading; this is a first clue for recognizing exclamative constructions. Subsection IA has also shown that the exclamative wh-element is licensed not as a clausal constituent but as an exclamative operator; we will consider the presence of this element as a second clue for recognizing exclamative constructions. A third clue is the presence of the spurious article een, although in this case we have to be more careful because it can also occur in other construction types, such as the interrogative wat-voor construction: cf. Wat voor een boeken heb je gekochtWhat kind of books did you buy?.

On the basis of these three clues, we can safely conclude that the examples in (475) are genuine exclamative constructions; indeed, it seems reasonable to assume that they are derived from the primeless examples in (474) by some form of ellipsis.

Other possible cases of exclamative constructions are given in the primeless examples in (476), which have a higher-than-expected degree or larger-than-expected quantity reading; although they actually allow both readings, we have indicated only the most salient one in square brackets. The fact that the noun phrases in these examples also contain the spurious article een can be taken as an additional clue that we are dealing with exclamative constructions.

If we are indeed dealing with exclamatives in (476), this raises the question of how the degree/quantity variable introduced by the adjective/spurious article is bound. Bennis (1998) has suggested that the ethical dative me is like the exclamative wat in that it can function as an exclamative operator (which, of course, would be consistent with the fact that the ethical dative always expresses the speaker’s emotional involvement). The fact illustrated in example (477) that the ethical dative blocks the use of exclamative wat can be used to support this: if the ethical dative is indeed an exclamative operator, then the addition of exclamative wat is blocked because it leads to vacuous quantification; it is not needed to bind the variable.

Bennis suggests that particles such as toch can also function as exclamative operators. This would be consistent with the fact that the primeless examples in (478) have a higher-than-expected degree or a larger-than-expected quantity reading, but the fact that toch does not block the use of exclamative wat may be a problem for this claim: if toch is indeed an exclamative operator, the addition of exclamative wat should be blocked because it would lead to vacuous quantification; the exclamative wat is not needed to bind the variable. In fact, the same problem occurs in Er staan me toch een (mooie) boeken in die kast!, where toch co-occurs with the ethical dative me.

Bennis solves the problem that the particle toch can co-occur with the exclamative wat and the ethical dative by assuming that toch can have other (adverbial) functions besides its use as an exclamative operator. Note, however, that it is not really clear a priori whether appealing to the presence of toch is necessary to account for the exclamative higher-than-expected degree or larger-than-expected quantity meaning of the primeless examples in (478). The reason for this doubt is that the examples in (479) also allow an exclamative reading: if the presence of the spurious article een is sufficient for assuming that we are dealing with exclamatives, these examples show that exclamative operators do not need to be realized phonetically. However, the assumption of a phonetically empty exclamative operator makes does not only make it unnecessary to appeal to the particle toch to account for the exclamative reading of the primeless examples in (478), but it also raises the question as to why we need the exclamative wat or the ethical dative in the first place.

For completeness’ sake, note that the spurious article can also be omitted in the examples in (479). The resulting structure in (480a) has the same higher-than-expected degree reading as (479a). The resulting structure in (480b), on the other hand, does not allow the larger-than-expected quantity reading, which is expected on the hypothesis that the quantity variable is introduced by een; the number sign indicates that a higher-than-expected degree reading is possible and can be attributed to a contextually determined gradable property that is left implicit.

Another type of exclamative construction without an overt exclamative operator is given in (481). The fact that (481a) contains the spurious article een can again be used for claiming that we are dealing with a syntactically marked exclamation. A plausible account of this example would be to assume that the noun phrase een boeken undergoes reconstruction into the object position of the verb hebben in the scope of an empty exclamative operator, as a result of which the degree/quantity variable can be properly bound; the details of such an analysis still need to be properly worked out. Although example (481b) does not provide a visible clue in favor of assuming exclamative status, we can assume it by analogy with example (481a).

The assumption of a covert exclamative operator in the examples in (481) can be supported by the fact that this would correctly predict that the exclamative operator what is excluded in this construction, as this would lead to vacuous quantification.

Note that the clauses following the noun and adjective are introduced by datthat; dat cannot be analyzed as a relative pronoun, because then it would not agree in number with its supposed antecedent; relative pronouns with a plural nominal antecedent appear as die, while relative elements with an adjectival antecedent appear as wat. We conclude that we are dealing with a complementizer: this conclusion will be relevant for our discussion in Subsection B.