2.1 Syntactic properties of the ‘quin’-clause

Quin is a wh-determiner; as such, it takes a noun phrase as its complement. Therefore, quin cannot select an adjective or an adverb, as shown in (3).

Other wh-words that can head wh-exclamatives in Catalan include que (‘how’), which combines with an adjective and quantifies over degrees, (4), and quant- (‘how much/many’), which combines with a noun phrase and quantifies over amounts of individuals, (5).

Quin also heads wh-interrogatives in Catalan, as shown in (6). The main syntactic difference with a quin-exclamative is that the quin-phrase in the latter can be followed by the complementizer que ‘that’, which is impossible in quin-interrogatives.Footnote ^[4]

Indeed, the sentence in (6) is a request to provide an answer as to the identity or the type of car that John got (e.g. this or that car, or a sports or a family car). It is not felicitously used to inquire about the degree of some contextually salient property (e.g. the car’s speed). Assuming that quin is the same lexical item in both wh-exclamatives and wh-interrogatives, it is not obvious how scalarity is conveyed in this particular type of wh-exclamative. Providing an answer to this is, in a nutshell, the goal of this paper.

Other Romance languages, such as Spanish, also include this type of wh-exclamative in their inventory. However, in Spanish, there is no overt distinction between the wh-specifier of a DegP wh-exclamative and a DP wh-exclamative, which in both cases surfaces as qué, as shown in (7).

I am hence using Catalan quin-exclamatives because they transparently show that there is a wh-determiner, namely quin ‘what, which’, that does not select a gradable adjective, but a noun phrase.

Two final cross-linguistic remarks are in order. First, let us take a look at the inventory of wh-words that may not head wh-exclamatives in Catalan, (8).

As noted in Castroviejo (Reference Castroviejo2006) for Catalan, interrogative wh-words (in contrast to phrases) cannot head a wh-exclamative unless the wh-word is the degree head com ‘how’ or quant ‘how much’. This is not the case in other languages, as pointed out in Chernilovskaya & Nouwen (Reference Chernilovskaya, Nouwen, Aloni, Kimmelman, Roelofsen, Sassoon, Schulz and Westera2012), Nouwen & Chernilovskaya (Reference Nouwen and Chernilovskaya2015).

Another relevant cross-linguistic difference regards embedding. While it has been shown in the seminal works on exclamatives (Elliott Reference Elliott1974, Grimshaw Reference Grimshaw1979) that wh-exclamatives only embed under factive predicates, (9), quin-exclamatives sometimes do not sound very natural when embedded under factive predicates whose only argument is the wh-clause (Castroviejo Reference Castroviejo2006, Gutiérrez-Rexach & Andueza Reference Gutiérrez-Rexach, Andueza, Capone and Mey2016) or else they are less felicitous than in other languages, (10).Footnote ^[5]

The natural way of expressing (10a) and (10b) in Catalan is by means of a how-clause, (11a) or a complex definite DP (a concealed exclamation, in Grimshaw’s (Reference Grimshaw1979) terms), (11b).

2.2 Semantic properties of ‘quin’-exclamatives

Building on Rett’s (Reference Rett2011) degree restriction, it can be shown that quin-exclamatives must be uttered to convey unexpectedness toward the high degree of a property held by some individual. This restriction says that only those wh-phrases that can range over degrees can introduce a wh-exclamative. In other words, both how and what in (12) range over degrees.

To prove this degree restriction, Rett proposes two scenarios, one which is only compatible with the degree restriction, and another one which is compatible with mere counter-expectation. Let us take (12b) as an example. Rett’s (Reference Rett2011: 418) scenario is as follows:

In such a scenario, the sentence in (12b) would be infelicitous. Alternatively, it would be felicitous if ‘it is used to exclaim that the desserts John baked instantiate some gradable property to a degree higher than the speaker expected’ (Rett Reference Rett2011: 418).

Let us now turn to a quin-exclamative example such as (2), repeated below for convenience.

The sentence (13) can be felicitously uttered if the speaker becomes acquainted with a car Joan bought, which she finds remarkable because of some gradable property that is held to a higher than expected degree (e.g. the car is very big, shiny, fast, or new). We can prove the claim that quin-exclamatives do not merely express counter-expectation – and thus observe Rett’s degree restriction – by adding that (13) cannot be felicitously uttered if the speaker expected Joan to buy a BMW instead of an Audi (note that we mention two top range brands, so it is clear that the surprise should be exclusively based on the difference in identity, rather than properties associated with each brand).

While it observes the degree restriction, a quin-exclamative cannot be felicitously used when the speaker expected a smaller amount of individuals to participate in an event or to hold some state, as shown in the following example:

To convey this content, the wh-word quant- ‘how much/many’ followed by a noun phrase can be used, as shown in (5) above, repeated below for convenience.

(15) conveys that the speaker is overwhelmed at the large amount of people who attended the concert.

The main thesis in Castroviejo (Reference Castroviejo2006) and Castroviejo (Reference Castroviejo and Puig-Waldmüller2007) is that wh-exclamatives in Catalan should be analyzed as a degree construction. This is basically motivated on two related facts. First, the subset of the inventory of wh-words that range over degrees can head wh-exclamatives (on this see also Rett’s (Reference Rett2011) degree restriction above). Second, even those wh-words that seem to range over individuals, namely quin, can have a separate overt degree phrase headed by tan ‘so’ or més ‘more’,Footnote ^[6] which select for degree properties. In this section, we focus on this last piece of data, illustrated below.

By mode of comparison, note that the adjective llampant ‘flashy’ follows tan/més, which, in turn, follows the noun cotxe ‘car’, while in English, flashy precedes the noun and follows a.

As observed in Brucart & Rigau (Reference Brucart and Rigau2002) and Castroviejo (Reference Castroviejo2006) a.o., adjectives (introduced by tan ‘so’ or més ‘more’) in quin-exclamatives must be gradable. Note that in (17), the relational adjective esportiu ‘sports’ is unacceptable.

As a matter of fact, the quin-phrase in a wh-exclamative cannot contain an overt gradable adjective unless it is preceded by més ‘more’ or tan ‘so’. That is, a sentence like (18) is only possible in a context where the speaker did not expect John’s fast car to have a contextually salient degree property to a particularly high degree (e.g. John’s fast car is extremely beautiful, modern or shiny).

If, in certain contexts, potential gradable properties for fast cars as a type of entity are easily accessible, then the sentence is acceptable. Imagine fast cars as a kind have properties that can be readily retrieved by speakers, for instance, beauty. Then (18) would sound acceptable, as a covert form of a sentence such as (19).

The borne out prediction is that the position of ràpid ‘fast’ in (19) can be occupied by a non-gradable adjective, such as polonès ‘Polish’ in (20).

As in the previous example, here the speaker expresses her amazement toward the degree of handsomeness of some Polish actor. It is not the case that she is surprised or admired at how Polish the actor is (the ethnic adjective is not coerced into being gradable). The sentence would be felicitous in a scenario where the speaker finds some specific actor more handsome than she expected. The reason why the set of actors is narrowed down to Polish ones has to serve some contrastive purpose. For instance, because the speaker wants to clarify that he is Polish and unexpectedly handsome given a set of international actors, or because he is unexpectedly handsome for a Polish actor (where the non-gradable adjective would make the comparison class explicit, cf. Klein Reference Klein1980, Kennedy Reference Kennedy2007b).

Again, as in (18), the tan/més-phrase can be omitted only under the condition that it can be easily retrieved why the speaker would be admired at a certain Polish actor (because he is very handsome, very professional, very tall, very rich, etc.).

2.3 ‘Més’ and ‘tan’ beyond ‘quin’-exclamatives

Before moving on to the previous and current analyses, let us complete the description of quin-exclamatives by adding relevant information about tan ‘so’ and més ‘more’, which precede the gradable adjective in a quin-exclamative. Up to this point, I have been glossing tan and més as ‘so’ and ‘more’, respectively. However, these degree expressions do not surface in the English translations. It should be pointed out that, while tan and més are interchangeable in the context of quin-exclamatives, this is not so in most contexts. More specifically, tan is the degree expression occurring in equative comparatives, (21a), and result clause constructions, (21b), whereas més is the comparative morpheme in superiority comparatives, (22).

As shown in e.g. Castroviejo (Reference Castroviejo and Puig-Waldmüller2007: 138), tan ‘so’ needs to combine with a gradable adjective (and hence the unacceptability of the sentences with a non-gradable adjective like quadrilàter ‘quadrilateral’), just like in quin-exclamatives, (23). The same holds for més ‘more’.

Such facts led Castroviejo (Reference Castroviejo2006, Reference Castroviejo and Puig-Waldmüller2007) to assume that tan ‘so’ and més ‘more’ are degree expressions that select a gradable adjective across these different constructions. Roughly – I defer the specifics of the proposal to Section 3 – these works assume that tan ‘so’ and més ‘more’ combine with a gradable adjective and establish a ${\geqslant}$ or ${>}$ relation, respectively, between two degrees. For ease of comparison with other proposals, here I assume the classical denotations for the comparative and the equative morphemes as degree quantifiers (i.e. as relations between sets of degree properties), following Bresnan (Reference Bresnan1973), Heim (Reference Heim, Jackson and Mathews2000) and many others after her, (24).Footnote ^[7] (21b) can be more perspicuously expressed as (25a), and (22) as (25b).Footnote ^[8]Footnote ^[9]

Since tan ‘so’ and més ‘more’ express different relations between degrees, the immediate question arises, how come they are interchangeable when they appear within the quin-phrase of a quin-exclamative? I will address this question in Section 4.4. Before that, let us mention that there is, in fact, another context besides quin-exclamatives in which tan and més do not seem to make a noticeably different contribution, namely sentences in which there is no overt standard of comparison (an as-, that- or than- clause or phrase) and a special intonation contour is used, which I translate with an exclamation mark, (26). Informally, this contour describes a truncated continuation,Footnote ^[10] so it has the shape of a rising contour.Footnote ^[11] Footnote ^[12]

Castroviejo (Reference Castroviejo2008) states that such truncated sentences are felicitously uttered by a speaker who is so emotional at the high degree of a certain property held by some individual, she cannot describe the standard of comparison. In other varieties, such as Québec French, an apparently similar structure is possible, according to Burnett (Reference Burnett, Colina, Olarrea and Maria Carvallo2010). She calls these sentences gradation exclamatives, illustrated below:

As argued for by Burnett, the semantics of the degree word assez ‘enough’ in Québec French is neutralized and limited to the expression of extreme degree when emphatic prosody applies to the degree word (as expressed by upper-case letters). It should be noted that the option of neutralizing the lexical meaning of the equivalent of assez ‘too’ in Catalan is not available, even in the presence of an emphatic prosody. Among the inventory of degree words in Catalan, only the ones for so and more are available both in quin-exclamatives and these truncated constructions.

To wrap up, tan ‘so’ and més ‘more’ are degree quantifiers that describe relations between sets of degrees. In particular, they assert the ${\geqslant}$ and ${>}$ relation, respectively, between the maximal degrees in such sets. This said, we have identified two contexts in which the different relation, ${\geqslant}$ vs. ${>}$, which elsewhere distinguishes equative from superiority comparative constructions, does not seem relevant. One construction is quin-exclamatives and the other one is truncated degree constructions with an exclamatory prosody.

2.4 Interim summary

Summing up, the properties that characterize quin-exclamatives and must be accounted for are the following, in a nutshell:

• ∙ Syntactically, quin-exclamatives are wh-clauses introduced by a wh-determiner that selects a noun phrase and which can have an overt complementizer que ‘that’. Moreover, the head noun can be modified by a degree phrase headed by degree quantifiers tan ‘so’ or més ‘more’.

• ∙ Semantically, they observe Rett’s (Reference Rett2011) ‘degree restriction’. More specifically, whenever the degree phrase headed by tan or més is overt, the speaker felicitously utters a quin-exclamative if she did not expect someone or something to hold the property denoted by the adjective preceded by tan/més to such a high degree. In the absence of the overt degree quantifier, even in the presence of an overt gradable adjective, the degree restriction forces the speaker to express unexpectedness toward the fact that someone or something holds some contextually salient property to a high degree.

• ∙ Tan ‘so’ and més ‘more’ are also the degree quantifiers that head equative and superiority constructions, respectively. While they convey different degree relations in such constructions, they are interchangeable in quin-exclamatives.

Quin-exclamatives are very similar to English what-exclamatives except for their building blocks, notably, the gradable property the speaker is surprised about has to be preceded in quin-exclamatives by degree quantifiers tan or més, which head a DegP that combines with the head noun, while in English a bare gradable adjective modifies the head noun after the wh-word what has moved leaving a degree-denoting trace behind (Rett Reference Rett2011). The goal of this paper is to explore the compositional consequences of this difference.

In what follows, we will take into account the laid-out properties of quin-exclamatives to critically assess Castroviejo’s (Reference Castroviejo2006) previous analysis.

4.1 Background: A degree property and an expressive illocutionary operator

As mentioned in Section 2.2, Rett (Reference Rett, Friedman and Ito2009) proposes a Degree Restriction that wh-exclamatives in English observe, according to which exclamatives can only receive degree interpretations. A direct consequence of this restriction is that only wh-words that range over degrees can head a wh-exclamative in English. This translates in the semantics as the need for the wh-clause to denote a degree property (of type $\langle d,\langle s,t\rangle \rangle$). In proposing this, Rett adopts another semantics that had already been proposed for certain wh-clauses (free relatives in Jacobson Reference Jacobson, Bach, Jelinek, Kratzer and Partee1995 and Portner & Zanuttini Reference Portner, Zanuttini, Elugardo and Stainton2005, and questions in Groenendijk & Stokhof Reference Groenendijk, Stokhof, Chierchia, Partee and Turner1988). Hence, this analysis takes into account the wh-component in wh-exclamatives.

In addition to this, this degree property feeds an illocutionary force operator Rett (Reference Rett, Friedman and Ito2009) names Degree E-force (but see also Gutiérrez-Rexach Reference Gutiérrez-Rexach, Garret and Lee1996 for an EXCL operator, and Castroviejo (Reference Castroviejo, McNabb, Bane, Bueno-Holle, Grano and Grotberg2010) and Grosz (Reference Grosz2012) for similar ideas), which is a function whose domain is a degree property and its range, an expressive speech act that introduces the notion of unexpectedness, here viewed as the expression of expectation contravention, as shown in (34).Footnote ^[14]

The interpretive effect of applying the Degree E-Force operator to a wh-exclamative denotation is the proposition that the speaker is surprised that a specific (high) degree holds of a contextually salient degree property. Importantly, this is not an asserted proposition, so it cannot be true or false, but rather it has to be considered expressively (in)correct, in the sense of Kaplan (Reference Kaplan1999).

Later on, in Rett (Reference Rett2011) a full-fledged semantic derivation is provided that shows step by step how a degree property is obtained through the semantics of what, (35), and the composition of a what-exclamative, (36).Footnote ^[15]

Note that, at the end of the derivation, when the two variables, $x$ and $d$, are lambda-bound, the individual argument undergoes existential closure. Note, too, that delicious desserts has combined through Predicate Modification (i.e. intersection) once what has moved leaving a trace behind of type $d$.

This theory manages to restrict wh-exclamatives to those wh-clauses that can denote a degree property, and it also makes a compositional proposal as to why they are not assertions, but rather expressive acts. These are components that I will implement in the analysis of quin-exclamatives with slight modifications. But before we get there, I will consider the semantics of the degree and the wh- components of quin-exclamatives separately.

4.2 The degree component

In this subsection I show that there are two ways in which adjectives can combine with the noun selected by quin, depending on whether they first combine with the positive pos morpheme (as in Cresswell Reference Cresswell and Partee1976, von Stechow Reference von Stechow1984, Kennedy Reference Kennedy1999, and many others after them) or with tan/més. To do so, I go back to some of the properties laid out in Section 2. Let us start with a basic example:

Remember that the gradable adjective is at first blush marginal without the degree head, (38a), unless the speaker wants to convey an interpretation compatible with an additional DegP headed by més/tan, (38b).

The sentence in (38b) is used to exclaim that the flashy car Laia bought instantiates beauty to a degree higher than the speaker expected. It is not used to exclaim that the car Laia bought instantiates flashiness or flashiness and beauty to a degree higher than the speaker expected. Thus, to make sense of example (38a), which does not include the degree phrase headed by tan/més ‘so/more’, the addressee needs to retrieve some salient gradable property that a certain flashy car instantiates to a degree higher than the speaker expected. If it is not easily retrievable which gradable property such a subset of objects (i.e. flashy cars) may have, the sentence sounds unnatural.

Consider again the Catalan sentence in (39) and the English translation, paying attention to the ordering of the adjectives. The key difference is that the adjective that immediately precedes the noun need not be gradable, while the adjective whose high degree the speaker did not expect needs to be gradable. This is granted in Catalan, because tan and més are degree quantifiers, which take sets of degrees as arguments.

To be slightly less informal, the NP actor polonès ‘Polish actor’ denotes the (characteristic function of the) set of individuals that are actors and Polish, (40), and cotxe llampant, the set of cars that are flashy (to a contextually determined standard degree), (41).Footnote ^[16]

That is, if the gradable adjective (of type $\langle d,\langle e,t\rangle \rangle$) has to modify the head noun, we have to assume that it first merges with the positive morpheme pos to yield a property of individuals that can intersect with the head noun denotation (via e.g. Heim and Kratzer’s (Reference Heim and Kratzer1998) Predicate Modification), as shown in (42).

Note that, before merging with quin ‘what’, the NP cotxe llampant ‘flashy car’ has the same type as actor polonès ‘Polish actor.’ That is, staying in an extensional semantics, we are left out with (the characteristic function of) a set of individuals. The degree argument of the adjective has been bound by the existential quantifier introduced by pos, so the DegP does not have any degree argument left to be bound by, e.g. an expressive operator. In other words, as we will see shortly, given this kind of composition between the DegP and the head noun, the sentence cannot be used to exclaim that the car Laia bought is flashier than the speaker expected.

Moving now to the composition of the noun with a DegP headed by tan/més, I propose that the existence of these degree quantifiers prevents the adjective from combining with pos. In the following, I spell out my assumptions regarding the denotation of these degree heads, and then I propose a semantic derivation.

In Section 2.3, it was shown that tan ‘so’ and més ‘more’ occur beyond the domain of quin-exclamatives. In an attempt to overcome one of the undesirable results of Castroviejo’s (Reference Castroviejo2006) proposal, here I assume a denotation for the two degree quantifiers that is viable for comparatives and equatives, as well as for quin-exclamatives. I repeat them below for convenience.

It was pointed out in Section 3 that as gives rise to the ${\geqslant}$ relation between reference and standard degree, whereas more yields the ${>}$ relation. Unlike in Castroviejo (Reference Castroviejo2006), I do not see the need to stipulate that tan and més have a specific denotation in the context of quin-exclamatives or to lexically specify that the standard degree is a salient high degree. Instead, I assume that the standard degree is an argument that can be provided either by a phrasal/clausal complement or context; and this latter option becomes available only when the degree is salient. An argument in favor of this idea comes from the fact that, if the standard degree of tan / més is grammatically encoded in the quin-exclamative, then the sentence is unacceptable, as shown in (44).

Since in the case of quin-exclamatives the standard is not pronounced as a phrasal/clausal complement, I will follow Heim (Reference Heim, Jackson and Mathews2000) in the denotation of examples such as (45) in assuming that the first argument of the degree head can rigidly denote a degree.

This is the same idea adopted by Burnett (Reference Burnett, Colina, Olarrea and Maria Carvallo2010) in her denotation for Québec French tellement …que (‘so …that’), as laid out in (46) below.

Returning now to the semantic composition of a noun which merges with a tan/més-headed DegP, I claim that the derivation of the quin-phrase in (47a) is as in (47b).

As it will be argued for in the next subsection, in (47b), the degree quantifier first combines with a standard (called $d^{\ast }$), which is taken from context, and then moves to adjoin to the CP leaving a trace of type $d$ behind, as is characteristically done in cases of Quantifier Raising. Hence, the derivation of the wh-phrase proceeds with a free degree variable until it is lambda-bound through predicate abstraction. The resulting degree predicate is the second argument of the degree quantifier. Once they have merged, at the very end of the derivation, the standard $d^{\ast }$ remains a free variable. By contrast, in (42), the adjective llampant ‘flashy’ has had its degree variable bound by the existential quantifier introduced by pos; in fact, cotxe llampant ‘flashy car’ denotes a function from individuals to truth values that is true only if these individuals are cars whose flashiness meets a standard degree calculated from a comparison class of similar individuals. So, the denotation of the NP does not include a free degree variable waiting to be eventually bound by a force operator.

Equipped with these denotations, we can go back to the semantic composition of the quin-exclamative.

4.3 The wh-component

As has been motivated in Section 2.2, quin-exclamatives are sensitive to Rett’s (Reference Rett, Friedman and Ito2009, Reference Rett2011) Degree Restriction. In Section 4.1, I adopted Rett’s translation of this restriction into the denotation of wh-exclamatives as degree properties. In what follows, I intend to show how a degree property can obtain from a quin-exclamative with a focus on the role of tan and més in making this possible.

As pointed out in Section 4.2, the key component in the derivation of a quin-exclamative is that the gradable adjective does not combine with pos. Instead, a degree quantifier is present, whose trace is left as a free degree variable until the very end of the derivation. The present proposal is that it is eventually existentially bound by an expressive operator, as in Rett (Reference Rett, Friedman and Ito2009) (on this see also Section 4.4).

Let us start with a step-by-step derivation of example (48).

Starting from the bottom, as shown in  and , from a semantic point of view, the que-clause is analyzed along the same lines as a relative clause and, thus, as involving predicate abstraction over an individual variable. Since tan/més have been analyzed as degree quantifiers, they leave a degree-denoting trace and adjoin to CP. As a result, the gradable adjective combines with a degree and the output is a predicate of individuals (), which intersects with the head noun denotation yielding a more restricted predicate of individuals (). Unlike Castroviejo (Reference Castroviejo2006), I am not analyzing quin as a generalized quantifier, but rather as a function from properties to properties, with the aim of obtaining a predicate of individuals, in line with Rett (Reference Rett2011) (cf. Section 4.1). As proposed by Rett, I assume that the individual argument undergoes existential closure. However, unlike English wh-exclamatives in Rett’s paper, here there is no lambda-bound degree variable. As shown in , the Quantifier Raising of the degree quantifier involves degree abstraction, so the free degree variable within the CP gets lambda-bound. This is the second argument of the degree quantifier, whose first argument is a salient degree in context. So, even in , when all the functions have been saturated, the output is an open proposition. The fact that the standard degree is treated as a contextually supplied variable permits that $d^{\ast }$ remain free until, as we will see in Section 4.4, it is bound by an expressive force operator.

Here, I follow the lead in Burnett (Reference Burnett, Colina, Olarrea and Maria Carvallo2010), who emphasizes the equivalence between an open proposition with a free degree variable and a degree property. As mentioned in Section 2.3, she studies gradation exclamatives, as illustrated in (51).

In her analysis, where she claims that gradation exclamatives denote a degree property that feeds Rett’s (Reference Rett, Friedman and Ito2009) DEGREE E-FORCE operator, she argues that the standard degree of these constructions is recoverable from context. At the end of the derivation, a denotation for (51) would be as in (52).

Since the value of the degree variable is taken from context, it depends on an assignment function $\unicode[STIX]{x1D6FC}$. Thus, an assertion of (52) will be true or false depending on the value of the degree argument provided by $\unicode[STIX]{x1D6FC}$. This, Burnett argues, makes (52) equivalent to (53).

Likewise, for the case of quin-exclamatives, which at the end of the derivation denote an open proposition, I want to assume that this denotation is equivalent to a degree property. That is – and sticking to extensions for simplicity – the output obtained in (50)  is equivalent to (54).

Recall from Sections 2.2 and 4.2 that the gradable property the quin-exclamative is about has to be preceded by degree quantifiers tan and més. If, on the other hand, pos occupies the position of the degree head, the degree interpretation of the exclamative has to concern a contextually supplied gradable property (on this, see Section 6). I repeat below the relevant example for the sake of convenience.

As noted before, a speaker cannot felicitously utter (55) if she becomes acquainted with a car Joan bought and she expresses surprise because she did not expect that Joan bought a car that can run at such a high speed. Instead, the only way to turn this sentence felicitous is by recovering a gradable property from context that fast cars can have to a high degree. (56) would be such a case.

Here the speaker is surprised that the fast car Joan bought has such a high degree of beauty. In the present analysis, this is the expected result, because the semantic derivation would proceed as follows:

What is important to note from this derivation is that the variable that remains unbound at this point of the derivation is not the degree of fastness of the car (which is specified to hold to at least a contextual standard determined through a comparison class of individuals), but rather the threshold met or exceeded by the degree to which the car is beautiful. As argued for above, the open proposition that obtains at the very end of the derivation is equivalent to a function from degrees to truth values (i.e. a degree property), which feeds the expressive operator, to which we now turn.

4.4 An expressive operator

Along with Rett (Reference Rett, Friedman and Ito2009) and Castroviejo (Reference Castroviejo, McNabb, Bane, Bueno-Holle, Grano and Grotberg2010), I assume there is an expressive force operator which contributes non-at-issue content,Footnote ^[17] and which can apply to either the denotation of a declarative clause, (58a), or a wh-clause, (58b).

Regarding the lexical semantics of this operator, which I call EXP-OP, I side with Rett (Reference Rett2011) in encoding the emotion expressed as the speaker’s unexpectedness. However, I follow Rett (Reference Rett, Friedman and Ito2009) in proposing two different lexical entries, one for propositional exclamations, where the speaker conveys that she did not expect the proposition denoted by the declarative cause to be true, (59), and one for wh-exclamatives, where the speaker conveys that there is a degree she did not expect the wh-clause denotation to apply to, (60).

As I have argued at the end of Section 4.3, the denotation of a quin-exclamative is an open proposition containing a free degree variable, which is equivalent to a degree property (type $\langle d,t\rangle$). Given the proposed semantics for EXP-OP$_{wh}$, when the quin-clause contains an unbound degree argument, this variable ends up being bound by the existential quantifier introduced by the expressive operator. This way, the (expressive) content conveyed is the existence of a degree $d$ (corresponding to the standard degree of tan or més) such that the speaker did not expect that a subject would have a certain property to degree $d$.

Let us go back to the flashy car example in (48), repeated in (61), to spell out the final output.

At the end of the derivation, the quin-clause denotes an open proposition of the following form: there is a car Laia bought whose degree of flashiness meets or exceeds a threshold degree that is retrieved from context. Since this degree receives a value through an assignment function, its truth value depends on the value of this degree, so the open proposition is equivalent to a function from degrees to truth values (cf. Section 4.3). Hence, the quin-clause denotation can feed EXP-OP$_{wh}$. The final output is the expression that there is a degree $d$ such that the speaker did not expect Laia to buy a car that is flashy to $d$.

This said, remember that in previous descriptive paraphrases, we would characterize a sentence such as (61) as conveying that the speaker did not expect the high degree of flashiness of some car. However, we have not introduced this bit of content in the aforementioned paraphrase. Does it follow from anything that has been discussed in the present proposal?

Bear in mind that the quin-exclamative in (61) is not felicitous in a context in which the car is less flashy than expected or has a funny degree of flashiness (if flashiness had a conventionalized system of measurement, it could have a weird number thereof, for instance). What the sentence conveys is that the actual degree of flashiness is unexpected and, starting from here, so would all the higher degrees. What Rett (Reference Rett, Friedman and Ito2009) does is state that $d$ exceeds a contextual standard in the sense of Kennedy (Reference Kennedy, Elliott, Kirby, Sawada, Staraki and Yoon2007a) (i.e. based on a comparison class of individuals, as yielded by the function standard). I, on the other hand, attempt to show that the expressive operator’s logical properties are responsible for this output.

More specifically, building on Nouwen (Reference Nouwen, Dekker and Franke2005, Reference Nouwen, Égré and Klinedinst2011), I assume that unexpectedness licenses upward-directed inferences in such a way that it behaves like a downward-monotonicity operator (Nouwen Reference Nouwen, Dekker and Franke2005).Footnote ^[18] That is, for any two potential competitor propositions that address the same Question under Discussion (QUD in the sense of Roberts Reference Roberts1996), $q,r$, if $r$ is unexpected and entails $q$, it follows that $q$ will also be unexpected (and thus potentially yielding the expression of an emotion). This is represented in (62).

Let us go back to imagining that there are different degrees of flashiness on a conventional scale, so we can think of $d_{1}$, $d_{2}$, $d_{3},\ldots d_{n}$ degrees of flashiness. As shown by e.g. Heim (Reference Heim, Jackson and Mathews2000), Nouwen (Reference Nouwen, Dekker and Franke2005), gradable adjectives are upward monotonic (i.e. they license downward-directed inferences).Footnote ^[19] So, if Laia’s car is $d_{3}$-flashy and $d_{3}>d_{2}>d_{1}$, then, Laia’s car is also $d_{2}$- and $d_{1}$-flashy. Importantly, when EXP-OP applies to the degree property, the sense of the entailment is reversed. So, if the speaker did not expect Laia to have a $d_{2}$-flashy car, this entails that she would not have expected Laia to have a $d_{3}$-flashy car (or any higher degree). That is, all the propositions that include a higher degree entail the ones including a lower degree. Therefore, unexpectedness follows for all the propositions that entail the one conveyed by the quin-exclamative once the degree argument is bound.

This monotonicity inference – which ensures that for any higher value of $d$ (however flashier the car may be), it holds that the proposition that the car is $d$-flashy is unexpected – is reminiscent of Nouwen’s (Reference Nouwen, Dekker and Franke2005) work on evaluative adverbs such as surprisingly in surprisingly tall. Here, too, evaluativity is tied with the higher values of a gradable predicate, not the lower or regular ones. The expression of unexpectedness is not directed toward a particular degree, either, but a set of degrees starting from a higher one, according to the speaker.

Viewed this way, there is no need to place any constraints that the threshold reached by the car’s flashiness must exceed a standard as introduced e.g. by pos (this would be similar to the road taken by Rett (Reference Rett, Friedman and Ito2009) in her DEGREE E-FORCE operator). Rather, the threshold in this approach corresponds to a degree that is unexpected to the speaker, and monotonicity ensures that unexpectedness is caused by a high degree (not low, not freakish).

Before concluding, let us go back to the question raised in Section 2.3 as to why both tan and més are interchangeable in certain contexts. It has been shown that these degree heads have different interpretations but are used interchangeably in quin-exclamatives and in truncated sentences with an exclamatory intonation revealing an emotional speaker, repeated below for convenience.

My take is that, in both constructions, the degree that is met or exceeded is an extreme degree in the sense of Morzycki (Reference Morzycki2012), i.e. one that goes beyond the salient range of values. In his own terms, the degree that is reached has gone “off the scale’ of contextually provided degrees.’ Morzycki uses this metaphor to represent the denotation of extreme adjectives such as gigantic, which builds on degrees of tallness that go beyond a so-called ‘perspectival scale’. In fact, the essential difference between big and gigantic is that the degree of bigness is in the perspectival scale in the former case, while it goes beyond the maximum of the perspectival scale in the latter. One of the puzzles that his theory addresses is the oddness of extreme adjectives in a comparative structure, as shown in (64).

In a few words – I will come back to this in Section 6 – the oddness of the sentence, viewed as a pragmatic rather than a semantic issue, has to do with attempting a comparison between two degrees that are not salient (because they are way above the perspectival scale). Coming back to the interchangeability of tan and més in certain contexts, we can assume that the degree that is met, which is an unexpectedly high standard, has similar properties as a non-salient degree, in the sense that the speaker did not conceive of some subject reaching such a high value. Given such an extreme measure, the distinction between ${\geqslant}$ and ${>}$ is too subtle to be distinguished.Footnote ^[20]

4.5 Summary

In this proposal, quin-exclamatives convey that the speaker did not expect a property to hold of some individual to such a high degree. Unlike in Castroviejo’s (Reference Castroviejo2006) analysis, here the wh-component is not ignored from the point of view of the semantic derivation. Following Rett (Reference Rett2008, Reference Rett, Friedman and Ito2009, Reference Rett2011), I treat the quin-clause as a degree property. A second improvement with respect to the previous proposal for quin-exclamatives is the full integration of the analysis of the degree expressions tan ‘so’ and més ‘more’ into the semantic derivation of the degree property; that is, I have not used an ad hoc denotation for the degree heads in this specific context, and I have been able to highlight the differences between the inclusion of pos and the inclusion of tan/més. Finally, I have replaced the notion of attitude toward a degree, which was not derived on syntactic or semantic grounds, with the unexpectedness conveyed through an expressive force operator.

Summing up, I follow Rett (Reference Rett2008, Reference Rett, Friedman and Ito2009, Reference Rett2011) in two important respects: First, wh-exclamatives are triggered by an expressive speech act operator. In fact, EXP-OP$_{wh}$ is a simplification of Rett’s (Reference Rett, Friedman and Ito2009) DEGREE E-FORCE. Second, quin-exclamatives are necessarily gradable in the sense that they are felicitous only when the speaker did not expect that the subject held a gradable property to a contextually salient degree. How to yield this result is not straightforward, though, because Catalan quin-clauses do not seem to range over degrees in contexts other than exclamatives. At the same time, quin-exclamatives do obey the Degree Restriction. I have thus proposed a technical implementation that yields an open proposition with a free degree variable at the end of the derivation of the quin-clause. The idea that wh-exclamatives denote high degree is ensured thanks to a monotonicity condition on EXP-OP, which licenses upward-directed inferences.

The analysis I have proposed is based on the main claim that whenever there is no DegP headed by tan ‘so’ or més ‘more’ preceding the gradable adjective, composition introduces the null morpheme pos turning the gradable predicate into a predicate of individuals. At the end of the derivation there is no degree variable to be bound by the exclamative operator, so this kind of quin-clause cannot be a good fit for the expressive speech act operator, EXP-OP. By contrast, tan and més are analyzed as degree quantifiers whose standard, a degree retrieved from context, ends up as a free variable that is ultimately bound by EXP-OP$_{wh}$.

In the following section, I evaluate the adequacy of alternative theories to account for the facts presented here (Section 5). In Section 6 I discuss the potential generalization of this proposal to DegP-less quin-exclamatives and the rest of possible quin-exclamatives in Catalan.

5.1 Zanuttini & Portner (Reference Zanuttini and Portner2003)

Zanuttini & Portner (Reference Zanuttini and Portner2003) is the first attempt to provide an analysis of exclamatives that maps syntax, semantics and pragmatics. The key ingredient of a wh-exclamative in Zanuttini and Portner’s analysis is the syntactic realization of factivity in a wh-clause. And the essence of an exclamative, its sentential force, is what they call ‘widening’: the formalization of a conventional scalar implicature that conveys that the true propositions in the set denoted by the wh-clause lie at the extreme end of a scale. Widening is the formal counterpart of descriptive terms such as surprise, admiration or mirativity.

In a nutshell, in their account, exclamatives have two main ingredients, a [+wh] feature (an integral part of wh-clauses, which involve an operator–variable relation) and a factive morpheme they call ‘FACT.’ Semantically, they follow other works such as Gutiérrez-Rexach (Reference Gutiérrez-Rexach, Garret and Lee1996) in claiming that wh-exclamatives have a classic question semantics as in Hamblin (Reference Hamblin1973) and Karttunen (Reference Karttunen1977); that is, they denote a set of alternatives that represent the true answers to the corresponding wh-interrogative. Take (65) as an example.

In a scenario where some friends are discussing about kinds of peppers a participant of the conversation eats, the denotation of the wh-clause is the set of propositions in (66).

Then, why are wh-exclamatives not wh-interrogatives? Because the presence of FACT is incompatible with inquiring. Whenever [+wh] and FACT co-occur, widening forces the standard set of answers to expand so as to include non-standard ones. On the other hand, factivity makes the propositions in the widened domain true. Following up on the previous example, widening amounts to considering extreme answers. For instance, ‘he eats güeros’, in (67), because this kind of pepper is too hot to be a regular pepper one would eat.

We have seen that in this account, [+wh] and FACT are the triggers of widening and, hence, the hallmarks of exclamatives. These ingredients rule out other exclamatory sentences that are not wh-clauses (or for which a [+wh] feature cannot be carved out). However, they do not rule out the possibility that certain wh-words can introduce wh-interrogatives, but not wh-exclamatives. To explain differences in the inventory of wh-words for interrogatives and exclamatives, Zanuttini and Portner appeal to their morphological make-up. More precisely, wh-phrases that are ‘E-only’ (i.e. introduce wh-exclamatives but not wh-interrogatives) contain an E-only morpheme that requires the presence of FACT. This explains that such wh-phrases cannot introduce interrogatives. Now, the possibility for a wh-phrase to include an E-only morpheme depends on its morphological make-up. In (68) we observe the difference between wh-phrases with and without the E-only morpheme.

While in English, the role of very is to encode the E-only morpheme, Zanuttini and Portner suggest that in Italian, this is the role of tanti (‘much/many’), as shown below. To be more specific, only t- encodes the E-only morpheme, while -anti expresses the measure.

Importantly, they assume that those wh-phrases without an overt E-only morpheme that can introduce wh-exclamatives contain a null E-only morpheme. This would explain that certain wh-phrases can introduce both wh-exclamatives and wh-interrogatives. This line of thought is also able to explain why Italian wh-heads like chi ‘who’ and cosa ‘what’ cannot introduce a wh-exclamative. In particular, Zanuttini & Portner (Reference Zanuttini and Portner2003) argue that such wh-phrases are morphologically complex words encoding wh and sortal, so the E-only morpheme cannot be inserted between the other two, which explains their absence in wh-exclamatives.

Let us go back to quin-exclamatives to test the predictions Zanuttini & Portner (Reference Zanuttini and Portner2003) make for them. For one, these authors expect them to have a question semantics, factivity being the reason why they do not have the sentential force of interrogatives. However, we have said at the outset that the evidence in favor of quin-exclamatives being factive is not as strong as in e.g. English, since wh-exclamatives in Catalan do not easily embed under factive predicates, so the motivation for the factive morpheme is weakened.

Additionally, in the set of propositions approach, the only way to account for the degree restriction is to appeal to the morphological make-up of certain wh-words. Following up on this strategy, an option would be to treat the DegP headed by tan/més as an E-only morpheme. The downside of this kind of strategy would be that tan and més would play the role of E-only morphemes, but this would not be informative about their actual lexical and compositional semantics, especially given that they occur in other degree structures (see Section 2.3). In other words, calling them E-only would be a descriptive label, but would not explain how they contribute to the overall denotation of quin-exclamatives. Hence, a proposal that incorporates the semantics of these expressions in the overall derivation of the quin-exclamative could be said to fare better.

5.2 Rett (Reference Rett2011)

As shown in Section 4.1, Rett (Reference Rett2011) diverges from Zanuttini and Portner’s account and argues that wh-exclamatives are a degree phenomenon. In terms of their semantics, these wh-exclamatives denote a degree property (type $\langle d,t\rangle$); that is, they involve lambda abstraction over the relevant variable, just like free relatives. This degree property feeds an illocutionary force operator Rett names E-Force, which is a function from propositions to expressive speech acts that introduces the notion of unexpectedness, here viewed as the expression of expectation contravention, (70).

Since the denotation of wh-exclamatives is a degree property, an operation other than functional application is needed to satisfy E-Force. Rett approaches this as a two-step process, (71). First, context provides an argument for the degree property. A proposition with an unbound variable results, (71b), which is the argument for E-Force. The $d$-variable gets eventually bound by existential closure, so when E-Force applies to it, it applies to a full-fledged proposition, (71c).

While I have adopted the degree property analysis as well as a version of the expressive operator, there are two reasons why this proposal cannot be straightforwardly implemented in the case of quin-exclamatives. First, I depart from the analysis of the wh-word as a degree operator. Specifically, I do not think there is enough motivation to analyze quin as a degree operator that moves away from the NP it selects, leaving behind a trace of type $d$, as in (36), repeated below for convenience.

While in English, such a mechanism is needed to allow for the intersection between the adjective (delicious) and the head noun (desserts), in the case of Catalan, as shown in Section 2, the head noun and the adjective are separated by tan ‘so’ or més ‘more’. The Catalan counterpart of What delicious desserts John baked! would be as in (73).

Observe that the gradable adjective delicioses ‘delicious’ and quin are not adjacent, as is the case in English. In Section 4.2 I have argued that tan and més are heads of a DegP that moves via Quantifier Raising and leave a degree-denoting trace. Hence, the head noun and the adjective can merge via Predicate Modification without quin intervening. The alternative analysis whereby quin directly modifies the gradable adjective and then moves upward to allow for the gradable adjective to combine with the degree-denoting trace the wh-word leaves behind does not seem to be justified. On the other hand, since quin seems to range over individuals of type $e$ in interrogatives, keeping the same analysis across clause types may be seen as a welcome output.

Second, in Rett (Reference Rett2011), high degree in wh-exclamatives is assumed to solely follow from existential closure of the degree variable. This is the same view Rett takes on the positive construction (e.g. John is tall), which is only felicitous in a context where this existentially bound degree is high. While this can stem from pragmatic reasoning (the non-evaluative reading would be trivial), in the case of E-Force, which plainly expresses unexpectedness toward a proposition which contains an existentially bound degree variable, it should be possible that this degree be unexpectedly low or a very funny number (on a similar issue, see Morzycki Reference Morzycki, Dölling, Heyde-Zybatow and Schäfer2008). Thus, the present analysis, where high degree is an entailment of the monotonic properties of EXP-OP, as proposed in Section 4.4, seems more explanatory.

5.3 Chernilovskaya & Nouwen (Reference Chernilovskaya, Nouwen, Aloni, Kimmelman, Roelofsen, Sassoon, Schulz and Westera2012)

Chernilovskaya & Nouwen (Reference Chernilovskaya, Nouwen, Aloni, Kimmelman, Roelofsen, Sassoon, Schulz and Westera2012) make the claim that wh-exclamatives are not necessarily a degree phenomenon, based on Dutch data, and evidence from German, Turkish, Russian and Hungarian. More specifically, they observe that, cross-linguistically, two classes of wh-exclamatives are available, depending on whether the speaker expresses noteworthiness toward the referent of the wh-expression or an entire proposition.Footnote ^[21]

Noteworthiness is defined as follows:

The difference between the two types of wh-exclamatives cross-linguistically is presented in (75) and (76), respectively.

Crucially, these two types of wh-exclamatives, with their own morpho-syntactic characteristics, have two different interpretations, depending on whether noteworthiness (the speaker’s exclamative attitude) is predicated of the referent of the subject or the proposition referenced in the exclamative. That is, the scope of the predicate noteworthy (whose domain may be either an individual or a proposition) is responsible for the two possible readings. (75) says that there is a song that John wrote that stands out when compared to other beautiful songs. (76) means that it is noteworthy that the speaker has seen precisely the person she has seen. Whereas what a is lexically specified to express noteworthiness toward the referent of the wh-word, wie ‘who’ is lexically specified to mark noteworthiness toward a proposition. Crucially, Type 1 exclamatives are scalar, in the sense that there has to be some property (in the example, beauty) that some object (the song) has to a high degree, so they obey Rett’s (Reference Rett2011) Degree Restriction. Even in the absence of the overt adjective, a Type 1 exclamative cannot convey that it is unexpected that John wrote a specific song. Instead, it must express that the song has some salient property to a high degree. By contrast, Type 2 exclamatives do not involve scalarity in the same sense, since degrees are not relevant. There still is scalarity, though, if we assume that noteworthiness as applied to a proposition involves evaluating the proposition in question with respect to a set of ordered alternatives.

In this cross-linguistic picture, quin-exclamatives would fall under Type 1, and the prediction that they are scalar is borne out. That is, in an example such as (77), the sentence cannot be used to express that John wrote one song instead of another one that the speaker expected. Rather, there has to be a salient property (e.g. beauty) that the song has to a high degree, which makes the song stand out in comparison to other songs.

The fact that Catalan (as well as other Romance languages) does not have Type 2 exclamatives (see Section 2, examples in (8)) is compatible with the typology presented by Chernilovskaya & Nouwen. However, there is a general worry, and then there are two aspects that seem a priori problematic if we were to extend this analysis to quin-exclamatives.

First, it is not clear how the inference of high degree follows from noteworthiness as applied to the referent of an individual. As far as I can tell, if an object stands out in comparison to other similar objects, the reason need not be related to scalarity. It could also be the case that an object stands out because it has a property that the rest of objects in the comparison class lack.

Second, in their formal rendering of (75), the comparison class to which the song John wrote is relativized contains beautiful songs. As Chernilovskaya & Nouwen (Reference Chernilovskaya, Nouwen, Aloni, Kimmelman, Roelofsen, Sassoon, Schulz and Westera2012: 276) put it, ‘The comparison class will normally be the class described by the wh-phrase.’ Importantly, it does not follow that the song John wrote, compared to a set of beautiful songs, is noteworthy because it holds a high degree of beauty. The formal expression permits that the song stands out in comparison to a set of beautiful songs because of a property this particular song has and the rest of beautiful songs do not have. For instance, it may be that the song John wrote is beautiful, but it stands out because it is also thematically weird. Conceiving the rest of objects to which the song is compared as beautiful does not seem to give rise to the scalar interpretation that the authors are after. More pressing for the case of quin-exclamatives, the presence of tan/més may not be easily integrated in such a theory, where the gradable adjective held by the subject is in the positive construction (the comparison class to which the subject is compared includes other subjects who also hold this property to a standard threshold). Remember from Section 4.2 that I have argued that tan/més and pos give rise to different semantic derivations, so we would want to rule out the possibility that the threshold degree of the adjective is existentially bound and established as a contextually determined standard based on a comparison class.

Finally, while the typology presented by Chernilovskaya & Nouwen seems empirically adequate, there is nothing in the analysis that explains why certain wh-elements can only give rise to Type 2 exclamatives and not Type 1 exclamatives. In the case of Catalan (and other Romance languages), nothing apparent seems to prevent that wh-heads such as qui ‘who’ or què ‘what’ introduce a wh-exclamative.

To wrap up, for different reasons, the potential implementations of the three alternatives we have reviewed to quin-exclamatives seem less successful than the proposal defended here. Let us now turn to considering the consequences of the current analysis for other quin-exclamatives.