A wh discourse particle: Dutch hoezo

. This paper provides an analysis of a special kind of why -questions in Dutch, formed with the wh -word hoezo (lit. ‘how so’). We propose that hoezo signals the speaker’s resistance to updating the discourse model, by asking the interlocutor to provide a reason for their preceding discourse move. Hoezo shares properties with both discourse particles (its not-at-issue content expresses the speaker’s attitude towards the current state of the discourse), and question words (it asks the interlocutor to provide a reason). Unlike waarom (the regular word for ‘why’), hoezo is not subject to wh -movement and does not bind a variable.

oretical analysis of hoezo, contrasting its syntactic, semantic and pragmatic properties with those of waarom.
The core idea underlying the analysis developed in this paper is that the (not-at-issue) contribution of hoezo is at the level of the discourse rather than at that of the truth-conditional sentence semantics.Hoezo challenges the interlocutor to provide a reason for their discourse move ("Why did you say that?").Even though hoezo is a wh-word just like why and waarom, at the same time it shares some of the properties of discourse particles.We thus propose the recognition of a novel lexical category, namely that of what we call a wh discourse particle.Because hoezo's main function is a pragmatic one, its syntax and semantics are "impoverished" compared to that of waarom/why.In contrast to waarom/why, hoezo is not a variable-binding operator.In the syntax this is reflected in the fact that hoezo does not undergo wh-movement but is base-generated in the left-periphery of the clause, more specifically in Spec-CP.This is in line with the common assumption in the literature that functional projections above the IP level are the locus of pragmatic and illocutionary operators, and hence the natural home for discourse particles (Grosz 2020;Wiltschko 2021).These syntactic and semantic differences between hoezo and waarom/why will be demonstrated through a battery of empirical tests.
Here we should mention from the outset that we do not exclude the possibility that waarom and why may have a pragmatic use that is similar (or identical) to that of hoezo, in addition to their 'regular' use as sentence-level variable-binding operators.In many, if not all, of the example with hoezo in this paper, it could felicitously be replaced by waarom.However, crucially, hoezo cannot function as a variable-binding operator and does not undergo wh-movement, as we will show in Section 5. Thus, whereas waarom/why may be ambiguous between being a sentencelevel variable-binding wh-operator or a pragmatic wh discourse particle, hoezo only has the latter interpretation.In fact, Ginzburg (2012) argues for such an ambiguity of English why, in his analysis of what he calls its 'metacommunicative' use (p. 308-316).Our analysis of hoezo is similar in spirit to his, but we will not make a detailed comparison in this paper.
The rest of the paper is organized as follows.In Section 2, we distinguish three types of utterances with hoezo based on formal criteria.Section 3 discusses its pragmatic uses, and Section 4 presents the core of our theoretical analysis.The consequences for the syntactic behaviour of hoezo are discussed in Section 5.The systematic differences with waarom observed there lend strong empirical support to our analysis.In Section 6 we further motivate our proposal to classify hoezo as a wh discourse particle, based on general characteristics of discourse particles commonly discussed in the literature.In Section 7 we summarize our analysis, and suggest some promising avenues for further research.
2. Syntactic configurations for hoezo.Using purely syntactic criteria, we can distinguish three types of hoezo-utterances, namely (a) hoezo occurring by itself (isolated hoezo), (b) hoezo plus a single word or phrase (hoezo + X), 4 and (c) hoezo as the initial question word in an interrogative 'Language window') of the popular monthly language magazine Onze Taal (n.d).Although this blog post is couched in prescriptive terms, it accords with our own native-speaker intuitions, pointing out that hoezo is less neutral than waarom ('why') and that it expresses 'surprise' or even 'indignation'.We will see that this claim is supported by our analysis, although we regard this aspect of hoezo as a more superficial pragmatic side-effect of its core meaning.There is no mention of hoezo in ANS (2021), the most comprehensive and up-to-date descriptive grammar of Dutch. 4 The analogous construction with why is known in the literature as 'why-Stripping'.Corver (2021) is a detailed discussion of the syntax of this construction in Dutch with waarom (but not hoezo).We'll have to leave a comparison between 'waarom-Stripping' and 'hoezo-Stripping' for another occasion.clause (CP) immediately followed by the inflected verb in the second position (hoezo + V2).In response to the assertion in (2), any of these constructions can be used, as illustrated in (3).In the hoezo + V2 construction, hoezo is syntactically integrated in the sentence in a way that is superficially parallel to other wh-words like wie ('who'), wat ('what'), and waarom ('why'), which have undergone wh-movement to the initial sentence position.Due to the well-known Verb Second (V2) constraint of Dutch syntax, the inflected finite verb then has to appear immediately following the wh-word.The standard analysis is that Dutch is underlyingly SOV, and that in finite main clauses the inflected verb moves to the C(omplementizer) position.The constituent preceding it (such as hoezo in (3c)) is in the Specifier of CP (Spec-CP).
In addition, there are hoezo-questions where the verb appears to be in the third position in the sentence, as in (5).It might be tempting to regard these as a fourth type of construction ("hoezo + V3"); however, we take it that these are actually instances of the hoezo + X construction, where X happens to be a full declarative main clause (CP) with V2 word order.
(5) Hoezo, HOEZO ons our plan plan is is totaal totally mislukt?failed 'What do you mean, our plan has totally failed?'Such cases differ from the hoezo + V2 construction in that hoezo is not syntactically and prosodically integrated with the rest of the sentence, requiring an obligatory pause after hoezo (indicated by the comma in (5)).There is no such pause in the hoezo + V2 construction. 5nother construction that should be carefully distinguished from hoezo + V2 is one in which hoezo is followed by a polar question.Because main-clause polar questions in Dutch have the inflected verb in clause-initial position (i.e., in C with Spec-CP being empty), such sentences superficially appear to involve V2.We might call this the "pseudo hoezo + V2" construction, but actually we take this to be another instance of hoezo + X, where X happens to be a V1 polar question: (6) Hoezo, HOEZO ben are je you onze our afspraak appointment soms may vergeten?forgotten 'Why?Did you forget our appointment?'Again, this is indicated by the lack of syntactic and prosodic integration of hoezo with the rest of the sentence, signalled by an obligatory pause.In this paper, we set aside the constructions in ( 5) and ( 6), leaving their further analysis for future research.
In corpus data, the isolated hoezo and hoezo + X constructions seem to be by far the most frequent.However, in this paper we will argue for a unified semantics and pragmatics of all three types of hoezo-questions.Since the hoezo + V2 construction is the closest parallel to regular whyquestions with waarom and is therefore the most interesting and challenging to analyze, many of the examples in the rest of the paper will be of that type.In the next section we focus on the pragmatics of hoezo-questions, but in Section 5 we will come back to syntactic issues, providing a more detailed analysis.The syntactic facts we discuss there provide important additional support for our analysis of hoezo and its contrast with waarom.
3. Pragmatic uses of hoezo.Hoezo can be used in conversation in a variety of ways.Perhaps most prominent is its use in response to the assertion of a proposition p in a preceding utterance by another interlocutor, as in ( 2)-(3).6However, hoezo can also be used to challenge speech acts other than assertions, such as questions (7) or commands (8).Hoezo can also target not-at-issue content, such as presuppositions and conversational implicatures, as illustrated in ( 9) and ( 10 Hoezo can even be used metalinguistically, for instance in order to object to the pronunciation of a preceding utterance, as in (11a-b), where the issue is the location of the lexical stress, or to reject the interlocutor's choice of sociolinguistic register, as in (12a-b):7  4. Core of the analysis.What ties all the different uses together that were surveyed in the preceding section?In each case, the speaker signals that they resist a discourse move on the part of the interlocutor by asking for a reason or motivation for that move.In what might be regarded as the prototypical cases, such as ( 2)-( 3), the speaker refuses to add a proposition p to the Common Ground (CG; Stalnaker 1978), where p is the at-issue truth-conditional content expressed by a preceding assertion on the part of the interlocutor.But p may also belong to the not-at-issue content of the interlocutor's utterance, such as a presupposition in (9) or a conversational implicature in (10).Generalizing over at-issue and not-at-issue content, we can say that hoezo signals that the speaker resists updating the CG in accordance with the interlocutor's preceding discourse move.This does not necessarily mean that the speaker rejects the interlocutor's contribution outright.Rather the speaker challenges the interlocutor to come up with a reason for why they made their discourse move.Presumably, if the interlocutor is able to provide such a reason to the speaker's satisfaction, the speaker will accept the interlocutor's discourse move after all, and update the CG accordingly.
But, as we saw in ( 7)-( 8), hoezo may also signal resistance to other illocutionary acts than assertions, such as questions or commands.Thus, in principle, hoezo can signal resistance to any type of discourse move by the interlocutor.We can capture this by saying that hoezo signals resistance to updating not just the CG, but also any other component of the discourse model, such as the QUD Stack for questions (Roberts 2012(Roberts [1996]]), or the To-Do List for commands (Portner 2004).Without going into any details, and staying theoretically neutral as much as possible with respect to different proposals in the literature, we assume a general discourse model consisting of various components (including, at a minimum, the CG, the QUD Stack, and the To-Do List), each of which may be updated following a discourse move by one of the interlocutors, depending on the uptake or response by the other interlocutor(s), in the spirit of, for instance, Farkas & Bruce (2010).
If our analysis is correct, hoezo's main pragmatic function therefore lies in "managing" the discourse, and, more specifically, whether and how the discourse model is updated (cf.Krifka's (2008) notion of Common-Ground Management).Of course, this function belongs to the domain of not-at-issue meaning.In this way, hoezo is very much like a discourse particle (Grosz (2020), among many others), in that its main function (as part of its not-at-issue content) is to relate the speaker's utterance to the current state of the discourse and to express the speaker's attitude as to whether and how the discourse model should be updated.At the same time, hoezo is a question word asking the interlocutor for a reason.In this respect it is similar to other wh-words, like Dutch wie ('who'), wat ('what'), and waarom ('why'), or their English counterparts.However, hoezo operates at the level of discourse rather than at that of the truth-conditional semantics of the sentence.In this way, hoezo differs from waarom.This has important implications for the syntactic behaviour of these two wh-words, as we will see in the next section.
5. Syntactic consequences: hoezo vs. waarom.When we consider occurrences of hoezo that are integrated syntactically in a full wh-clause (i.e., the hoezo + V2 construction) it appears at first glance to be syntactically analogous to other wh-words such as waarom.However, a closer look reveals that its syntactic properties differ substantially.Direct comparison between hoezo and waarom shows a number of syntactic differences that can be attributed to the pragmatic function of hoezo.We claim that hoezo is different from waarom in that hoezo is not subject to whmovement and is base-generated in the left periphery (Spec-CP) of an interrogative main clause. 8he deeper reason for this syntactic difference is that, semantically, hoezo is a pragmatic, discourseoriented operator, and as such does not bind a variable in the semantic representation of the sentence, unlike its apparent near-synonym waarom.
Putting aside some of the semantic and pragmatic intricacies of why-questions, we can characterize the regular wh-word waarom/why as a semantic operator that binds a variable ranging over reasons (see, e.g., Shlonsky & Soare 2011; Cox 2021): (13) What is the reason x such that . . .because of x . . .?
In ( 13), the wh-morpheme binds the variable x, which syntactically corresponds to the trace left by wh-movement.Hoezo, however, is a pragmatic operator that acts at the discourse level and in that sense has more in common with discourse particles and illocutionary operators.Semantically, hoezo does not bind a variable and syntactically it does not participate in wh-movement but is base-generated in its surface position (Spec-CP).
Our claim that hoezo is unable to create a binding configuration predicts that hoezo will be illicit in several types of wh-questions that require movement and variable-binding, in which other wh-words are able to appear, including waarom.We give a number of examples that show that this prediction is confirmed, which provides strong empirical support for our analysis.
First of all, in contrast to other wh-words, because hoezo does not undergo wh-movement, it always has to be interpreted locally and cannot participate in long-distance relations (unbounded dependencies).As a result, in interrogative sentences that contain an embedded clause, hoezo can only receive a root-clause construal,9 in contrast to waarom, which can also be interpreted as having originated in the embedded clause, and may therefore result in a structural ambiguity: Secondly, a regular wh-interrogative such as waarom can appear in situ in multiple wh-questions and echo questions.Under common assumptions, both constructions are interpreted by means of wh-movement operating at the level of LF. 10 If hoezo is base-generated in surface position and is not a variable-binding operator, we predict that it cannot appear in situ in multiple wh-questions or echo questions, and that is exactly what we find: Since hoezo cannot head an embedded CP, it follows that sluicing with hoezo is also impossible. 10 Alternatively, in theories that assume a semantic analogue of LF movement, the interpretation of in situ or multiple interrogatives will result in a variable-binding configuration, which will likewise be ruled out for hoezo. 11There may be an another (or additional) reason why hoezo cannot appear in an echo question.Interestingly, Csipak (2023) observes that hoezo cannot even be used in an echo question where the wh-word appears in Spec-CP, as shown in (i).She relates this observation to the exceptional stress on the first morpheme of compound wh-words in echo questions, and the fact that, unlike all other Dutch wh-words, hoezo lacks a demonstrative counterpart.
(i) a. Speaker A: Hotze bought a yacht because mumble.To sum up, while hoezo acts as a (pragmatic) question word, we have seen that in its syntactic distribution and semantic interpretation, it differs markedly from regular wh-words such as waarom.Specifically, the data in this section show that hoezo is restricted to a position at the left edge of the root clause.Since this is the only position in which hoezo appears, we conclude that hoezo is base-generated in Spec-CP (or another projection within the left-periphery, cf.Rizzi 1997; Wiltschko 2021), which is a natural location for discourse operators.We illustrate this syntactic difference with the schematic trees in ( 19)-( 20 6. Hoezo as a discourse particle.Taking stock of the syntactic patterning and pragmatic behaviour of hoezo, we have seen that hoezo is certainly not like regular wh-words in Dutch.By questioning a preceding discourse move by the interlocutor, and by signalling resistance to updating the Common Ground or other components of the discourse model, hoezo contributes to Common Ground Management in the sense of Krifka (2008).More specifically, we argue that hoezo acts like a discourse particle.
There is a huge and highly diverse literature on discourse particles, and there is no consensus by any means about their definition, or about a set of empirical criteria for identifying them (Thoma 2016).There are many different subtypes of discourse particles, and there is no generally agreed-upon classification scheme.And, of course, across different languages, discourse particles do not necessarily all behave the same way.Nevertheless, there are a number of common characteristics that are shared by most, if not all, lexical items that are usually classified as discourse particles.How does hoezo compare to standard discourse particles on such diagnostics?
On the one hand, several typical characteristics of regular (non-wh) discourse particles that are also shared by hoezo are listed in Table 1.These include the property of both hoezo and nonwh discourse particles that they do not contribute to truth conditions and do not express at-issue content (Kratzer 1999;Gutzmann 2015).They also situate a proposition p in the CG (Grosz 2020: see ex. ( 17)) and dispute the CG content, marking p as controversial/contradictory (Grosz 2010).
On the other hand, Table 2 outlines a few common characteristics of (non-wh) discourse particles for which it is more questionable whether they also apply to hoezo.Perhaps most striking is the fact that non-wh discourse particles cannot be used independently -i.e., they must be positioned within a clause or phrase (Grosz 2020).However, hoezo can obviously be free-standing, since it often occurs in isolation, as we saw in (3a).While this is the case, it should also be mentioned that some bona fide discourse particles have homophones (modulo stress) that can be used as free-standing utterances, such as German ja and doch.These are often treated as belonging to a different category than that of discourse particles (for instance, response particles), but this raises the question of whether this homophony is indeed completely accidental and a whether a more unified analysis might be possible.In any case, not allowing a free-standing use may not be such a clear cut criterion as is sometimes thought.
Secondly, discourse particles are often observed to be sensitive to the sentence type they occur in (e.g., declarative vs. interrogative), yet they themselves are unable to introduce a question.(Zimmermann 2004).When hoezo occurs in a clause (in Spec-CP) that clause of course has to be an interrogative and cannot be declarative or imperative.Whether this should be described as sensitivity to sentence type is debatable, since it may be better to say that the presence of hoezo is what makes the sentence an interrogative.But this is a rather fine distinction to make.At the very least, the presence of hoezo is tightly connected to the interrogative sentence type.
Thirdly, it has been argued that while discourse particles can mark a proposition p as controversial or uncontroversial, they do not directly challenge p (Grosz 2020).But although hoezo may be said to challenge p, arguably this challenge is only an indirect one.What hoezo does is to challenge the interlocutor to provide a reason for why they asserted (or presupposed or implicated) p.
Finally, discourse particles have been characterized as deficient adverbs, i.e., as weak or clitic-like sentence adverbs (Grosz 2020).Hoezo does not seem to fall in such a category, given its independent use that was discussed above and is illustrated in (3a).However, as we just noted, several common discourse particles have homophonous, but stressed, counterparts, and it's not clear where to draw the line, or whether this should be regarded as a definitional property of all discourse particles.Table 2. Some differences between hoezo and discourse particles Note that the criteria listed in Table 1 center on hoezo's pragmatic function, which we take to be at the heart of what makes something a discourse particle, whereas the criteria in Table 2, for which the evidence is more equivocal, mainly involve distribution and form.Given that several core pragmatic and semantic characteristics of (non-wh) discourse particles are shared by hoezo, we think that it is entirely plausible to categorize hoezo as a (wh) discourse particle, especially given the large variety among the recognized discourse particles in different languages and the lack of consensus in the literature about definitions.But clearly this is an issue that merits further investigation and debate.Our analysis of hoezo extends the typology of discourse particles with a new subcategory, that of wh discourse particles.In the final section of the paper we give some suggestions for other lexical items that are potential members of this new subcategory.We hope that our proposal to will contribute to the wider debate about the definition and empirical diagnostics for discourse particles.b.It does so by asking the interlocutor to provide a reason for their earlier discourse move; in that sense, hoezo is a (pragmatic) question word.

Non-wh discourse particle
c. Hoezo functions as a discourse particle, because as (part of) its not-at-issue meaning it expresses the speaker's attitude towards the current state of the discourse model and whether and how it should be updated.
d. Unlike canonical wh-words such as waarom/why, hoezo is not a variable-binding operator.
e. Syntactically, hoezo is base-generated in Spec-CP, and does not undergo wh-movement.
There are several issues that would be worth exploring in future research.The first is identifying other members of the class of wh discourse particles, and comparing them to hoezo.Some plausible candidates include English so what and its Dutch equivalent wat dan nog.In contrast to hoezo, these seem to be asking not for a reason, but rather for the implications or consequences of the interlocutor's discourse move.In a very interesting recent paper, Wiltschko (2023) points out a certain use of plain English what (which is shared by Dutch what), in response to what she calls a "summons", in a way that seems very similar in spirit to our analysis of hoezo in this paper.Another interesting case study might be that of the English expression what of it.And what about how so, the literal translation of hoezo?
A second promising area for further research is the implications of our analysis for the typology of why-questions.We already mentioned that waarom and why may have pragmatic uses similar to hoezo (cf.Ginzburg 2012).The differences between why and how come could be explored further as well (cf.Zwicky &Zwicky 1973 andCollins 1991).Important questions could also be asked (and have been) about parallels between why-questions and how-questions (cf.Schwarz &Simonenko 2018 andSandoval &Morzycki 2023).Can it be a coincidence that hoezo contains the morpheme hoe ('how')?(cf.Corver 2023).
A third issue for further research is the connection between why and the connective because.As is well-known, a because-clauses can be used pragmatically to express the reason why the speaker is making the discourse move that they do, as in (22):15 (22) When's dinner, because I'm starving.
Interestingly, Dutch has a special connective for this discourse-oriented function of because, namely want ('for, since').Want contrasts with the usual Dutch counterpart of because, omdat, not only in that the latter lacks the pragmatic use illustrated in ( 22), but also syntactically: whereas omdat is a subordinating connnective, with the verb in clause-final position, want is coordinating, with V2 word order (ANS 2021): We conjecture that want corresponds to hoezo, and omdat to waarom, in that the former two are both discourse-oriented, whereas the latter two operate at the level of the truth-conditional sentence semantics.This could potentially account for their syntactic properties as well, since hoezo and want both appear to be located high up in the syntactic spine, at or above the CP layer.But of course the details remain to be worked out.
To end this paper, we want to point out a final, intriguing fact about hoezo.It's possible to use hoezo in response, not to a discourse move, but to a surprising (and unpleasant) non-linguistic event, typically with an undertone of anger or frustration: (24) Context: Speaker opens the blinds and looks out the window.Hoezo It may be possible to extend our analysis to account for such cases, if we think of them as resistance on the part of the speaker to updating the CG not with a proposition that was linguistically expressed by an interlocutor, but with a new fact that the speaker just discovered or noticed.This resistance to updating the CG could account for the sense of incredulity associated with such examples.But we will leave it to a future occasion to try and work out that idea.
you mean, our plan has totally failed?'In the hoezo + X construction, X can be a word or constituent of any syntactic category, as shown in (4): (4) a. Hoezo mislukt?b.Hoezo totaal mislukt?c.Hoezo totaal?d.Hoezo plan?e. Hoezo ons plan?
you mean, that dude?That's the pastor!' : 'What is the reason that the police said the suspect was arrested?' * Embedded construal: 'What did the police say the reason is that the suspect was * : 'What is the reason that the police said the suspect was arrested?' * Embedded construal: 'What did the police say the reason is that the suspect was * arrested?'

7.
Conclusion and outlook.The main points of our analysis of hoezo are summarized in (21): (21) a. Hoezo signals resistance on the part of the speaker to updating the discourse model (including the CG, the QUD Stack, and the To-Do List).
time are we going to eat, because I'm hungry.At what time are we going to eat, because I'm hungry.' heck is it raining again?'; 'I can't believe its raining again!' you throw out my stamp collection?'; 'I can't believe you threw out my stamp collection!'