The article explores German discourse particles (DiPs) in rhetorical wh-questions (wh-RQs). While schon (roughly ‘unexpectedly’) only marks rhetorical wh-questions, denn (roughly ‘I wonder’) marks contextually arising information-seeking or rhetorical Questions under Discussion (QuDs), with or without schon. Since ja (roughly ‘unquestionably’) marks shared information, it is incompatible with questions by itself, but occasionally occurs in wh-RQs left of DiPs like schon instead of denn. The results of two acceptability judgment experiments confirm that ja is strongly dispreferred in RQs, the presence of schon improves RQs with and without ja, and denn has no effect on acceptability. A follow-up study further indicated the rhetorical reading of our target questions to prevail independently from DiPs. We conclude that ja in RQs operates on the information contributed by elements like schon, denoting roughly that the issue in question arises ‘unquestionably against expectations’. Our contexts were neutral regarding the discourse functions of ja and denn (side remarks vs. QuDs), unlike the contexts of the findings, from which we deduce that the marked ja schon-RQs, while grammatical, require specific felicity conditions. A first attempt to confirm this experimentally was globally unsuccessful and could only reveal potential hints in an exploratory analysis.
This paper explores a case of interaction between the German discourse particles (DiPs) ja (literally ‘yes’, roughly ‘uncontroversially/unquestionably’), denn (roughly ‘I wonder’), and schon (literally ‘already’, roughly ‘against expectations’) in rhetorical questions (RQs). While the meaning of schon is only compatible with RQs, denn is common in both RQs and information-seeking questions (ISQs). ja is generally not acceptable in questions but occasionally occurs inside RQs, under specific contextual conditions and never without other DiPs, sometimes adverbs, in its scope.
Native speakers agree that ja-RQs are significantly degraded without these extra elements, although spontaneous judgments on the natural findings differ. We suggest that ja operates on the meaning of schon, forming a complex roughly meaning it is uncontroversial that the answer to Q follows from the Common Ground (CG). This is a non-standard interpretation of the particle combination ja schon, but applying its meaning to information independent from the at-issue meaning of an interrogative is generally the only way in which ja can occur in questions.
Since ja does not occur in RQs or other questions on its own, we chose RQs for testing if DiPs can modify each other, a suspicion raised before with regard to other combinations of DiPs in different environments. If so, a DiP like schon in the scope of ja should lead to a measurable increase in the acceptability of a ja-RQ as compared to the ja-RQ without schon. We thus designed two acceptability experiments to determine the relative acceptability of RQs with single, combined and no DiPs, one for ja and schon, one for denn and schon. The first experiment, although the effect is subtle, confirmed that the presence of schon improves the acceptability of ja-RQs. The second replicated the effect for schon but did not detect one for denn. A third experiment showed the rhetorical reading to prevail in the materials independently from DiPs.
In light of the natural findings, we conclude that a particle like schon is necessary to license ja in ja schon-RQs but not sufficient to make it unconditionally acceptable. A comparison of ja schon- and denn schon-RQs in context reveals that the two types of RQs have different discourse functions. Since (in)appropriate contexts have independently been known to influence the acceptability of ja substantially, we attribute the fact that the improvement of ja-RQs by the presence of schon is comparatively small to the fact that the experimental ja schon-RQs were not accompanied by contexts fulfilling the very specific felicity conditions of such questions: as our contexts were neutral to the different discourse functions of schon-RQs with ja vs. denn, denn was insignificant and ja or the particle combination ja schon were uncalled for. These suspicions are supported by the results of a fourth experiment, in which the same experimental items as in the first two experiments were presented in different contexts, suggesting that this might improve the acceptability of ja schon-RQs.
Our experiments are a first attempt at detecting an interaction between DiPs by exploring the relative acceptability of different combinations in a 2×2 design. That the effect was measurable yet smaller than initially expected is an indication that the fulfillment of a DiP’s felicity conditions is a baseline requirement for its use in both standard and non-standard environments. This is an important insight for future research.
Section 2 is an introduction to DiPs, DiPs in RQs, and the special case of ja in RQs. After an introduction to DiPs in general in Section 2.1, and ja, denn, and schon in particular in Sections 2.2–2.4, we turn to the interaction problem in Section 2.5. Section 2.6 reviews some facts about RQs and DiPs in RQs before ja in RQs is introduced in Section 2.7. We justify our decision to focus on ja schon-RQs in Section 2.7.1 and sketch an analysis in Section 2.7.2. Section 3 reports on the first experiment (on ja and schon) and Section 4 deals with the use conditions of ja schon-RQs by contrasting the findings with denn schon-RQs. Section 5 reports on the experiment on denn and schon, Section 6 on the first follow-up experiment conducted to verify that the target questions were rhetorical, and Section 7 on the second follow-up experiment, exploring the influence of context on ja schon-RQs. Section 8 concludes the article.
Many issues about German DiPs are still unsettled, e.g. their meaning type, if they are clause-type sensitive, if they interact, to what extent they behave uniformly, what individual particles mean and if they should not rather be labeled modal particles. We focus in Sections 2.1–2.5 on aspects relevant for the discussion of our data. While RQs are a controversial matter, too, we are primarily interested in them as hosts for German DiPs, see Section 2.6, although the distribution of German DiPs in RQs is instructive for the debate on RQs as well (Sections 3.4 and 4). Section 2.7 is devoted to ja schon-RQs from an interactional point of view.
German has a class of about 20 non-inflecting lexical items that generally contribute specific non-at-issue meanings when occurring in clause-medial position (see for instance Zimmermann 2011 for an overview). In this function, these items are known as modal or discourse particles. We adopt the latter term, to prevent undue assumptions on the role of these particles for sentence modality, see the discussion in Grosz (2016b), and to emphasize on their function in discourse management. This terminological choice is widespread, e.g. Bayer and Struckmeier (2017), Grosz (2016a), Egg (2013), Rojas-Esponda (2013), Zimmermann (2011), and appropriately close but not identical to the segment-initial discourse markers (like aber ‘but’) in the sense of Fraser (2006: 189). We assume that both mark relations between propositions, see Viesel (2015) on ja, Egg (2013) on doch (roughly ‘as you should know’) and Egg (2012) on schon, only DiPs do not require their antecedent argument to be left-adjacent to their host sentence, see Egg (2012: 315).
Negligence to felicity has led to problematic claims about DiPs being ungrammatical in specific environments, e.g. Viesel (2017) vs. Coniglio (2008) on adnominal modifiers and adverbial clauses; Jacobs (2018) vs. Jacobs (1991) on licensing by illocutionary operators. The effect of missing or inappropriate contexts is especially clear with DiPs in unusual positions as speakers will not reconstruct marked prosodic patterns indicative of marked information structure unless contextually required (see Féry 2006). Hence, certain structures are only accepted in context (e.g. Viesel 2015) while even canonical cases are rejected in ill-fitting contexts, see Section 2.2.
The precise semantic nature of DiPs is unclear, see Section 2.5. It is under debate if DiPs are expressive types or presupposition triggers, see Grosz (2016b). DiPs are not well-behaved expressives, e.g. they defy predictions about structural restrictions on their occurrence, see for instance Kaufmann (2010) vs. Kratzer (1999) on the absence of intervention effects for DiPs scoping over quantifier-bound pronouns. Kaufmann concludes that either the contribution of DiPs like ja is a presupposition, or the notion of expressive meaning needs to be refined.
Another issue is if there is a structural relation between DiPs in clause-medial position and the C system, the left periphery, of the root clause, where force features are localized. Following Grosz (2016b), Frühauf (2018), Viesel (2017, 2015) and Jacobs (2018), we do not assume that DiPs are structurally connected to the left periphery, but that they operate on the material in their syntactic scope, i.e. VP, relating this information to the discourse context. Accordingly, the option of employing a DiP in a particular clause depends on the compatibility of its meaning with the discourse status of the information modified and the type of its VP-complement. Consequently, ja is ruled out in questions even assuming that it is not clause-type dependent.
The DiP ja, quite distinctly geared to assertions and difficult to embed, is standardly described as a marker of shared information (e.g. Kratzer 2004). According to Gutzmann (2009), the felicity of ja hinges on the speaker’s belief that p is common knowledge. However, Zeevat (2004: 102) acknowledges that “repeating old information is not useful by itself” but “allows further causal or other connections based on that.” That information modified by ja should be known to the hearer or at least unproblematic, and that therefore ja-utterances are used to substantiate explicit or reconstructable conclusions is already observed in Franck (1980: 232–233). Viesel (2015) stresses this function, arguing that the felicity of ja is dependent on discourse-contextual conditions rather than epistemic ones. Thus ja establishes a causal explanatory claim-relation (CEC relation) between its first, semantic and syntactic, argument p, and its second semantic argument q, which is present or at least accommodatable in the preceding or following linguistic or extralinguistic context. Viesel demonstrates that q is usually the information that a speech act is made – a sharedly recognized, manifest, event, see Stalnaker (2002: 708):
According to this approach, concessive relations as in (1-b) are not special, unlike Franck (1980: 233) suggests, and neither are non-assertive speech acts (1-c). (2) paraphrases the CEC relations that ja hints at in the non-at-issue dimensions of (1-a)–(1-c) in a unified manner:
According to Viesel (2015), for the ja-utterance in (1) to be felicitous, it is of minor concern if p is believed to be common knowledge prior to the utterance. Important is that a statement of the ja-utterance in (1) makes the hearer expect that something else is said that has to do with the bed being large. Not providing a suitable antecedent or continuation results in infelicity (1-d). For more pertinent examples, see Viesel (2015).
While Viesel (2015) integrates the context-dependence of ja in its lexical entry, it might instead be derived by a Need a Reason-implicature (NaR implicature) as proposed by Lauer (2014) with phenomena other than DiPs in mind. This mandatory conventional implicature plausibly springs into action if speakers, expressly consciously, assert information known by and salient to the hearers. In short, speakers must have a reason to assert p other than informing the hearer about p, as when p supports another, informative utterance. Thus a very basic meaning can be retained for ja itself, see (3), and it will still result in the requirement that a relation between p and q hold, as paraphrased in (2). Recall that the ordinary meaning of ja(p) is p.
Altogether, the role of context for the acceptability of ja can hardly be overestimated. This even holds for main clauses with ja, which are always judged acceptable, i.e. (4-a) is a prototypical utterance with ja, see (1), but only as long as it is not paired with the context in (4-b):
The DiP schon (roughly ‘against expectations’) can occur in declaratives, imperatives and interrogatives, takes the lowest position in most combinations with other DiPs and requires two semantic arguments according to Egg (2013, 2012). The first one, p, is associated with the schon-utterance, the second one, q, contextually retrieved. In assertions, the relation between the propositions has been described as a defeasible inference:
In line with Egg (2013, 2012), Féry (2010) and Ormelius-Sandblom (1997), we assume schon(p) to contrast with , which expresses very strong to rather tentative agreement with p and must be relevant in the context. As Zimmermann (2014: 15) points out, schon is not necessarily discourse-anaphoric, but there must be “circumstantial fact supporting ”.
Whenever the DiP schon occurs in wh-questions, only a rhetorical reading is possible with either a negative or positive answer expected (see Egg 2012). Like Egg (2013, 2012), we are convinced that schon in assertions and questions should receive a unified treatment. Unfortunately, a comprehensive unified account of this particular DiP is still missing. Egg (2013, 2012) proposes to associate p, in wh-questions, with the rhetorical interpretation of the question:
RQs as in (7) and (8) show that CG information need not be activated in the minds of all interlocutors (Egg 2012: 319). Very roughly, the use of schon by speaker B in (7) and (8), according to Egg (2012), indicates that A’s utterances suggest (more precisely: preparatory conditions of A’s utterances defeasibly entail) , viz. that it is not known information that Günther will invite specific people, or that B could not have done anything.
Although schon is only compatible with wh-questions that are rhetorical, we are reluctant to adopt the assumption that it operates on the rhetorical interpretation itself (much less on the answer to the RQ). We know of no other case in which a DiP has been suggested to operate on meaning that is conveyed non-lexically. Besides, if the paraphrases of p in (7-b) and (8-b) were available for schon, we might expect them to be available to ja as well. Especially (8-b) looks like the perfect candidate. We would expect to find at least some occurrences, perhaps where the hearers’ previous utterances suggest that the answer to the RQ is activated in their mind.
We would prefer a modification of (5) such that schon operates on a proposition or a set of propositions, as in wh-questions. Thus schon and denn would modify the same semantic object in denn schon-RQs, unlike in Egg’s approach. Roughly, the use of schon in a wh-question should indicate that a fact q defeasibly entails that the issue raised in the RQ should not come up in conversation according to the CG (yet it does – entailment defeated!). This issue is brought up in form of a VP denoting a set of propositions, which we identify as the complement of schon of type . Preferably, the implicit answer to an RQ, which Egg includes in his paraphrases of p, would be identified as the q-argument of schon or as entailed by the q-argument.
This way, hearers are triggered to search for q in their minds and realize for instance that their bringing up the issue, either literally as in (7) or implicitly as in (8), was incongruous. If schon in assertions marks an unexpected fact, as Egg describes p, we might say that schon in wh-questions marks an unexpected issue. For instance, the issue arising by enquiring about speaker B’s options to help Hans in (8) is unexpected if the CG contains that B was sick at the relevant time (q) and could not have done anything (the answer to the RQ). In (9), the answer is not just deducible from some CG information but can itself be identified as q for being CG information. Bringing up the issue of who is perfect would therefore not be expected, judging from what we already know, but it is still brought up for contextual reasons:
Nevertheless, we will basically just treat schon as an RQ marker for the current purpose:
(10) is quite close to Egg’s (2013, 2012) account without referring to a specific answer. The paraphrase is designed to capture the common observation that “[i]n -questions, schon yields a rhetorical question” (Bayer et al. 2016: 594), but keep in mind that (10) is a stopgap solution in that it follows from rather than defines the meaning contribution of schon. The data in Section 2.6.2 shows that schon, unlike the paraphrase in (10), is not compatible with all kinds of rhetorical wh- or polar questions, which we would expect if a special “rhetorical schon” were available (see Meibauer 1986). Hence, to summarize, schon seems able to select complements of type , but also , in which we are interested. Crucially, our working paraphrase of schon does not refer to the rhetorical interpretation of the question or its implicit answer.
The DiP denn is perceived as the interrogative DiP par excellence. Bayer (2018) glosses the meaning of denn in questions – the standard case – as ‘I’m wondering’. Its clitic form n has been described as semantically no more than an obligatory marker of interrogativity in Bavarian, e.g. Bayer (2012). As observed by König (1977), denn-questions cannot be asked completely out of the blue but require contextual justification, in a very general sense. Both ja and denn standardly occupy the highest position in combination with other DiPs and relate their complements to the discourse context, but arguably fulfill opposing discourse functions, with ja(p) justifying information in the context and denn(Q) justified by information in the context. Assuming, as for ja in (3), that the requirement for contextual justification can be inferred from a basic meaning, we suggest the paraphrase in (11). Nothing hinges on the exact wording:
This working definition, designed to capture the somewhat intangible context reference of denn, is applicable to denn in RQs as well, see Section 2.6.1.
The question if combined DiPs interact is as controversial as the related problem of their semantic type. As expressives, several DiPs should apply simultaneously (see Zimmermann 2011): in same-scope, or additive analyses of combined DiPs, each DiP targets only the descriptive meaning in its scope (12-a). However, as Gutzmann (2015) shows, there are good reasons to assume that expressive, or in his terms use-conditional, items can themselves be the arguments of use-conditional modifiers. Hence, detecting an interaction between DiPs will not necessarily settle the dispute about the meaning type of DiPs.
The most important argument for different-scope approaches(12-b) is the relatively strict order of most combined DiPs. Rinas (2007) observes that combined DiPs need not be restricted to the intersection of the sets of sentence types that each DiP can occur in, as same-scope analyses predict. Hence in Rinas’ presuppositional analysis, DiPs can take scope over each other.
Zimmermann (2011: 2030–2031) points out that a different-scope analysis yields the right results for certain combinations, whereas other combined DiPs seem to apply simultaneously. For the combination ja doch, which sometimes occurs in the reverse order, see Müller (2017), Lindner (1991: 196) already suggested that both DiPs are interpreted independently of one another. Müller (2017) proposes that the prevailing, in some cases obligatory, order ja schon is a grammatical reflex of iconicity with different discourse functions of utterances. Considering the high number of combinations of DiPs, however, she also casts doubt upon the idea that one analysis captures all cases. In effect, both interpretational mechanisms might exist side by side.
Most suggestive of an interaction between DiPs are cases in which a combination occurs in an environment incompatible with one of the DiPs on its own. Rinas (2007) points out that doch nicht etwa (nicht etwa: roughly ‘not unexpectedly’) needs to receive a different-scope analysis because its occurrence is less restricted than that of nicht etwa. This relates to observations by Thurmair (1989: 223, 83), who deems doch and etwa essentially non-combinable from a distributive perspective. She notices that etwa can only appear in imperatives and declaratives following nicht ‘not’. Similarly, doch cannot appear in bare root infinitivals with a deontic interpretation by itself even though it is licit in imperatives:
Gärtner (2017: 119) mentions that this might be explained by assuming “that doch directly attaches to and operates on the particles following it.” To our knowledge, there have been no attempts to gather experimental evidence for such an interaction, so we will make a start by looking at a configuration similar to (13). In Section 2.7, we will elaborate on our test case, ja schon-RQs, make a suggestion for the semantic treatment of the particle combination and explain why we are indecisive about its syntax.
RQs are interesting for us insomuch as they are able to host certain DiPs. Some general background knowledge on RQs is presented in Section 2.6.1. Section 2.6.2 turns to DiPs as markers of rhetoricity.
Sadock (1974) observes that RQs, like assertions and unlike ISQs, are eligible arguments for discourse connectors such as after all. Han (2002) even treats RQs as assertions (of negative polarity) semantically. However, the clause type of RQs, the verbs able to select rhetorical subordinate clauses (15), and the German DiPs that can be encountered in RQs (see Section 2.6.2) indicate that RQs are syntactically and semantically questions (see Meibauer 1986).
As Han’s (2002) analysis falls short of wh-RQs with positive answers, Caponigro and Sprouse (2007) take the literal meaning of RQs and ISQs to be identical, the difference being that hearer and speaker are aware that both know the answer to an RQ. According to Meibauer (1986), RQs are primarily indirect assertions, i.e. the primary illocution of any rhetorical utterance is assertive. Importantly, RQs cannot be identified with the assertion of their answer. Meibauer (1986: 183) clarifies that “what is meant by a rhetorical question is not the answer […] but an illocution.” [Translation Y.V.] Despite knowing the answer, speakers use RQs instead of assertions in order to make hearers think about the question (Meibauer 1986: 169). The secondary illocution of an RQ, a question, is thus preserved and speakers aim at convincing hearers of what is indirectly asserted by exploiting the felicity conditions of questions in a specific way (Meibauer 1986: 173). Biezma and Rawlins (2016) point out that RQs signal that the answer is in the CG and are felicitous where an assertion of the answer would be felicitous. Both Biezma and Rawlins (2016) and Caponigro and Sprouse (2007) observe that RQs do not have the same discourse impact as assertions: Caponigro and Sprouse demonstrate that RQs, unlike assertions, allow for the same answers as questions, and Biezma and Rawlins (2016) show that RQs disallow replies that would work for assertions.
These descriptions chime in well with the discussion in Section 2.3. While it seems indisputable that RQs signal that the speaker has an answer in mind, which he/she considers to be (entailed by) CG information, examples like (7) and (8) show that this answer does not have to be salient, known, or easily deducible to the hearer (see also Section 4). The interrogative character of RQs is especially obvious in RQs used to start discussions on controversial issues. Admitting that he does not see a real advantage for smaller companies in introducing a comprehensive personnel management software, a personnel manager initiates a lively debate posting the RQ in (16) in the Xing section ‘IT and personnel management’.
To initiate a debate rather than just voice his opinion, a speaker using an assertion instead of an RQ would have to clarify that the assertion is put up for discussion, or explicitly challenge his readers to convince him otherwise, which is unnecessary in (16). This highlights the differences between RQs and assertions at the illocutionary level, matching their semantics and syntax.
Whether hearers agree with or know the speaker’s intended answer to an RQ, they should recognize the RQ as such. There are different ways to distinguish RQs from ISQs, e.g. only RQs can contain strong negative polarity items (NPIs, see Caponigro and Sprouse 2007: 123), negative RQs do not license NPIs (see Han 2002: 205), only RQs can be introduced by the discourse marker after all (see Sadock 1974), and RQs and ISQs also differ phonologically (see Wochner et al. 2015). German wh-questions containing the DiP schon can also only be rhetorical, no matter if there is a positive (7) or negative answer (8), see Meibauer (1986). However, schon is still ambiguous between a DiP and a temporal adverb in the following wh-question:
In the reading as a temporal adverb,(17) can receive a reading as an ISQ or RQ. Thus, schon is neither necessary nor sufficient for marking rhetoricity in wh-RQs, and Meibauer (1986: 81) rejects the assumption of a special RQ clause type based on a selection of DiPs. Like Meibauer, we have decided not to assume a specific literal meaning for the DiP schon for the time being. While treating it as a mere marker of rhetoricity in wh-RQs, however, we are reluctant in adopting Meibauer’s (1986: 124) claim that the DiP schon creates rhetoricity. We commit to saying that, by virtue of its still undefined abstract meaning, the DiP schon is only compatible with rhetorical wh-questions and thus has a strong disambiguating effect, see Section 2.3.
If “rhetorical schon” were distinct from the DiP that occurs in declaratives, creates a strong bias in declarative questions and is at least marked in (necessarily biased) polar questions, “rhetorical schon” should be found in rhetorical polar RQs. Moreover, schon and auch, between which Meibauer does not differentiate, are not distributed exactly alike. Both occur in wh-questions in their functions as DiPs only when the wh-questions are rhetorical, unlike schon-RQs, auch-RQs may receive only a negative answer:
Thurmair (1989: 158–159) notes that auch-RQs typically ask for a reason where there is none:
Crucially, although schon is usually possible with negative answers, it is not possible in (19):
Without pinpointing the semantics of these DiPs, we conclude that schon and auch have individual meanings on which their function as markers of rhetoricity is based. Since these lexemes, as DiPs, never occur in non-rhetorical wh-questions, to dub them ‘rhetorical DiPs’ helps to distinguish them from DiPs that fulfill the same function in wh-RQs and wh-ISQs, e.g. wohl (literally ‘presumably’) and denn. Without schon, a reading of (21) as ISQ is possible (e.g. in a conversation about students and their results in a multiple choice test):
Meibauer (1986: 119) considers denn to play its usual role, which he describes as establishing a relation to a preceding utterance. Since denn only marks (contextually motivated) questions, we conclude that RQs can be marked by different DiPs as rhetorical utterances and as questions simultaneously. It might seem as if there was a slight tension between the meaning of denn, see (11), and the meaning of schon, see (10), but it is no more contradictory to express with denn schon that a question requires an answer that follows from the CG than to ask a question and signal with schon that its answer follows from the CG. In addition, according to Egg (2012: 300), the CG is not closed under deduction. Therefore, a hearer aware of all information from which the answer to a question follows may not have drawn the relevant inferences that a speaker strives to elicit by using a denn schon-RQ.
According to the distribution of ja and denn, RQs are questions (see also Thurmair 1993). An anonymous reviewer points out that the ban on ja in RQs was used as an argument against the view that DiPs are licensed at a pragmatic level. Thurmair (1993: 34) considers it at least possible that a DiP like ja might appear in an RQ if DiPs were dependent on the illocutionary type or communicative purpose of an utterance. However, recall that the discussion in Section 2.6.1 showed that the illocutionary resemblance between RQs and assertions is rather limited, and even assuming that ja operates at the speech act level (e.g. Bayer 2016), it need not be expected to operate on the indirect assertions that make RQs eligible arguments for discourse connectors such as after all (see Sadock 1974), rather than on the secondary illocutions, questions.
Additionally, denn but not ja was argued to be compatible with VPs denoting sets of propositions of type . In short, there are pragmatic as well as structural reasons why ja is impossible in RQs, even without reference to a mismatch between ja and syntactic features or semantic operators in the clausal C system. Clauses like (22) are intuitively clearly unacceptable, and the empirical evidence of RQs with just ja is non-existent, as far as we can tell. We conclude that ja in combination with other DiPs can only be found in RQs because it can be interpreted differently than when it occurs on its own.
Although illicit in RQs on its own, ja is exceptionally found in RQs combined with other DiPs (and adverbs). Intuitions on the data out of context differ, but speakers agree that ja is more tolerable above pertinent adverbs and DiPs:
Schon is not the only DiP that occurs with ja in RQs. Other DiPs, and possibly adverbs, can play the same crucial role for the overall acceptability of a ja-RQ:
The intuition that the unattested variants with just ja are much worse seems robust even among speakers who do not readily accept the attested RQs with ja and other DiPs, i.e. ja cannot occur in any kind of questions, including RQs, on its own, but it improves in the company of schon (and other DiPs). A same-scope analysis of combined DiPs cannot explain the findings. If, however, a mechanism is available to interpret a DiP as modifying another DiP or adverb, then speakers should make use of this mechanism when encountering a DiP in a combination with other items in an environment in which the DiP cannot occur on its own. Combined ja schon should hence be measurably more acceptable in RQs than just ja.
We chose schon to test for an interaction between ja and another DiP in our experiments because schon is relatively frequent. Testing auch would have complicated matters as all RQs containing auch seem to convey irony and/or sarcasm, see (24). As an anonymous reviewer points out, unknown factors like the type of wh-word (asking for subject or object) or expected answer (negative or positive) might affect the acceptability of ja. If so, we suspect that this will most likely be due to differences in the contextual conditions under which different wh-RQs are used. For instance, unlike who-RQs like (23) and (24-b), why-RQs might significantly more often occur in situations in which speakers confront addressees about their unreasonable behavior than in situations in which speakers want to express shared agreement among the interlocutors. If our suspicion is correct, then the impact of the type of wh-word on the acceptability of ja is not a question of grammaticality. Either way, it is desirable to keep the conditions under which to test for the potential interaction between ja and another DiP both plausible and stable, and even if only very specific RQs are able to display the phenomenon we are concerned about, we can still test for it in this environment. In order to test for the interaction between two DiPs, we therefore modeled our target sentences after the simplest and most natural findings like (23).
Regarding a potential interaction between combined DiPs, ja schon-RQs seem ideal in that ja schon resembles the combinations of DiPs for which Zimmermann (2011: 2030–2031) suggests combined meanings, i.e. ja wohl and doch wohl, which consist of DiPs that occur at different heights of attachment (see Coniglio 2008: 115), with wohl occurring relatively low. schon is even lower than wohl, i.e. schon does not precede ja (or even only wohl) in combinations:
If the difference between ja schon and *schon ja was due to iconicity in the sense of Müller (2017), we might expect some exceptions to the rule, as with ja doch and the more marked doch ja, but if any exist, they are so rare that we could not come up with any. Alternatively, (25) might be attributed to a difference in the type of argument that each DiP takes (with ja having been assumed to take an assertion, i.e. a speech act, as its semantic argument, instead of a proposition or set of propositions like the lower DiP wohl, see Zimmermann 2011). However, Zimmermann (2011: 2031) considers this approach problematic in general, given the variable distributional pattern of DiPs like doch and seeing that an isomorphism between the surface order of DiPs and their interpretation at the sentence-level still requires stipulation, as even ja appears below scrambled elements in IP and nowhere near the speech act level in CP. Thus we should consider the possibility that ja operates on schon. This appears to be the only plausible explanation for the existence of ja schon-RQs. The data chime in with the data in Section 2.5, for which an interaction between DiPs suggests itself because certain DiPs would be illicit without others.
Generally, ja can occur in questions only if it does not modify the proposition that is enquired about in the question, but instead modifies information independent from the at-issue meaning. In (26), ja appears inside a determiner phrase (DP) above an adjective phrase (AP) that is non-restrictive and arguably constitutes a local force domain (Viesel 2017: 181):
In such cases, the availability of ja in questions depends on information which is not in the scope of the matrix illocutionary operator. Interestingly, in subordinate clauses without root properties and APs, ja is rarely found without other DiPs or adverbs in its scope either, and deleting the latter reduces the acceptability, see Viesel (2017, 2015).
Furthermore, ja interacts at least syntactically, possibly semantically, with other expressive items, e.g. the adverb leider ‘unfortunately’. For instance, combined leider ja ‘unfortunately [JA]’ can appear in the pre-V2-position:
The combination leider ja looks similar to a small particle phrase as described in Bayer (2018, 2016), Bayer and Trotzke (2015) and Bayer and Obenauer (2011). A possible interpretation might be it is uncontroversially unfortunate that p. Similarly, Patrick Grosz (p.c.) provides the authentic finding in (28-a) along with the paraphrase in (28-b), suggesting that ja takes narrow scope over vielleicht ‘perhaps’ since it cannot scope over the propositional content of the non-factual conditional:
Like with leider, ja is able to occur in the prefield with vielleicht:
The data show that ja schon-RQs and the suspected interaction between DiPs are part of a range of related phenomena yet to be explored. Testing ja schon-RQs is a start.
An improvement of ja-RQs due to other expressive elements in the scope of ja strongly suggests an interaction and a different-scope approach. We propose that ja, when it occurs with schon in wh-RQs, operates on the specific contribution of schon because ja requires a propositional argument and in the wh-RQs with ja, only a DiP like schon contributes an argument of type . To specify, we distinguish several variants of the (rhetorical) question in (23) with and without DiPs. As (30-b) shows, the at-issue meaning is the same for any version of the question in (30-a):
According to the working definition of schon in (10), we can paraphrase the non-at-issue meaning of the variant in (31-a) as in (31-b), with schon operating on the at-issue meaning of (31-a), i.e. the object of type in (30-b):
The variant featuring denn works completely parallel to (31) because denn selects the same argument as schon – the object of type in (30-b), the at-issue meaning (32-a). Assuming the meaning of denn in (11) yields (32-b):
If both denn and schon occur in the same question, they have the same scope and make two separate contributions in the non-at-issue dimensions:
For ja schon-RQs, however, a same-scope analysis is impossible because ja cannot modify objects of type , see the definition of ja in (3). Employing ja by itself in an RQ as in (34-a) results in a type mismatch, see (34-b):
Hence, for any ja schon-RQ, a different-scope reading is the only option:
As argued in Section 2.3, schon is not a mere marker of rhetoricity, although its meaning is only compatible with wh-RQs. If we assumed that the information conveyed by schon was merely identical to the rhetorical reading of the question, we might want to find a way to explain why ja cannot modify the rhetorical reading if it is indicated by other means. An explanation could be that expressives cannot take scope over non-lexically encoded information, behaving as described by Potts (2005), whereas lexical expressives may be visible to other expressives in the sense of Gutzmann (2015).
If DiPs are expressives, the analysis sketched for ja schon-RQs in (35) treats ja as an expressive modifier in the sense of Gutzmann (2015). Thus, ja operates on the highly abstract propositional information contributed (conventionally) by schon, which, like any other expressive, contributes conventionally implicated meaning with “the same semantic force as a main clause assertion” (Potts 2005: 68), the standard environment for ja. An anonymous reviewer points out that the assumption of ja operating on the meaning contribution of schon remains problematic: Gutzmann’s (2015) system is more liberal than Potts’ (2005) or McCready’s (2010) and expressive propositions are not isolated at once, so that they remain available as arguments of expressive modifiers in principle, but there are no types in Gutzmann’s system that take expressive propositions as arguments.
First, we agree that we have only sketched a proposal, as the goal of this section was to make a suggestion for how to deal with data that cannot be accounted for at the moment. Our suggestion should be seen as a preliminary to our first experiment, which aims at determining how acceptable ja can become in the presence of schon in RQs. Furthermore, the natural findings of ja-DiP-RQs are highly marked, which is likely indicative of interpretive difficulties. Any analysis would presumably reflect this. In addition, we consider the expressions that Gutzmann presents exemplary although we cannot provide a more detailed discussion for reasons of space.
Importantly, the question whether Gutzmann’s system is fit to capture ja schon-RQs arises only if DiPs are expressive items and contribute conventionally implicated meaning. A comparison by Grosz (2016b) of use-conditional and presuppositional approaches to DiPs shows the evidence to be inconclusive. All things considered, we conclude that any proof of an interaction between ja and schon will not help to decide between the two approaches because we do not consider such an interaction as insurmountably problematic to Gutzmann’s (2015) system.
Finally, while we are interested in the acceptability of a very specific phenomenon, the problems with its theoretical analysis are much the same for other comparable data, only a fraction of which was mentioned in Section 2.5. It would be a pity if the most intriguing data were hardly even presented due to theoretical difficulties. Thus we offer a first attempt at clarification, hoping to spur further debate. In a nutshell, we explored the interesting phenomenon of ja schon-RQs experimentally, detected a subtle yet measurable effect of schon on the acceptability of ja, can consequently report on noteworthy difficulties in substantiating an interaction between DiPs, found, again, that context seems to influence the acceptability of DiPs, and were able to describe subtypes of RQs that are not often considered as such in the pertinent literature.
The goal of this experiment was to determine if the presence of schon detectably affects the acceptability of ja in RQs.
We tested 30 participants (age 19–48 years, mean 24 years; 5 male; 1 left-handed), all self-declared German native speakers and students of the University of Konstanz. Participants received a reimbursement of 5 €. The materials consisted of 28 experimental items interspersed with 34 fillers (24 from a different experiment) resulting in 62 stimuli per participant. Each experimental item consisted of a context clause followed by the target RQ. We systematically varied the occurrence of schon and ja in a 2×2 design (36-c):
All target clauses conformed to the following structure: All were introduced by the conjunction aber ‘but’, the subject interrogative pronoun wer ‘who’, and a finite verb in the present tense, for example hat ‘has’. Next came the slot with the DiP(s) according to the experimental conditions, followed by an object argument consisting of a determiner, an adjective and a noun, followed by a PP. The form of the target clauses aber wer … ja schon was motivated by corpus findings, see (23)/(46) and (50). The implicit answer to the RQs was always negative (Nobody!).
The procedure was a rating task with a reference sentence (anchor) and an open scale, similar to thermometer judgment (Featherston 2009, Featherston 2008). The current task design was as follows: We presented the participants a single reference sentence (37), a declarative with a center-embedded object clause which is possible but dispreferred to the right-extraposed variant. This reference sentence was associated with a score of 0. We asked participants to judge the acceptability of the target sentences in comparison to the reference sentence. We described it as sentences being better or worse, easier or harder to understand. We stated explicitly that we are not asking for prescriptive rules or “good” writing style. We also provided some examples with respective proposals for scores.
If a target sentence was worse than the reference sentence, participants should associate a negative score. If it was better, they should assign a positive score. The number range was open but only integers allowed.
The procedure was programmed in the python experiment suite PsychoPy (Peirce 2007) and run in a psycholinguistics lab. The experimental stimuli were randomized according to the Latin square design, so each participant saw each of the 28 items in exactly one of the four experimental conditions. The total list of stimuli was randomized by the experiment software. The experiment took approximately 15 minutes.
The condition without any DiPs (∅-∅) should be acceptable. The ∅-schon condition should be equally good or even better since it marks an RQ explicitly. We expect the ja-∅ condition to receive significantly lower ratings than the two aforementioned ones because ja is virtually impossible in questions. The most interesting predictions concern the condition with both DiPs (ja-schon): Under the hypothesis that DiPs can modify the meaning of other DiPs, we expect an interaction of the two conditions that might yield ratings as in the ∅-∅ condition. This is motivated by corpus findings. Under the hypothesis that multiple DiPs are interpreted in parallel we would expect that the presence of ja leads to the same decrease in acceptability with or without schon.
We analyzed the data using the statistical software package R (R Core Team 2015). Some data cells had to be removed because they contained no/missing data. We used linear mixed-effects models of the lme4 package (Bates et al. 2015) to analyze the z-transformed response scores. Following Barr et al. (2013) we tried to fit a maximal random effect structure for the LME models, but had to simplify the random effect structure to obtain model convergence. We performed a model criticism as recommended in Baayen and Milin (2010) by excluding absolute standardized residuals exceeding 2.5 standard deviations. We report the results of the criticized model, taking effects as significant if the respective t-value exceeds an absolute value of 2. The fixed effects of the final model for this experiment were the presence vs. absence of the particles ja and schon as well as their interaction. The random effect structure of the model included random intercepts for participants and items and the two fixed effects as random slopes for both participants and items.
A graphical summary of the results is given in Figure 1 and the model output of the fixed effects of the criticized and raw models is given in Table 1. The statistical analysis revealed main effects for ja ( ) and schon ( ) indicating that ja led to lower and schon to higher acceptability judgments, as visible in Figure 1. The negative value of the interaction term of the raw model indicates that the difference between ja-∅ and ja-schon is larger than between ∅-∅ and ∅-schon which is also visible in the plot. However, that the values for this interaction term decreased considerably in the model criticism indicates that the difference is mostly driven by outliers.
|Criticized model||Raw model|
In interpreting our results, note that the value 0.0 on the Z-score scale in Figure 1 does not mark the threshold of grammaticality but the mean level of acceptability in relation to the reference sentence given in (37). Negative values do not indicate ungrammaticality, just lower acceptability than the mean acceptability. To give the reader an impression of the realm of acceptability, we included the judgments for independent fillers in Figure 1, of which three are given in (38), all of which are prescriptively grammatical. The underlined target utterance of filler item f1 in (38-a) has received considerably lower judgments than the ja-∅ condition even though (38-a) is not ungrammatical. The distribution of the fillers also shows that neither high nor low judgments for the experimental conditions are constrained by ceiling or floor effects. Hence the results only allow statements regarding the relative acceptability of the experimental conditions.
Against this background we can assess if the effects for the individual DiPs conform to our predictions: Taking the bare ∅-∅-RQ as center of reference, we can state that ja led to a large decrease in acceptability and schon led to a small increase in acceptability. Clearly most interesting are the results for the combination, i.e. the ja-schon condition. Against our predictions, schon did not cancel out the negative effect of ja nor remain ineffective. Instead, schon mitigated the negative effect of ja but left the ja schon-RQ still less acceptable than the bare ∅-∅-RQ, which falls in line with the fact that findings of ja schon-RQs are rare and controversial. It was suggested to us that speakers might decide to disregard ja in schon-RQs, but the decrease for ja-∅ and ja-schon show that speakers notice ja in RQs.
Note that the statistical analysis revealed that the simultaneous presence of both ja and schon does not yield an effect that is significantly different than the sum of the effects of adding each DiP individually. However, this does not imply that the decrease of adding ja to a bare RQ and to a schon-RQ is due to the same cause, despite its comparable effect size. While the effects of ja and schon need individual explanations, it is a theoretical matter whether the effect of ja schon can be assumed to be the combination of these individual effects or must be attributed to a different cause. We assume the latter possibility because a simple combination is theoretically implausible, as discussed shortly.
First, the negative effect of ja is unanimously expected under all accounts although for different reasons. As laid out in detail in Section 2.7, we assume that the main reason for the decrease in acceptability is that the object in the scope of ja is of the wrong semantic type. The positive effect of schon was unexpected, but we assume that it is caused by the explicit marking of the RQ as rhetorical, thereby making the interpretation more easily available. Furthermore we interpret the higher acceptability rating of ja schon-RQs in contrast to ja-∅-RQs as an indication that ja must modify the information conveyed by the latter DiP, as proposed in light of the authentic findings in Section 2.7. This explains why we never found ja in ISQs, and in the small number of RQs, it was never the only DiP. We further assume that the reason why ja schon-RQs are more degraded than bare ∅-∅-RQs is due to the specific felicity conditions of those RQs that have not been satisfied in the experimental design. A detailed illustration of these felicity conditions will follow in Section 4.
Turning now to alternative interpretations of the statistical results, assuming an additive effect of the two DiPs would mean that schon improves both ja-RQs and ∅-RQs to the same degree for the same reason. This would require a highly implausible model of acceptability judgments. It would imply that speakers compute two judgments for one utterance, a quite good one for schon(Q) and a quite bad one for ja(Q), and then instead of being disturbed by the mismatch, sum it up as not so bad. Put differently, adopting the above explanations for the individual effects, according to which schon makes the rhetorical reading more easily available and ja is incompatible with questions, why should hearers be more tolerant if the reading with which ja is not compatible is more readily available? Hofmeister et al. (2014) show in a series of acceptability experiments that multiple grammatical violations may add up, as well as multiple processing difficulties. Crucially, however, processing difficulties only affect acceptability judgments in the absence of grammatical violations, i.e. grammatical violations overrule processing effects. Transferred to our data this leaves two possibilities: Either, the difference between bare RQs and schon-RQs results from the violation of a requirement according to which an RQ must be marked by schon, a view which we believe is indisputably unwarranted. Or this difference indicates that schon facilitates the interpretation of the utterances. In this case, a decrease caused by the absence of schon should not add up with a decrease caused by the presence of an incompatible lexeme. If the DiPs were interpreted independently, as assumed under a same-scope approach, we would rather expect equally low ratings for ja in any RQ, with schon or without, i.e. the negative impact of the grammatical violation caused by ja would prevail.
An additive interpretation of the effects leads to even more questionable consequences: We would have to conclude that the natural findings are the product of speakers making use of the mitigating effect of schon in order to be able to use ja despite the mismatch with the semantics of a question. Alternatively, if we assumed that the meaning of ja does not apply to schon, but that it was somehow possible to apply it to the meaning of the interrogative after all, it is unclear why the presence of ja would be tied to the presence of DiPs like schon at all. If speakers desire to make use of the discourse function of ja in an RQ, and if the effect of schon is really negligible, we would expect to find ja in RQs on its own as well, contrary to the facts. In consideration of these inconsistencies of an additive interpretation and the strict order of multiple DiPs, we interpret the observed improvement as indicating an interaction of ja and schon, favoring a different-scope approach.
Taking stock, we have a number of natural findings in which ja seems rather acceptable in a position to the left of other DiPs, and we were able to measure that the presence of schon improves the overall acceptability of RQs with ja. We cannot take ja as operating on the VP meaning because ja cannot apply to a set of propositions, so we take it as operating on the information conveyed by the DiPs in its scope. These DiPs are thus necessary to make sense of ja, which explains the slight improvement, but they are not sufficient to make ja completely acceptable in the RQs. In the words we used to ask for the speakers’ judgments, ja (schon)-RQs are worse or harder to understand than the versions without ja, but ja schon-RQs are better or easier to understand than ja-∅-RQs.
Finally, the overall acceptability of ja schon-RQs cannot be reduced to grammaticality. Recall that we have argued in Section 2.2 that the acceptability of DiPs, especially ja, is heavily dependent on contextual conditions, that this is especially crucial when dealing with less canonical findings such as ja in semantically embedded environments, but that the wrong context renders even standard cases of assertions with ja unacceptable, see (4). Hence, we need to take a closer look at the discourse function of the authentic ja schon-RQs in their natural contexts and compare it to other RQs, specifically denn schon-RQs.
The effect of schon on ja is more subtle than anticipated in our experiment, plausibly because the overall acceptability of ja-RQs is influenced by factors not targeted by the experimental design. Unlike the experimental data, the natural occurrences are production data, i.e. speakers planned their utterances as RQs with ja and do not run into any interpretive traps that would decrease the acceptability ratings for making the interpretation difficult. In the following, we look at contexts in which speakers might not possibly use ja, see Section 4.1, and those in which they decided to utter ja schon-RQs, see Section 4.2. Section 4.3 contrasts instances of wer weiß ja schon p ‘who knows [JA SCHON] p’ and wer weiß denn schon p and their contexts. As acceptability subsumes felicity and the latter has featured prominently in other non-canonical ja-utterances, we suspect that the intuitions on the natural ja schon-RQs are not wrong and they are actually more acceptable than the experimental data, see Section 4.4.
Given the rather small number of natural findings, speakers seldom express by way of ja schon-RQs that it is uncontroversial that the answer to a certain question follows from the CG. As RQs have been described as questions to which both speaker and hearer know the answer and are aware of the shared knowledge, see Section 2.6.1, this is a bit surprising. We would either expect ja to appear in RQs more often, as it seems to match the notion of shared awareness of the answer, or we might expect it not to occur at all for being redundant (on top of difficult).
On closer inspection, many of the RQs in Section 2.6.1 are uttered in contexts where the speaker signals that the answer is in the CG in the sense of Biezma and Rawlins (2016) but the hearer is ignorant of the answer or the issue is controversial. In fact, run-of-the-mill RQs as in (18) above or (39) are not captured by definitions that require shared awareness (of the shared awareness) of the answer, cf. Caponigro and Sprouse (2007: 129). An RQ picking up on a prior ISQ, for instance, suggests that the hearer should be aware of the answer in spite of a context signaling that this is not the case. Accordingly, the speakers in (18-a) above or (39) do not give the hearers more time to think again but proceed to provide the answer themselves right away:
In such cases it is not uncontroversial that the answer to the question follows from the CG. Crucially, we would not predict ja in such cases:
That speaker and hearers share an awareness of the answer is obvious in the truism in (44), but it arguably also holds for (45) (by an electrician asking other experts for advice). However, shared awareness of the answer is not unusual in RQs of a different type than (40) and (41). To use ja, it should additionally be useful to point out that the answer is uncontroversially shared. This expectation is borne out in the findings. Consider (44) and (45) in their natural contexts:
The context of (23)/(44), see (46), is a report on a café-restaurant in a park. The author is annoyed by the esoteric aspects of the venue (a circle of Celtic horoscope trees), which she only mentions as feature attractions (in the wider, omitted context) while praising the natural beauty and physical lighting conditions of the wintry ambience. With ja she stresses that she considers the readers to be aware of the shared answer to the issue. Unlike in (8)/(40) and (39)/(41), the issue does not arise from an utterance by the hearer, but only circumstantially, from the location.
Recall from Section 2.2 that ja relates the information in its scope to other information in the context. Paraphrasing the causal-explanatory discourse relation of information in the scope of ja to contextual information is trickier in the case of RQs than in the case of two assertions, but we suggest that ja still marks the CEC relation paraphrased in (49).
When ja scopes over subsentential information, it often relates this information to the superordinate sentence. It therefore seems patent to try to relate the information in the scope of ja to information outside the RQ proper. In (46), the RQ is in the scope of aber, so the speaker’s raising of objections might be a good candidate for the q-argument of ja, in parallel to (1) and (2) above. With aber ‘but’ and the RQ, the speaker challenges the basic premise of the horoscope that always needing more selects a subset of people. The discourse function of ja in (46) might thus be paraphrased roughly as in (48). (30-b) is the at-issue meaning of the RQ in (46):
However, it is unclear if this paraphrase motivates the use of ja, as a mere schon-RQ in the scope of aber ‘but’ would receive essentially the same interpretation. Moreover, since ja does not operate on the VP meaning of the RQ in (46) directly but on the contribution of schon, it is already embedded in the RQ. Therefore, we may not have to search for the second semantic argument of ja outside the RQ. If ja, when modifying another DiP, behaves like other instances of ja at the subclausal level, its second semantic argument can be found one level below the connector aber ‘but’:
If the meaning of schon is not identical to the rhetorical reading, (49) is not a tautology. Still, according to this paraphrase, the use of ja makes the RQ self-referential. Since the issue of complete contentment is brought up by the horoscope information about people who always need more than they already have, the use of ja in effect signals that an RQ suggests itself in this context. This analysis may have to be refined, but it shows that an analysis based on the ordinary semantics of ja seems feasible. Contrary to denn, ja in the RQ also signals that no answer is needed between hearer and speaker, and thus the RQ in (46) is the closing remark to dismiss the subject of tree horoscopes altogether.
Unlike in (40) and (41), the issues expressed in the ja schon-RQs in (46) and (47) are not at the center of the discourse, but, assuming a question-based discourse model such as Roberts’ (2012), terminal nodes in the discourse tree. The RQ in (47) is even one of two remarks in parentheses. Here, the issue expressed in the RQ does not arise from the extra-linguistic context as described by the speaker in (46), but it does not arise from an utterance by the hearer either, unlike in (40) and (41). Rather, the speaker preempts comments that might arise from his own prior request. The example may have to be analyzed differently from other ja schon-RQs because of the additional adverb schließlich ‘after all’, so we will not deal with it in detail. However, it is interesting that (47) suggests that the speaker wants to inform hearers that he himself is aware of the answer to the RQ, rather than express that he thinks the hearers should be aware of the answer (40)/(41) or are aware of the answer (46).
Another natural ja-RQ presupposing shared awareness that issue P is solved is (50). In the context of the RQ, a Hessian speaker (HES) complains in Hessian about another speaker’s (STG) previous remark that Hessian is still the most unpopular dialect in Germany. STG’s posts are in standard German and do not reveal other users where he is from. The obvious answer to HES’s RQ is that only STG himself knows where he is from:
Note that the string wer weiß ja schon is suspiciously frequent in the natural findings, as illustrated by data like in (53):
RQs like in (53) resemble the truism in (46) in that it is obvious between speaker and hearer who knows p (nobody), and the issue of who knows p is not going to be discussed further. The primary purpose of these RQs is not to solve the issue of who knows something (in a rhetoric sense of solving, i.e. agreeing, or making the hearer acknowledge the answer), but to express a general lack of certainty about the issue in the subordinate clause. (Consequently, the embedded issues cannot be solved and need not be discussed further either.) The denn schon-RQs in (54) contrast very clearly with the ja schon-RQs in (53):
The hearers in (54) are not presumed to be aware that nobody knows p, and as in (40) and (41), the RQs in (54) are controversially discussed. The interlocutors in (54-a) are partners in a footrace from Los Angeles to New York in 1931 and debate if the hard fight is worthwhile. Hugh’s two questions cast doubt on whether Doc’s enthusiasm is justified. By using RQs, Hugh signals that he considers it obvious that nobody knows or cares about how victorious they are according to Doc, but he also calls Doc’s attention to this in the first place, as Doc in his ardor is apparently not aware of it. Doc’s reply then indicates that he disagrees with Hugh on the matter.
Similarly, the RQ in (54-b) is an article headline, i.e. the issue is of concern to the author and of public interest. The RQ sets up a topic for discussion and is repeatedly referred to in the following. The South Tyrolean author argues that the question of South Tyrolean identity, with which South Tyroleans are obsessed, can ultimately not be answered by anybody (außer vielleicht dem Schützenkommandanten ‘except perhaps the rifle commander’), and if it could be answered, it might not be to the South Tyroleans’ liking. Hence, as in (54-a) and unlike in (53), the matrix question is of argumentative relevance in the discourse and not just a phrase to express uncertainty about and then drop the issue expressed in the embedded clause.
The lack of shared awareness (about nobody knowing p), the controversy and the ensuing debate distinguishes the cases in (54) from ja schon-RQs, i.e. it seems impossible to replace denn felicitously with ja here:
However, we would not necessarily exclude denn schon-RQs from contexts like (53). denn schon-RQs are less marked and have less specific felicity conditions than ja schon-RQs in that contextual justification for bringing up an issue under denn, see definition (11), is a very unspecific requirement and even RQs on uncontroversial side issues can be regarded as momentarily central to the current discourse. For instance, a speaker aiming for explicit consent might employ denn, rather than ja, in cases like the ones in (53), whereas with ja, the RQs have a more musing flavor by expressly presupposing tacit approval, although the effect is subtle:
However, while the use of denn is always possible in principle, note, finally, that replacing ja with denn, or even just leaving ja out, might in some cases cause misunderstandings, at least in the absence of prosody in written data, due to the presence of temporal adverbials as in (53-c), which seem to promote a reading of schon as an aspectual adverb. This need not change the rhetorical reading of the question, but it has the potential to do so:
The desire to be absolutely unambiguous with regard to the rhetorical intention might thus be another driving factor for the use of ja in RQs. To summarize, ja schon-RQs are rare, but they are possibly just as rare as the contexts they occur in, so that the low acceptability ratings of isolated instances are probably more adequately attributed to markedness, due to interpretive complexity, rather than ungrammaticality.
The contexts in our experimental items do not contradict a reading according to which the RQs are dismissive side remarks as described in Sections 4.2 and 4.3 due to prior shared awareness that the answer follows from the CG. However, there is no obvious motivation, comparable to the speaker’s bewilderment about esoterics in (46) or the speaker’s wish to preclude hearer comments by admitting to the peculiarity of a previous request in (47), to point the shared awareness out. We interpret the subtle improvement of the ja schon-RQs in our experiment as an indication that it should be discernable to hearers why a speaker forces them to stretch their interpretive capabilities by pointing out shared awareness by using ja in a question, on top of pointing out by using schon that the answer follows from the CG. The low scores of ja schon-RQs are reminiscent of other marked constructions that are only acceptable in specific contexts (see Viesel 2017 on ja in restrictive adnominal modifiers, adverbial and factive clauses). In conclusion, DiPs in the scope of ja are indispensable in RQs, but so are the felicity conditions of the resulting utterance.
Recall that our target RQs were presented only with preceding contexts designed to ensure the rhetorical reading, in order to test if ja is licensed by schon. There are no contexts evolving around different QuDs, unlike e.g. in (53-a), where the speaker goes on to discuss disadvantages of the insurance after barely mentioning the motivation for one in the RQ. Our contexts would work with continuations developing the issues arising in the RQs, especially since issues like the perfect fall weather (36) are not as uncontroversial as truisms like nobody’s perfect contentment, and neither is it obvious in our experiment that an imagined hearer or the participant are aware that the answer is deducible from CG information. Moreover, in Section 4.3 we observed that denn can be replaced for ja in any RQ: while denn schon-RQs are more than side remarks, it is enough if they are (perceived as) pressing questions at the time of utterance, but it is not necessary for the entire discourse to evolve around the question like in (54-b).
The goals of this experiment were twofold: First we aimed to replicate the acceptability-improving effect of schon from experiment 1. Second, by replacing ja with denn, we wanted to trigger a contrasting effect of combining schon with another DiP in our RQs.
We tested 30 participants (age 18–40 years, mean 23 years; 3 male; 1 left-handed). All were self-declared native speakers of German and received a reimbursement of 3 €. The materials consisted of 28 experimental items interspersed with 34 fillers (24 from a different experiment, 10 identical to experiment 1) resulting in 62 stimuli per participant. The experimental stimuli were the same as in experiment 1 except that ja was replaced by denn. The experimental items appeared in one of the four versions illustrated in (58):
The experimental procedure and the data analysis were identical to experiment 1. The fixed effects of the final model for this experiment were the presence vs. absence of the particles denn and schon as well as their interaction. The random effect structure of the model included random intercepts for participants and items as well as the two fixed effects and their interaction as random slopes for both participants and items.
|Criticized model||Raw model|
We expect the same effect for schon as in experiment 1, higher acceptability of the ∅-schon than of the ∅-∅ condition. As denn is compatible with all contextually justified interrogatives, we expect acceptability ratings in the denn-∅ condition to be at least as high as in the ∅-∅ condition. In the denn-schon condition, we expect acceptability ratings as high as in the ∅-schon condition.
The results are summarized in Figure 2 and Table 2. The statistical analysis revealed only a main effect of schon ( ). The presence of schon led to higher acceptability judgments. No main effect of denn and no statistical interaction between denn and schon could be detected.
The effect of schon in experiment 1 could be replicated,denn has no apparent effect. Our predictions are borne out: Judging from their preference for the explicit schon-marking of questions that would otherwise be just as unambiguously rhetorical (see Section 6), we might expect speakers to strongly prefer denn in questions that are more than subordinate QuDs, e.g. set a pressing issue up for discussion, like (54). Since labeling our RQs with denn as RQuDs in need of an answer did not affect acceptability, we conclude that our contexts are not tailor-made for denn, but indifferent to the discourse functions of both ja and denn and thus suitable to test only the effect of schon on ja in isolation, which they were designed for.
The experiment was conducted to ensure that the rhetorical reading of the target questions in experiment 1 and 2 was enforced by the preceding context independently of the DiPs.
We tested 45 participant (age 18–32 years, mean 23 years; 7 male; 4 left-handed), all self-declared native speakers of German. One was bilingual with German and English. The experiment was conducted in a session with another experiment. Participants received 8 € for the whole session. The materials consisted of 30 experimental items interspersed with 35 fillers resulting in 65 stimuli per participant. We systematically varied the occurrence of the three DiPs ja, denn, and schon in a 2+2+2 design, see (59). The materials were basically identical to experiment 1 and 2. Because of the larger number of conditions, we created 2 additional items with the same structure. Each of the 30 items, such as (59), was associated with a pair of continuations, as in (60), one of which rendered a rhetorical interpretation of the target RQ more plausible, the other an information-seeking reading.
The procedure was a decision task. For every target, the two continuations were displayed in the lower part of the screen in left and right, respectively. The positions were varied so that for half the items and fillers, the RQ continuation appeared on the left, for the other half on the right. Participants were asked to decide which continuation they would prefer. The experiment lasted approximately 15 minutes.
We predict an overall preference for the continuation disambiguating for the RQ reading. Assuming that DiPs specify the illocutionary force of utterances, we expect subtle differences in the preference patterns that ideally correlate with the acceptability judgments: schon might cause speakers to favor rhetorical continuations, the interrogative/open issue particle denn might do the opposite without schon, and ja only works with an RQ reading and should trigger a strong preference for these continuations. If the availability of an RQ reading is independent from DiPs, all conditions should show similar preference patterns.
The analysis procedure was identical to the previous experiments except that we used generalized logistic mixed-effects models due to the elicited categorial response. The generalized mixed effects model did not detect a significant effect or interaction for the experimental conditions. The strongest tendency was detected for schon ( , , ). In line with our predictions, exact binomial tests revealed the preference for the continuation implying an RQ reading to be significantly higher than chance level in all conditions (Table 3).
|Percentage of RQ reading||p value|
Variability within participants as well as items was unexpectedly high. Figure 3 displays the responses per item. All but two stimuli show an overall preference for the predicted rhetorical continuation. Figure 4 shows the responses per participant. Most participants favor the RQ continuation for the majority of items but some disprefer them quite strongly.
The continuations for RQ readings were preferred across all conditions. This indicates that an RQ reading was available in all conditions and even more prominent than an ISQ reading, so the experimental materials were overall well fitted to ensure the desired RQ interpretation. Further, the between-condition analysis did not indicate a significant correlation between DiPs and the rate of choosing an RQ reading. According to experimental logic, from this null result we cannot conclude that there is no correlation between the presence of DiPs and the availability of an RQ reading. Rather, if the presence of DiPs influences the availability of an RQ reading in similar sentences, then our setup is probably not suited to detect this difference because the effect is either too small or of a different nature. Considering that the task seems suited to detect the overall preference for an RQ reading, the influence of the DiPs is likely too small. This renders it unlikely that the acceptability differences that we found in experiment 1 and 2 are a consequence enforcing or blocking a specific reading. Crucially, this weakens a potential argumentation that ja improved with schon because schon enforces the rhetorical reading.
To understand why the preference for the predicted continuation ranges around 65 % and not close to 100 %, we take a look at the variability between items and participants. Remember that we did not ask participants explicitly to decide whether an item was an RQ or not. Instead, we asked participants to choose one of two continuations. These continuations have been created to strongly favor the two different readings. The ISQ continuations always contained a nominal expression such as diejenigen ‘they’ in (60-b), intended to point to an individual to trigger a sharp contrast with the answer to the RQ, namely nobody. However, (60-b) can also follow an RQ under the following interpretation: Who (in the world) has perfect weather? If there are any, they should write a (damn) report! The availability of such an interpretation for the ISQ continuation seems to be strongly context-dependent, which is indicated by the almost linear continuum in Figure 3. The same logic might explain part of the participant variation. Participants may differ in their preference for such (more emotional) discourse turns. For lack of space, we do not discuss more potential factors influencing this variation. Important is that the variability affects participants or items across conditions, i.e. in each of the six configurations of DiPs, so that our conclusions on their relative acceptability are not compromised.
We conducted another acceptability experiment to explore if the felicity conditions of context may influence the acceptability judgments of ja schon-RQs. The experimental procedure and data analysis were identical to experiments 1 and 2. We tested 49 participants (age 19–37 years, mean 23 years; 9 male; 4 left-handed), all self-declared native speakers of German (5 bilinguals) and students of the university of Cologne receiving either course credits or 4 € for participation.
The materials consisted of the same 28 experimental items as in experiments 1 and 2, interspersed with 52 fillers (36 from a different experiment), resulting in 80 stimuli per participant. The target sentences were identical to the ja-schon and ∅-schon condition of experiment 1 (61-c). They followed one of two contexts manipulated to indicate shared awareness of the RQ content (61-a), or remain neutral about the knowledge state of the addressee (61-b).
In the neutral context, we expect to replicate the difference between the ja-schon and ∅-schon condition from experiment 1. According to the felicity conditions for ja schon-RQs, we expect that the mutual context will increase acceptability for ja schon-RQs but not schon-RQs. The final model was comparable to Experiment 2, i.e. both factors (mutual vs. neutral context and presence vs. absence of ja) and their interaction were included as fixed effects and as random slopes for both participants and items.
The results are summarized in Figure 5 and Table 4. The statistical analysis revealed a main effect of the DiPs ( ) with higher acceptability judgments for schon-RQs, replicating the results of experiment 1. The context manipulation did not turn out to be significant.
Subsequently, we performed an exploratory analysis to estimate how the influence of the contexts varies across the items and whether an effect might be found in an new experiment with slightly changed material. For each item, we computed the difference between context conditions (mutual-neutral) in the ja-schon and ∅-schon condition, respectively. These differences are displayed in Figure 6a and illustrate, first, that two thirds of the ja schon-RQs (solid line) show a positive difference indicating that the mutual context condition was rated higher than the neutral context condition. Second, the context influence of the ja schon-RQs (solid line) appears not to correlate with the differences for the schon-RQs (dashed line), as confirmed by Pearson’s product-moment correlation ( , ). Applying the same LME model as above to a subset of the data, by excluding four items with the most negative difference (14 % of the data, marked area in Figure 6a), yields the predicted results (Figure 6b). The analysis still shows a robust main effect for particle but also a significant interaction, indicating higher acceptability judgments for the mutual context in the ja-schon condition but not in the ∅-schon condition.
|Criticized model||Raw model|
|Intercept (ja schon-neutral)||−0.15||0.07||−2.13||−0.17||0.07||−2.35|
|particle (∅ schon)||0.73||0.08||9.48||0.75||0.08||9.46|
The results replicated again the contrast in our previous experiments. Figure 5 shows the judgments for the schon-RQs in all three experiments to be nearly identical, indicating that the additional context in this experiment did not affect overall acceptability of the RQs. The ja schon-RQs in this experiment seem significantly more acceptable compared to the judgments of experiment 1. However, the filler items f1 and f2 also show higher absolute Z-scores, indicating that the increase is not (solely) caused by the prolonged context of the experimental items but most likely by the absolute acceptability of other stimuli in this experiment.
The context manipulation did not result in the predicted effect. The exploratory analysis indicated that there is potential of finding a context effect in future experiments by minimally adjusting the material. Note the exploratory effect of context was comparably small in contrast to the penalty of the presence of ja. However, the difference score in Figure 6a is only due to context. Participants were instructed to judge only target items, which are identical for both context conditions. The difference score shows that participants might be influenced by the context and that the potential effect of context influence spans 0.8 scores (Figure 6b), which is approximately the difference between ja schon- and schon-RQs. Additionally, the contexts did not uniformly increase acceptability for a single stimulus but seem to interact with the particle combination, as expected under our assumptions of the specific felicity conditions for ja schon and schon alone.
We did not succeed in creating homogeneous supporting contexts. This was already conjectural during the construction of the items. As mentioned in Section 6.4, the original material was ill-suited for this manipulation although we used it to retain comparability to the other experiments. Contextual felicity is inherently complex, operating on multiple levels. By satisfying mutual knowledge about the answer to the RQ, we might violate other constraints. Still, the results signal that felicity is important in interpreting the acceptability data of DiPs. As shown already by Crain and Steedman (1985), context is very influential for language processing and interpretation processes and can even induce or prohibit garden path effects. Crucially, they illustrate that a null context is as influential in these matters as an assumedly ill-fitting overt context.
Our primary goal was to explain why ja appears sporadically in RQs if accompanied by another particle or adverb. Due to its basic function of marking common knowledge, ja is incompatible with questions, including RQs, which raises the issue how an accompanying particle could cancel the incompatibility. We focused on combined ja and schon in RQs, the most frequent found pattern. The predictions of our theoretical analysis were tested in four psycholinguistic experiments, yielding two main insights: ja operates on the meaning of schon, and ja schon-RQs must meet very specific contextual conditions to be (fully) felicitous and acceptable.
We first provided a detailed analysis of the meaning contributions of ja and schon individually and in combination. In experiment 1, we tested these predictions, expecting to confirm the intuitions on the natural findings of ja schon-RQs according to which the acceptability is heavily dependent on the presence of schon. Although schon did improve the acceptability of ja-RQs, the detected effect is rather subtle, which is why we deem schon necessary but insufficient to license ja. We took a closer look at the contexts of ja schon- and denn schon-RQs and concluded that a marked grammatical mechanism like the one that we argued to be at work in the natural findings of ja schon-RQs is unacceptable if contextually unwarranted. Our overall conclusion is that there is an interaction between the DiPs in ja schon-RQs.
Contrasting natural findings of RQs with ja and denn in context, we demonstrated that natural ja schon-RQs are identifiable as dismissive side remarks and are contextually motivated in that the speaker has a reason to indicate that it is uncontroversial, i.e. there is a shared awareness between speaker and hearer (as expressed by ja), that the answer to the RQ follows from the CG (as expressed by schon). Denn schon-RQs are (at least momentarily) QuDs around which discourses may evolve and which may be hotly debated although the speaker signals the answer to follow from the CG. Unlike ja schon-RQs, denn schon-RQs are not marked for, alternatively: do not presuppose, shared awareness of the answer’s following from CG information. In experiment 2, we replaced ja with denn: the partial null result for denn supported our predictions concerning the meaning contribution and the felicity conditions of denn in contrast to ja.
We reported two follow-up experiments strengthening our interpretations. Experiment 3 confirmed that participants interpreted the experimental stimuli as RQs, independently of DiPs. Although experiment 4 failed to show the expected contrast across all items, an exploratory analysis revealed that felicitous contexts may selectively increase the acceptability of ja schon-RQs without affecting the acceptability of schon-RQs. In sum, the experimental results support our theoretical analysis that schon is a necessary precondition to license ja in RQs, as are the contextual felicity conditions. The experiments also revealed that schon improves the acceptability of bare RQs and denn-RQs. The different effects that we attribute to the combination of ja vs. denn with schon can be connected to the discussion of same-scope vs. different-scope approaches to multiple DiPs (Section 2.5). We tentatively assume the following heuristic: If the higher particle is dependent on another element, they constitute a different-scope construction, as for ja schon-RQs. If both elements may appear in the respective environment on their own, a same-scope analysis seems preferable, as for denn schon-RQs. Whether this heuristic can be upheld for a wider set of combinations of DiPs must be decided by future research.
Defining the meanings of the DiPs, we took into account that DiPs surface in the midfield of German sentences above VP and select complements of type (ja) or (denn) or both (schon). Thus ja is excluded from RQs without reference to a mismatch with the illocution or clause type of the utterance. We justified this approach in light of recent research on ja and other DiPs in non-assertive environments. However, whether a structural dependence between DiPs and the clausal C system is assumed or not, ja must operate semantically on information other than the at-issue meaning, as this is generally the only way for it to occur in any question. Yet without an agreement relation between ja and the clausal left periphery, ja can be assumed to merge in its usual position with the schon-modified VP and not cause an agreement violation. Finding the object in its syntactic scope to denote a set of propositions in the at-issue dimension, ja would then necessarily apply to the meaning of schon, a semantic object of type . Alternatively, ja schon could be analyzed as a syntactic complex, but it remains unclear what would trigger the syntactic association with schon, due to the information-structural differences between schon and stressed phrasal constituents that ja occasionally attaches to (Section 2.7.1).
The distribution of DiPs in German RQs is informative to the ongoing debate on RQs in general insofar as German RQs are marked as questions rather than assertions, but a shared awareness of the obviousness of the answer is only present in a subset. Our approach to combined DiPs feeds into current theoretical debates on DiPs specifically and expressives generally. Having proposed that ja in RQs must be operating on non-at-issue information contributed e.g. by schon, we argued that this does not necessarily mean that DiPs are not expressives, given that Gutzmann’s (2015) framework for use-conditional meaning already provides a type of use-conditional modifier. If ja is able to operate on schon as proposed above, this mechanism could be applied to other lexical items and environments, some of which we presented above.
We especially thank Josef Bayer, María Biezma, Maribel Romero, Hans-Martin Gärtner, Umesh Patil and two anonymous reviewers for discussion and general support. Remaining mistakes are all ours.
The second author gratefully acknowledges the funding from the European Union’s Seventh Framework Programme for research, technological development and demonstration under grant agreement no. 613465 within the AThEME project.
Barr, Dale J., Roger Levy, Christoph Scheepers & Harry J. Tily. 2013. Random effects structure for confirmatory hypothesis testing: Keep it maximal. Journal of Memory and Language 68(3). 255–278.10.1016/j.jml.2012.11.001 Search in Google Scholar
Bates, Douglas, Martin Maechler, Ben Bolker & Steve Walker. 2015. Fitting linear mixed-effects models Using lme4. Journal of Statistical Software 67(1). 1–48. doi:10.18637/jss.v067.i01. Search in Google Scholar
Bayer, Josef. 2012. From modal particle to interrogative marker: A study of German denn. In Laura Brugé, Anna Cardinaletti, Giuliana Giusti, Nicola Munaro & Cecilia Poletto (eds.), Functional heads, 13–28. Oxford: Oxford University Press. Search in Google Scholar
Bayer, Josef. 2016. Why doubling discourse particles? Ms., University of Konstanz. Search in Google Scholar
Bayer, Josef. 2018. Criterial freezing in the syntax of particles. In Jutta Hartmann, Marion Jäger, Andreas Kehl, Andreas Konietzko & Susanne Winkler (eds.), Freezing: Theoretical approaches and empirical domains, 225–263. Berlin & Boston: de Gruyter Mouton. Search in Google Scholar
Bayer, Josef, Jana Häussler & Markus Bader. 2016. A new diagnostic for cyclic wh-movement: Discourse particles in German questions. Linguistic Inquiry 47(4). 591–629.10.1162/LING_a_00224 Search in Google Scholar
Bayer, Josef & Hans-Georg Obenauer. 2011. Discourse particles, clause structure, and question types. The Linguistic Review 28(4). 449–491. Search in Google Scholar
Bayer, Josef & Volker Struckmeier (eds.). 2017. Discourse particles: Formal approaches to their syntax and semantics. Berlin & Boston: de Gruyter. Search in Google Scholar
Bayer, Josef & Andreas Trotzke. 2015. The derivation and interpretation of left peripheral discourse particles. In Josef Bayer, Roland Hinterhölzl & Andreas Trotzke (eds.), Discourse-oriented syntax, 13–40. Amsterdam: Benjamins. Search in Google Scholar
Biezma, María. 2014. The grammar of discourse: The case of then. In Todd Snider, Sarah D’Antonio & Mia Weigand (eds.), Proceedings of SALT 24, 373–394. Washington, D.C.: LSA and CLC Publications. Search in Google Scholar
Biezma, María & Kyle Rawlins. 2016. Is this a rhetorical question or what? Ms., University of Konstanz, Johns Hopkins University. Search in Google Scholar
Caponigro, Ivano & Jon Sprouse. 2007. Rhetorical questions as questions. In Estela Puig-Waldmüller (ed.), Proceedings of Sinn und Bedeutung 11, 121–133. Barcelona: Universitat Pompeu Fabra. Search in Google Scholar
Coniglio, Marco. 2008. Die Syntax der deutschen Modalparktikeln: Ihre Distribution und Lizensierung in Haupt- und Nebensätzen. Venezia: Università Ca’ Foscari dissertation. Search in Google Scholar
Crain, Stephen & Mark Steedman. 1985. On not being led up the garden path: The use of context by the psychological syntax processor. In David R. Dowty, Lauri Karttunen & Arnold M. Zwicky (eds.), Natural language parsing: Psychological, computational, and theoretical perspectives, 320–358. Cambridge: Cambridge University Press. Search in Google Scholar
Doherty, Monika. 1985. Epistemische Bedeutung (Studia grammatica 23). Berlin: Akademie. Search in Google Scholar
Egg, Markus. 2012. Discourse particles at the semantics-pragmatics interface. In Werner Abraham & Elisabeth Leiss (eds.), Modality and theory of mind: Elements across languages, 297–333. Berlin & Boston: de Gruyter. Search in Google Scholar
Egg, Markus. 2013. Discourse particles, common ground, and felicity conditions. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives: Explorations in use-conditional meaning, 125–149. Leiden: Brill. Search in Google Scholar
Featherston, Samuel. 2008. Thermometer judgments as linguistic evidence. In Claudia M. Riehl & Astrid Rothe (eds.), Was ist linguistische Evidenz?, 69–90. Aachen: Shaker. Search in Google Scholar
Featherston, Samuel. 2009. A scale for measuring well-formedness: Why syntax needs boiling and freezing points. In Samuel Featherston & Susanne Winkler (eds.), The fruits of empirical linguistics: Volume I: Process, 47–74. Berlin & New York: de Gruyter Mouton. Search in Google Scholar
Féry, Caroline. 2006. Laute und leise Prosodie. In Hardarik Blühdorn, Eva Breindl & Ulrich H. Waßner (eds.), Text – Verstehen: Grammatik und darüber hinaus, 164–183. Berlin & New York: de Gruyter. Search in Google Scholar
Féry, Caroline. 2010. Information structure of schon. In Thomas Hanneforth & Gisbert Fanselow (eds.), Language and logos: Studies in theoretical and computational linguistics, 160–175. Berlin: Akademie Verlag. Search in Google Scholar
Franck, Dorothea. 1980. Grammatik und Konversation. Königstein: Scriptor. Search in Google Scholar
Fraser, Bruce. 2006. Towards a theory of discourse markers. In Kerstin Fischer (ed.), Approaches to discourse particles, 189–204. Amsterdam: Elsevier. Search in Google Scholar
Frühauf, Felix. 2018. On discourse particles in adjective phrases and the representation of speech acts. Berlin: Humboldt-Universität zu Berlin master thesis. Search in Google Scholar
Gärtner, Hans-Martin. 2017. Root infinitivals and modal particles: An interim report. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 115–143. Berlin & Boston: de Gruyter. Search in Google Scholar
Gast, Volker. 2008. Modal particles and context updating: The functions of German ja, doch, wohl and etwa. In Ole Letnes, Eva Maagerø & Heinz Vater (eds.), Modalität und Grammatikalisierung, 153–177. Trier: Wissenschaftlicher Verlag. Search in Google Scholar
Groenendijk, Jeroen A. G. & Martin J. B. Stokhof. 1984. Studies on the semantics of questions and the pragmatics of answers. Amsterdam: Universiteit van Amsterdam dissertation. Search in Google Scholar
Grosz, Patrick G. 2016a. Information structure and discourse particles. In Caroline Féry & Shin Ishihara (eds.), The Oxford handbook of information structure, 336–358. Oxford: Oxford University Press. Search in Google Scholar
Grosz, Patrick G. 2016b. Discourse particles. Ms., University of Tübingen, https://semanticsarchive.net/Archive/GQ4ZDU0N/grosz-SemCom-discourse-particles.pdf (27.06.2019). Search in Google Scholar
Gutzmann, Daniel. 2009. Hybrid semantics for modal particles. Sprache und Datenverarbeitung 33(1–2). 45–59. Search in Google Scholar
Gutzmann, Daniel. 2015. Use-conditional meaning: Studies in multidimensional semantics. Oxford: Oxford University Press. Search in Google Scholar
Hinterhölzl, Roland & Manfred Krifka. 2013. Modal particles in adverbial and adnominal clauses. Ms., Università Ca’ Foscari Venezia, Humboldt-Universität zu Berlin. Search in Google Scholar
Hofmeister, Philip, Laura Staum Casasanto & Ivan A. Sag. 2014. Processing effects in linguistic judgment data: (Super-)additivity and reading span scores. Language and Cognition 6(1). 111–145.10.1017/langcog.2013.7 Search in Google Scholar
Jacobs, Joachim. 1991. On the semantics of modal particles. In Werner Abraham (ed.), Discourse particles: Descriptive and theoretical investigations on the logical, syntactic and pragmatic properties of discourse particles in German, 141–162. Amsterdam: Benjamins. Search in Google Scholar
Jacobs, Joachim. 2018. On main clause phenomena in German. Linguistische Berichte 254. 131–182. Search in Google Scholar
Kaufmann, Stefan. 2010. Strong and weak presupposition: German ja under quantifiers. Ms., Northwestern University. Search in Google Scholar
Keller, Frank. 2001. Experimental evidence for constraint competition in gapping constructions. In Gereon Müller & Wolfgang Sternefeld (eds.), Competition in syntax, 211–248. Berlin & New York: de Gruyter Mouton. Search in Google Scholar
Klein, Wolfgang & Alexander Geyken. 2010. Das Digitale Wörterbuch der Deutschen Sprache (DWDS). Lexicographica: International Annual for Lexicography 26. 79–96. Search in Google Scholar
König, Ekkehard. 1977. Modalpartikeln in Fragesätzen. In Harald Weydt (ed.), Aspekte der Modalpartikeln: Studien zur deutschen Abtönung, 115–130. Tübingen: Niemeyer. Search in Google Scholar
Kratzer, Angelika. 1999. Beyond Oops and Ouch: How descriptive and expressive meaning interact. Paper presented at the Cornell Conference on Theories of Context Dependency, Cornell University, 26 March 1999. Search in Google Scholar
Kratzer, Angelika. 2004. Interpreting focus: Presupposed or expressive meanings?: A comment on Geurts and van der Sandt. Theoretical Linguistics 30. 123–136. Search in Google Scholar
Kratzer, Angelika & Lisa Matthewson. 2009. Anatomy of two discourse particles. Paper presented at SULA 5 – Semantics of Under-represented Languages in the Americas, MIT/Harvard University, May 15–17, 2009. Search in Google Scholar
Krifka, Manfred. 2011. Questions. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics: An international handbook of natural language meaning, vol. 2, 1742–1785. Berlin & New York: de Gruyter Mouton. Search in Google Scholar
Lauer, Sven. 2014. Mandatory implicatures in Gricean pragmatics. In Judith Degen, Michael Franke & Noah D. Goodman (eds.), Formal and experimental pragmatics: Workshop proceedings of ESSLLI 26, 21–28. Tübingen: Universität Tübingen. Search in Google Scholar
Lindner, Katrin. 1991. Wir sind ja doch alte Bekannte: The use of German ja and doch as modal particles. In Werner Abraham (ed.), Discourse particles: Descriptive and theoretical investigations on the logical, syntactic and pragmatic properties of discourse particles in German, 163–201. Amsterdam: Benjamins. Search in Google Scholar
McCready, Eric. 2010. Varieties of conventional implicature. Semantics and Pragmatics 3(8). 1–57. Search in Google Scholar
Meibauer, Jörg. 1986. Rhetorische Fragen. Tübingen: Niemeyer. Search in Google Scholar
Mroczynski, Robert. 2012. Grammatikalisierung und Pragmatikalisierung: Zur Herausbildung der Diskursmarker wobei, weil und ja im gesprochenen Deutsch. Tübingen: Narr. Search in Google Scholar
Müller, Sonja. 2017. Combining ja and doch: A case of discourse structural iconicity. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 203–240. Berlin & Boston: de Gruyter. Search in Google Scholar
Ormelius-Sandblom, Elisabet. 1997. Die Modalpartiken ja, doch und schon: Zu ihrer Syntax, Semantik und Pragmatik. Stockholm: Almqvist & Wiksell International. Search in Google Scholar
Potts, Christopher. 2005. The logic of conventional implicatures. Oxford: Oxford University Press. Search in Google Scholar
R Core Team. 2015. R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. https://www.R-project.org/ (27.06.2019). Search in Google Scholar
Rinas, Karsten. 2007. Abtönungspartikel-Kombinationen und Skopus. Sprachwissenschaft 32. 407–452. Search in Google Scholar
Roberts, Craige. 2012. Information structure in discourse: Towards an integrated formal theory of pragmatics. Semantics and Pragmatics 5(6). 1–69. Search in Google Scholar
Rojas-Esponda, Tania. 2013. The roadsigns of communication. In Raquel Fernández & Amy Isard (eds.), DialDam: Proceedings of the 17th Workshop on the Semantics and Pragmatics of Dialogue, 131–139. Amsterdam: University of Amsterdam. Search in Google Scholar
Sadock, Jerrold M. 1974. Toward a linguistic theory of speech acts. New York: Academic Press. Search in Google Scholar
Schulz, Petra, Emanuela Sanfelici, Alex Thiel & Corinna Trabandt. 2017. On the semantics and syntax of relative clauses in acquisition. Paper presented at the closing conference of the Research Unit “Relative Clauses”, Goethe-Universität Frankfurt, May 4, 2017. Search in Google Scholar
Searle, John R. 1975. Indirect speech acts. In Peter Cole & Jerry L. Morgan (eds.), Syntax and semantics: Volume 3: Speech acts, 59–82. New York: Academic Press. Search in Google Scholar
Thurmair, Maria. 1989. Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Search in Google Scholar
Thurmair, Maria. 1993. Äußerungsform oder Äußerungsfunktion?: Zu den Bedingungen für das Auftreten von Modalpartikeln. Deutsche Sprache 21. 22–43. Search in Google Scholar
Trabandt, Corinna. 2016. On the acquisition of restrictive and appositive relative clauses. Frankfurt: Goethe-Universität dissertation. Search in Google Scholar
Viesel, Yvonne. 2015. Discourse structure and syntactic embedding: The German discourse particle ja. In Thomas Brochhagen, Floris Roelofsen & Nadine Theiler (eds.), Proceedings of the 20th Amsterdam Colloquium, 418–427. Amsterdam: ILLC, University of Amsterdam. Search in Google Scholar
Viesel, Yvonne. 2017. Discourse particles ‘embedded’: German ja in adjectival phrases. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 173–202. Berlin: Gruyter. Search in Google Scholar
Vogel, Ralf. 2006. Degraded acceptability and markedness in syntax, and the stochastic interpretation of Optimality Theory. In Gisbert Fanselow (ed.), Gradience in grammar: Generative perspectives, 246–269. Oxford: Oxford University Press. Search in Google Scholar
Wochner, Daniela, Jana Schlegel, Nicole Dehé & Bettina Braun. 2015. The prosodic marking of rhetorical questions in German. INTERSPEECH 2015. 987–991. Search in Google Scholar
Zeevat, Henk. 2004. Particles: Presupposition triggers, context markers or speech act markers. In Reinhard Blutner & Henk Zeevat (eds.), Optimality theory and pragmatics, 91–111. Basingstoke: Macmillan. Search in Google Scholar
Zimmermann, Malte. 2011. Discourse particles. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics: An international handbook of natural language meaning, vol. 2, 2011–2038. Berlin & New York: de Gruyter Mouton. Search in Google Scholar
Zimmermann, Malte. 2014. Wird schon stimmen! A degree operator analysis of schon. Ms., Universität Potsdam, https://www.ling.uni-potsdam.de/~zimmermann/papers/MZ2016-schonPaper.pdf (27.06.2019). Search in Google Scholar