Bias in tag questions

Chapter 5 Bias in tag questions Cory Bill Leibniz-Zentrum Allgemeine Sprachwissenschaft (ZAS) Todor Koev Ben-Gurion University of the Negev We study various kinds of reverse-polarity tag questions in English, arguing that the speaker biases that such questions convey differ across three dimensions: optionality, strength, and polarity. We propose that the bias profile in each case primarily depends on the shape of the tag, while pointing at the central role of polarity focus and answer salience in generating these biases. 1 Introduction Tag questions are composed of two elements: a declarative anchor and a tag. While such questions require a response from the addressee, they also seem to convey some prior belief on the part of the speaker as to what the true answer is. Specifically, they seem to convey a belief or bias on the part of the speaker that the proposition expressed by the anchor is true.1For example, (1–3) suggest that it is indeed raining. 1Henceforth, when we talk about a question conveying ‘bias’, we are referring to this speakeroriented type of belief (i.e., ‘epistemic bias’ or ‘original bias’). Other kinds of question bias that have been discussed in the literature include ‘contextual bias’, or bias that has to do with evidence available in the context (Büring & Gunlogson 2000, Romero & Han 2004, Sudo 2013, Northrup 2014, Domaneschi et al. 2017), as well as ‘answer bias’, or bias about which answer the addressee is going to provide (cf. Krifka 2015, Malamud & Stephenson 2015, AnderBois 2019). While we have some ideas about how these biases could ultimately be derived from speaker bias, a proper presentation and exploration of these ideas goes beyond the scope of this paper. Cory Bill & Todor Koev. 2025. Bias in tag questions. In Tue Trinh, Anton Benz, Daniel Goodhue, Kazuko Yatsushiro & Manfred Krifka (eds.), Biased questions: Experimental results and theoretical modelling, 141–174. Berlin: Language Science Press. DOI: 10.5281/zenodo.17158184 Cory Bill & Todor Koev (1) It’s raining, isn’t it? (2) It’s raining, right? (3) It’s raining, eh? While, as (1–3) show, there are a variety of forms the tags can take, we will be focusing here on questions with final rising boundary tones and so-called reversepolarity tags, like (1). These are questions with a syntactically interrogative tag that is of the opposite polarity to its declarative anchor. We will refer to the variants with negative tags as negative-tag questions and the variants with positive tags as positive-tag questions.2 As noted in Ladd (1981), reverse-polarity tag questions have been associated with two kinds of intonation patterns: nuclear and post-nuclear. The nuclear intonation pattern involves a short pause after the anchor and a separate pitch accent on the tag. We will signal this intonation contour by placing a ‘||’ between the anchor and the tag, and by capitalizing the auxiliary verb in the tag, as shown in (4). In contrast, the post-nuclear intonation pattern involves no clear pause after the anchor and no separate pitch accent on the tag, with the pitch contour on the tag merely being a continuation of the contour of the anchor. This pitch accent will be signalled by placing a ‘=’ between the anchor and the tag, as shown in (5).3 (4) You don’t believe in Santa Claus || DO you? (nuclear) (5) You don’t believe in Santa Claus =do you? (post-nuclear) As presented in Table 1, these two axes of variation (tag polarity and intonation pattern) generate four distinct kinds of tag questions, all of which we will be exploring in the current paper. For each of the tag questions in Table 1, we have two goals. The first is to identify the bias profile associated with it. The second is to propose an analysis that derives this profile. We claim that the bias profiles for each of these questions are composed of three elements: polarity (positive vs. negative), strength (weak vs. strong), and optionality (optional vs. obligatory). We will now present each of these elements in turn, including examples of their different settings. In doing so, 2For recent discussions regarding how falling boundary tones and/or matching-polarity tags affect the meaning conveyed by tag questions, see Reese & Asher (2009), Malamud & Stephenson (2015), and Krifka (2015). 3From a more theoretical point of view (Selkirk 2005), these two prosodic patterns reflect the fact that nuclear tags form their own IntP, while post-nuclear tags are included in the IntP associated with the anchor. 142 5 Bias in tag questions Table 1: Typology of tag questions. positive-tag negative-tag post-nuclear Timmy can’t swim =can he? Timmy can swim =can’t he? nuclear Timmy can’t swim || CAN he? Timmy can swim || CAN’T he? we will also present the diagnostic tests that we will use throughout this paper to identify the bias profiles of our targeted tag questions. We start with bias polarity, which has two possible settings: positive and negative. In this paper, we will not be using any special diagnostics to determine the polarity of the bias associated with our targeted tag questions. This is because the bias polarity is intuitively clear and uncontroversial, and moreover this polarity is revealed by one of our diagnostic tests for bias strength (outlined below). Notice also that, as shown in (6) and (7), the bias inference has a polarity that matches the polarity of the anchor and is the opposite of the tag’s polarity. (6) You like pasta, don’t you? (positive bias) ⇝The speaker believes that you like pasta. (7) You don’t like pasta, do you? (negative bias) ⇝The speaker doubts that you like pasta. Now for bias strength, which we claim can be weak or strong. We will use two novel diagnostic tests to determine the strength of the bias; we call these the Follow-up Test and the Weighted Coin Test. In the Follow-up Test, a question is followed up with an epistemic statement which ostensibly reports the strength of the bias conveyed by the original question. Infelicity is expected to be generated when there is a mismatch between the strength of the bias conveyed by the question and the strength of the epistemic follow-up expression. For example, the question in (8) is felicitous when combined with a weak epistemic expression like suspect but less felicitous when combined with a strong epistemic expression like be sure.4The pattern presented in (9) is the opposite. This indicates that postnuclear negative-tag questions convey a weak bias, whereas nuclear negative-tag questions convey a strong bias.5Here and below, in paired examples we will use 4We chose suspect and be sure because these expressions are commonly used in English and because they clearly differ in modal strength. More generally, we hypothesize that epistemic expressions of similar strengths give rise to similar judgments. 5Notice that (8b) and (9a) are unacceptable for two different reasons. (8b) is bad because it overstates the bias of the tag question, a Quality violation, while (9a) is bad because it understates that bias, a Quantity violation. The same applies to the Weighted Coin Test, discussed below. 143 Cory Bill & Todor Koev ‘#’ to mark a contrast in acceptability, without committing to how strong the claimed contrast is. (8) This is your book =isn’t it? That is to say, ... (weak bias) a. I suspected it was. b. #I was sure it was. (9) This is your book || ISN’T it? That is to say, ... (strong bias) a. #I suspected it was. b. I was sure it was. Our second diagnostic for bias strength is the Weighted Coin Test. This test involvesa context with a coin that is weighted to land on heads/tails to some degree. The degree of this weighting dictates the strength of the prior belief, which is expected to match the strength of the bias conveyed by the question. When there is a mismatch between the strength of that prior belief (dictated by the weighting of the coin) and the strength of the question bias, we expect the question to be infelicitous. In contrast, when there is a match between these two elements, the question should be felicitous. For example, the post-nuclear negative-tag question in (10a) is felicitous only when combined with the weaker prior belief (i.e., a more weakly weighted coin), indicating that this question conveys a weak bias. In contrast, the nuclear negative-tag question in (10b) is more felicitous when combined with the stronger prior belief (i.e., a more strongly weighted coin), indicating a strong bias. (10) Mary and John are playing with a coin that they both know is designed so that it lands on tails N% of the time. Mary tosses the coin and it lands on the other side of John’s legs, such that only John can see the result. John looks confused, causing Mary to say: a. 70% / #99%: It landed on tails =didn’t it? (weak bias) b. #70% / 99%: It landed on tails || DIDN’T it? (strong bias) Finally, we claim that there is also variation with regard to the optionality of the bias conveyed by a tag question, which can be optional or obligatory. To determine the optionality features, we will use the By Any Chance Test, proposed in Sadock (1971). This test is based on the observation that the discourse marker by any chance only combines with questions that can receive a neutral interpretation, or questions for which any associated bias was only optionally conveyed. This means that an infelicitous combination of by any chance and a question can 144 5 Bias in tag questions be taken as evidence that the question is obligatorily biased. For example, the felicity of the post-nuclear positive-tag question in (11) indicates that the bias it may convey is optional, while the infelicity of the post-nuclear negative-tag question in (12) indicates that the bias conveyed is obligatory. (11) By any chance, Julia isn’t here =is she? (optional bias) (12) #By any chance, Julia is here =isn’t she? (obligatory bias) Our analysis of tag questions is predicated on the assumption that the bias properties of such questions primarily follow from the properties of the tag itself, as previously proposed in Romero & Han (2004). That is, we analyze tag questions as complexes consisting of a declarative and an elliptical question, where the latter conveys a bias with the very same features as would the corresponding independent non-elliptical question. This means that, for example, the post-nuclear negative-tag question in (13) and the high negation question in (14) share the same bias profile. (13) Julia is here =isn’t she? (14) Isn’t Julia here? Against this general parallelism between tags and independent questions, we will show how our analysis can capture the polarity, strength, and optionality settings for each of our chosen tag questions. We will argue that the trigger of bias in tag questions is ‘polarity focus’, i.e., focus marking on some polar element, such as negation or a covert verum operator. Specifically, we will claim that polarity focus triggers a bias that is obligatory and weak, while the semantics of the focused operator determines the polarity feature and may further boost the strength of the bias. In the following sections of this paper, we will present and apply the pieces of our analysis gradually, going through the following steps: (i) present a piece of our analysis, (ii) present the profile of the bias conveyed by one or more tag questions, and (iii) explain how the analytical pieces unveiled so far can account for this bias profile. Specifically, in Section 2 we present a general analysis of tag questions and show how this analysis can capture the bias profile of post-nuclear positive-tag questions. In Section 3, we provide some short background on polarity focus and show how its effects raise the salience of one of the answers and derive the bias profile of post-nuclear negative-tag questions. Section 4 introduces the phenomenon of verum accent and brings together all of the machinery to capture the bias properties of nuclear tag questions. Section 5 evaluates previous accounts and Section 6 is the conclusion. 145 Cory Bill & Todor Koev 2 Tag questions We start by laying out a general analysis of question tagging and then show how our analysis can be fruitfully applied to post-nuclear positive-tag questions. 2.1 General analysis It has been noted by many that tag questions of the form discussed here appear to be hybrid sentences. That is, they combine a declarative and a (VP-elliptical) interrogative clause in a single structure. It is for this reason that previous work dubbed them ‘queclaratives’ (Sadock 1971), ‘part statement and part question’ (Rando 1980), ‘double-barreled speech acts’ (Ladd 1981), ‘complex speech acts’ (Reese 2007), ‘speech act disjunctions’ (Krifka 2015), etc. Although the details of these characterizations and their associated analyses differ, the important point is that tag questions encode a regular proposition and a question partition, and these meanings are not further combined into a single semantic object. The core motivation for this claim seems to be that, if the two parts were to be collapsed into a single meaning, the output would either be a regular question (thus losing the bias) or a regular proposition (thus losing the interrogative force). In order to preserve both of these effects, we assume that the link between these two parts of tag questions is mediated by a covert tag operator. What this operator does is take the anchor meaning and the tag meaning as arguments, and create out of them a complex ‘dot’ object of the form ♢𝑝•𝑄, where 𝑄corresponds to the tag question and ♢𝑝corresponds to the anchor proposition prefixed by an epistemic possibility operator. An appropriate meaning for tag with these properties is given in (15).6 (15) JtagK=𝜆𝑄𝜆𝑝.♢𝑝•𝑄 Such metalinguistic dot operators have been previously used as separators between two aspects of meaning that are associated with the same linguistic structure (see Pustejovsky 1996; Potts 2005; Asher 2011). What we intend the dot operator to do for us is ship the (modalized version of the) anchor proposition and the tag question partition to the pragmatic component, where these are treated as engendering two separate speech acts, i.e., a modal assertion and a polar question. With this basic analysis of tag questions in place, we will now show how it can account for the bias profile of post-nuclear positive-tag questions. 6One plausible line is that the epistemic possibility operator introduced by tag is contributed by the rising boundary tone on the question tag. We leave the issue of the provenance of this operator to future work. 146 5 Bias in tag questions 2.2 Post-nuclear positive-tag questions Post-nuclear positive-tag questions, like (16), are composed of a negative anchor plus a positive tag, and contain no perceivable pause after the anchor and no separate pitch accent on the tag. As with all the tag questions we will explore, its bias profile is characterized by its optionality, polarity and strength settings. (16) You don’t eat fish =do you? Starting with optionality, as mentioned in Section 1, we diagnose its setting with the By Any Chance Test. As shown in (17), post-nuclear positive-tag questions can be felicitously combined with the discourse marker by any chance, indicating the possibility of a neutral reading of this question. In other words, the bias conveyed by post-nuclear positive-tag questions is optional. (17) By any chance, you don’t speak Romanian =do you? Additional evidence for this neutral reading comes from Reese & Asher (2009), who point out that in a context like (18) the tag question does not convey any bias. This optionality of bias in post-nuclear positive-tag questions was also noted in Sadock (1971) and Ladd (1981). (18) A and B are trying to complete a task at which neither is proficient, but at which Julie is known to be. A: We need someone who has consulted for us before. B: Julie isn’t here =is she? While the bias conveyed by such questions is optional, we would still like to identify the features it has when it is present. As already mentioned, we will employ the same diagnostic tests to identify both the strength and the polarity settings of our targeted biases, starting with the Follow-up Test. As the contrast in (19) shows, the bias conveyed by post-nuclear positive-tag questions is negative and weak. (19) Mark isn’t a body-builder =is he? That is to say, ... a. I suspected he wasn’t. b. #I was sure he wasn’t. Our second diagnostic test for strength, the Weighted Coin Test, provides further support that the bias conveyed by post-nuclear positive-tag questions is weak. As shown in (20a), such questions are infelicitous when the speaker’s prior belief that the coin would not land on tails is very high. The cause of this infelicity as 147 Cory Bill & Todor Koev coming from the strength of the bias is confirmed by the felicity that occurs when the chance of the coin landing on heads is decreased significantly, as shown in (20b). (20) Mary and John are playing with a coin that they both know is designed so that it lands on tails N% of the time. Mary tosses the coin and it lands on the other side of John’s legs, such that only John can see the result. John looks confused, causing Mary to say: a. 1%: #It didn’t land on tails =did it? b. 30%: It didn’t land on tails =did it? In sum, the bias profile of post-nuclear positive-tag questions is optional, negative, and weak. We will now show how this profile can be accounted for with the elements of our analysis introduced so far. On our analysis, post-nuclear positive-tag questions are the most basic form of tag question. That is, as presented in (21), they are composed of a modalized proposition in the anchor and an elliptical positive polar question. (21) Mary doesn’t eat fish =does she? a. [[CP Mary𝑖not eat fish] [tag [CP q she𝑖eat fish]]] b. ♢𝜆𝑤.¬𝑒𝑎𝑡𝑤(𝑚𝑎𝑟𝑦,𝑓𝑖𝑠ℎ)•{𝜆𝑤.𝑒𝑎𝑡𝑤(𝑚𝑎𝑟𝑦,𝑓𝑖𝑠ℎ), 𝜆𝑤.¬𝑒𝑎𝑡𝑤(𝑚𝑎𝑟𝑦,𝑓𝑖𝑠ℎ)} This analysis makes two good predictions about post-nuclear positive-tag questions. The first good prediction is that it does not say that the anchor proposition is plainly asserted. Instead, we merely predict that the possibility of the anchor proposition is asserted. If the anchor proposition was plainly asserted, we would create something like an illocutionary contradiction, where the speaker is both certain about the truth of the anchor proposition (by the norm of assertion) and ignorant about it (by the norm of questioning). Indeed, such a sequence of discourse moves would not be felicitous under normal circumstances (cf. #Mary doesn’t eat fish. Does she eat fish?). The second good prediction our analysis makes is that post-nuclear positivetag questions need not convey any bias. This is because the anchor merely asserts the possibility of the relevant proposition, a very weak statement. Moreover, we analyze the tag as an (elliptical) positive polar question, which is the canonical non-biased polar question form. Since post-nuclear positive-tag questions present a combination of a (negative) possibility and a plain positive polar question, it is unsurprising that such questions need not convey a bias at all. That 148 5 Bias in tag questions said, the fact that the negated proposition in the anchor is presented as a possibility may suggest that the speaker is slightly biased in this direction. But because this kind of pragmatic triggering is not directly linked to the semantic properties of the tag, the bias is cancelable. In sum, our general analysis of tag questions presented here straightforwardly captures the bias profile (optional, negative, weak) of post-nuclear positive-tag questions. 3 Polarity focus Another crucial piece of our analysis is ‘polarity focus’, or focus applied to a polar element. Therefore, we will start by providing a short background on focus as a general phenomenon, followed by a discussion of its effect when applied to polar elements, especially in post-nuclear negative-tag questions. 3.1 Background on focus interpretation A prominent theory of focus, known as ‘alternative semantics’, models focus as a feature 𝐹that marks syntactic constituents and generates relevant alternatives (Rooth 1985, 1992, 1997; see also Jackendoff 1972, Hamblin 1973, Kratzer 1991, Selkirk 1995, Schwarzschild 1999, Beaver & Clark 2008, Büring 2019; a.o.). According to this theory, each linguistic expression has two semantic values: ‘ordinary’ and ‘focus’. The ordinary semantic value of an expression 𝛼is rendered as J𝛼K𝑜 and corresponds to its usual denotation. The focus semantic value of 𝛼is rendered as J𝛼K𝑓and is always a set, although the nature of its content depends on whether the expression is 𝐹-marked or not. When 𝛼is not 𝐹-marked, its focus value is the singleton set comprised of the ordinary value of 𝛼. In contrast, when 𝛼is 𝐹-marked, its focus value is the set comprised of all alternative objects that are of the same semantic type as the ordinary value of 𝛼. When it comes to complex expressions, the focus semantic value is derived compositionally from the focus values of the immediate constituents, and so focus alternatives project up the tree. Formally, this process is generated via the recursive procedure shown in (22–23). (22) a. Non-focused lexical items J𝛼K𝑓={J𝛼K𝑜} b. Focused expressions (lexical or complex) J𝛼𝐹K𝑓={𝑥∈𝐷𝜏|J𝛼K𝑜∈𝐷𝜏} 149 Cory Bill & Todor Koev prior belief that the coin would land on tails is very high. The cause of this infelicity as coming from the strength of the bias is confirmed by the increase in felicity that occurs when the chance of the coin landing on tails is decreased, as shown in (38b). (38) Mary and John are playing with a coin that they both know is designed so that it lands on tails N% of the time. Mary tosses the coin and it lands on the other side of John’s legs, such that only John can see the result. John looks confused, causing Mary to say: a. 99%: ?It landed on tails =didn’t it? b. 70%: It landed on tails =didn’t it? In sum, the bias profile of post-nuclear negative-tag questions is obligatory, positive, and weak. We will now show how this profile can be accounted for using the elements of our analysis introduced so far. The structure that we assume for post-nuclear negative-tag questions is presented in (39). We have our general tag question shape, here consisting of a declarative anchor and an elliptical high negation question. Recall from Section 2.2 that a tag on its own does not necessarily convey bias, as displayed by the fact that post-nuclear positive-tag questions are only optionally biased. We argue, therefore, that the obligatory nature of the bias in post-nuclear negative-tag questions is coming from the tag, an (elliptical) high negation question. And as we presented in Section 3.2, the bias associated with high negation questions is weak, positive and obligatory, exactly the same as the bias associated with post-nuclear negative-tag questions. Therefore, we propose that the bias in such questions is generated in precisely the same manner. That is, it arises because the only possible antecedent that contrasts with the negative focus alternative is the positive alternative in (39c). And since this alternative entails the positive cell of the question partition denoted by the tag, the salience of that cell is raised and we get the intuition of a positive bias. And again, this salience mechanism indicates a mere preference, so the generated bias is weak. Moreover, the fact that this focus is derived from the necessary structure of the tag means that the bias is obligatory. This is based on the same explanatory mechanism as with all high negation questions. (39) Phillip rides to work =doesn’t he? a. [[CP Phillip𝑖ride to work] [tag [CP q [FocP not𝐹[TP he𝑖ride to work]]𝜙∼𝐶]]] b. ♢𝜆𝑤.𝑟𝑖𝑑𝑒.𝑡𝑜𝑤(𝑝ℎ𝑖𝑙𝑙𝑖𝑝,𝑤𝑜𝑟𝑘)•{𝜆𝑤.𝑟𝑖𝑑𝑒.𝑡𝑜𝑤(𝑝ℎ𝑖𝑙𝑙𝑖𝑝,𝑤𝑜𝑟𝑘), 𝜆𝑤.¬𝑟𝑖𝑑𝑒.𝑡𝑜𝑤(𝑝ℎ𝑖𝑙𝑙𝑖𝑝,𝑤𝑜𝑟𝑘)} 156 5 Bias in tag questions c. 𝐶=𝜆𝑤.𝑟𝑖𝑑𝑒.𝑡𝑜𝑤(𝑝ℎ𝑖𝑙𝑙𝑖𝑝,𝑤𝑜𝑟𝑘) d. J𝜙K𝑜=𝜆𝑤.¬𝑟𝑖𝑑𝑒.𝑡𝑜𝑤(𝑝ℎ𝑖𝑙𝑙𝑖𝑝,𝑤𝑜𝑟𝑘) e. J𝜙K𝑓={𝜆𝑤.𝑟𝑖𝑑𝑒.𝑡𝑜𝑤(𝑝ℎ𝑖𝑙𝑙𝑖𝑝,𝑤𝑜𝑟𝑘), 𝜆𝑤.¬𝑟𝑖𝑑𝑒.𝑡𝑜𝑤(𝑝ℎ𝑖𝑙𝑙𝑖𝑝,𝑤𝑜𝑟𝑘)} f. 𝐶∈J𝜙K𝑓3,𝐶≠J𝜙K𝑜3 Next, we will introduce another important element of our analysis, the verum operator. We will follow by an explanation of how, with this additional element, we can capture the bias conveyed by nuclear tag questions. 4 Verum The phenomenon of ‘verum accent’ involves a pitch accent on the finite auxiliary and – in the case of a declarative sentence – has the effect of emphasizing the truth of the expressed proposition (Höhle 1992). Thus, by uttering Oliver IS from Australia, the speaker stresses their belief that it is indeed true that Oliver is from Australia. This section presents the core data on verum accent and our account of it, and then discusses the role of verum accent in deriving the bias profiles of nuclear tag questions. 4.1 Core data on verum accent There are certain restrictions on the occurrence of verum accent that any account of it should capture. As Gutzmann et al. (2020) point out, verum accent is felicitous in two kinds of contexts: ‘contradictory’ and ‘affirmative’. Contradictory contexts are more common and arise when there is some dispute about whether the prejacent is true or false, as in (40). (40) A: Oliver is not from Australia. (contradictory context) B: He IS from Australia. In turn, affirmative contexts come about when the speaker and the addressee agree on the prejacent. We note that this use typically involves ‘extreme’ adjectives, like amazing,awesome,excellent, etc. (Cruse 1986, Paradis 2001, Rett 2008, Morzycki 2012). An example of such a context is presented in (41).13 13In an affirmative context, verum accent is also possible with regular predicates, although the result is once again an “extreme” interpretation. For example, if It IS raining has been uttered as a reaction to It’s raining, it would suggest a heavy rain and not just a light drizzle (cf. Umbach 2011 on extreme verbs). 157 Cory Bill & Todor Koev (41) After a colloquium talk: (affirmative context) A: Paula is an amazing linguist. B: She IS an amazing linguist. Crucially, a verum accent is not possible in neutral contexts, e.g., when a new issue has been raised by a neutral polar question (Wilder 2013, Samko 2016, Goodhue 2018, Gutzmann et al. 2020). This is illustrated in (42). (42) Out of the blue: (neutral context) A: Is it raining outside? B: #It IS raining. That is, in order for a verum-marked declarative to be felicitous, the issue must have already been discussed in prior discourse, as in (40–41) above. Just like in declaratives, when a verum accent features in polar interrogatives, we typically get the intuition of some kind of bias (Romero & Han 2004). For example, the question in (43) seems to convey a negative bias. (43) IS Oliver from Australia? ⇝The speaker doubts that Oliver is from Australia. Importantly though, the bias associated with verum accent in polar interrogatives is optional, as it can disappear in certain contexts. One such context is (44), where evidence for and against the prejacent has been provided by other parties and the speaker herself does not take a stand. The examples in (45) and (46) are drawn from the literature and make the same point. (44) DID Mary join the team? Because some say she did, others say she didn’t.  ⇝The speaker doubts that Mary joined the team. (45) A: Did Karl kick the dog? (Gutzmann et al. 2020: 41) B: No, Karl didn’t kick the dog. C: No, he DID kick the dog. A: Which is it? DID he kick the dog?  ⇝The speaker doubts that Karl kicked the dog. (46) B wants to know whether Jill will be at a meeting for members of a club. But B lacks an opinion about whether Jill is a member. (Goodhue 2019: 473) B: Will Jill be at the meeting? A: If she’s a member, she will. 158 5 Bias in tag questions B: IS she a member?  ⇝The speaker doubts that Jill is a member. The By Any Chance Test gives rise to the same result, as shown in (47), providing further evidence that the bias conveyed by a verum accent in polar questions is optional. (47) DID Mary join the team, by any chance? Though optional, notice that the bias triggered by verum accent is strong. This is attested by the Follow-up Test, as shown in (48). (48) IS Oliver from Australia? That is to say, ... a. ?I suspected he wasn’t. b. I was certain he wasn’t. We will now present our analysis of verum accent and show how it is able to capture the effects of this accent in declarative and polar interrogative sentences. 4.2 Verum as a covert operator There are two main approaches to analyzing verum accent. The “focus approach” posits that verum accent involves focus on a polarity head and manifests itself as a pitch accent on some element in the left periphery of the sentence (Laka 1990, Wilder 2013, Samko 2016, Goodhue 2018). This approach analyzes verum accent in essentially the same manner as we have polarity focus in Section 3.2, with focus being placed on a syntactically realized polarity head. In turn, the ‘operator approach’ contends that a verum accent signals the presence of a covert operator with certain conversational properties (Romero & Han 2004, Repp 2013, Goodhue 2019, Gutzmann et al. 2020). For reasons that we explore in detail in Bill & Koev (2021), we favor an explanation that is more in line with the latter approach.14 We propose that verum accent manifests the presence of a purely presuppositional verum operator that requires an epistemic conflict regarding the prejacent proposition in the given context. This is stated in (49). (49) JverumK𝑜𝑐(𝑝)=𝑝, provided that there is conflicting evidence about 𝑝in 𝑐 We assume that conflicting evidence about 𝑝involves two mutually exclusive pieces of evidence: a piece of evidence for 𝑝and a piece of evidence against 𝑝. 14On the empirical side, the strongest argument comes from Gutzmann et al. (2020), who argue that verum is overtly lexicalized in various typologically unrelated languages. 159 Cory Bill & Todor Koev Notice that contrasting evidence alone does not suffice, as such evidence need not produce an epistemic conflict and verum may not be licensed. Thus, if the positive and the negative pieces of evidence are presented as mere possibilities, a verum-marked sentence is degraded, as shown in (50).15 (50) A: It’s possible that Oliver is from Australia. B: It’s also possible that he is from New Zealand (though). C: ?No, he IS from Australia. Moreover, note that the strength of the two pieces of evidence does not need to be equal. For example, as shown in (51), it is possible for one side of the evidence to be strong and the other weak, provided the outcome is that they conflict. (51) A: Oliver is from Australia. B: I think he might be from New Zealand, actually. C: No, he IS from Australia. In sum, we claim that verum accent indicates the presence of a verum operator which contributes no at-issue content but rather a conflicting evidence presupposition. We will now show how this simple semantics can account for the distribution of verum in declaratives and polar questions. Starting with contradiction contexts, recall from example (40), repeated below as (52), that the prototypical use of verum accent is as a denial, targeting negative utterances. (52) A: Oliver is not from Australia. B: No, he IS from Australia. In this case, the conflicting evidence presupposition conveyed by verum is satisfied as follows: the negative evidence comes from the previous utterance, while the positive evidence has two possible sources. One option is that this evidence may be due to a prior positive utterance that A’s negative utterance is itself responding to. After all, one would generally not utter a negative sentence if the positive alternative had not been uttered or raised in some way. Even in the absence of such prior utterance, the conflicting evidence presupposition can be accommodated from the fact that the verum-marked sentence is being asserted by B and thus it is strongly supported by the evidence. Either way, the conflicting evidence presupposition is satisfied and verum is licensed. 15Notice that C’s utterance in (50) is not entirely out. The reason, we suggest, is that strong positive evidence can be accommodated from C’s (verum-marked) assertion, thus deriving the required conflict with B’s utterance. 160 5 Bias in tag questions As for affirmation contexts, we noted earlier that such uses typically involve extreme adjectives (or, more generally, extreme readings of predicates). The example in (53) is a repetition of (41) from earlier. (53) After a colloquium talk: A: Paula is an amazing linguist. B: She IS an amazing linguist. Morzycki (2012) proposes that extreme adjectives make use of the far end of the scale associated with the respective regular adjective. Following up on this idea, we can say that in (53) the extreme adjective amazing is parasitic on the regular adjective good, as it refers to extreme degrees of goodness. This derives the required epistemic conflict as follows. Let us assume that ⟨𝑔𝑜𝑜𝑑,𝑎𝑚𝑎𝑧𝑖𝑛𝑔⟩forms a Horn-scale, such that a sentence with amazing naturally invokes the respective alternative with good. In (53), A’s initial utterance of Paula is an amazing linguist will invoke the weaker alternative Paula is a good linguist. Now, if we allow that this latter alternative be strengthened to Paula is a good but not an amazing linguist by some standard scalar mechanism, we get an alternative that contradicts B’s verum-marked utterance She IS an amazing linguist. In other words, the use of an extreme adjective creates an implicit contraction within the same scale by splitting it into two non-overlapping regions. As a result, the conflicting evidence presupposition is met and verum is licensed once again. Finally, our semantics for verum straightforwardly derives the observation that verum accent is out in neutral contexts. That is, since such contexts lack conflicting evidence about the prejacent, the presupposition of verum is not satisfied and so a verum-marked sentence is out. As for the effects of verum accent in polar questions like (54), the bias that is generated is strong, negative, and optional, as already established in Section 4.1. We will derive this profile by appealing to the semantics we have presented for verum combined with the effects of polarity focus we laid out in Section 3.2. Specifically, we will claim that the polarity is dictated by polarity focus, whereas the strength and optionality are contributions of verum. (54) DID Mary join the team? ⇝The speaker doubts that Mary joined the team. In order to derive the optionality of the bias, we propose that verum-marked polar interrogatives may be associated with two homophonous Logical Forms, one with and another without focus marking. While both forms contain verum and thus require conflicting evidence about the prejacent, only the variant in 161 Cory Bill & Todor Koev which verum is 𝐹-marked conveys a bias. That is, we propose that (54) is ambiguous between (55a) and (55b). (55) a. [CP q [PolP verum [TP Mary join the team]]] (unbiased) b. [CP q [PolP verum𝐹[TP Mary join the team]]𝜙∼𝐶] (biased) The ordinary meaning of (55a) is the usual question partition that is comprised of the prejacent proposition and its complement. Since this structure also contains verum, it generates the presupposition of conflicting evidence about the prejacent. This is illustrated in (56). (56) a. [CP q [PolP verum [TP Mary join the team]]] b. JCPK𝑜𝑐={𝜆𝑤.𝑗𝑜𝑖𝑛𝑤(𝑚𝑎𝑟𝑦,𝑡𝑒𝑎𝑚), 𝜆𝑤.¬𝑗𝑜𝑖𝑛𝑤(𝑚𝑎𝑟𝑦,𝑡𝑒𝑎𝑚)}, provided that there is conflicting evidence about 𝜆𝑤.𝑗𝑜𝑖𝑛𝑤(𝑚𝑎𝑟𝑦,𝑡𝑒𝑎𝑚)in 𝑐 Notably, no part of the evidence needs to originate from the speaker and it can stem from other contextual sources entirely. This accounts for the optionality of the bias associated with verum-marked polar questions. In turn, (55b) gives rise to the same question denotation and conflicting evidence presupposition. However, in this case verum is focus-marked and thus requires an antecedent. Given the contrastive focus interpretation, the only antecedent that meets the squiggle-imposed condition in (26) is the negative question alternative, as shown in (57). (57) a. [CP q [PolP verum𝐹[TP Mary join the team]]𝜙∼𝐶] b. 𝐶=𝜆𝑤.¬𝑗𝑜𝑖𝑛𝑤(𝑚𝑎𝑟𝑦,𝑡𝑒𝑎𝑚) c. J𝜙K𝑜𝑐=𝜆𝑤.𝑗𝑜𝑖𝑛𝑤(𝑚𝑎𝑟𝑦,𝑡𝑒𝑎𝑚), provided that there is conflicting evidence about 𝜆𝑤.𝑗𝑜𝑖𝑛𝑤(𝑚𝑎𝑟𝑦,𝑡𝑒𝑎𝑚)in 𝑐 d. J𝜙K𝑓𝑐=JCPK𝑜𝑐={𝜆𝑤.𝑗𝑜𝑖𝑛𝑤(𝑚𝑎𝑟𝑦,𝑡𝑒𝑎𝑚), 𝜆𝑤.¬𝑗𝑜𝑖𝑛𝑤(𝑚𝑎𝑟𝑦,𝑡𝑒𝑎𝑚)}, provided that there is conflicting evidence about 𝜆𝑤.𝑗𝑜𝑖𝑛𝑤(𝑚𝑎𝑟𝑦,𝑡𝑒𝑎𝑚)in 𝑐 e. 𝐶∈J𝜙K𝑓3,𝐶≠J𝜙K𝑜3 The presence of polarity focus in verum-marked polar interrogatives derives the negative speaker bias in the same manner as the other questions with polarity focus. That is, the negative focus antecedent makes salient the negative cell of the question partition, resulting in the generation of negative bias. 162 5 Bias in tag questions Taking stock, we have derived both the optionality and the negative direction of the speaker bias in polar interrogatives with verum. The optionality follows from the assumption that verum, qua polar operator, may (though need not) carry focus marking. The negative direction is due to the fact that when such marking is present, the contrasting antecedent will be resolved to the negative focus alternative. The final element of the bias conveyed by questions with a verum accent is that it is strong. We can account for the strength of the bias conveyed by verum-accented questions as follows. Focus marking on a polar element (e.g., negation or really) only conveys a preference for one of the question partition cells, and the generated bias is expected to be weak by default. However, verum also introduces the presupposition that the context is conflicted about the prejacent, so the bias gets a boost. That is, in a conflicted context, conventionally the level of certainty required to make a contribution is higher than in a neutral context. For this reason, if biased at all, polar interrogatives with verum are strongly biased.16 4.3 Nuclear tag questions Nuclear positive-tag/negative-tag questions, like (59) and (58), are composed of a negative/positive anchor and an opposite polarity tag. Crucially, the prosodic contour of these questions is such that there is a clear break after the anchor, and there is a pitch accent on the auxiliary verb. (58) You haven’t watched Star Wars || HAVE you? (59) You have watched Star Wars || HAVEN’T you? Let us identify the bias profiles of these tag questions. Starting with optionality, as shown in (60) and (61), nuclear tag questions of both polarities are infelicitous when combined with by any chance. This indicates that a neutral reading of these questions is not possible. That is, the bias conveyed by both positive and negative nuclear tag questions is obligatory. 16One might wonder how our analysis would go accounting for questions containing both a verum accent and a focused polar element, like (i–iii). (i) DO vegetarians REALLY eat fish? (ii) DO vegetarians NOT eat fish? (iii) DON’T vegetarians eat fish? Basically, such questions would be analyzed as having a structure that contains both verum and polarity focus – this time not on verum, but on the other polar element, i.e., really or not. By applying a parallel reasoning to that above, we correctly predict that the resulting speaker biases are strong, obligatory, and of the opposite polarity to that of the focus domain. 163 Cory Bill & Todor Koev (60) #By any chance, you don’t like dancing || DO you? (61) #By any chance, you like dancing || DON’T you? Now, we will explore the strength and polarity settings of their biases, starting with the Follow-up Test. As shown by the preference for the strong epistemic follow-ups in (62) and (63), the bias conveyed by nuclear tag questions is strong and of the opposite polarity to the tag polarity. (62) Susan doesn’t hate exercise || DOES she? That is to say, ... a. #I suspected she didn’t. b. I was sure she didn’t. (63) Susan hates exercise || DOESN’T she? That is to say, ... a. #I suspected she did. b. I was sure she did. The Weighted Coin Tests in (64) and (65) confirm that the biases are strong. As shown, these tag questions are infelicitous when the speaker’s prior belief regarding the prejacent is relatively weak. The cause of this infelicity as coming from the strength of the bias is confirmed by the fact that we get felicity when the chance of the coin landing on tails is increased/decreased to near certainty one way or the other. (64) Mary and John are playing with a coin that they both know is designed so that it lands on tails N% of the time. Mary tosses the coin and it lands on the other side of John’s legs, such that only John can see the result. John looks confused, causing Mary to say: a. 30%: #It didn’t land on tails || DID it? b. 1%: It didn’t land on tails || DID it? (65) Mary and John are playing with a coin that they both know is designed so that it lands on tails N% of the time. Mary tosses the coin and it lands on the other side of John’s legs, such that only John can see the result. John looks confused, causing Mary to say: a. 70%: #It landed on tails || DIDN’T it? b. 99%: It landed on tails || DIDN’T it? In sum, the bias profile of nuclear positive-tag/negative-tag questions is obligatory, negative/positive (respectively), and strong. We now consider how the machinery of our analysis can capture the biases conveyed by nuclear tag questions. 164 5 Bias in tag questions To do so, we need to bring together all the different pieces of our analysis, including our general analysis of tag questions, polarity focus, and verum. To start with, we claim that the verum operator is necessarily present in nuclear tag questions, due to the prosodic contour associated with the tag. That is, the pitch accent on the auxiliary verb in the tag signals the presence of verum and its semantic effects. As for polarity focus, in the case of nuclear negative-tag questions, it is necessarily generated as a result of the high negation structure. For example, the tag question in (66) receives the analysis shown below. (66) Paul goes to church || DOESN’T he? a. [[CP Paul𝑖go to church] [tag [CP q [FocP not𝐹[PolP verum [TP he𝑖go to church]]]𝜙∼𝐶]]] b. ♢𝜆𝑤.𝑔𝑜.𝑡𝑜𝑤(𝑝𝑎𝑢𝑙,𝑐ℎ𝑢𝑟𝑐ℎ)•{𝜆𝑤.𝑔𝑜.𝑡𝑜𝑤(𝑝𝑎𝑢𝑙,𝑐ℎ𝑢𝑟𝑐ℎ), 𝜆𝑤.¬𝑔𝑜.𝑡𝑜𝑤(𝑝𝑎𝑢𝑙,𝑐ℎ𝑢𝑟𝑐ℎ)}, provided the context contains conflicting evidence for and against 𝜆𝑤.𝑔𝑜.𝑡𝑜𝑤(𝑝𝑎𝑢𝑙,𝑐ℎ𝑢𝑟𝑐ℎ) c. 𝐶=𝜆𝑤.𝑔𝑜.𝑡𝑜𝑤(𝑝𝑎𝑢𝑙,𝑐ℎ𝑢𝑟𝑐ℎ) d. J𝜙K𝑜=𝜆𝑤.¬𝑔𝑜.𝑡𝑜𝑤(𝑝𝑎𝑢𝑙,𝑐ℎ𝑢𝑟𝑐ℎ), provided the context contains conflicting evidence for and against 𝜆𝑤.𝑔𝑜.𝑡𝑜𝑤(𝑝𝑎𝑢𝑙,𝑐ℎ𝑢𝑟𝑐ℎ) e. J𝜙K𝑓={𝜆𝑤.𝑔𝑜.𝑡𝑜𝑤(𝑝𝑎𝑢𝑙,𝑐ℎ𝑢𝑟𝑐ℎ), 𝜆𝑤.¬𝑔𝑜.𝑡𝑜𝑤(𝑝𝑎𝑢𝑙,𝑐ℎ𝑢𝑟𝑐ℎ)} f. 𝐶∈J𝜙K𝑓3,𝐶≠J𝜙K𝑜3 The strong, positive and obligatory bias in (66) arises from the characteristics of the tag, which mirror those of the verum-accented question in (57), except that here the polar element focused is negation, instead of verum. Recall that the statement made by the anchor is very weak in that it only posits the possibility of the relevant proposition. On the other hand, the tag is comprised of a focused high negation and a verum operator. Therefore, as with all the other questions containing polarity focus, an obligatory bias that is of the opposite polarity to that of the focus domain (i.e., a positive bias) is generated. Moreover, the satisfaction of the conflicting evidence presupposition introduced by verum increases the strength of the bias, due to the higher certainty requirements conventionally associated with conflicted contexts. As for nuclear positive-tag questions, their bias is generated in precisely the same manner, except that in their case the focused polar element is verum. The details are presented in (67). 165 Cory Bill & Todor Koev Höhle, Tilmann. 1992. Über Verum-Fokus im Deutschen. In Joachim Jacobs (ed.), Informationsstruktur und Grammatik, vol. 4 (Linguistische Berichte Sonderheft), 112–141. Opladen: Westdeutscher Verlag. Jackendoff, Ray S. 1972. Semantic interpretation in generative grammar. Cambridge: MIT Press. Jamieson, E. 2018. Experimental evidence for the contextual acceptability of biased questions and tag questions. In ConSOLE 26: : Proceedings of the 26th conference of the student organization of linguistics in Europe, 331–349. Leiden: Leiden University Centre for Linguistics. Katz, Jonah & Elisabeth Selkirk. 2011. Contrastive focus vs. discourse-new: Evidence from phonetic prominence in English. Language 87(4). 771–816. http: //www.jstor.org/stable/41348860. Kratzer, Angelika. 1991. The representation of focus. In Arnim von Stechow & Dieter Wunderlich (eds.), Semantics: An international handbook of contemporary research, 825–834. Berlin: de Gruyter. Kratzer, Angelika. 2004. Interpreting focus: Presupposed or expressive meanings? A comment on Geurts and van der Sandt. Theoretical Linguistics 30(1). 123–136. DOI: 10.1515/thli.2004.002. Kratzer, Angelika & Elisabeth Selkirk. 2020. Deconstructing information structure. Glossa: A Journal of General Linguistics 5(1). 113. DOI: 10.5334/gjgl.968. Krifka, Manfred. 2015. Bias in commitment space semantics: Declarative questions, negated questions, and question tags. In Sarah D’Antonio, Mary Moroney & Carol-Rose Little (eds.), Semantics and Linguistic Theory (SALT) 25, 328–345. Ithaca: CLC Publications. Ladd, D. Robert. 1981. A first look at the semantics and pragmatics of negative questions and tag questions. In Chicago Linguistic Society (CLS) 17, 164–171. Laka, Itziar. 1990. Negation in syntax: On the nature of functional categories and projections. MIT. (Doctoral dissertation). Malamud, Sophia A. & Tamina Stephenson. 2015. Three ways to avoid commitments: Declarative force modifiers in the conversational scoreboard. Journal of Semantics 32(2). 275–311. DOI: 10.1093/jos/ffu002. Morzycki, Marcin. 2012. Adjectival extremeness: Degree modification and contextually restricted scales. Natural Language & Linguistic Theory 30(2). 567– 609. DOI: 10.1007/s11049-011-9162-0. Northrup, Oliver Burton. 2014. Grounds for commitment. Santa Cruz, CA: University of California Santa Cruz. (Doctoral dissertation). Paradis, Carita. 2001. Adjectives and boundedness. Cognitive Linguistics 12(1). 47– 65. DOI: 10.1515/cogl.12.1.47. 172 5 Bias in tag questions Potts, Christopher. 2005. The logic of conventional implicatures. Oxford: Oxford University Press. Pustejovsky, James. 1996. The generative lexicon. Cambridge: MIT Press. Rando, Emily. 1980. Intonation in discourse. In Linda R. Waugh & Cornelis Hendrik van Schooneveld (eds.), The melody of language, 243–278. Baltimore: University Park Press. Reese, Brian & Nicholas Asher. 2009. Biased questions, intonation, and discourse. In Malte Zimmermann & Caroline Féry (eds.), Information structure: Theoretical, typological, and experimental perspectives, 139–173. Oxford: Oxford University Press. Reese, Brian Jon. 2007. Bias in questions. The University of Texas at Austin. (Doctoral dissertation). Repp, Sophie. 2013. Common ground management: Modal particles, illocutionary negation and verum. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Expressives and beyond: Explorations of conventional non-truth-conditional meaning, 231–274. Leiden: Brill. Rett, Jessica. 2008. Degree modification in natural language. Rutgers, The State University of New Jersey. (Doctoral dissertation). Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of grammar: Handbook in Generative Syntax, 281–337. Dordrecht: Springer Netherlands. DOI: 10.1007/978-94-011-5420-8_7. Rochemont, Michael. 2013. Discourse new, F-marking, and normal stress. Lingua 136. 38–62. DOI: 10.1016/j.lingua.2013.07.016. Romero, Maribel & Chung-Hye Han. 2004. On negative yes/no questions. Linguistics and Philosophy 27(5). 609–658. Rooth, Mats. 1985. Association with focus. University of Massachusetts. (Doctoral dissertation). Rooth, Mats. 1992. A theory of focus interpretation. Natural Language Semantics 1. 75–116. Rooth, Mats. 1997. Focus. In Shalom Lappin (ed.), The handbook of contemporary semantic theory, 1st edn. (Blackwell Handbooks in Linguistics), 271–298. Oxford: Blackwell. Sadock, Jerrold M. 1971. Queclaratives. Papers from the Regional Meeting of the Chicago Linguistic Society (CLS) 7. 223–231. Samko, Bern. 2016. Syntax & information structure: The grammar of English inversions. Santa Cruz, CA: University of California Santa Cruz. (Doctoral dissertation). 173 Cory Bill & Todor Koev Schwarzschild, Roger. 1999. Givenness, avoidF and other constraints on the placement of accent. Natural Language Semantics 7. 141–177. DOI: 10 . 1023 / A : 1008370902407. Selkirk, Elisabeth. 1995. Sentence prosody: Intonation, stress and phrasing. In John A. Goldsmith (ed.), The handbook of phonological theory, 550–569. Cambridge: Blackwell. Selkirk, Elisabeth. 2005. Comments on intonational phrasing in English. In Sónia Frota, Marina Cláudia Vigário & Maria João Freitas (eds.), Prosodies: Selected papers from the Phonetics and Phonology in Iberia conference, 11–58. Berlin: Mouton de Gruyter. Selkirk, Elisabeth. 2008. Contrastive focus, givenness and the unmarked status of “discourse-new”. Acta Linguistica Hungarica 55(3-4). 331–346. DOI: 10.1556/ aling.55.2008.3-4.8. Sudo, Yasutada. 2013. Biased polar questions in English and Japanese. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives: Explorations in use-conditional meaning, vol. 28 (Current Research in the Semantics/Pragmatics Interface (CRiSPI)), 275–296. Leiden: Brill. Umbach, Carla. 2011. Intensifiers and measure phrases combined with verbs. In Proceedings of workshop “Scalarity in Verb-based Constructions”. HeinrichHeine-Universität Düsseldorf. Wilder, Chris. 2013. English ‘emphatic do’. Lingua 128. 142–171. DOI: 10.1016/j. lingua.2012.10.005. Woods, Rebecca & Tom Roeper. 2021. The acquisition path of “high” negation in English. In 45th Boston University Conference on Language Development, 761– 775. Somerville: Cascadilla Press. 174