Indirect truth marking via backgrounding: evidence from Bantu

1 Introduction

The term verum focus was coined by Höhle (1988, 1992) to refer to the subtype of focus scoping over the truth of a proposition, which in German and English is marked by a pitch accent (1). An early example in the Africanist literature is discussed in terms of polarity focus (a subtype of auxiliary focus), which is marked morphologically in the Grassfields Bantu language Aghem (Hyman and Watters 1984), illustrated in (2).

In the now-standard model of focus scope, verum focus is subsumed under operator focus within the notion of predicate-centred focus (PCF) (see e.g., Dik 1997: 331; Güldemann 2009; Zimmermann 2016), as presented in Figure 1.

Figure 1:

Diagram of focus scope, showing polarity focus as subtype of predicate-centred focus (Van der Wal 2021, based on Dik 1997: 331; Güldemann 2009; Zimmermann 2016; Güldemann and Fiedler 2022: 540).

More recently, the association between emphasis on the truth of a proposition and the notion of focus has been challenged, with Gutzmann et al. (2020) explicitly arguing that verum should be divorced from focus. They support this claim on the basis that the languages Gitksan (Tsimshianic), Bura (Chadic) and South Marghi (Chadic) have different marking for verum contexts to (other) focus contexts, concluding that the link between the use of pitch accent in (1) and in other focal contexts in German and English is in fact “superficial”:

we think that the concept of ‘verum focus’ should be abandoned, as it was motivated by the superficial similarity between verum and focus marking in languages like German and English. […] what is called ‘verum focus’ is not focus, but just a way to mark verum. (Gutzmann et al. 2020: 42 [emphasis added])

Thus, instead of invoking the notion of focus, Gutzmann et al. (2020) propose that markings such as (1) and (2) are due to the presence of a pragmatic operator VERUM, which functions to downdate the question about the truth of a proposition (?p) from the discourse (working under a Question Under Discussion (QUD) framework; Roberts 1996 et seq.). In other words, verum is used to refer to a situation whereby a speaker “fixes the polarity left uncertain in the question, and in addition adds emphasis on the polarity of the answer” (Gutzmann et al. 2020: 11). Building on earlier work by Romero and Han (2004), Gutzmann and Castoviejo-Miró (2011), and Repp (2009, 2013), they provide the following semantics for the VERUM operator:

Here, we see that VERUM is a use-conditional operator that applies when the speaker not only asserts the truth of a proposition p, but moreover wants to prevent the reverse polarity ¬p (not p) being added to the common ground. The negotiation of the question under discussion is therefore argued to be key to VERUM, rather than the polarity value of the proposition.

This separate conceptualisation of focus is formalised as the lexical operator thesis (LOT) by Gutzmann et al. (2020), where verum marking arises due to the presence of the VERUM operator. The approach stands in contrast with the focus accent thesis (FAT), whereby verum marking is a form of focus marking. As the languages that Gutzmann et al. (2020) bring forward have dedicated strategies for the verum contexts, i.e., strategies that are not found in polarity focus contexts, they propose a two-way typology:

The existence of Type II is used to provide evidence against the FAT in motivating a separation between verum and focus. The authors argue that the LOT applies universally, in that verum should be seen as separate from polarity focus, but has been previously unnoticed due to the “superficial” overlap between verum marking and focus in Type I.

It is therefore clear that verum can (at least in some languages) be distinguished from polarity focus, and we agree with Gutzmann et al. (2020) that this is indicative of a conceptual difference. However, in this paper we argue against the larger conclusions of Gutzmann et al. (2020) that the two concepts are unlinked, instead showing that verum is still dependent on focus (and background) marking. In the rest of this paper, we make a terminological distinction between verum (the pragmatic notion related to QUD management) and polarity focus (the traditional notion of focus on the polarity of a proposition), thus avoiding the term verum focus. We use the term truth as a cover term that includes both polarity focus and verum.

Empirically, the difference between polarity focus and verum can be identified by means of different discourse contexts, as provided in Matthewson and Glougie’s (2018) update of Zimmermann and Hole (2008), shown in (5) (see also Matthewson 2021).^[1] For example, a neutral answer to a yes/no question (ii) can be said to contain polarity focus, but not verum (‘Did you find the keys?’ ‘Yes, I found them’/’#I DID find them’).

Here, the difference between a verum context and a polarity focus context is that the former not only requires salient alternatives but also an “open conflict between salient alternatives” or “the final settlement of a question (regarding salient alternatives)” (Gutzmann et al. 2020: 14). For example, in a negatively biased yes/no question, the alternative p = 1 is not only salient but also in conflict with the assumptions of the person asking the question (namely, p = 0).

With this terminological background in place, we can return to the empirical picture. Gutzmann et al. (2020) acknowledge that their claims are based on a small number of languages, and therefore encourage the investigation of verum in more languages: further studies “have to look at languages that do not use accents as their means to mark focus” (Gutzmann et al. 2020: 15). In this paper we therefore investigate the morphosyntactic expression of verum and polarity focus in eleven Bantu (Niger-Congo) languages, as part of the broader cross-linguistic investigation of truth marking conducted by the authors in this volume. We find that the Bantu languages in our sample use a wide variety of grammatical strategies to express truth, with only one language having a dedicated strategy that directly marks verum (Type II). However, the other ten Bantu languages in our sample do not use focus accent (or other dedicated focus marking) to express truth (as in Type I). Instead, we show that verum can be marked indirectly by means of backgrounding strategies (in addition to universally-available lexical strategies), which we will add to the cross-linguistic typology as Type III.

Importantly, our data cover the expression of polarity focus and verum (truth) within the broader context of information structure, taking an onomasiological approach: by which linguistic means is truth expressed? This allows us to show more clearly whether and how polarity focus and verum are linked to focus more generally, in contrast to other work in the literature that has focused on the semantics and pragmatics of a particular strategy (i.e., the semasiological approach). As we will discuss, such consideration of the wider information-structural system is crucial for the evaluation of the relationship between truth and focus.

The remainder of this paper is structured as follows. Section 2 introduces our methodology and the languages in our sample, Section 3 illustrates the different strategies used to mark verum, Section 4 argues for distinctions in degree of underspecification through backgrounding, Section 5 relates the Bantu results to the cross-linguistic and theoretical picture, and Section 6 concludes.

2 Methodology

The starting point for our investigation is the question as to whether and how the expression of truth interacts with focus expression. To answer this question, we need to have a broader look at where truth is in the ecology of information structure in various languages. The Bantu languages are particularly interesting for such a study, as they are known to use morphosyntactic strategies to express various aspects of information structure (see e.g., Downing and Marten 2019; Van der Wal 2015).

In this study we draw upon the data on the expression of information structure in nine Bantu (Niger-Congo) languages that were systematically gathered by the members of the Bantu Syntax and Information Structure (BaSIS) research project, using a specially-designed methodology for the project (Van der Wal 2021). These languages are Tunen (A44, Cameroon), Teke-Kukuya (B77a, Republic of Congo), Kîîtharaka (E54, Kenya), Kirundi (JD62, Burundi), Rukiga (JE14, Uganda), Kinyakyusa (M31, Tanzania), Makhuwa-Enahara (P31E, Mozambique), Changana (S54, Mozambique) and Cicopi (S61, Mozambique).^[2] We provide additional data from a short field study on Gusii (JE42, Kenya) and additional work on Cinyungwe (N43, Mozambique). The map of BaSIS languages plus Gusii is shown in Figure 2. This set of languages was chosen as a convenience sample based on the expertise of our collaborators and languages of focus of the PhD students on the project, and therefore is not a balanced sample (specifically, there is over-representation of eastern Bantu). Furthermore, as the Bantu languages number approximately 550 varieties (Hammarström 2019), such a study can only be an initial investigation, rather than an exhaustive account of truth expression in Bantu. One point we can however already remark on is that the Southern Bantu languages in the Nguni cluster were not included in our study, but Bloom-Ström and Zeller (this volume) interestingly report similar findings with respect to the strategies used for truth expression.

Figure 2:

Map of ten Bantu languages in the BaSIS sample [plotting in R using lingtypology (Moroz 2017) with co-ordinates from Glottolog (Hammarström et al. 2022)].

The BaSIS project methodology (Van der Wal 2021) builds on earlier work on the investigation of information structure resulting in the Questionnaire on Information Structure (QUIS) methodology (Skopeteas et al. 2006). Diagnostic contexts for focus in the BaSIS methodology include function-to-form tests for the different focus scopes in Figure 1. Additionally, different types of focus are investigated, from information focus (tested via question-answer pairs) to more contrastive focus types including contrastive focus (an alternative is explicitly contrasted with another), corrective focus (an alternative is provided in correction of a previously asserted alternative), and mirative focus (an alternative is expressed that is high on a scale of surprisal) (see e.g., Cruschina 2021).^[3] Form-to-function tests are also used for further investigation of interesting strategies found in individual languages. Importantly, the elicitation tests involve explicit introduction of a discourse context, either by verbal or visual means, which controls for the focus interpretation of the answer. Finally, natural speech data were collected for each of the 9 BaSIS languages in order to observe focus expression in more natural speech contexts, allowing for example for study of discourse-initiality and interactive speech.

In terms of polarity focus and verum, the BaSIS methodology provides a more limited list of contexts than those in (5), due to having been designed prior to this investigation into the expression of truth cross-linguistically. However, investigation of polarity focus versus verum is still possible from our databases, as we have data from both unbiased yes/no questions (polarity focus context) and corrective focus tests (i.e., dispute context; verum context), in addition to supplementary material from the natural speech data. We have also collected additional judgements from our native-speaker linguist collaborators (whom we thankfully acknowledge). Examples where the source is not explicity indicated all come from the databases in the project; further information on these databases is provided at the end of the paper.

3 Verum strategies in Bantu languages

In conducting this study, we documented all of the strategies found in polarity focus and verum (jointly referred to as ‘truth’) contexts for each of the ten languages in our sample (see Table 1). By working from function to form, we show that each language has multiple means of truth expression. At the same time, in considering the broader information-structural system of each language (as detailed in Van der Wal ms. and chapters therein) we observe that none of the strategies found are dedicated to truth marking. We therefore see that there is no 1-to-1 mapping between the expression of truth and a given strategy. This multifunctionality of the strategies found in truth contexts will be important in our discussion in Section 4, where we link it to the notion of indirect focus marking via backgrounding. First, though, we will go through each of the strategies in turn: lexical strategies, unmarked verb form, marked verb form, dislocation of terms from comment, predicate doubling, and OM-doubling. We then turn to data from one additional language (Cinyungwe) that shows a dedicated verum marker. Due to space constraints, we will not exemplify each strategy for each language; we refer the reader to Table 1 for an empirical summary and to the Appendix for a full list of supplementary material.

4 Degrees of underspecification through backgrounding

Apart from lexical markers such as the word for ‘truth’ (Section 3.1) and the dedicated verum marker (Section 3.7), the other strategies presented in Section 3 show multifunctionality: they can all be used in more than one context and/or with more than one meaning. Therefore, the marking in these strategies only indirectly expresses truth, as the marking does not directly (i.e., does not by itself) indicate truth. Instead, we believe that the marking in these strategies consists of the backgrounding of what should not be interpreted as the focus, leaving what should be in focus open to be resolved via the discourse context. What is (directly) marked is therefore neither focus nor verum, but non-focus (background).

In previous work on information structure in African languages, Güldemann (2016) describes the backgrounding of non-focus as an alternative strategy to (direct) focus marking with potentially the same effect as direct marking of focus: when everything else apart from the focused constituent is marked as backgrounded, this essentially also highlights the focus. Güldemann refers to this strategy as maximal backgrounding, on account of it leaving “all but one potential focus host from the assertion domain” (Güldemann 2016: 577). However, as we shall see below, we believe that while the truth expression involves backgrounding, it is less helpful to phrase this as maximal backgrounding, in the sense that multiple sizes of focus scope can be identified (as also acknowledged by Güldemann 2016: 578). Our use of the term backgrounding instead of maximal backgrounding moreover links the African language data more clearly to earlier work on backgrounding as an indirect focus marking strategy in Germanic languages and the related idea of Givenness deaccenting (see e.g., Schwarzschild 1999).

We analyse the multifunctional potential of the strategies as varying in degrees of underspecification, corresponding to different degrees of restriction on the scope of what is left as the possible focus. Specifically, we distinguish four levels of underspecification, as illustrated in Figure 3: (i) no restriction (full underspecification), (ii) restriction to the predicate (partial backgrounding), (iii) restriction to operators (further backgrounding), and (iv) restriction to verum (no underspecification). We discuss and illustrate each below.

Figure 3:

Four degrees of underspecification via different degrees of backgrounding (dark shading), varying in what remains as scope of focus domain (unshaded).

The first degree of underspecification in our investigation is full underspecification, i.e., the use of the canonical SVO (or SOV in the case of Tunen) word order with an unmarked verb form (cf. Section 3.2). We see this unmarked strategy appearing in a topic-comment structure, where the comment may or may not contain narrow focus on an argument, adjunct, or the predicate. The multifunctionality of this maximally underspecified structure is clearly visible in example (34), where the canonical order and verb form can be used in a whole range of contexts, including polarity focus, and only excluding subject focus.^[11]

The second degree of underspecification has a moderate restriction, with backgrounding only of terms.^[12] By marking arguments and adjuncts (terms) as non-focus, the predicate and all its aspects are left as the potential focus (cf. Figure 1). We propose that this second level of underspecification is the case in our sample for dislocation, marked verb forms, pronominalisation and object marker doubling. We discuss these in turn.

As illustrated in Section 3.4 above, dislocation of arguments and adjuncts is used to indicate that the referents of these phrases are given information and therefore cannot form the focus in the current sentence. A typical effect of dislocation is the verb ending up in clause-final position. When the object is pronominalised as an object marker on the verb as part of a non-doubled object construction, it also indicates givenness. Although intonation does not seem to play a major role in the expression of information structure in the languages under study, there may still be an association between (prosodic) phrase-finality and focus. We can link this to the tendency in other languages to locate the focus element in the prosodically most prominent sentence position, either by moving the focused element to the position, or by evacuating non-focal material away from the position (Buell 2006, a.o.; Szendrői 2003; Zubizarreta 1998).

In some of the languages in our sample, the verb form can also indicate what is or is not in focus. We illustrate this with data from Makhuwa, which has the conjoint/disjoint verbal alternation. The conjoint verb form indicates that the element following the verb is in exclusive focus (Van der Wal 2011), whereas the disjoint form indicates that there is no such focus and must be used when the verb is clause-final. The disjoint form can thus be used in various predicate-centred focus interpretations, indicated in (35), but not for object focus, where the conjoint form must be used, as in (36). Note that in (35), the lexical object ehopa ‘fish’ can either precede or follow (but not both), and can easily be omitted altogether. Importantly, even when the lexical object follows, the disjoint form has to be used in the indicated contexts Q1–4.

Makhuwa (database Van der Wal, and Zanaira N’gamo p.c.)

The same goes for OM-doubling in languages where object marking is otherwise pronominal (not agreement): the generalisation is that the object marker defocuses the object, i.e., excludes it from the focus domain via backgrounding. Such defocusing merely indicates that the doubled object is not the focus – if another constituent is present in the postverbal domain, such as the adverb bwaangu ‘quickly’ in (37), this is now interpreted as the focus.

Lubukusu (Sikuku and Diercks 2021: 14)

Therefore, OM-doubling only restricts the possible focus by indicating what is not in focus, leaving the possibility of other focus scopes within predicated-centred focus open – including verum (in the right context, as seen in e.g., (28) above).^[13]

The third degree of underspecification illustrated in Figure 3 is found when not only terms are defocused but the predicate itself too. In this case, focus cannot be on a referential value, i.e., an actual referent in the case of arguments, a time or location in the case of adverbs, and the intentional meaning of an event in the case of predicates. Instead, the only aspects that are left to form the new information are the TAM value or truth of the statement. This leads to a restriction to operator focus, including verum, which therefore is still underspecified but less so than in the second level. The backgrounding of both terms and predicate is encountered in predicate doubling, specifically in topic doubling.^[14] The predicate here is marked as topical by placing an infinitive in the left periphery, which may also be marked by a contrastive topic marker (ko in (39) and kwé in (40) below). Logically speaking, the interpretation is underspecified in that it allows TAM focus, neutral polarity focus, or emphatic verum (biased context). However, we are unsure whether all these theoretically possible interpretations are actually possible for topic doubling in our sample. This is because, first, we do not know empirically whether topic doubling is actually used for TAM focus; and second, since an unmarked sentence is also acceptable as an answer in the same context, the extra marking must (through Gricean reasoning) have an added dimension of interpretation with respect to polarity focus, as illustrated in (38) and (39).^[15]

Kinyakyusa (database Lusekelo et al., and Amani Lusekelo p.c.)

Moreover, topic doubling can have additional pragmatic flavours of intensity, depreciation, and mirativity, as illustrated in the contexts for (40). Note, however, that these contexts all build on the contrastive topic interpretation of the infinitive (as argued by Jerro and Van der Wal 2022).

Rukiga (Asiimwe and Van der Wal ms.)

The final degree of underspecification is the complete lack of underspecification, i.e., the use of a marker that is dedicated to verum (direct truth marking). This strategy cannot be used in a neutral polarity question but only in a context of debate/doubt, or a biased question (see the diagnostic in (5viii) above). Examples are the particles in Gutzmann et al.’s (2020) data from Bura, South Marghi, and Gitksan, and our example of the Cinyungwe enclitic =di seen in Section 3.7 above.

In summary, we have seen that the Bantu languages in our sample use strategies that are underspecified to various degrees with respect to the scope of focus. The context by itself can be sufficient to trigger a verum interpretation from a completely unmarked form, while further restrictions (to the predicate, more narrowly an operator, or even more narrowly verum) provide “extra” indications to help the addressee identify the intended interpretation. In the second and third degrees of underspecification, the languages mark (to a lesser or greater degree) what is not in focus, rather than directly marking the focal content. By explicitly backgrounding non-focal information, these strategies indirectly mark what is in the scope of focus, leaving underspecified where the focus is precisely. For these strategies, any verum interpretation is therefore a pragmatic effect, in line with what Gutzmann and Castroviejo-Miró (2011) suggest when they say verum is a use-conditional rather than a truth-conditional operator.

Adding these Bantu data to the picture, we thus arrive at a new typology of truth-marking strategies, as represented in Figure 4. Note that, in contrast to the direct strategies, the strategies under ‘indirect’ mark the backgrounded information, rather than the focal information.

Figure 4:

A typology of truth-marking strategies (with examples in italics).

With these insights about indirect marking of verum, we can return to the main research question about the relation between verum and focus.

5 Discussion: (how) does verum relate to focus?

Returning to Gutzmann et al.’s (2020) comparison of how focus and verum relate in various languages, we have seen in the introduction that they only consider those (strategies in) languages where verum is marked directly as focus, and those where verum is marked separately. On the basis of the Bantu languages in our sample, we propose an updated typology in (41), adding the type in which verum may be marked indirectly via backgrounding.

Thinking back on how languages with Type I strategies led to proposals of the Focal Accent Thesis (FAT) and how the attestation of Type II strategies led to the Lexical Operator Thesis (LOT), we can ask what Type III tells us about the link between focus and verum. In order to do that, we first consider whether Type I can actually be reinterpreted as Type III. Gutzmann et al. (2020) in the very last paragraph of their article suggest the following, discussing whether the use of the same strategy for verum and focus marking in Type I strategies is a case of accidental homonymy:

the parallel realisation [between focus and verum in English and German, EK & JW] may be due to more general constraints on the prosodic systems of the languages. Suppose that (i) there must be a nuclear pitch accent in every utterance; (ii) this pitch accent must not be realised on given material (Schwarzschild 1999); and (iii) the core proposition is always given under verum and may therefore not carry accent. A similar idea, suitably adapted, could extend to non-accent languages, and could account for the parallel realisation of verum and focus in some languages without resorting to accidental homonymy. Confirmation of this idea, and further cross-linguistic investigation of this important question, will have to await future research.

(Gutzmann et al. 2020: 43 [emphasis added])

The shift of the focal accent in verum contexts could thus be due not to the marking of verum, but to the deaccenting of given material. With our investigation of 11 Bantu languages, we have confirmed that the main motivation for 4 out of our 6 grammatical strategies for verum marking is the backgrounding (or defocussing) of other potential targets of focus. Type I could thus be reinterpreted as or subsumed under Type III.^{[16] , [17]} We leave the possible reclassification of Type I to one’s analytical preferences regarding the relationship between focal accent and givenness deaccenting. Importantly for our current purposes, Types I and III show that it is untenable to maintain the complete independence of focus and verum. The emphatic verum interpretation is crucially only available in contexts where the truth of the proposition is on the table (i.e., ?p is the QUD). In other words, a potential VERUM operator can only apply in sentences in which the focus is neither on a term, nor on (the lexical content of) the predicate. This is a necessary precondition at the conceptual level, independent of the expression of verum and/or focus in individual languages. The use-conditional verum interpretation then builds on the information-structurally restricted set of potential focus targets.

Our results therefore give rise to what we call the Backgrounding and Underspecification Thesis (BUT), capturing the intuition that through the backgrounding of non-focal material, the scope of focus can be narrowed to ?p to a lesser or greater degree (42).

If direct focus marking of verum as in Type I can be reanalysed as an accent shift due to deaccenting of given constituents (i.e., analysing Type I as Type III), then the BUT can replace the FAT to subsume both languages with accent marking and languages with morphosyntactic marking. We may wonder, however, if the BUT can replace the LOT. It is clear that languages like Cinyungwe (and Bura, Southern Marghi, and Gitksan) formally distinguish between polarity focus on the one hand, and the emphatic verum interpretation on the other. Therefore, as mentioned, we agree with Gutzmann et al. (2020) when they claim that “verum is, semantically, not just focus” – the (use-conditional) semantics of verum and focus are indeed different. But how the verum reading is theoretically modelled is then another question. In languages that show a dedicated marker for verum (Type II), this marker could be seen as the spellout of the proposed VERUM operator (Gutzmann and Castroviejo-Miró 2011; Gutzmann et al. 2020; Repp 2009, 2013; Romero and Han 2004), which can be taken as evidence for the existence of such an operator. However, there is no empirical evidence in languages without a grammaticalised marker for the existence of a VERUM operator. This leaves only the conceptual proposal that since all languages can express verum, there must be a universal VERUM operator. Goodhue (2022: 153) in reaction to Gutzmann et al. (2020) “urge[s] caution before deciding in favor of an operator account for Germanic on the basis of evidence from other languages. After all, it is possible for different languages to make use of different grammatical constructions to achieve similar functional ends”. We agree with Goodhue, but in the end, it comes down to how much one would want to formalise pragmatics: should every shade of interpretation be captured as a feature or operator, or do we leave space for underspecification within the pragmatics? We leave this question open, and conclude here that while polarity focus can exist without the emphatic verum interpretation, the reverse is not true: verum needs a situation in which the truth of the proposition is under discussion, and it needs a sentence in which no other potential focal targets are marked. Empirically, the cross-linguistic findings of Type I and III strategies provide important evidence for the link between verum and predicate-centred focus. It is therefore misleading to completely separate the concept of verum from the concept of polarity focus.

While we leave the precise semantic/pragmatic modelling for further theoretical work, we note here that our proposal of indirect truth marking via backgrounding overlaps conceptually with the Unalternative Semantics approach of Büring (2015, 2016, 2019), in which crucially what is not available for focal interpretation (the so-called unalternatives) is key, rather than what is directly marked as focal. In other words, what is relevant is not the production of alternatives (in the sense of Rooth 1985, 1992) but rather the restriction of focus alternatives, i.e., the restriction of potential focus targets (Büring 2019: 14, 16). Because backgrounding is a means of limiting the scope of focus by removing possible targets (as we illustrated in Figure 3), such backgrounding strategies could thus be formalised as a means of marking unalternatives, rather than as means of marking focus (in contrast to the FAT-style approach). Note here in relation to the formalism that the Bantu language data show backgrounding via morphosyntactic means rather than prosodic deaccenting as found in Germanic languages, requiring a formalisation in non-prosodic terms (cf. Assmann et al. 2023 for an extension of a similar formalism to languages which mark focus morphologically).

6 Conclusions

In this paper, we have presented new data from ten Bantu languages to show that lexical as well as grammatical means are employed in the expression of truth focus, and that the grammatical strategies mark backgrounded, non-focal constituents, resulting in different degrees of underspecification of the scope of focus. We have also added data from one additional Bantu language showing evidence for a dedicated verum strategy that cannot be used in a polarity focus context. On the basis of these data, we extended the typology for truth marking strategies, adding indirect marking via backgrounding as a type (Type III) alongside the use of direct focus marking strategies (Type I) and dedicated verum marking (Type II). The notion of indirect focal marking via backgrounding builds on Güldemann’s (2016) notion of focus marking via maximal backgrounding but more explicitly highlights that strategies can vary in the degree of backgrounding, as understood in terms of focus scope (from maximal underspecification through generalised predicate-centred focus to focus on nothing but the truth value).

Our results have implications for the theoretical debate about the relation between verum and focus. Considering our typology, we see that the VERUM operator of Gutzmann et al. (2020) is well-motivated for Type II, but there is no evidence for it in the indirect strategy cases. While such an operator may be assumed to be covertly present in Type I and III strategies, we nevertheless argue that verum must be connected to focus, both on conceptual grounds and due to the non-trivial number of languages that have empirical overlap in markings – either of focus or of non-focus (i.e., backgrounding). We propose, then, that the emphatic verum reading, whether or not dependent on the presence of a VERUM operator, is dependent on having an information-structural context in which nothing other than the polarity value of the verb (i.e., ?p) is the QUD, and that the sentence expressing this may explicitly mark what is non-focal. The overlap in marking between verum and backgrounding in many languages is therefore non-superficial. We formulate this as the Backgrounding and Underspecification Thesis (BUT).

Since our study began as a research project on the expression of information structure in Bantu, it could be extended in further research by investigating truth expression in more detail. Further field investigation could systematically check more of the verum contexts of Matthewson and Glougie (2018) and Matthewson (2021), for example embedding, which is taken by Goodhue (2022) as crucial data for evaluating the lexical operator thesis in Germanic. Finally, while this paper was motivated by empirical data from Bantu languages, we also expect that indirect truth marking via backgrounding is available in other language families. This can be investigated further as part of studies on other languages, where we advocate for considering the (possible) multifunctionality of a given strategy in terms of the broader information-structural system of the language.

Abbreviations and symbols

Numbers refer to noun classes unless followed by sg or pl, in which case they refer to persons. High tones are indicated by an acute accent; low tones are generally left unmarked but are sometimes marked as a grave accent. Words in ALL CAPS in the translation indicate a pitch accent.

^!

downstep

*

ungrammatical

#

infelicitous in given context

?

questionable acceptability

{}

constituent in either position (but not both)

A

Changana determiner (unclear function)

appl

applicative

aug

augment

cj

conjoint

cm

contrastive marker

conn

connective

cop

copula

dem

demonstrative

dm

discourse marker

dist

distal

dj

disjoint

dur

durative

excl

exclusive

foc

focus

f.pst

far past

fut

future

fv

final vowel

^H

floating high tone

hab

habitual

ic

immediate causative

inf

infinitive

ipfv

imperfective

loc

locative

neg

negation

n.pst

near past

om

object marker

p/prep

preposition

pass

passive

pfv

perfective

poss

possessive

pot

potential

prs

present

prl

predicative lowering

pro

(independent) pronoun

prox

proximal

pst

past tense

q

question marker

QUD

question under discussion

QUIS

Questionnaire on Information Structure

rel

relative

rem.pst

remote past

sbjv

subjunctive

sm

subject marker

stat

stative

subs

subsecutive

ypst

yesterday past