Bura is a Chadic language belonging to the Biu-Mandara branch. It is spoken by approximately 250,000 speakers in the Nigerian states of Borno and Adamawa (estimation by Ethnologue in 1987). The only systematic linguistic description of Bura is Hoffmann’s grammar from 1955. The focus system of Bura is described in Hartmann & Zimmermann (2012). The data in the present article represent the variety of Bura spoken in Garkida, a city in Adamawa State, Nigeria. The data were mainly elicited from Chris Mtaku, a highly educated Bura speaker born in 1963 in Garkida. All the examples were confirmed by another speaker of Bura, Talatu Wakawa, from the same city.

South Marghi is also a Biu-Mandara language of the Chadic family. It is spoken by about 166,000 speakers (estimation by Ethonologue in 2006) in roughly the same regions as Bura. It has several dialects, which differ considerably from each other (see Hoffmann 1963). The language is basically undescribed. Hoffmann’s grammar from 1963 analyses Central Marghi, a related variety. A sketch of some basic grammatical properties of South Marghi is presented in Hartmann (2013). The data presented in this article are from Hajara Njidda, a resident of the city of Maiduguri.

Gitksan is an Interior Tsimshianic language that is spoken in northwestern British Columbia, Canada. The language is highly endangered, currently having approximately 500 fluent speakers according to Dunlop et al. (2018). Gitksan and its close relative Nisga’a form a chain of mutually intelligible dialects. The data in this paper come primarily from fieldwork with three speakers, representing three different dialects: Barbara Sennott (from Ansba’yaxw (Kispiox)), Vincent Gogag (from Gitanyaaw (Gitanyow)) and Hector Hill (from Gijegyukwhla (Gitsegukla)). Some data have been checked in addition with Ray Jones (from Prince Rupert and Gijegyukwhla) and Louise Wilson (from Ansba’yaxw, and seasonally Prince Rupert).

Kwak’wala belongs to the Northern branch of the Wakashan family. It is spoken in British Columbia, Canada, on northern Vancouver Island, adjacent islands, and the mainland opposite. It is critically endangered. Data and discussion in this paper are drawn from Littell (2016) and from Patrick Littell (p.c.).

Some notes on terminology, before we move on. Since the main aim of this paper is to get rid of the term verum focus, because we think its head focus is misapplied, we do not want to use that expression in the following. Instead, we use the term verum marking to refer to the observable linguistic phenomenon, verum accent to talk about the particular stress pattern used in German or English for verum marking, and just verum if we want to talk about about the concept itself.

The idea that the verum accent used in German and English for verum marking amounts to a focus accent is already the position Höhle (1992) takes in his original paper. We label this way of thinking about the verum accent the focus accent thesis or FAT. The two core assumptions of the FAT, which Höhle (1992) makes explicitly, are as follows.2

In order to make the assumption in (4a) viable, there must be something independent that is being focused. This is where assumption (4b) kicks in. Höhle assumes that sentences host a covert VERUM predicate which marks the proposition expressed by the sentence as true. The basic structure of the FAT can thus be stated as follows.

Given that in many cases, verum is realized by an accent on an element in C, a first assumption could be that the verum predicate is located there. However, as Höhle (1992) shows, this cannot be the case, because, under certain conditions, verum can also be realized in I (in which case the finite verb in final position in German is accented) or, in the case of relative clauses in German, in Spec,CP. For that reason, Höhle (1992: 137–138) assumes that the verum predicate is not represented in the syntactic tree, but instead is introduced during the semantic translation (see Lohnstein (2016) for an overview and discussion), and an accent on certain elements can then focus the non-segmentally introduced VERUM predicate.3 Thus, even without assuming a syntactic location for VERUM (which is nevertheless often assumed, see again Lohnstein (2016) for discussion), the general structure of the FAT holds.

In order to maintain that the verum accent is a focus accent on the (not segmentally located) VERUM predicate and does not, on its own, provide the verum reading, Höhle assumes that the VERUM predicate is present in the logical representation of every sentence, even ones which do not show verum marking. For a simple declarative with a focus accent elsewhere as in (5a), we would thus get a logical form along the lines of (5b), which can be paraphrased in natural language as in (5c).

In such an analysis, a sentence containing a verum accent involves a focus accent that focuses the VERUM predicate and thereby emphasizes the truth of the proposition.

What is the semantics of the VERUM predicate? Given that proponents of the FAT assume that VERUM is present in every sentence – whether there is verum accent or not – VERUM must not make an actual contribution to the meaning of a sentence if it is unfocused. That is, in the absence of any verum accent, the following equivalence should hold.

The paraphrase for VERUM given in (5c) fulfills this. If the propositional argument p is true, then It is true that p is also true. If p is false, then It is true that p is false. Hence, as suggested for instance by Zimmermann & Hole (2008: 5), VERUM has to be rendered as an identity function on truth values (i.e., the reverse of negation) or, when speaking intensionally, propositions.

This ensures that the truth value (or the proposition) denoted by a sentence is not altered by the presence of VERUM.

Given that VERUM is an identity function, one may wonder where the typical discourse effects induced by verum come from: for instance, that a verum utterance is infelicitous out of the blue and should address the question under discussion (Gutzmann & Castroviejo Miró 2011). This can be achieved by the standard machinery of alternative semantics for focus (Rooth 1992; Büring 1997). The main idea is that each expression has not just an ordinary semantic value ⟦·⟧^o but also a focus semantic value ⟦·⟧^f which represents the alternatives to the extension of that expression. That is, if Alex is focused, it invokes the contextually salient alternatives to Alex, say Blair and Chris, as well as Alex. Its ordinary semantic value remains unchanged.

If an expression is not focused, its focus value is just the singleton set of its ordinary semantic value.

The focus value of complex expressions is compositionally derived from the focus values of their immediate constituents. That is, if Alex is focused, the focus value from (9b) composes with the focus values of the other expressions in the sentence – all of which are singleton sets – to provide the alternatives for the entire sentence.

In the case where VERUM – an identify function on truth values – is focused, we only have three possible alternatives.4

Of those, however, only negation (12b) amounts to an actual linguistic alternative to VERUM, since the other two functions are a bit too useless to surface as natural language expressions. That is, the two focus alternatives to VERUM (and negation, for that matter) are VERUM and negation.

With this in place, a declarative sentence with verum accent has as its focus values just the proposition it expresses as well as its negation.

To see how this focus value can help derive the restrictions on the use of verum accents, we employ a general context condition (adapted from Büring 1997: 43) which links the focus value to the so-called question under discussion (QUD) of a context c (Roberts 1996/1998).

The QUD can be viewed as the immediate guidance of the discourse; it is the most recent issue (or discourse topic) that the interlocutors are trying to address. Technically, the QUD is just a semantic question modeled as a set of propositions. It is part of the larger discourse context and can be thought of as being for questions what the common ground is for assertions: Just as the discourse effect of an assertion can be modeled as an update of the context by contributing its propositional context to the common ground, the discourse effect of a question can be modeled as setting the current QUD. According to this idea, the QUD guides the discourse by determining felicitous discourse moves. For instance, assertions should address the QUD by (at least partially) answering it (Roberts 1996/1998). Note that the QUD does not have to be set by an explicit question, but is often given implicitly, in which case it has to be recovered, e.g. Büring (2003); Zimmermann (2014).

Coming back to a verum-marked utterance like (14), the condition in (16) ensures that it is only felicitous in contexts in which the polar question corresponding to the propositional content of the utterance is the QUD. That is, since the focus value of (14) corresponds to the semantic value of the question whether Alex loves Blair, (16) ensures that the verum-marked utterance is infelicitous if the question whether Alex loves Blair is not the QUD. For instance, (14) is infelicitous in an out-of-the-blue context, where the QUD is more like ‘What happened?’. (14) is likewise infelicitous in a context induced by the question of who Alex loves, since here also, the QUD is not equivalent to the focus value of the assertion.

This sketches a very straightforward view of the FAT, based on an approach that directly links focus to the QUD. However, there are some approaches to focus marking that are stronger than the direct QUD-based approach.5 One of the core ideas of these approaches is that the alternatives to the focused expression must be salient in the discourse context. These approaches thus disentangle the information-structural categories of focus and newness, which (more or less) fall together under the question based-approach. The easiest and, for our purposes, most relevant case to illustrate this is the difference between simple polar yes/no-questions and polar alternative questions.

In terms of the question being raised, (17a) and (17b) are equivalent as they both give rise to the question whether it rains or not. Under the QUD-based FAT sketched so far, we would therefore expect an answer to both (17a) and (17b) to license verum, as it selects the true alternative out of the two propositions that it is raining and that it is not raining. However, verum as an answer to (17a) seems weird in a neutral context, whereas it seems much more natural in an answer to (17b).

A version of the FAT that is based on the idea that alternatives must be salient in discourse can easily account for the difference between (18) and (19). While both answers provide new information (and relevantly address the QUD), only the context in (19) licenses focus. This is because while both questions introduce the question whether it is raining, only the explicit alternative question makes the alternative that it is not raining salient. Adopting recent ideas from inquisitive semantics (e.g. Roelofsen & Gool 2010; Farkas & Roelofsen 2015; Onea 2016), we may say that while both (18) and (19) introduce the question p,¬p, only (19) makes both alternatives salient; the simple polar question in (18) only makes the positive proposition salient. Using a box to indicate highlighting/salience of propositions, we can illustrate the difference between the polar and the alternative question as follows.6

The salience-based view of focus, and hence of the FAT, is stronger than the mere question-based view, since, as just illustrated, not every question is equally able to license focus in an answer. On the flip side, every context that licenses focus under the salience-based view should also license focus under the weaker question-based view. That is, focus contexts under the salience-based view are subsets of the focus contexts of the question-based view. Therefore, let us call the two versions of the FAT that correspond to these two views of focus the ‘weak FAT’ and the ‘strong FAT’ (using FAT when we do not wish to discriminate between the two). Even if we do not want to delve into the question which of the two approaches is more suited to deal with focus in general, we will have to discuss the two if we want to evaluate the FAT.

This concludes a brief outline of two ways in which the FAT may be spelled out more concretely, which we will use for comparing it to the alternative, the LOT. Before turning to the LOT, let us briefly reflect on why the FAT (in both versions) seems so attractive. For languages like German or English, the FAT does not have to assign any special status to an accent that looks like an ordinary focus accent. The only stipulation it has to make is to assume the presence of the VERUM predicate defined in (8b). Beyond that, the ordinary machinery of focus semantics should derive the contribution the verum accent makes to an utterance. That is, the attractiveness of the FAT is based on two assumptions. First, that the verum accent is just a focus accent and, second, that the predictions made by a focus analysis are all on the right track. We are going to question the validity of both these assumptions in this paper. Before we do so, let us first introduce the alternative contender to the FAT.

The competing thesis to the FAT is not based on the assumption that the verum accent is a focus accent, or that the contribution it makes is related to focus alternatives (neither in terms of relating to the QUD, as in the weak FAT, nor in terms of relating to salient alternatives, as in the strong FAT). Instead this approach assumes that the verum accent is a way to realize a lexical verum expression, which is responsible for the special discourse conditions verum puts on the felicitous use of an utterance. We call this the lexical operator thesis or LOT. Instead of assuming that every (positive) sentence involves a VERUM predicate with a trivial meaning, which in interaction with a focus feature gives rise to special discourse restrictions, the LOT builds the contribution of verum directly into a semantic operator which is only present in the semantic representation of a sentence if there is actually verum marking, i.e. verum accent in case of German or English.

While the FAT is explicitly stated by its proponents, the LOT is not usually as explicitly argued for. However, there are various proposals that are based on the LOT as they assign special semantics to the verum accent, often in the form of a contentful conversational operator that directly relates the propositional content of the sentence to the QUD (see, among others, Romero & Han 2004; Romero 2005; Gutzmann & Castroviejo Miró 2011; Repp 2013; Romero 2015) for different realizations of this idea). Again, as this is crucial, the verum predicate is only present if it is overtly marked (e.g. by verum accent). That is, the verum accent has “lexicalized intonational meaning” (Potts 2004). Despite the common term, the LOT therefore assumes that verum focus is no focus at all! It just happens that in German or English, this operator is realized by the same pitch accent that marks focus. This is then ultimately a case of homonymy (although it need not be merely accidental homonymy; see Section 8 for discussion).

Another difference between the FAT and the LOT is that the latter is much more flexible when it comes to analysing the semantic and pragmatic contribution of verum marking, since it is neither tied nor limited to whatever a focus-based view is able to derive. However, as a baseline, both versions of the FAT are rather easy to derive within the LOT. We just have to directly build the context condition into the lexical VERUM predicate. However, it is crucial that this contribution is not made at the truth-conditional layer, because that would lead to unwarranted truth-conditions (Romero 2005; Gutzmann & Castroviejo Miró 2011; Gutzmann 2012). Instead, it has to be located at a not-at-issue (Potts 2005), expressive (Potts 2007), or use-conditional (Gutzmann 2015) layer. That is, besides a sentence’s ordinary truth-conditions, which remain unaffected by the verum accent, the VERUM predicate realized by the accent puts conditions on the felicitous use of an utterance. We indicate this using a superscribed u for the use-conditional dimension and indicate felicity with the checkmark symbol. For replicating the weak FAT, which ties the verum effect to the QUD, we have something along the following lines.

The stronger, alternative-based view of the FAT can be rebuilt using the LOT as well. The following is not meant to be a perfectly worked-out analysis, but it shows how the strong FAT can be modeled by the LOT.

However, and this is crucial for later argumentation, the LOT is also free to assign more specific semantics to the VERUM predicate, if both the semantics given in (21) and (22) turn out to be insufficient to capture the facts. This is a move that is not available to proponents of the FAT, because they have to work with just what is independently established for focus phenomena. That is, it would be an argument against the FAT and for the LOT if it can be shown that a semantics along the lines of (21) and (22) is not sufficient to capture the restrictions imposed by verum. Previewing the results, we will see that neither the fact that the propositional content of a verum utterance corresponds to the QUD (= weak FAT) nor the fact that ¬p is a salient alternative to p in the utterance context (= strong FAT) are sufficient conditions to license verum. Therefore, we need something more. And crucially, the LOT is able to implement such a something that goes beyond the requirement of both alternatives being highlighted. That is, while the alternative-based, strong FAT can be understood as capturing a subset of the contexts that the question-based, weak FAT captures, the specific LOT we have in mind can, in turn, capture a subset of the contexts captured by the strong FAT. What we will propose later on is the following.

Figure 1 illustrates the relation between the weak FAT, the strong FAT, and the LOT we have in mind.

To summarize, the FAT assumes that the verum accent is a focus accent that focuses a trivial ever-present VERUM predicate. Together with an independently motivated focus semantics, this gives rise to discourse restrictions on the use of verum accents. The connection between verum accent and the discourse effects of verum is indirect and mediated by focus semantics. Since there is more than one theory of focus interpretation, there are corresponding varieties of the FAT, of which we sketched two here. The weaker version is based on the idea that the focus value of the sentence should correspond to the QUD. The stronger version in addition requires that the focus alternatives are contextually salient.

In contrast to the two versions of the FAT, the LOT assigns a lexical meaning to the verum accent. While the link to salient alternatives and the QUD is indirect in FAT approaches, under the LOT view, it is the accent itself which introduces the verum predicate into the semantic representation, and this operator alone is responsible for the verum effect, without any appeal to focus mechanisms. Figure 2 illustrates the architecture of the two approaches.

Despite the fact that the focus accent thesis is a thesis about the verum accent pattern in languages like English or German – namely that verum is, semantically and phonologically, just an ordinary focus – we like to think of it as a more universal thesis: according to the FAT, verum is a special case of focus (e.g. focus on the truth value, or something similar). And as a consequence, it is marked by a focus accent in German or English. In a similar vein, we take the LOT also to be a universal thesis: verum is, semantically, not just focus, not even in the languages in which it happens to be marked by means that otherwise are used to mark focus. That is, we argue for a universal semantic claim, namely that the semantic (or pragmatic) phenomena of verum and focus are separate, instead of verum being a special case of focus. Hence, our claims are not so much about verum and focus having different forms in some languages, but about their semantics being different. However, we will to some extent use the form of marking as evidence for semantics. In this way we will show that the flip-side of the reasoning that has led to the FAT for German or English (“If it doesn’t look like focus, it isn’t focus”, instead of “If it looks like focus, it is focus”) leads to a different result when applied to different languages.7

In order to diagnose the presence of verum in our language sample, we will use specific contexts where the expression of verum is either expected or excluded. The first contexts to be considered are so-called out-of-the-blue contexts. Here, the QUD is either non-existent, or a very broad ‘What happened?’. Accordingly, the realization of verum is awkward in such contexts.

Both versions of the FAT and the LOT predict that verum marking is infelicitous here, since the propositional content of the utterance does not correspond to the (maximal) QUD in such contexts. The even stronger condition that there are salient alternatives (e.g. that Carl is not writing a book in (24a)) is not fulfilled either of course.

There are different contexts that may help to distinguish the different approaches, as we discuss further below. These are cases where verum marking is optional and has an additional emphatic effect when it is present. The first of these context-types concerns the affirmation of a preceding truth value. In (25), the positive polarity is affirmed, and in (26) it is the negative polarity. The expression of verum is not obligatory as confirmed by the neutral replies. Stress on the auxiliary verbs in English leads to the interpretation of an emphatically confirmational statement. Therefore, the verum variants in (25B′) and (26B′) are only felicitous in a richer context.

Another context involves answers to yes-no questions. Verum fixes the polarity left uncertain in the question, and in addition adds emphasis on the polarity of the answer. The first two answer possibilities in (27) are non-emphatic answers; only the last answer contains emphasis on the truth value. This, again, requires a richer context than given here. For example, the verum-marked version of (27B) would be appropriate if the speaker expects that somebody might doubt her ability to sing. Again we will argue that this type of context distinguishes the competing theories.

A further type of context we use for diagnosing verum are opposite polarity contexts where the truth value of a previous utterance is corrected or denied; for illustration, see the Dutch example in (28) where verum is expressed by the particle wel.

Finally, to see whether the strong version of the FAT is sufficient, we should consider explicit alternative questions, that make both the positive and the negative alternative salient and therefore, according to the strong FAT, directly license verum marking. Consider the following variant of the dialog in (25).

The alternative question in (29A) makes both alternatives salient and verum in the answer feels completely natural, in contrast to (25), which indicates that the stronger context condition is more adequate than the pure question-based view. However, it should be noted that this observation does not mean that the contextual salience of alternatives is enough to license verum. In fact, we think there is more going on. As we will outline in more detail below, the alternatives must not only be salient, but there must have been some dispute going on as to which alternative should be accepted as the true one (see also Repp 2013). These seem to be precisely the contexts in which an explicit alternative question like in (29) would be licensed as well; in out-of-the-blue contexts, they are odd questions to ask. That is, verum may be licensed in the answer here not because the alternative question highlights both possible alternatives, but rather because of the context that licenses the alternative question in the first place. We will keep this in mind.

Having introduced a few different contexts that allow verum, let us now briefly discuss the distinction between the weak version of the FAT, its stronger variant, and the LOT, in a bit more detail.

The strong FAT, which requires that the context provides salient alternatives in order for verum to be used, correctly rules out verum in answers to polar questions in neutral contexts. Let us consider a couple of examples in order to make this observation clear. First consider a neutral context again:

However, if the question already contains some bias, using verum marking becomes perfectly fine.

This suggests that verum seems to require more than just addressing the QUD (as predicted by the weak version of the FAT). The strong FAT has the potential to perform better here, as it requires that both alternatives be salient.

A similar observation can be made with respect to responses which agree with assertions. As shown in (25) and (26), verum is not obligatory here, but, if employed, requires a more specific context than the version without verum. This poses a problem for the weak, purely question-based version of the FAT. The assertion of a proposition p makes the question whether p into the question under discussion. This is the case because the assertion awaits acceptance or rejection. (Stalnaker 1978; Farkas & Bruce 2010). Therefore, the context condition of the weak FAT is satisfied by the assertions in (25A) and (26A) and hence verum in the reaction should be fine. Under the assumption that the focus of an utterance should always match the question under discussion, verum would even be predicted to be obligatory in such contexts, contrary to fact. And even if one took a leaner position and allowed for non-optimal accentuation placement (Schmitz 2008), the use of verum in such cases should not have an additional discourse effect. However, the intuition is that the verum variants in (25B′) and (26B′) are only well-formed in a context in which the question whether Katie was looking good was already being discussed.

Turning to the strong FAT, which requires that both alternatives are salient, we note that the data in (25) and (26) are also challenging for this version. The problem is that in both contexts, only one polar alternative is salient: p in (25) and ¬p in (26). Hence, the strong FAT would require no verum accent in either case. However, the LOT, with its possibilities for finer-grained discourse conditions, has an avenue to account for the difference between the two variants in (30) and (31), without saying that the use of verum is required or ruled out. See § 3.3 for more discussion of the difference between the FAT and the LOT with respect to whether verum is obligatory.

For a slightly different line of evidence that verum marking requires more than ?p being the QUD, consider verum marking inside a yes-no question. As (32) shows, this is infelicitous in a question context in which there are no salient alternatives.8

The strong version of the FAT correctly rules out verum in such contexts, as the negative alternative is not salient in the context. But as already indicated, we think there is more going on than the mere salience of the alternatives. Consider the following context in which the speakers are pondering linguistic terminology.

In this context, the two alternatives (morphemes are part of syntax and morphemes are not part of syntax) are salient in the discourse context. However, the use of verum seems too strong and not completely felicitous. The reason, and this will be elaborated later, is that verum invokes a feeling of emphasis, something that would be unsuitable in the context of (33) in which speaker B is unsure about the answer. However, the contextual requirement of the strong FAT is still met, which shows that that is not enough, otherwise (33B) should be completely fine. See also Romero & Han (2004) for the observation that speaker certainty is a factor for the realization of verum.

Another context that illustrates that the mere presence of salient alternatives is not enough to license verum is the following. Imagine a committee of four people who must decide about a Mars mission. Only a unanimous vote will get the mission started. Even if the first three members have already voted for the mission, only with the last vote is the issue settled. Interestingly, if all previous votes have been positive, verum focus is licensed in the last vote, but in none of the previous ones.

In this example, the alternatives do not become more or less salient during the first three votes. However, it is the final vote that settles the issue of whether the Mars mission should be started or not. This, we believe, is what licenses verum in an example like this. We think that either an open conflict between salient alternatives or the final settlement of a question (regarding salient alternatives) is what licenses verum, something that goes beyond the mere salience of alternatives, as the strong FAT requires. This is the “something extra” we alluded to in Section 2.2 and what finally sets the LOT apart from the strong FAT, because the LOT is free to incorporate issues like the final settlement of a question into the semantic contribution of the verum expression, whereas the FAT is confined to what a theory of focus interpretation can derive. We will elaborate on this near the end of this paper and suggest a modified lexical semantics for the verum predicate that goes beyond the version given in (22) that was built after the strong FAT.

The first difference in predictions made by the FAT and the LOT concerns the means by which focus and verum are marked in a language. If verum ultimately is a focus phenomenon, as the FAT states, then the default assumption would be that the same means are used to mark focus and verum. This is obviously the case for German and English (ignoring the do-insertion), as this was the main motivation for the FAT in the first place. In contrast, if – as the LOT assumes – verum accent is not related to focus but introduces a use-conditional conversational operator, we should not expect a systematic overlap between verum and focus marking strategies. We will present cross-linguistic data to show that, if we consider a more diverse set of languages, the tight connection between verum and focus in German and English is not a general pattern.

To be clear, we do not claim that, in order for the FAT to be true, verum should always be realized by a (focus) accent. As formulated, the FAT is a thesis about languages like German or English in which verum is marked – like focus – by an accent. However, if the FAT were to hold cross-linguistically, similar parallelisms between verum and focus marking would be expected, even if focus is not realized by an accent to begin with. That is, the FAT lets us expect that if a language uses, say, a particle for focus marking, it also uses the particle for verum marking. Likewise, if a language has the option to not mark focus at all, verum focus should also be unmarked. We will discuss a Chadic language, Bura, which has unmarked VP-focus, but uses an obligatory particle to express verum. We take this fact as strong evidence for the correctness of the LOT.

To strengthen our point, we therefore have to look at languages that do not use accents as their means to mark focus. Of course, it is logically possible that in some languages verum and focus just happen to be realized differently – just like, for instance, predicate focus is often realized differently from ordinary term focus (Zimmermann 2016). Finding languages in which verum marking differs from focus marking is therefore admittedly not a knock-down argument against the FAT, especially if there are morpho-syntactic reasons why verum cannot be realized by the same means as ordinary alternative focus.9 However, it would be surprising if verum were a focus phenomenon but only looked like one in those languages that use pitch accents for focus. Therefore, finding differences between verum and focus marking in languages that do not use accents for them can be indicative for a not-as-close connection between verum and focus as the FAT assumes.

If the FAT were correct and verum were an instance of focus (on a verum predicate), then verum marking should obey the same restrictions that hold for focus phenomena in a given language. In particular we should observe a correlation between the possibility of having multiple foci and the possibility for verum to co-occur with another focus in the same sentence. That is, if a language exhibits multiple foci, then the FAT predicts that verum and focus can also co-occur, and if a language prohibits multiple foci, verum and focus must not co-occur. Since the LOT does not treat verum as a focus-based phenomenon, we do not expect a correlation between the possibility of multi foci and co-occurrence of verum and focus. That is, if we square the two factors multiple foci and verum+focus, the FAT predicts that only two of the four possible combinations are attested cross-linguistically, while the LOT in principle allows all four possibilities. This is illustrated in Table 1. The lighter gray cells are the combinations that are possible under both theses, while the darker ones are only expected if the LOT is true.

The third difference between the FAT and the LOT concerns whether the use of verum is obligatorily mandated by certain discourse conditions, or whether it is optional. We already touched on this in § 2.3. To give some background to this discussion, recall that in question-answer pairs, it is required that the focus in the answer matches the target of the preceding question. If there is a mismatch, the answer becomes infelicitous, as the following examples illustrate. (35A) establishes a context in which the subject is new information and thus focus on the subject is required in the answer as in (35B). In (35B′), the object is focused and hence this stress pattern is infelicitous in the context of (35A).

Now, if verum accent is focus on the verum predicate (and thus on the truth value), which contrasts with negation, then the weak FAT expects that verum accent should be required in answers to yes-no questions. In contrast, this is not necessarily the case under the strong FAT, because a simple yes-no question does not make both alternatives salient and therefore does not, without further contextual factors, fulfill the contextual requirement of the strong FAT. However, in answers to explicit alternative questions that combine a positive sentence with its negative counterpart, verum should be expected under the strong FAT. This contrasts with the LOT, according to which the use of verum marking is entirely optional, as it adds additional content to the use-conditional dimension. Furthermore, as already alluded to above, if the conditions that license verum are even richer than those that license a focus-based interpretation, this could easily be implemented by the LOT, while it is hard to see how the FAT could do that without stipulating extra content to verum focus (which would turn it into a special LOT-variant).

Table 2 summarizes the different predictions of the three approaches regarding the obligatoriness/optionality of verum. With simple polar context we refer to contexts of assertion or polar questions in which only one side of a polarity is salient in discourse, while we use salient polarities context to refer to contexts in which both alternative polarities are salient in the dicsourse context, as is the case, for instance, with alternative questions.

While the weak FAT renders verum obligatory in both contexts, the strong FAT only lets us expect verum to be used in the contexts of salient polarities. The LOT, in contrast, says that verum can optionally occur in both contexts, depending what use-conditions it assigns to verum and whether these are fulfilled. That is, the actual predictions of whether verum can be used in a particular context depends on the actual meaning the LOT assigns to the lexical verum operator. However, in any case, the LOT does not predict verum to be obligatory.

In this section, we considered three areas in which the FAT and LOT lead to different empirical predictions. The FAT leads us to expect a tight connection between verum and focus, and makes the following predictions.

In contrast, under the LOT there is no direct connection to focus and its interpretation. This leads to the following predictions.

The different predictions made by the FAT and the LOT are also summarized in Table 3.

While it is not easy to come to different conclusions regarding these predictions, and thus to differentiate between the two competing theses, on the basis of German or English alone, languages that differ from the well-studied ones can help to evaluate the analyses. In the next three sections, we will present data from two Chadic languages and the Tsimshianic language Gitksan. The data suggest that it is the LOT which is on the right track.

The first case where the FAT and the LOT make different predictions is in whether ordinary focus and verum marking are realized in the same fashion. In intonation languages like German or English, both focus and verum are realized by the same pitch accent. This is shown in (38a) for object focus, (38b) for verum marking on the auxiliary, and (38c) for subject focus:

As pointed out in Section 2.1, the identity in the expression of focus and verum essentially motivated the FAT. However, the formal equivalence between focus and verum marking does not hold cross-linguistically, not even among European intonation languages (see e.g. the contributions in Breitbarth, De Clercq & Haegeman 2013). For illustration, consider Dutch. Besides also allowing for a pitch accent realization on the finite verb (similar to German), emphasis on a contrasting truth value is expressed by the multifunctional particle wel. In (39), repeated from (28), wel expresses the denial of the propositional content of the previous utterance.

Affirmative polarity is expressed by a different particle, namely inderdaad ‘indeed’:

Note that the particles wel and inderdaad are obligatorily stressed (Beata Moskal and Laurette Artois, p.c.). Thus, verum marking – both opposite and affirmative – in Dutch can be realized by preverbal particles, yet in combination with prosodic emphasis.10

These facts do not follow straightforwardly from the FAT, according to which the verum predicate (which, in Höhle’s approach, is not syntactically present hence not segmentally located) is focused by accent. In order to explain the particle insertion in Dutch, the FAT would have to assume that in some languages, one (or sometimes maybe the) way to focus the null expression is by making available an additional element to put stress on. However, the FAT does not explain why a language would choose this option, especially if – as is the case in Dutch – stressing the finite verb can be employed for this as well.

This already sheds some doubt on the idea that verum is focus, even though, as mentioned above, a discrepancy in the realization of verum and focus does not logically preclude the idea that they are nevertheless the same. However, the fact that in Dutch, the normal method of marking focus could easily be employed to mark verum, but is not, is unexpected given the FAT and should skew our expectations more away from the FAT compared to when we were only looking at German.

In the following sections, we will find even more indications of a disconnect between focus and verum when we discuss non-intonational languages. As alluded to above, if the FAT were true, we might expect similarities between verum and focus marking even in languages in which no accent is used for focus. However, this is not what we find. The LOT, which does not predict any correlation between focus and verum, seems more likely given such differences in marking strategies.

The formal realization of constituent focus and verum differs in many Chadic languages. In this section we address verum expression in two Central Chadic languages (Bura, cf. Hartmann, Jacob & Zimmermann 2008; Hartmann & Zimmermann 2012), and South Marghi (cf. Hartmann 2013). Both languages have SVO as their base order. They are tone languages and differentiate a high and a low tone. In the following, only high tones are represented, by accents on the vowels.

The Chadic languages do not behave alike with respect to the availability of multiple focus. Whereas Bura allows for multiple foci, South Marghi is more restricted in this respect. However, neither of the Chadic languages under discussion allows for the co-occurrence of wh/focus-marking and the expression of verum. We start the discussion with Bura, which falls into the top-right cell in Table 1.

Bura exhibits multiple wh/focus. Just like in German and English, only one wh-/focus phrase may be fronted; all additional wh/focus phrases remain in-situ. This is shown in (82).

A crucial difference between Bura and German or English consists in the fact that wh- and constituent focus marking in Bura is incompatible with the expression of verum. Under the FAT, which considers verum as an instance of focus, one would as a default expect the availability of verum focus in multiple focus constructions. This is not borne out for Bura. Neither a wh-question (83) (Hartmann, Jacob & Zimmermann 2008: ex. (50c)), nor a focus construction (84) (Hartmann, Jacob & Zimmermann 2008: ex. (51)) may contain the particle kú.

South Marghi also blocks verum in focus constructions, but unlike in Bura, it does not license multiple focus in general. It thus falls into the bottom-right cell in Table 1 above. Example (85) shows that multiple wh-questions are marginal in South Marghi. A potential reason for the marginality of (85) could be that South Marghi does not license in-situ focus; cf. Section 4.2.2.

South Marghi exhibits an interesting blocking effect: The realization of constituent focus is incompatible with verb raising, which was identified as the means to mark verum in this language in Section 4.2.2. Thus, verb raising is excluded in wh-questions, as illustrated in (86). It is also blocked in answers to wh-questions; see (87). In both examples, the ungrammaticality results from combining wh/focus and raising of the verb usa ‘to greet’ to the subject agreement head (-ja). See Hartmann (2013) for more discussion.

To summarize, although the two Chadic languages investigated here show different strategies to express verum, they are similar in that constituent focus marking is incompatible with the respective expression of verum.21

Since Bura allows for multiple focus constructions, these results are in contrast to what is predicted by the FAT. South Marghi does not have multiple focus in the first place, which makes the South Marghi results compatible with both the FAT and the LOT.

Gitksan differs from the two Chadic languages discussed in that it allows the co-occurence of verum marking and focus. It therefore appears on the lefthand side of Table 1 above. Gitksan also allows multiple foci. This makes Gitksan parallel to German and English and places it in the top-left cell in our table.

As evidence that Gitksan allows multiple foci, first we discuss multiple wh-questions. Important background to the discussion is that Gitksan is a language with wh-indefinites. These are not polarity items, but freely appear in all argument positions, usually in combination with the domain-widening element ligi.

Wh-words are obligatorily fronted in wh-questions, and only one may be fronted. A second wh-word must remain in-situ. This leads to an ambiguity, whereby the same string can be interpreted either as a wh-question containing an indefinite, or as a multiple wh-question. This is shown in (89) and (90), where the same string was offered by consultants to convey the two different meanings.

Multiple wh-questions can receive pair-list answers, as shown in (91).

Further evidence for multiple foci in Gitksan comes from the example in (92).22

Now we turn to the evidence that verum marking can co-occur with focus in Gitksan. This is shown in (93) and (94) for verum co-occurring with constituent focus (marked by fronting/clefting).

Further evidence for the co-occurence of verum with focus in Gitksan comes from the fact that the verum marker k’ap can, at least marginally, appear inside wh-questions. An example of this is given in (95).23

The data in this section have shown that Gitksan patterns just like German and English with respect to multiple wh/focus and the co-occurence of verum and focus – it allows both. These facts are compatible with either the FAT or the LOT.

So far we are missing a language from the final cell in our table, the bottom left: a language which does not allow multiple foci, but does allow verum marking to co-occur with focus marking. The existence of such a language is predicted by the LOT, but not by the FAT. In this sub-section we briefly bring in a sixth language, Kwak’wala, which appears to fill this gap in our typology.

In Kwak’wala, the primary means of marking constituent focus are clefting and nominal predication. These are illustrated in (96) and (97) respectively. In (97B), the nominal ‘dog’ is the main predicate of the sentence.

Multiple marking of constituent focus appears to be disallowed in this language. In support of this we note that ‘“multiple WH questions cannot be constructed” (Littell 2016: 225; see also Littell 2016: 362).24 If it is correct that multiple focus is disallowed in Kwak’wala, the FAT would predict that verum marking cannot co-occur with focus marking. However, the verum marker ʔm can co-occur with explicit marking of focus, as shown in (98).25

We therefore tentatively conclude that Kwak’wala is a bottom-left-cell language: it disallows multiple focus, but allows verum to co-occur with focus.

In this section, we provided further evidence for the independence of verum and focus realization across languages. Investigating multiple focus marking, we showed that the languages in our sample vary on whether verum may co-occur with focus. We found evidence for all possible combinations. Among those languages that have multiple focus constructions, German, English and Gitksan allow focus and verum to combine, whereas Bura excludes this combination. South Marghi does not have multiple focus and disallows the co-occurrence of focus and verum. Our typology also predicts the possibility of languages which do not exhibit multiple focus but allow for the combination of constituent focus and verum. We were able to tentatively identify Kwak’wala as such a language. Table 3 summarizes the results; we see that Bura and Kwak’wala are the languages which provide evidence for the LOT over the FAT.

We conclude this section by stressing once again that the FAT fails to predict the existence of the languages in the dark gray cells in Table 3. There is no reason within the FAT analysis why a language with multiple focus should disallow the combination of focus and verum, or why a language without multiple focus should allow the combination of focus and verum.

Verum marking is optional in Bura in answers to yes-no questions. This is shown in (105), repeated from Section 4.2.1, ex. (50). The presence of the verum marker kú leads to an emphatic interpretation, expressed by the use of the adverbial really in the English interpretation.

A further example is given in (106). Again, the presence of the verum marker in the yes-no question and in the answer conveys an emphatic interpretation, which is compatible with the LOT but not with the FAT.

The verum marker is also optional in affirmative confirmations of a preceding utterance, cf. (107), repeated from example (46) above. The verum marker, which corresponds to the insertion of the stressed auxiliary in English, puts additional emphasis on B’s utterance.

In the next subsection we show that the optionality of verum marking is also observable in Gitksan, further supporting the LOT which predicts optionality in these contexts.

Just like in English, German and Bura, in Gitksan verum marking is optional in responses to yes-no questions, as predicted by the LOT but not the (weak) FAT. This is shown in (108). The consultant’s comments on the k’ap-version of (108B) indicate that the addition of k’ap conveys a verum interpretation, as we expect.

(109) similarly shows that k’ap is optional in answers to yes-no questions, and that its presence conveys additional meaning.

The consultants’ comments show that verum marking with k’ap is not neutral and adds additional emphasis that would not be expected if k’ap were just a way to realize ordinary alternative focus.