2.1 Previous approaches

What all contrastive contexts have in common is that they concern alternatives, their membership in a set, and often the exclusion of other alternatives, as has been pointed out by a number of authors (e.g. Chafe 1976; Rooth 1992: 79–82; Vallduví & Vilkuna 1998; Umbach 2004; Selkirk 2008).

Approaches to contrast mainly differ in the degree to which they are restricted to explicit sets of alternatives and exhaustive alternatives. Narrow approaches, e.g. those of Chafe (1976); É. Kiss (1998), only view contexts with an explicit and closed set of alternatives as contrastive. A typical contrastive selection context in that sense is shown in (2a) and (2c), an alternative question (Q-ALT) with its answer (SEL-ALT). From the explicit set of alternatives coffee_A1 and tea_A2 given in (2a), one alternative, coffee_A1, is selected in the answer in (2c).

An even narrower approach to contrast is adopted by Krifka (2007), who argues that the contrasted alternatives are part of the Common Ground content, which corresponds to the set of propositions which the discourse participants mutually agree on and commit to for the purposes of the discourse (going back to Karttunen 1974; Stalnaker 1974). Thus, following this approach to contrast, the answer in (2c) would not be contrastive, because even though the set of alternatives (set_{A}) is closed and explicit, it is part of a question and thus not (yet) part of the Common Ground; only corrections as in (1) would qualify as contrastive context. These narrow approaches of contrast share that a WH-question (Q-WH) such as the one given in (2b) is not considered to be contrastive as it does not encode explicit alternatives. Instead, it introduces a variable which is filled by the alternative given in its answer in (2c). We treat such answers as SELECTION contexts (SEL-WH), since they contain the selected alternative that fills the variable introduced by the WH-question.

A much broader approach to contrast is argued for in Vallduví & Vilkuna (1998). The authors define “kontrast” as generating a membership set (the members are determined by ontological and contextual restrictions) for the chosen alternative (Vallduví & Vilkuna 1998: 84). This makes WH-questions as in (2) and their answers (SEL-WH) fall under contrastive contexts; the WH-question makes available an implicit set_{A}, from which one explicit (set of) alternative(s) is then chosen in the answer.

Another dimension along which approaches to contrast differ is exhaustivity, i.e. the exclusion of alternatives other than the current one. Traditionally, it is assumed that contrast involves the chosen alternative at the exclusion of all (e.g. É. Kiss 1998) or at least one other alternative (e.g. Molnár 2006: 220) to hold true. This corresponds to a more restrictive view of exclusion in contrast. A less restrictive position is presented in e.g. Umbach (2004: 164), where exclusion is treated as an integral part of certain types of contrast but as irrelevant for other types. For instance, corrective contexts such as example (1) are exhaustive by entailment in that they necessarily reject and thus exclude another explicit alternative. Other types of contrastive contexts may implicate exhaustivity but do not necessarily entail it, since the exhaustivity-effect can be cancelled (Umbach 2004; Repp 2016). Example (3) illustrates this with a question-answer pair; while the WH-question in (3a) itself is non-exhaustive, the exhaustive implicature in the answer in (3b), namely that the speaker met only John_A1 can be cancelled without contradiction by adding the information that she also met Paul_A2.

Apart from such question-answer contexts (Q-WH & SEL-WH, Q-ALT & SEL-ALT), further types of contrastive discourse relations distinguished in Repp (2016) are SIMILARITIES (SIM), OPPOSITIONS (OPP), and CORRECTIONS (CORR).² Contexts of SIMILARITY feature alternatives that make a similar contribution to the current question under discussion (QUD) (cf. Repp 2016). QUDs are constructs that help to structure discourse, to determine the relations between sentences within a coherent discourse, and to relate sentence topics with more general discourse topics (cf. Roberts 1996; Benz & Jasinskaja 2017).³ As such, the current QUD corresponds to an implicit or explicit question that the current utterance is assumed to answer. An example of SIMILARITY as a contrast-relevant discourse relation is given in (4): the current QUD could refer to the kinds of food that invited guests prepared, for which both clauses in (4) provide possible answers to the current QUD. SIMILARITIES can (but do not have to be) exhaustive by implicature, as the hearer can assume that the speaker mentions all alternatives for which the proposition holds true (following from the Gricean maxim of quantity).

An OPPOSITION compares the current alternative to another explicit or implicit background alternative, with both alternatives making opposing contributions to the current QUD (Repp 2016). Such a context is shown in example (5), where the alternative of John mowing the lawn_A1 is given first, and then juxtaposed with the second alternative of Pete pruning the roses_A2 through the use of but.⁴

Repp (2016: 277) proposes the label of OPPOSITION for such discourses, because but in this case implies a violated expectation, namely that John and Pete would mow the lawn together. In OPPOSITIONS, both alternatives make different contributions to the current QUD and both can hold true. If, on the other hand, the two propositions are presented as in (6), no such expectation is violated, and we can speak of a SIMILARITY discourse relation, as both clauses make similar contributions to the QUD.

As OPPOSITIONS and SIMILARITIES are not always formally distinguishable but can, depending on the context, be interpreted as either, we will not make a categorical distinction between SIMILARITIES and OPPOSITIONS when applying them to Tima in Section 3.

In contrast to OPPOSITIONS, a CORRECTION explicitly denies another alternative A1 to hold true in addition to the current one. The current alternative A2 either rejects A1 because certain background assumptions for the felicitous use of A1 are not met, or because the propositions of making dinner associated with A2 and A1 cannot both be true in the evaluation world (Repp 2016):

Another approach building on a violation of discourse expectations is to relate contrast to the unexpectedness or special status of the chosen alternative (e.g. Halliday 1967; Frey 2006; 2010). Contrastivity in this sense does not rely on explicit sets of alternatives or on exhaustivity, but on additional pragmatic effects of the chosen alternative (cf. Repp 2016: 273–274 for an overview); it serves to draw the hearer’s attention to the special status of the selected alternative.

Unexpectedness and surprising information has also been argued to account for overt focus marking. For instance, Hartmann & Zimmermann (2007; 2008) show that morphological and syntactic focus marking in Hausa is not obligatory in focus contexts, i.e. that focus can be left formally unmarked. They argue that focus marking in that language is pragmatically rather than semantically motivated, as it is found in those contexts in which the speaker would assume the information in focus to contain a certain degree of unexpectedness or noteworthiness for the hearer. Zimmermann (2007) extends this kind of analysis to a number of Chadic languages, relating an unexpected current alternative to contrastive focus. He argues that “contrastivity in this sense means that a particular content or a particular speech act is unexpected for the hearer from the speaker’s perspective” (Zimmermann 2007: 148). In a similar vein, using the label of “mirativity”, Brunetti (2009), Cruschina (2012; 2016; 2019), and Cruschina & Remberger (2017) argue that the unexpectedness of the current alternative can also explain syntactic effects of focus marking in various Romance languages; Matić (2003: 287–296) finds a similar effect for fronted foci in Serbo-Croatian, Albanian, and Modern Greek.

Under the label of “drawing attention”, Skopeteas & Fanselow (2011: 1698) make a similar argument for focus fronting in German: “[…] this operation is not associated with a truth-conditionally relevant property, but it has the effect of attracting the hearer’s attention to that portion of the utterance that may not be in line with the hearer’s expectations”.⁵ Yet another similar observation is made in Abeillé et al. (2008) concerning preposed NPs in French. The authors argue that “the partition of the sentence serves to make the utterance more dramatic, highlighting that part of the situation that seems particularly worth of notice […]” and that the construction “is typically associated with speaker’s attitudes (surprise, admiration, disgust, justification etc.)” (Abeillé et al. 2008: 13–14).

Even though these studies use a number of different labels (attention drawing, noteworthiness, unexpectedness, surprisal, mirativity), we follow this line of work in arguing that the unexpectedness of the current alternative makes a context contrastive in that, according to the speaker, the alternative contrasts with the hearer’s belief or the Common Ground. CORRECTION contexts are inherently unexpected in this sense, since the alternative in the correction is by definition rejecting another alternative that the hearer was committed to. Also OPPOSITION contexts contain an unexpected alternative, implying that previous expectations are violated. Other contexts mentioned above, such as SIMILARITY and SELECTION, can but do not necessarily involve an unexpected alternative. WH-questions and alternative questions could be argued to aim at drawing attention towards the lack of knowledge of the speaker with the implicit request to fill this gap. While Hartmann & Zimmermann (2007: 390) argue that answers to WH-questions and alternative questions are never unexpected being the default discourse move following a question, we view the alternative in answers to WH-questions (SEL-WH contexts) as necessarily unexpected – by default, the speaker utters the WH-question so that their knowledge gap is filled by the addressee. On the other hand, we treat the answers to alternative questions (SEL-ALT contexts) as expected alternatives, since they are part of the preceding alternative question.

Table 1 shows the different contexts that are viewed as contrastive in different approaches to contrast. The narrow approaches (e.g. Chafe 1976; É. Kiss 1998) would only include the contexts with an explicit set_{A} and the selected alternative being exhaustive, with the most restrictive approach in Krifka (2007) only counting in CORRECTION contexts. A broader account of contrast, as proposed in Vallduví & Vilkuna (1998), also considers the remaining contexts to be contrastive, since all involve a set_{A}. Only CORRECTION contexts are exhaustive in that they exclude the proposition to hold for any other alternative. OPPOSITIONS and other contexts for which exhaustivity is marked in brackets in Table 1 are exhaustive by implication, i.e. they involve an exhaustivity effect restricting the proposition to hold true for the current alternative only, which can be cancelled. Table 1 also shows how the various contrastive contexts relate to unexpectedness: only CORRECTION and OPPOSITION contexts are inherently unexpected; for all other context types, unexpectedness is an orthogonal property (again indicated by brackets).

An important question concerning contrast emerges from the properties of contrast outlined above, namely whether contrast should be taken as a categorical or as a gradient concept. In this study, we follow (Molnár (2001; 2006); Paoli (2009); Calhoun (2010); Repp (2016) in arguing that contrast is gradient, meaning that an element is not necessarily either contrastive or non-contrastive, but can be more or less contrastive. For instance, Molnár (2001: 108) distinguishes between three degrees of exhaustivity which are relevant for different degrees of contrast: non-exhaustive (several alternatives can hold true), some-exclusion (a single other alternative cannot hold true), and all-exclusion (all alternatives other than the selected one cannot hold true). This gradient nature of contrast can also be seen in the discourse relations introduced above. SIMILARITIES are only weakly contrastive, as they include alternatives that make a similar contribution to the current QUD. OPPOSITIONS are more contrastive than SIMILARITIES, because the alternatives make opposite contributions to the current QUD. As argued in Umbach (2004: 171–173), the discourse relations of OPPOSITION and CORRECTION can be distinguished on the basis of different degrees of exhaustivity; CORRECTIONS can therefore be thought of being more contrastive than OPPOSITIONS.⁶

2.2 Tima: evidence for a broad and gradient approach to contrast

The narrow and broad approaches to contrast, paired with the question of contrast being of categorical or gradient nature, lead to the following four predictions as to how contrast can be marked in grammar:

As we will show in Section 3, Tima provides evidence for a broad and gradient approach to contrast, since all contexts listed in Table 1 (except for Q-ALT) are systematically marked by one of two markers (or both) along the scale from less to more contrastive contexts. Therefore, we follow Vallduví & Vilkuna (1998: 83–84) in defining contrast as follows:

Following from the properties of different contrastive contexts discussed in Section 2.1, gradience in contrast, i.e. stronger and weaker contrastive contexts, can be found along three dimensions: the explicitness of the set_{A}, exhaustivity, and the unexpectedness of the current alternative, as shown in (10).⁷

The Tima data provides evidence for contrast being gradient along the dimensions of exhaustivity and unexpectedness, while the explicitness of the set_{A} does not play a role. In Section 3, we discuss the two morphological markers of Tima, the selective and focus marker, and show in detail how their distribution aligns with weaker and stronger contexts of contrast. Their distribution in contrastive contexts is summarised below in Table 2.

Table 2 shows that there is no consistent relation between explicitness and the use of the focus and selective marker; both occur in contexts with and without explicit sets_{A}. For instance, when comparing answers to WH-questions (SEL-WH) with implicit sets_{A} and answers to alternative questions (SEL-ALT) with explicit sets_{A}, Table 2 shows that they can be both marked in the same way: the focus marker is the main signal of contrast in both cases, and the selective marker is compatible with such contexts in the sense that it can but does not have to be used in addition to the focus marker. Furthermore, we distinguish between explicit and implicit SELECTION contexts, marked as SEL-EXPL and SEL-IMPL in Table 2, respectively. In SEL-IMPL contexts, the selective marker is the only marker of (weaker) contrast. Without this marker, the context will not be interpreted as contrastive (in the sense of opening a set of relevant alternatives) anymore. The brackets in Table 2 indicate the optional use of the selective marker in SEL-EXPL and SIM/OPP contexts.

Contexts that are exhaustive (by entailment or implicature) require focus marking. In these cases, the selective marker is compatible although it is not obligatory. Non-exhaustive contexts such as WH-questions are incompatible with the focus marker but require the use of the selective marker. The dimension of exhaustivity thus establishes the focus marker as a signal of stronger contrastive contexts, and the selective marker as a signal of weaker contrastive contexts. The evidence for unexpectedness as another relevant dimension comes from the fact that the focus marker can also be used in non-exhaustive contexts. As we show in Section 3.2, the focus-marked and non-exhaustive alternative is always unexpected and noteworthy.

For focus marking in Tima, we adopt a definition of focus as a pragmatic notion, following Chafe (1976) and Vallduví & Vilkuna (1998) in their definition of focus as the rhematic part of an utterance: it conveys the information that is not yet asserted or part of the Common Ground, and acts as Common Ground management. This corresponds to what Krifka (2007: 21) notes about the pragmatic use of focus, namely that it “does not have an immediate influence on truth conditions, but it helps in guiding the direction into which communication should develop […]”. In the remainder of this paper, we will only discuss the conditions of the use of the Tima focus marker, not focus contexts in general.

The second morphological marker, expressing weaker contrast in Tima, is the selective marker, which signals the availability of a set of alternatives and, in contexts other than WH-questions, marks the current alternative as selected against the background of the set_{A}. It may be tempting to analyse the selective marker as another focus marker in the language, given that it opens a set of alternatives to the current one. We do not treat the selective marker as a focus marker for the following reasons: although being compatible with contexts typically associated with focus marking (SEL-WH, CORRECTION), it is not systematically used in such contexts. It does not convey unexpectedness or noteworthiness of the current alternative, and it is not restricted to discourse-new, not yet asserted content. As we will show in Section 3.3, it also occurs in the context of contrastive topics.

3.1 Preliminary remarks

Tima, spoken by roughly 7,000 speakers mainly in the Nuba Mountains in Sudan, is one of three languages which constitute the Katloid family within the Niger-Congo phylum (see e.g. Dimmendaal 2018). Tima is a tonal language with both lexical and grammatical tones. Since a systematic analysis of tone in Tima is still pending, tone marking in the examples reflect their surface realisation. As is the case with other languages in the Nuba Mountains (see Rose 2018), Tima has ATR (advanced tongue root) harmony (see Bashir 2013: 272).

The basic word order in Tima is AVO, case marking being absent with this order. By default, the sentence-initial position is reserved for the sentence topic, i.e. the A argument in AVO constructions. The alternative word order OVA can occur when the object occupies the sentence-initial position; in this case, the object is the sentence topic and the postverbal subject A requires ergative marking. Additional word orders are possible in focus constructions.

With regard to information structure in Tima, we find – in addition to the generally morphologically unmarked topic – marked constituent focus as well as selective marking.⁸ The two markers differ with regard to their effect on word order. The focus-marked constituent in a focus construction must occur in a preverbal position, i.e. either sentence-initially or in second position. We thus find AO_FOCV, OA_ERG.FOCV, A_FOCVO, or O_FOCVA_ERG word orders, with the latter two constructions maintaining the word order of the non-focus constructions AVO and OVA_ERG. Thus, depending on the type of constituent that occurs with focus marking and which constituent is the sentence topic, a change of word order is required. The selective marker, on the other hand, does not involve any word order changes and the selective-marked constituent remains in its default position.

The data used in this study is primary data obtained during numerous fieldwork trips between 2007 and 2019. Most of the material was gathered from storytelling and through prompted elicitation.⁹ Certain elicited material relevant for this study was inspired by Malchukov (2004); Umbach (2004); Davies (2012); Repp (2016).

3.2 Focus marking in Tima

Constituent focus in Tima is expressed by the exponents shown in Table 3. We can distinguish three types of exponents: a singular exponent, a plural exponent, and a third exponent that is used with proper names, obliques, ergative subjects, and singular 1st and 2nd person pronouns. The ATR-feature of the vowel in all exponents depends on the ATR-feature of the preceding root vowels. In addition, the markers =e/=ɛ and =ʌ/=a require an initial glide if their host ends in a vowel.¹⁰

Diachronically, the focus marker in Tima is related to the predication marker used in non-verbal predicative contexts, shown in example (11). This analysis is based on the synchronic observation that the same marker is obligatorily used for non-verbal predication in equative constructions – including classifying, identifying, and specifying constructions (for more details see Schneider-Blum 2018).¹¹ This is supported by the fact that the development from copulas to focus markers is cross-linguistically very common (cf. Heine & Kuteva 2002: 109-110; Green 2007: Ch. 5 for Hausa; Creissels 2020).

Synchronically, the marker is a priori functionally ambiguous in such equative constructions.¹³ For instance, the statement kʌ̀pʌ́nꜜdílí ‘She is a teacher.’ does not only fit into the context as described in (11) but would also be a grammatically as well as pragmatically correct answer to the alternative question Is your friend a nurse or a teacher? or to the WH-question Who is this?, which are typical contexts for focus. Note also that non-verbal sentences with either predicate or focus function of the marking option follow the same tonal and prosodic patterns. Therefore, a formal difference between focus marking and predicative marking cannot be established in such constructions, and the marker’s function can only be determined by the context.

Example (12) illustrates that an expression containing focus in a SEL-WH context is formally equivalent to the construction that we find in a CORRECTION context, as in (12d).

Whereas focus/predicate marking is obligatory in equational sentences without a verbal predicate, this is not the case in verbal sentences. Compare example (13) without focus marking and (14) with focus marking. Focus marking is only acceptable, and in fact mandatory, under the conditions described below in detail, e.g. as an answer to a WH-question (‘What is the woman raking?’) in (14). The more general prompt to describe the picture in Figure 1 did not result in focus marking in (13).

Figure 1

Stimulus of example (13).

As demonstrated in (14), the focus-marked constituent needs to occur in a preverbal position.¹⁴ The marker attaches to the rightmost element of the focused noun phrase (cf. (46) and (47) in Section 3.4). Where the A argument is focus-marked, no changes in word order are needed as A is already preverbal by default (cf. example (22) below). However, if the non-focused A argument (kʌ̀húnénnʌ́ ‘this woman’) remains in the sentence-initial position, as in (14), the focus-marked object (yʌ́ꜜkɨ́rɛ̀ ‘remainders’) must immediately precede the verb. Furthermore, if a focus-marked object occurs in the sentence-initial position, the A argument follows the verb and receives ergative marking (cf. final clause in (21b)).

Figure 2

Stimulus of example (14).

Example (14) strongly suggests that the constructions discussed here are not cleft sentences. Cleft constructions typically dislocate and separate the clefted material from the rest of the sentence so that it appears in the sentence-initial (most commonly) or sentence-final position (see also Hartmann & Veenstra 2013: 5; Güldemann et al. 2015: 170; Creissels 2020: Section 2). This is clearly not the case in Tima; in (14) and similar examples, the focused object is neither in an edge-position, nor does a prosodic break cleave the focused constituent from the rest of the sentence.¹⁶

Concerning the distribution of the focus marker, its use is obligatory in the following contexts that are generally considered as typical focus contexts (see e.g. Gundel & Fretheim 2006; Krifka 2007; van Putten 2014): SEL-WH, SEL-ALT, and CORRECTIONS.¹⁷ In addition, its occurrence may signal unexpectedness. We will now illustrate the use of the focus marker in each of the aforementioned contexts, beginning with possible answers to the WH-question. This prototypical use of the focus marker is what we see in (15b).

If alternatives are coordinated in the answer, only the first is focus-marked, independent of whether the alternatives are expressed as coordinated noun phrases with a single predicate as in (16b) or whether they are expressed in two separate clauses as in (17b). This also holds for coordinations with no overt coordinator in non-verbal predications (cf. example (20b) below).

Focus marking in Tima encodes exhaustivity by implicature rather than by entailment. This becomes especially clear in examples such as (17b). The exhaustivity effect results from the first part of the answer, kàpɪ́ŋà ùkúmùn mɨ́nʌ́ ‘she saw Kaping’, but can be cancelled by the additional segment ùkúmúnyáŋ ìt̪ʌ̂ŋ ‘she also saw Ithang’.

As opposed to (17b), where exhaustivity is merely implied, exhaustivity is entailed in (18b) with the addition of the contrast-sensitive lexeme tɪ̀ɪ́n ‘only’. Thus, the exhaustivity has to be denied and corrected if the set_{A} for which the relevant proposition holds is expanded, as seen in (18c).

As was mentioned in Section 2, the explicitness of the set_{A} in a question is irrelevant for focus marking in answers in Tima. In other words, focus marking can occur with both explicitly given and implicit sets_{A}. The same holds for the selective marker, as we will show in Section 3.3.

An example showing an explicit set_{A} can be seen in the alternative question in (19a). In Tima, this kind of question allows for two types of quite natural answers: a shortened one with the constituent only, as in (19b), and a longer one containing the constituent in question as well as the verb (while not mentioning the object), as in (19c).¹⁸ In both cases, focus marking is necessary on the constituent that selects the alternative for which the predication holds true (kʌ́ʌ́dɨ̀h ‘monkey’) from the set_{A} given in the question (monkey vs. cat). The focus-marked alternatives in (19b) and (19c) are not unexpected. As we will argue later in this section, focus marking in Tima usually occurs with unexpected alternatives. SEL-ALT contexts seem to be the only type of contexts in which it is used to mark an expected alternative. That the focus marker is nevertheless required in such contexts is due to the exhaustivity (effect) of the selected alternative.

A WH-question such as in (20a) leaves the alternatives implicit, so that the set_{A} is only constrained by contextual plausibility. A possible answer to that question is shown in (20b). We see that the first selected alternative ìnt̪ùkúbúk ‘groundnuts’ requires focus marking, while the second one, ìt̪ùk ‘bread’, cannot be focus-marked (cf. example (16b)).

A typical example of a CORRECTION is given in (21), where the presupposition Kokuung ate soup is rejected and corrected by the speaker:

The last context in which the focus marker is used involves the unexpectedness, surprise, and noteworthiness of the chosen alternative (cf. Zimmermann 2007; Cruschina 2019). We take this function to be a (diachronic) extension of focus marking in CORRECTION contexts, as the latter includes an unexpected alternative by default.

Thus, a focus construction like (22) is felicitous if the alternative is intended to be marked as unexpected for the hearer or, as was confirmed by our consultant, when answering the question Who/What was in the street yesterday? with the less natural complete sentence. The more natural answer would consist of the focused noun phrase alone, i.e. ɪ̀hwááꜜná ɪ̀háꜜwʊ́kɪ́yɛ́ ‘many people’. If the speaker does not intend to convey unexpectedness but wants to transmit neutral information, the construction as shown in (22) is infelicitous. Instead, the sentence in (23) would be used, employing the copula ɲ̀cɛ̀:

Note that in non-verbal sentences the copulas ɲ̀kɔ̀ (SG) and ɲ̀cɛ̀ (PL) replace the focus marker in contexts where focus marking is not appropriate. As is shown in (23), the focus marker cannot be used in addition to ɲ̀cɛ̀ (see also example (34)) and vice versa, i.e. the copulas ɲ̀kɔ̀/ɲ̀cɛ̀ and focus markers are in complementary distribution in such contexts.

We further probed the unexpectedness condition using tasks designed by Skopeteas et al. (2006). The consultant was shown different series of three pictures, each of which reflects a coherent short story. When being shown the series, e.g. the one in Figure 3, and asked What happened? (24a), the speaker always uttered replies that featured an initial focus-marked constituent, as in (24c) (Condition A3, Task 15 in Skopeteas et al. 2006). In example (24c), it is also unexpectedness, rather than exhaustivity, that accounts for the presence of the focus marker. In the context of all new information or thetic statements (Sasse 1987) or sentence focus (Lambrecht 2000), such kind of focus marking in African languages is not exceptional, as outlined by van Putten (2014: 121). She further explains: “The rationale behind this is that in most sentences, the main information update is encoded in the predicate, whereas the subject tends to encode uncontroversial information to which the new information is linked […]. Marking focus on the subject is important, because it makes clear to the listener that the sentence does not have this expected information structure.” Similarly, Skopeteas & Fanselow (2011: 1693) suggest that “[t]he use of a marked construction violates an expectation of the human parser that can be used to control the listener’s attention and to signal that a non-default interpretation of the utterance is attended”. Furthermore they “expect to find evidence that the interpretation of excluding alternatives is not a necessary concomitant of constructional focus” (Skopeteas & Fanselow 2011: 1699). This is exactly what we find in example (24c) as the answer to (24a) – the ‘fire’ is not related to alternatives but discloses what is non-expected, non-predictable, and thus surprising information.

Figure 3

Stimulus of example (24).

Also with the more precise question (24b) What happened to the house? (Condition B3, Task 15, Skopeteas et al. 2006), we received an answer almost identical to the one in (24c). However, with regard to other series, answers to the question What happened to X? did not always occur with focus marking. For instance, with the pictures shown in Figure 4 and the accompanying question What happened to the cars? (Condition B2, Task 15, Skopeteas et al. 2006), the speaker began his description with ɪ̀ʊ̀rʊ̀mbɪ́l ‘cars’ (25b). As ‘cars’ had already been established as the discourse topic in (25) and hence as an expected referent, focus marking did not occur in (25b). This difference in the structures of (24c) vs. (25b) suggests that the focus marking on cɪ̀ŋɪ́ ‘fire’ in (24c) signals that the ‘fire’ is an extraordinary referent to be mentioned (whilst occupying the sentence-initial position) when the question concerns the house.

Figure 4

Stimulus of example (25b).

As was argued for examples (22) and (24c), the focus marker does not require an exhaustive context when signalling unexpected and noteworthy information in order to draw the hearer’s attention to the current alternative. This is also the case in example (26). It is the first sentence of a description of a picture series after the request to ‘tell me the story the pictures show’ (Task 15, Condition C4, Skopeteas et al. 2006).

Sentences such as those in (26) resemble to some extent an example from the Eastern Cushitic language Boni, presented as ex. (79) in Sasse (1987: 542) in his elaboration on thetic (vs. categorical) statements. In Boni, as well as in Tima, it is the “characterising/identifying copula” – identical to the focus marker in Tima – which is employed in these constructions, starting with an entity-central statement (for the difference between entity-central and event-central theticity see Sasse 1987: 526–527).²⁰ When answering the question ‘What happened?’, constructions with focus marking are the structure of choice for Tima, and, as in (26), can also occur at the beginning of a story. As Sasse (1987: 566) contends, “[i]n languages having grammatical subjects it [i.e. a noun denoting an entity] is necessarily also a ‘would-be subject’, so that it must be marked in such a way as to preclude its interpretation as a predication base” and thus as to differentiate a thetic from a categorical statement. Lambrecht (2000: 625), dwelling on the matter of “detopicalization”, offers as an answer to the question of how a subject argument can “be coded in such a way that it will not be interpreted as a topic” that one could provide “the subject constituent with grammatical properties which are conventionally associated with focus marking”.²¹ This is exactly what we observe in Tima.

A comparable distribution and function of focus marking to that found in Tima is also seen in Tuu languages (Khoisan). Güldemann (2010: 87) suggests that these “split structures […] are polyfunctional in that they serve to encode contrastive term focus as well as entity-central theticity” (whilst also referring to Sasse 1987). The split manifests itself in the separation of the term/central entity “by a kind of pivotal marker” (Güldemann 2010: 72) from the rest of the sentence and “can achieve at least two things. First, it disrupts the canonical syntactic subject-predicate relation which is associated with the topic-comment configuration of categorical statements. Second, it cancels the assertivity of the predicate as the central expression of the state-of-affairs” (Güldemann 2010: 88). Both statements can apply equally to Tima. As examples (22), (24c), and (26) illustrate, focus marking in thetic sentences functions to indicate “surprising or unexpected events” and are found with “background descriptions (local, temporal, etc. setting)” (Sasse 1987: 566; cf. also Güldemann 2010: 87 for his findings in the Tuu languages). Note that in Tima, the establishment of new participants in discourse can also be achieved by using the existential, presentative copula ŋ̀kwíyʌ̀, as seen in (27).

While ŋ̀kwíyʌ̀ is used in verbless thetic sentences which “describe the existence or appearance of their subjects” (Sasse 1987: 525), the identificational copula (= focus marker) occurs in entity-central verbal thetic sentences in Tima.

A cleft or cleft-like analysis like the one posed by Güldemann (2010) for the split constructions in Tuu languages is a possible interpretation of the Tima data, given that the focus marker previously originated from a predication marker. However, Tima has split constructions which are different in that they exhibit a prosodic break and, in addition, a rising-tone contour on the split-off constituent. Furthermore, a focus-marked constituent must occur preverbally but not necessarily in the sentence-initial position (see (14)). As such, although we acknowledge the relationship of Tima focus constructions to clefts on morphological grounds, we think their prosodic properties and the position of focus-marked elements other than S/A speak against such an analysis.

In summary, this section showed that the focus-marked constituent must occur preverbally, either in sentence-initial position or immediately preceding the verb. Focus marking in Tima is generally exhaustive by implicature (except for in CORRECTION contexts); however, if focus marking occurs in the context of unexpectedness, exhaustivity seems to no longer be a necessary criterion for its use.

3.3 Selective marking in Tima

The other contrast-expressing marker in Tima is the selective marker =(G)i/=(G)ɪ. The exponents of the selective marker are morpho-phonologically conditioned, following the same regularities as described for focus marking with regard to the choice of the glide and the ATR-feature of the vowel. The selective marker generally signals the existence of a set_{A} out of which the current alternative is selected. It marks referents that are expressed as nouns with modifiers and cliticises to the modifier, which can be an adjective, another noun, or a clausal modifier such as relative clauses. Elliptically used modifiers may also feature the selective marker. Unlike the focus marker, the selective marker does not affect word order and it is not restricted to noun phrases in a preverbal position (see e.g. (28)).²² The selective marker also occurs in subject and object WH-questions, where it shows a slightly different syntactic behaviour. In the following paragraphs, we will discuss the distribution of the selective marker, focusing on its function to signal and open a set_{A}, while keeping in mind different types of contrast-relevant contexts.

Example (28), taken from a description of slaughtering an animal, shows the typical use of the selective marker, which appears on the modifier kwààɽɘ̀k ‘God’ following the head noun kùrtú ‘house’.²³ It indicates the existence of alternative houses apart from the ‘house of God’. These alternatives are not specifically mentioned in this text. The selective marker, and thus the opening of a set_{A}, is optional, as numerous other examples corresponding to the structure ‘house of X’ show. Here, optionality refers to the pragmatic intention of the speaker of whether or not to make alternatives available. If the selective marker is absent, the noun phrase is descriptive by nature without pointing out that there may be other types of houses to be considered, i.e. without the selective marker, the notion of contrast is absent.

As we will argue in this section, even though one could think of the selective marker as a kind of focus marker, a contrastive topic marker, a backgrounding marker, or a clausal determiner, it does not fit well into either of these categories. Therefore, we treat it as a different type of marker and refer to it as “selective” marker, following Schneider-Blum & Hellwig (2018) and Casaretto et al. (2020).

The prototypical syntactic context in which the selective marker occurs is on nominal modifiers, e.g. adjectives (29), adnominal possessive markers (31), or nouns that modify a head noun, e.g. in (28). In addition, the marker occurs to mark WH-questions, shown in e.g. (41). In its function to open a set_{A} and to draw attention to the selected alternative, we can distinguish the following types of contrast-sensitive contexts in which the selective marker occurs: SIMILARITIES/OPPOSITIONS, SELECTIONS, WH-questions, each of which is discussed in turn.

Example (29) shows a SIMILARITY/OPPOSITION context with explicit alternatives, as ‘this female child’ is opposed to ‘this male child’ in a parallel construction (also called contrastive topic construction, cf. Büring 2003). Both modifiers are marked by the selective marker. Note that the other pair of alternatives that are opposed in example (29), ‘dog’ and ‘cat’, cannot receive selective marking, since they consist of simple nouns without any modifier, even though, semantically, they would qualify as two opposed alternatives and hyponyms to the category of animals as the higher taxonomic level. Similarly, selective marking would be infelicitous on both occurrences of cíbʌ̀nʌ̀ ‘the child’ if unmodified, i.e. *cíbʌ̀nʌ̀=yí. Also, focus marking for any of the participants in this SIMILARITY/OPPOSITION context is not allowed. Thus, neither ‘dog’ nor ‘cat’, nor the ‘female child’ or the ‘male child’ in (29) can receive focus marking.

Figure 5

Stimulus of example (29).

Example (29) displays contrast enforced by the parallel structure used to juxtapose the two events. That is, in such pairing events with explicit alternatives, the contrast is greater than in events signalling implicit alternatives. However, the difference in degree of contrast cannot be ascribed to the selective marker itself, and parallel structures and selective marking have to be kept apart. Thus, each of the two clauses in (29), if used alone, is a grammatically and pragmatically complete sentence and would then indicate implicit alternatives.

Example (30) shows the use of the selective marker in the presence of nominal modifiers in another SIMILARITY/OPPOSITION context. We also see that the selective marker is not sensitive to exhaustivity: the utterance is an enumeration of feasts (organised by different Tima clans) which all make a similar contribution to the current QUD, and the list could be extended further. The superordinate set_{A}, i.e. the events celebrated by the Tima people, is left implicit.

The selective marker is often, but not necessarily always, used in the presence of adnominal possessives as in (31). Usually, possessives rarely occur with explicitly mentioned alternatives, (31) being an exception to the rule. It shows the use of the selective marker in a SIMILARITY/OPPOSITION context.

That a syntactically parallel structure is not required for the use of the selective marker is illustrated in (32). This sentence is part of a description about the use of a number of different local plants which can be taken as the QUD of the discourse. It concludes the discourse segment whose topic is the sausage tree, before starting to explain what the baobab tree is used for. The explicitly mentioned kwʌ̀yʌ́wù ‘sausage tree’ in (32) occurs together with the selective marker =wí in order to associate the sausage tree with the other plants that were the topics of different discourse segments within that conversation. It marks a SEL-IMPL relation on the discourse level, selecting the alternative ‘sausage tree’ from the implicit set_{A} of plants that were mentioned during the conversation. Selective marking is hence not restricted to utterances but also used to relate larger discourse units.

In Tima, adjectives and quantifiers can generally be used without a head noun. In such cases, the selective marker occurs on the nominalised modifier itself, e.g. kʊ́kɔ́lɔ́ŋ ‘the big (one)’ in (33), which shows a SELECTION context with explicit alternatives.²⁴

Example (34) shows another explicit SELECTION context with the preceding question containing an explicit set_{A} (‘these people’), of which the chosen alternative forms a part. One could also argue that (34) is a CORRECTION context. We do not treat it as such, however, because it does not entirely reject the proposition to hold true for the first alternative mentioned (‘these people’) but restricts it to a subset of the alternative (‘the male children’).

Figure 6

Stimulus of example (34).

If the same picture stimulus of (34) following the same question was described by using the Tima lexeme for ‘men’ instead of ‘male children’, neither selective marking nor focus marking would be felicitous.

Focus marking, as was tested, is not allowed in (34) (cf. also example (25b)), but if the (implied) question was instead Who is wearing hats?, as in (35), the focus marker would have been required and the copula ɲ̀cɛ̀ would have been inappropriate.

The examples provided so far all showed the selective marker attaching to noun phrases with phrasal modifiers. However, it can also occur on clausal modifiers, namely on relative clauses (36), temporal when-clauses (37), and reason-clauses (38). In such cases, the selective marker occurs at the right edge of the entire subordinate clause (indicated by square brackets). We classify such contexts as implicit SELECTIONS, since the function of the selective marker in such contexts still is to draw attention to the selected alternative (the referent of the modified head noun) out of other possible alternatives.

The use of the selective marker with relative clauses (36) can be explained as an extension of its use with nouns that have phrasal modifiers to nouns that have clausal modifiers. Its occurrence with the temporal clause in (37) can be explained by the subordinator àyɪnà ‘when’ having a nominal base, formed by an oblique head and the demonstrative clitic, i.e. a=y=ɪ=na ‘SRC=EP=PL=DEM.PROX’. Similarly, ŋ̀kɔ́ɔ́nɔ́ ‘because/since/as’, likely originated from ŋ̀=kɔ́ɔ́=nɔ́ ‘INS=walking=DEM.PROX’, accounting for the presence of the selective marker in (38). Thus, the subordinate clauses that occur with the selective marker structurally resemble complex noun phrases in which the nominal head is modified by a demonstrative clitic and a clausal modifier to which the selective marker cliticises.

The use of the selective marker, especially in clause-final position in relative clauses and the other two types of adverbial clauses, is reminiscent of the backgrounding marker in Chadic, e.g. in Ngamo (Grubic & Renans 2016; Grubic et al. 2019), Barayin (Lovestrand 2018), Buwal (Viljoen 2015), Gemzek (Scherrer 2012), Muyan (Smith 2003), and of clausal determiners in Kwa and Gbe languages, e.g. in Akan (Saah 2010; Amfo 2010; Bombi et al. 2019), Gã (Korsah 2017: 151–165), Ewe (Ameka 1991: 265–298), Fongbe (Lefebvre & Brousseau 2002: 481–502), Gungbe (Aboh 2005), Logba (Dorvlo 2008: 185).²⁵ However, we argue that there are substantial differences between those markers and the selective marker in Tima, which is why we do not classify it as a clausal determiner.

First, the clausal determiners in the languages cited also function as (some sort of) definite or anaphoric determiner of noun phrases. This is not the case in Tima, where the selective marker does not formally mark definiteness (nor specificity, givenness, nor any other referential value). Even though the referents of the noun phrases marked by the selective marker are mostly also definite or given, this is not necessarily the case. With regard to givenness, this can be seen in example (28) above, where the selective marker occurs on a salient but newly-introduced referent. Example (39) shows that definiteness is not a criterion for the occurrence of the selective marker either:

Second, it also is difficult in Tima to establish a relation between the selective marker and the demonstratives (=na, =yaa, =weeŋ) as a potential diachronic source of the former. The only hint that could indeed point towards a deictic origin of the selective marker are the proximal demonstratives in the related languages Katla and Julut. For proximal demonstratives, Katla uses =gi/=gı (SG), =ii/=ıı (PL) (Birgit Hellwig, p.c.), and Julut has =nggi (SG) and =ii (PL) (Nüsslein 2019: 76). However, without better diachronic evidence, the reconstruction of a deictic source of the selective marker remains rather speculative.

Third, the use of the selective marker in subordinate clauses also differs from the use of clausal determiners in two ways. As previously discussed, the selective marker in Tima only occurs with the three types of subordinate clauses shown in (36) to (38), which consist of overt nominals (even if semantically bleached) with clausal modifiers. With clauses beginning with other subordinators as in (40), it is not attested. The selective marker is therefore still a nominal determiner rather than a clausal one.

Also, clausal determiners such as the one in Ewe, for instance, often only occur in subordinate clauses when the latter precede the main clause, since they convey background information, set the frame for the information in the main clause, and mark the information in the subordinate clause as given. Ameka (1991) illustrates this backgrounding function for the clausal determiner in Ewe in great detail. Similar patterns with a certain degree of variation are attested with the backgrounding markers as well as with the clausal determiners. In Tima, however, the position of the subordinate clause does not affect the use of the selective marker, which can occur in a subordinate clause preceding the main clause, as in (37), or following it, as in (38).

Fourth, both backgrounding markers and clausal determiners appear within the topical, given background part of a bipartite cleft structure in the clause. The selective marker can occur in such positions as well; however, as we show in Section 3.4, it is not restricted to the background part of a sentence and can co-occur with the focus marker on the same constituent.

Thus, even though the selective marker does share a number of distributional and functional properties as well as potentially its deictic origin with clausal determiners and backgrounding markers, it differs from them in the ways outlined above, which is why we analyse it as a related, yet different type of marker.

One more type of context in which the selective marker appears still remains to be discussed: its systematic occurrence in WH-questions.²⁶ In (41) and (42), it marks a subject and an object question with an implicit set_{A}, whereas (43) shows that it is also used with questions that have an explicit set_{A}.²⁷ The explicitness is contextually conditioned, as the teacher is asking two students which one of them lied. While the selective marker is obligatory in questions such as (41)–(43), the focus marker is not acceptable in such constructions – neither on the WH-word, nor on any other constituent.

Note that the selective marker does not occur on the WH-word directly but systematically in the clause-final position instead. This differs from its use with modified nouns and suggests a different development, in which the subject and object WH-constructions containing a clause-final selective marker could go back to an earlier bi-clausal relative structure. While there is no strong evidence from the Tima data itself to support this analysis, we find a number of unrelated or extremely distantly related languages with a similar ex-situ construction for subject and object questions. Such constructions usually originate from earlier clefts with relative clauses such as What is it that X did?, and languages may have retained relative-clause markers even though the WH-question may synchronically no longer be bi-clausal. For instance, in Chadic (cf. Hartmann & Zimmermann 2008 for Hausa), Bantu (cf. Sabel & Zeller 2006 for Zulu and Nguni, Mchombo 2004 for Chichewa, Wasike 2006 for Lubukusu), Coptic (Reintges 2008), or Algonquian (cf. Déchaine 1997 for Plains Cree), relative markers occur on the verb in WH-questions with (ex-situ) WH-words. Rose et al. (2014) have also shown for Moro, a Kordofanian language spoken in the Nuba Mountains of Sudan, that ex-situ subject and object WH-questions show formal similarities with relative clauses. In Tima, such relative marking corresponds to the clause-final selective marker. Thus, the hypothesis that WH-questions in which the selective marker occurs go back to bi-clausal structures involving relativisation can in principle account for the occurrence of the selective marker in the clause-final position. However, since there is no strong evidence from Tima that would point towards such an earlier bi-clausal stage of WH-questions, this hypothesis has to remain rather speculative.

To conclude, we showed in this section that the selective marker occurs in different contexts of weaker contrast in which alternatives are signalled, namely in SIMILARITIES/OPPOSITIONS, SELECTIONS, WH-questions. All contrast-relevant contexts of the selective marker are non-exhaustive, but, as we will show in Section 3.4, the selective marker is nevertheless compatible with exhaustive contexts.

3.4 The interaction of the focus and the selective marker

Having discussed the distribution and function of the focus marker in Section 3.2 and the selective marker in Section 3.3, we now turn to a brief discussion of constructions in which both markers co-occur. Example (44) is a variant of the utterance presented in (33). While the structure in (33) was uttered by a consultant without any specific preceding question, the one presented in (44) below is the answer to the question Which giraffe died? or Did the old or the young giraffe die?. It features both the selective marker =ɪ́ and the focus marker =lɪ́.²⁸ Thus, the selective marker may occur in answers to WH-questions and alternative questions (under the conditions described in Section 3.3), but only together with the focus marker. The latter is required in the answer to those questions (see Section 3.2).

The use of selective marking in (44) indicates that kʊ́kɔ́lɔ́ŋ ‘the old one’ is a subset of the aforementioned explicit set_{A}, whilst focus marking is triggered by the SELECTION of one of the alternatives given in the preceding question.

In example (45), focus marking has a corrective function. The context has to be imagined as follows. A discourse participant assumed that the younger of two giraffes had died. This was not true and his interlocutor, i.e. the speaker of (45), expressed a CORRECTION which requires the use of the focus marker. In addition, the selective marker is present on both alternatives to signal the relationship between the current alternatives (the old vs. the young giraffe).

Example (46) shows another CORRECTION context. The speaker rejects the proposition Your sister gave birth. and corrects it with My aunt gave birth. The two contrasted constituents wɛ̀ɛ́n lɛ́ɛ́nɪ́ ꜜkɘ́tɛ́ʔɛ́n ‘my (younger) aunt’ and lùwì lɛ̀ɛ̀nɪ́ ‘my sister’ are focus-marked in both (46a) and (46b), similarly to the contrasted elements in example (45). The only difference between (46a) and (46b) is in the use of the selective marker on the nominal modifier ‘small’. Generally speaking, the selective marker can be attached to any one of the modifiers, some of them, or all of them, in case the noun has more than one modifier. In example (46a), the selective marker =ɪ́ attaches to lɛ́ɛ́n ‘my’ and thus precedes the second modifier kɘtɛʔɛn ‘small’. In (46b), on the other hand, the selective marker =ɪ́ occurs on both modifiers, i.e. on lɛ́ɛ́n ‘my’ and kɘtɛʔɛn ‘small’.

This difference in usage between examples (46a) and (46b) entails a semantic difference. In (46a), where the selective marker only occurs with the possessive pronoun lɛ́ɛ́n ‘my’, the speaker indicates that there are mothers other than his own. This is due to the fact that the term wɛ̀ɛ́n in Tima does not only refer to a person’s biological mother, but also to all the sisters and female cousins of the mother and the father, in addition to the mother’s co-wives. Moreover, it may be used to refer to any respected woman who is older than the speaker. Thus, to narrow down potential referents, kinship terms like wɛ̀ɛ́n are usually additionally specified. This is achieved first and foremost by possessives, but also by reference to the age of the person, here kɘ́tɛ́ʔɛ́n ‘small’. The phrase wɛ̀ɛ́n lɛ́ɛ́nɪ́ ꜜkɘ́tɛ́ʔɛ́n, literally, ‘my small mother’, excludes any non-relatives and refers to an ‘aunt younger than the mother’. In (46a), kɘ́tɛ́ʔɛ́n ‘small’ does not receive selective marking. Thus, the ‘aunt’ is simply described as being younger than the mother. In example (46b), on the other hand, the selective marker is also used on kɘ́tɛ́ʔɛ́n ‘small’, which signals existing alternatives on the level of that modifier, i.e. one or more aunts who are older than the biological mother.

Another example with both selective and focus marking on the same phrasal constituent was presented in Section 3.2, repeated here for convenience as example (47). The two markers are attached to the last element, i.e. ɪ̀háꜜwʊ́k ‘many’ of the phrase ɪ̀hwááꜜná ɪ̀háꜜwʊ́k ‘lots of people’. This example shows their co-occurrence in a context in which the focus marker is used to signal unexpectedness rather than exhaustivity, while the selective marker indicates that there were other people who were not in the street. In the two phrases ínʌ̀ ɪ̀háꜜwʊ́k=ɪ́ ‘a lot/most of them’ and ɪ̀bɛ̀ʔɛ́ŋ=ɪ́ ‘a few of them’, selective marking refers to subgroups of the people in the street. In the former phrase, the quantifier modifies a demonstrative pronoun, in the latter, the selective marker appears again on a nominalised modifier, i.e. ɪ̀bɛ̀ʔɛ́ŋ ‘some, a few’.

To sum up: both the selective and the focus marker share the function of indicating alternatives. However, the co-occurrence of the two markers in a sentence, especially when they occur on the same constituent, is evidence for those two markers having distinct, though overlapping domains.²⁹ Focus marking, at its core, singles out the current alternative, either exhaustively (by implicature), or as an unexpected, surprising, and noteworthy alternative. The selective marker signals alternatives that are available on the basis of nominal modifiers, which semantically specify the referent and relate the chosen alternative to others which are similar, yet distinct. The examples from this section are also evidence against an analysis of the selective marker as a topic or backgrounding marker, given that its use with focus-marked constituents is felicitous.