Jiu-conditionals in Mandarin Chinese: thoughts on a uniform pragmatic analysis of Mandarin conditional constructions

1 Introduction

While conditionals are a vital area in linguistics, they are not straightforward to define. For instance, Bhatt and Pancheva (2006: 641) take conditionals “as structures involving an adverbial clause interpreted as stating the conditions under which the proposition expressed by the main clause is true (or its truth is relevant, as in the case of relevance conditionals)”. This covers conditionals formed by the use of a conditional connective (CC) in the antecedent – for example, English if and German wenn/falls – or related prepositional phrase – for example, English on condition that, German unter der Bedingung, dass. Although CCs are a common strategy for forming conditionals, they are not obligatory crosslinguistically.

Even if we take the most studied language, English, for example, we have conditional sentences without explicit marking in the antecedent but formed by and, called conditional conjunction, as illustrated by the examples in (1).

Bhatt and Pancheva (2006) also report data in Bengali (cf. Comrie 1986) and Hindi, where the presence of the conditional particle (CP) ‘then’ is obligatory instead of the presence of the CC, as shown in the Hindi example in (2).^[1]

In comparison, conditionals in Mandarin Chinese, the object language of this paper, have at least the following properties: Firstly, Mandarin does not use ‘and’ to form conditionals, unlike English. Secondly, CCs such as ruguo ‘if’, yaoshi ‘if’, and wanyi ‘if, in case’ (see Liu and Wang 2021) in the antecedent are optional. Mandarin conditionals can be formed without markers in the antecedent or the consequent, as seen in (3). Thirdly, CPs such as jiu (which is often translated as ‘then’, although we will argue against this in the paper) are also optional, as seen in (4). Furthermore, as shown in (5) and (6), the modal adverb hui ‘will’ or the additive particle ye ‘also’ seem to give rise to conditional meanings as well.

Examples (3)–(6) taken together indicate that a Mandarin sentence in the form of P, Q is ambiguous between a conjunctive reading and a conditional reading. The question of whether the explicit marking for the latter is always optional is debatable. For example, Comrie (1986) claims that in the absence of any overt conditional marking, a sentence like (4) is ambiguous between a variety of relations holding between the two clauses, such as conditional, temporal, or causal. From here, we will argue that this generalizes over the cases in (3)–(6). In this paper, we will focus on jiu-conditionals (P, jiu Q) and then present a uniform pragmatic analysis of Mandarin conditional constructions in general. In Section 2, we argue that jiu is ambiguous between an unstressed, nonexclusive, left associating element (jiu ₁) and a stressed, exclusive, right associating element (jiu ₂), and that jiu-conditionals involve jiu ₁. In Section 3, we review previous accounts in the literature focusing on the conditional conjunction analysis in Liu (2017) and the scalar analysis of jiu versus German schon ‘already’ (Wimmer 2020). In Section 4, we propose an analysis of jiu ₁ in terms of scalarity involving temporal, conditional, or spatial alternatives. In Section 5, we present a uniform pragmatic account of Mandarin conditional constructions relying on Giannakidou’s (1998) nonveridicality framework. Section 6 concludes the paper.

2 The ambiguity of jiu

Jiu in Mandarin is ambiguous and shows different distributional properties. While some researchers assume there is a series of various (i.e., polysemous) types of jiu (Lü 2013 [1999]: 315), others argue for two basic types (Lu 1984; Xu and Yang 2010). Following the latter view, we assume that jiu can have left or right focus associates, with jiu ₁ being unstressed and jiu ₂ stressed. Furthermore, jiu ₁ has a nonexclusive meaning whereas jiu ₂ has an exclusive meaning, which indicates that we are not dealing with the same particle with different focus (F) associates in (7). Thus, we can infer in (7b) that Zhangsan ate nothing other than a chicken, but we cannot infer that nobody other than Zhangsan ate a chicken in (7a). The examples in (7) also show that jiu is different from then in English conditionals.

While jiu can be used as a conditional marker, it is ill-formed with a preposed consequent, as seen in (8). This contrasts with the use of English then (which can be preposed, though not sentence-initially, e.g., *Then I will come home, if John leaves; see Bhatt and Pancheva 2006) or German dann, in (9).

This further shows that jiu is different from then. We will argue in the following sections that unlike then, which has a deictic or anaphoric use, Mandarin jiu in conditionals mainly contributes scalarity without exclusive force.

3 Jiu-conditionals: previous accounts

Mandarin jiu-conditionals without CCs have raised the question of how the conditional reading is derived and what contribution jiu makes. In this section, we review two recent formal studies.

4 Jiu-conditionals: current proposal

We propose a uniform analysis for jiu ₁ and jiu in conditionals, henceforth, simply jiu ₁ for both, based on the data in (16).

Our analysis has the following components. Syntactically, we assume that the antecedent CP (expressing the proposition P) in jiu-conditionals is base-generated in the consequent TP (Q), as is the AdvP or the PP in (16a, b): CP/AdvP/PP, [_TP jiu …] from [_TP CP/AdvP/PP-jiu …]. We leave it open as to whether it is the specifier of a functional projection headed by jiu ₁, an adjunct to VP, or if it occupies a higher position (see Pan and Paul [2018], among others, on related syntactic issues). Secondly, the CP/AdvP/PP can undergo movement to the left periphery, possibly yielding a (contrastive) topic (see Haiman 1978; Kaufmann and Whitman 2021; Pan and Paul 2018).

Regarding the semantics and pragmatics, we propose that jiu ₁ triggers a scalar inference (SI), which is computed based on the alternatives of its left associates, similar to Kellert’s (2018) treatment of Italian già ‘already’ as a trigger of scalar presuppositions.

The SI of jiu ₁ is similar to Wimmer’s (2020) analysis of jiu in relation to German schon. The scalar analysis can explain why jiu is fine in some kinds of conditionals (e.g., predictive, epistemic, subjunctive) but not in others, such as biscuit conditionals (e.g., There are biscuits on the shelf, if you want some) or speech act felicity conditionals (e.g., If I may say so, the wine is delicious; see a comprehensive list of different kinds of conditionals based on the literature in Hamm and Franke [2010]). In the latter, the antecedent cannot function as the LOW alternative, incompatible with jiu, as minimality presupposes relevance, but in biscuit conditionals, the consequent is true independently of the antecedent. However, it remains to be further investigated whether the SI and the MS analysis are the same; one possibility is to use a context involving two equally minimal alternatives and test whether any scalar inference arises (Magdalena Kaufmann, pers. comm.).

A minor aspect where our analysis differs from Wimmer (2020) is that we take into consideration the involvement of spatial alternatives as in (19). A more important difference, however, is that our analysis is multidimensional, as we propose explicitly that in a jiu-conditional, the conditional meaning is solely the semantic core whereas the scalar meaning is a non-at-issue meaning. This is shown in (21), for (19), and (22), for (20): in the negated, modalized, hypothesized, or questioned variants of the above sentences, the SI survives or projects (Simons et al. 2010), confirming its non-at-issue nature.

The last question that remains is how the conditionality is derived for CC-less jiu-conditionals. While we do not have a fully developed answer, any following account needs to consider the fact that in Mandarin, not only jiu but also cai ‘only’ and ye ‘also’, as well as modal adverbs, seem to trigger conditionality without CCs. Instead of treating these as cases of conditional conjunction (see Section 3.1), we propose a uniform pragmatic account in Section 5.

5 A uniform pragmatic account of Mandarin conditionals

Our account is based on the following patterns of conditionals in Mandarin, where CPs include jiu, ye, cai, and modal adverbs.

Unlike Klinedinst and Rothschild (2012) and Liu (2017), we do not attribute the ambiguity of (23a) to that of a covert or overt conjunction. Instead, we argue that it is the subjective (non)veridicality (Giannakidou 1998) property of the first clause P that determines the reading of the sentence. If P is nonveridical with regard to the speaker’s epistemic state – that is, it is not entailed or presupposed – we receive a conditional reading. For example, in (24), the NPI move a finger indicates that P is not entailed; that is, it forces a nonveridical reading of the first clause, also triggering a conditional reading (cf. (10)). Or, in a context where the speaker does not know whether the addressee is male or not, the sentence (25) will then have a conditional reading. If the speaker knows that the addressee is male, the sentence gets a conjunctive reading. In addition, the sentence can also get a counterfactual reading in contexts where it is presupposed that the addressee is not male; we will not discuss this reading for reasons of space.

If P is entailed or presupposed by the context, we get a conjunctive reading. In contrast, the presence of the factive adverb (Liu 2012) in (26) creates a veridical environment so that the first clause is entailed, triggering a conjunctive sentence. Without support from a broader discourse context or the narrow linguistic context, such sentences are ambiguous.

In addition to common knowledge, polarity items, and adverbs, in spoken language, the intonational and stress patterns of P and jiu can also help. In (27), we show that without contextual or intonational cues, the sentence is ambiguous. The same mechanism applies to conditionals licensed by cai or ye, as in (28) and (29). The common crucial difference between readings 1 and 2 across (27)–(29) is the (non)veridicality property of the first clause.

Furthermore, we believe that the basis for the selection of P as a condition (when P is nonveridical) or cause (when P is veridical) – from a list of possible conditions/causes – for Q, is “the extra-linguistic principle of relevance” (Haiman 1978: 586).

What about (23b), where the conditional contains a CC? In our analysis of (23a) and (23c), it follows naturally that the lack of ambiguity in (23b) is due to the CC creating a nonveridical context. That is, the discourse function of CCs in deriving conditionality is similar to that of NPIs or pragmatic context. The core is the property of nonveridicality, not the lexical ambiguity of an overt or covert and. But how does this account deal with factual conditionals, such as (30a)?^[3] We show that B’s use of jiu-conditionals is only compatible when the first conjunct is subjectively nonveridical, that is, nonveridical with regard to B’s epistemic state; see (30b). This might apply to factual or premise conditionals in general, an empirical question to be tested.

6 Conclusion

In this paper, our discussion has focused on jiu-conditionals in Mandarin Chinese. We assume that jiu is ambiguous and relate jiu in conditionals to unstressed jiu ₁ with left associates and a nonexclusive meaning. We reviewed existing accounts in the literature and argued against a conditional conjunction analysis assuming an ambiguous covert conjunction for jiu-conditionals or the like. Instead, we propose a multidimensional account for jiu-conditionals where jiu contributes scalar inferences. Furthermore, we propose a uniform account of Mandarin conditionals in the form of conjunction with or without connectives or particles, relying on the notion of (non)veridicality. While the details still need working out, we believe that multidimensionality and (non)veridicality are two crucial notions for understanding Mandarin conditionals, in addition to scalarity and additivity, as showcased by jiu and other particles.

There are a number of related issues that we cannot discuss here. Mandarin can use name (‘that way’, i.e., ‘then’) in conditionals, and like jiu this can co-occur with overt CCs. Name and jiu can also co-occur with each other; see (31a). Both are bad in biscuit conditionals, as shown in (31b). What is the function of name in Mandarin conditionals, in comparison to Mandarin jiu and English then (see the contrast in (32))?

Other related expressions are cai and scalar dou or ye (ambiguous between additive and scalar meanings). Cai seems to trigger biconditionality like only if, that is, with exclusive force, and the latter two are concessives. Regarding these two, we think nonveridicality is responsible for the conditionality, with the scalar highness for the concession meaning component.

These related matters await further investigation. While conditionals in Mandarin Chinese are too complex a topic to be dealt with in such a short paper, we hope that the analysis we sketch here will be useful for a better understanding of them.