1. Introduction
Cartography is a research project that is devoted to the precise characterisation of the fine details of the universal functional sequence fseq that makes up each syntactic phrase. In particular, the highly articulated nature of the structural maps drawn is a corollary of the following assumptions:
Since its conception around the 1990s, cartographic research has featured a rich variety of languages, including East Asian languages like Mandarin (Paul Reference Paul2005, Reference Paul2014, Tsai Reference Tsai2008, Reference Tsai2018, Stepanov & Tsai Reference Stepanov and Tsai2008, Badan & Del Gobbo Reference Badan, Del Gobbo, Benincà and Munaro2011, Si Reference Si, Si and Rizzi2021, Reference Si2023; see also the papers collected in Tsai Reference Tsai2015b and Si Reference Si2017) as well as languages from other families. In this squib, we observe that multiple modals belonging to the same semantic class, especially those encoding epistemic meaning, may co-occur within a single sentence in Mandarin, and explore the implications of such an observation for cartography in general. As will become clear, the present study yields novel evidence for the existence of recursive functional heads.
This squib is organised as follows. Section 2 introduces Tsai’s (Reference Tsai2010, Reference Tsai and Shlonsky2015a) cartographic analysis of Mandarin modals. Section 3 demonstrates that epistemic modals may in fact co-occur in a multiple-modal sentence. Section 4 then explores how the stackability of epistemic modals can be accommodated under the cartographic framework. Section 5 discusses the implications of the present findings, and Section 6 concludes.
2. On the cartography of Mandarin modals
Tsai (Reference Tsai2010, Reference Tsai and Shlonsky2015a) examines the topography of Mandarin modals and arrives at the following ordering restrictions:
Adopting cartographic assumptions, Tsai takes such rigid orderings to reflect the hierarchical arrangements of the functional projections introducing different semantic classes of modals, whereby MPEpi dominates MPDeo, which in turn dominates MPDyn. Incidentally, Tsai also notes that these projections each occupy a distinct zone under the familiar tripartite view of clausal organisation.
Tsai recognises the existence of two syntactic classes of modals in Mandarin, namely adverbs and auxiliaries. Examples of the former include yiding ‘surely’, dagai ‘probably’ and bixu ‘obligatorily’, and examples of the latter include ken ‘be willing to’, hui ‘be capable of’ and yao ‘be going to’ (Tsai Reference Tsai2010: 214). Furthermore, modal adverbs are analysed as specifiers of dedicated functional projections, whereas modal auxiliaries are analysed as (non-affixal) heads of these projections. Under the antisymmetric view of syntax (Kayne Reference Kayne1994) which cartographic studies standardly assume, the fact that the ordering restrictions that hold of modal auxiliaries exactly match those that hold of modal adverbs – as seen in (2) – falls out.
It is clear that cartographic accounts like Tsai (Reference Tsai2010, Reference Tsai and Shlonsky2015a) yield considerable mileage in making sense of the observed ordering restrictions. It is hardly surprising, therefore, that the domain of modals is offered as a prime example of how cartography can inform Chinese syntax (and vice versa) in a recent overview article on the topic (see Tsai Reference Tsai2019: 31–32).
3. Co-occurrence of modals
An interesting fact that has received increasing attention in the literature (Tsai Reference Tsai2010, Reference Tsai and Shlonsky2015a, Lin Reference Lin2012, Lai & Li Reference Lai and Li2024) concerns the possibility of co-occurrence of multiple modals. In the following example, for instance, the deontic modal bixu co-occurs with the dynamic modal hui:
While such multiple-modal data as (4) fall squarely under Tsai’s (Reference Tsai2010, Reference Tsai and Shlonsky2015a) cartographic account, what the current studies have failed to explicitly highlight is that multiple modals of the same semantic class may also co-occur within a single sentence. The following naturalistic examples all involve the dagai–keneng sequence, a combination not noted in prior literature:
Both dagai and keneng belong to the class of epistemic modals. Crucially, the authoritative BCC corpus (http://bcc.blcu.edu.cn) yields 81 examples containing the string ‘dagai keneng’, which thus indicates that such a combination is representative. Some attested examples from the corpus are reproduced below:
Before proceeding further, it might be tempting to assign a biclausal structure to the above examples, such that the epistemic modals dagai and keneng, despite appearances, are located in separate clauses:
However, it is worth noting that dagai is commonly assumed to be a modal adverb in the literature (see Li Reference Li1990: 150, Tang Reference Tang1998: 164, Law Reference Law2006: 118, Yang & Ku Reference Yang and Ku2006: 427 among many others) – in fact, we know of no existing work that departs from such an assumption. Notice that dagai does not appear in A-not-A form (i.e. *da-bu-dagai), on a par with clear examples of adverbs like xiaoxinde ‘carefully’ (i.e. *xiao-bu-xiao-xin-de). Accordingly, the biclausal analysis as sketched in (10) can simply be dismissed, because adverbs do not select (clausal) complements.
4. Stackability in cartography
The potential co-occurrence of epistemic modals in Mandarin, as seen in (5)–(7), now raises the non-trivial question of how it can be accommodated under an analysis that incorporates cartographic assumptions (see Section 1). In this section, we will demonstrate that an adequate analysis of the current facts necessitates the proliferation of MEpi heads.
4.1. Head recursion
The syntactic status of keneng ‘probably’ is a controversial issue, and both the XP and X
$ {}^0 $
views have been proposed. For example, Li (Reference Li1990) notes that ‘keneng can be an adverb’ (ibid.: 150). Accordingly, multiple epistemic modals can be introduced in the specifiers of separate functional projections:
Under this analysis, it is clear that the MEpi head is recursive (i.e. MEpi*).
On the other hand, several existing studies take keneng to be a head instead (Lin & Tang Reference Lin and Tang1995, Lin Reference Lin2011, Reference Lin2012, Chou Reference Chou2013). The co-occurrence of epistemic modals may then be dealt with as follows:
Interestingly, the proliferation of MEpi heads is an inescapable conclusion even under such an alternative analysis. To see this, it is necessary to ascertain what Tsai (Reference Tsai2010, Reference Tsai and Shlonsky2015a) initially takes to be instantiations of the MEpi head. Although no clear answer can be drawn from Tsai (Reference Tsai2010), the subsequent work of Tsai (Reference Tsai and Shlonsky2015a) makes explicit that the modal hui that co-occurs with an epistemic modal (henceforth ‘irrealis hui’) is one such element, as in the following example:
The analysis in (12) thus predicts the modal keneng and the irrealis hui to be in complementary distribution, which is however incorrect:
As such, we are back to the proliferation of MEpi heads.Footnote 1
The skeptical reader might wonder why the modal hui in examples like (13) should instantiate MEpi
$ {}^0 $
instead of some lower functional head (e.g. T
$ {}^0 $
). Indeed, the Mandarin word hui is multi-functional, and the observed linear order itself is not incompatible with such possibilities. In this respect, Tsai (Reference Tsai and Shlonsky2015a) offers an argument for the hypothesised position of the irrealis hui. In Mandarin, wh-phrases are not inherently interrogative expressions, but may receive a non-interrogative interpretation in certain contexts (Li Reference Li1992). The crucial observation is that whereas the irrealis hui may license a wh-indefinite, the future hui may not:
According to Tsai, the contrast follows because the irrealis hui is high enough to trigger existential closure at the TP-level (Heim Reference Heim1982), as (17) exemplifies. On the other hand, the future hui, which is taken to occupy T
$ {}^0 $
, is too low to trigger such an operation.
The fact that the wh-phrase in examples like (18) may be interpreted as indefinite, therefore, corroborates the analysis shown in (15), where the modal hui instantiates an X
$ {}^0 $
outside TP.
4.2. Multiple specifiers?
Taking a step back, however, one might wonder if dagai and keneng may both be introduced by one and the same functional head MEpi, which can itself be realised by the irrealis hui:
However, for the current analysis to truly have force, its proponents must demonstrate that the epistemic modal keneng can never be a head (pace Lin & Tang Reference Lin and Tang1995, Lin Reference Lin2011, Reference Lin2012, Chou Reference Chou2013), since if it can, MEpi will again be recursive when keneng is a head. Notice also that elements such as focalised phrases readily intervene between the two epistemic modals:
These facts are not immediately expected if dagai and keneng are specifiers of one and the same MEpi, since there would be no space for a distinct functional head that introduces a focalised phrase in between.
More generally, as mentioned in Section 1, one hallmark of cartography (Rizzi Reference Rizzi1997; Cinque Reference Cinque1999 among many others) is the adoption of some version of Kayne’s (Reference Kayne1994) Linear Correspondence Axiom, a linearisation scheme that has the effect of ruling out multiple specifiers. In more recent work, Rizzi (Reference Rizzi2017) also explicitly denies multiple-specifier representations ‘on grounds of restrictiveness’ (ibid.: 184), now based on comparative evidence (see also Rizzi & Bocci Reference Rizzi, Bocci, Everaert and van Riemsdijk2017: 19–20 for relevant discussion). For all these reasons, therefore, the current analysis will not be further entertained.
4.3. Preposing?
Let us also consider the possibility that preposing is involved in deriving the multiple-modal sentences of interest (Lai & Li Reference Lai and Li2024; cf. Tsai Reference Tsai2010, Reference Tsai and Shlonsky2015a):
In fact, this is irrelevant, because the crucial question remains as to what the representation would be for these epistemic modals prior to movement. In this connection, it is worth noting that the dagai–keneng sequence is not limited to pre-subject position. We have already seen an attested example in (8) where the sequence appears after the surface subject, repeated as (23) below for convenience, together with two additional examples.
Again, preposing does not help escape from the conclusion that the proliferation of MEpi heads is necessary.
4.4. Interim conclusion
Taken together, we conclude that the domain of Mandarin modals constitutes a novel piece of evidence that one and the same functional head can be recursive.
Recently, Cinque (Reference Cinque2017) proposed the idea of ‘paired functional projections’ (see Tescari Neto Reference Tescari Neto2022 for further support for such a claim). In essence, the idea is that a functional projection FP consists of lower and higher sub-projections. As an illustration, the ‘unitary’ functional projection MPEpi will accordingly comprise the sub-projections MPEpi(Low) and MPEpi(High):
What matters for present purposes is that the splitting of projections this way does not make available additional structural positions. In Cinque’s (Reference Cinque2017) conception, the lower sub-projection FP(Low) encodes the ‘core functional notion’, whereas the higher sub-projection FP(High) represents an ‘adverbial modification of the same functional notion’ (ibid.: 522).Footnote 2 Hence, the current conclusion will remain unscathed should we adopt the recent idea of paired functional projections: depending on whether keneng is a phrase or a head, we will end up having the representation in (27) or (28), which still involves recursive ‘MPEpi”s:
5. Implications
The overall conclusion from Section 4.4 is significant in the following sense. In cartography, one open question is whether a functional head may truly iterate. In the latest overview summarising the major findings of cartographic studies (Rizzi & Cinque Reference Rizzi and Cinque2016), the detailed maps of the IP-, DP-, PP- and AP-zones all indicate that each functional head within a zone is unique and occupies a dedicated position. Concerning modal heads, Rizzi & Cinque (Reference Rizzi and Cinque2016: 149) recognise the existence of five such heads on the functional hierarchy, namely Mepistemic, Mnecessity, Mpossibility, Mvolition and Mobligation, which, again, are all unique.
Tempting as it is to conclude that each functional head is thus unique, there exists a well-known (in fact, the sole) exception, namely the Top head. Rizzi & Cinque (ibid.: 146) offer the following fine structure of the clausal left periphery:Footnote 3
Nonetheless, more in-depth studies have revealed the existence of different kinds of topics associated with distinct interpretive properties (Benincà & Poletto Reference Benincà, Poletto and Rizzi2004, Frascarelli & Hinterhölzl Reference Frascarelli, Hinterhölzl, Schwabe and Winkler2007, Bianchi & Frascarelli Reference Bianchi and Frascarelli2010), as Rizzi & Cinque (Reference Rizzi and Cinque2016: 146) themselves also acknowledge. Specifically, Frascarelli & Hinterhölzl (Reference Frascarelli, Hinterhölzl, Schwabe and Winkler2007) identify three sub-types of topics, namely aboutness topics, contrastive topics and familiar topics. Accordingly, the representation of a multiple-topic sentence may involve no ‘true recursion’ in the technical sense of the term. In fact, one of the main claims of Benincà & Poletto (Reference Benincà, Poletto and Rizzi2004) is precisely that ‘recursion is not an option [
$ \dots $
] recursive in the sense that there is a virtually infinite set of totally identical Topic phrases’ (ibid.: 53, our emphasis).
Recently, Rizzi (Reference Rizzi2017) briefly offers the following Italian example to support his claim that true recursion exists. As shown below, the example involves six topic phrases:
In this respect, the current analysis of the novel Mandarin data adds a new dimension to the debate. First, it moves the relevant discussion from the familiar domain of topics to the seldom-discussed domain of modals. Second, the data offer a clear piece of evidence for the possibility of true recursion. In the current case, note that both the modals dagai and keneng belong to the class of epistemic modals. There is no detectable interpretive difference between the two, and hence no justification for assigning them to two separate semantic categories.Footnote 4
Certainly, one might still insist to posit distinct functional projections such as dagaiP and kenengP (despite what has just been said), and hierarchically arrange these projections in a specific way. To further strengthen the claim that true recursion is indeed at stake, the following examples from the BCC corpus are informative:
Similar to the logic of Rizzi’s (Reference Rizzi2017) example reproduced in (30), these data show that as many as four epistemic modals can co-occur in a sentence (in fact, more). More importantly, they demonstrate that the reverse order keneng
$ > $
dagai is likewise attested (with 62 such examples in the BCC corpus). Both facts further support the view that MPEpi is recursive in Mandarin.Footnote
6
6. Conclusion
In this squib, we noted for the first time that epistemic modals may actually co-occur within the same clause in Mandarin. A detailed examination of the relevant data revealed that their structural representations necessarily involve recursion of the same functional projection. Accordingly, the domain of Mandarin modals constitutes novel and clear evidence for the claim that a functional head may iterate (at least for some heads in some languages), a finding that directly bears on the proper understanding of the ‘One Feature – One Head’ principle as standardly assumed in cartography.
Acknowledgements
We thank the editors and the three anonymous referees for their constructive comments. We are also grateful to Stephen Matthews for his assistance with glossing the Italian example. The usual disclaimers apply.
Open access
