The Puzzle of Anaphoric Bare Nouns in Mandarin: A Counterpoint to Index!

Mandarin is known to lack an overt definite article, expressing definiteness via bare nouns (e.g., Chao 1968, C. Li and Thompson 1981, Cheng and Sybesma 1999, Yang 2001) (1a). The ability of bare nouns to function as definites is also illustrated by (1b), in which the linguistic context brings out their anaphoric use (e.g., Dayal 2004, 2011, Jiang 2012, 2020, X. Li 2013).

Jenks (2018) modifies this generalization on the basis of examples like (2a–c). According to him, the anaphoric potential of bare nouns does not extend beyond the subject position. The bare nouns in (2b–c) are infelicitous in both direct object and indirect object positions, and demonstratives must be used instead.

Jenks further argues that only uniqueness-based definites, or weak-article definites in the sense of Schwartz 2009, are realized with bare nouns in Mandarin and that demonstratives can occur in these contexts only with contrastive interpretations. He examines three environments that require unique definites, using the classification in Hawkins 1978. In addition to bare nouns functioning as immediate-situation definites, as shown in (1a), they also function as larger-situation definites, demonstrated in (3a), and as associative definites in part-whole bridging contexts, as shown in (3b) (for more on associative anaphora, see Clark 1977, Hawkins 1978, Schwarz 2009).

Jenks offers the following analysis to account for the contrast between Mandarin bare nouns and demonstrative expressions. First, he assumes that Mandarin bare nouns project NPs (4a), whereas demonstrative expressions project DPs, consisting of at least three distinct nominal projections: DP > ClP > NP (4b). Second, he adopts Trinh’s (2011) analysis of common nouns and numeral classifiers and treats Mandarin bare nouns as predicative; he further adopts the analysis that Mandarin bare nouns achieve their definite interpretation via the type-shifting operator iota ɩ (e.g., Yang 2001) (5a).¹ Third, he follows Schwarz’s (2009) account of weak vs. strong definites, which are distinguished by the presence of an index only on strong anaphoric definites. Specifically, he proposes that Mandarin demonstratives are like strong-article anaphoric definites in German in taking an extra semantic argument, which in most contexts is satisfied by an index ɩ^x (5b). He assumes that the index is interpreted as an indexical property relativized to an assignment function λx[x = g(1)] and that the indexical argument of ɩ^x is represented as a DP adjunct.²

Jenks follows the literature on bare nouns in interpreting the Mandarin common noun as a kind-denoting term. He posits a covert iota type shift in the case of bare noun arguments. Details aside, we get the expected meaning for the noun phrase in (4a): a presupposition that there is a unique student in the situation and that, when defined, the NP refers to that individual. In the case of the demonstrative in (4b), a classifier phrase intervenes.³ Again, as is standard, classifiers take a kind term and deliver a set of atomic or singular individuals (AT_obj). To this the demonstrative, qua strong definite, adds the indexical property. Here Jenks departs slightly from Schwarz 2009 but the essential idea is the same. Though iota is still implicated, its uniqueness requirement is defined on two properties, the property of being a student and that of being identical to a previously introduced individual (x = ɡ(1)). As in Schwarz’s original account, this distinction is meant to separate the contexts in which uniqueness of the common noun is at stake from the contexts in which uniqueness rides on indexicality. The exponent for the first type in German is the weak article and the exponent for the second type in German is the strong article.⁴ On Jenks’s account, the exponent for the first type in Mandarin is the bare noun and the exponent for the second type in Mandarin is the demonstrative.

According to Jenks, then, Mandarin patterns with German in making a principled distinction between weak (unique) and strong (anaphoric) definites. It differs from English, which uses a single form for both types of definites. The resulting crosslinguistic picture is given in table 1.

Jenks’s (2018) core proposal is a principle called Index!, which requires Mandarin to explicitly represent indices whenever possible.

Since the claim is that Mandarin demonstrative expressions include an index that is absent in definite bare nouns, Index! would require demonstratives to be used whenever they can. In the case of (2b–c), demonstrative expressions are available since they are anaphoric to expressions in the antecedent clause, and therefore they must be used; as a result, bare nouns become unavailable in those contexts.

Jenks makes one exception to Index!. As seen in (1b), subject bare nouns can be anaphorically linked to expressions in antecedent clauses. To account for this, he proposes that anaphoric bare nouns in subject position are continuing topics in Mandarin. The pragmatic function of topic marking overrides and neutralizes the effect of Index! in such environments since topics are salient members of the question under discussion and do not need to be indexed.

This, in brief, is Jenks’s (2018) account, which combines the view that Mandarin bare nouns are kind-denoting terms with the view that definiteness is composed of a uniqueness-based subtype and an anaphora-based subtype.

The data presented by Jenks (2018) certainly change the picture of Mandarin bare nouns that was earlier assumed in the literature. However, the conclusions that he draws are not quite supported when more facts are taken into account. In this section, we take a closer look at the two empirical generalizations on which Jenks’s proposal is based: the behavior of demonstratives in contexts that favor weak-article definites and the behavior of bare nouns in contexts that favor strong-article definites. In each case, we fill in some crucial gaps in the paradigm in order to lay a more solid foundation for claims about the lexical exponents of definiteness in Mandarin.⁵ See also Bremmers et al. 2021 for similar facts related to Jenks’s description of Mandarin.

Schwarz (2009, 2013) bases his division of definites on languages like German, Lakhota, and Akan that show a lexical distinction in behavior, and he argues that the distinction between strong and weak definiteness holds crosslinguistically.⁶ In English (and certain other languages) this distinction is not apparent, but Schwarz holds that this is because, in the case of English, a single definite article (the) ambiguously encodes both weak and strong definiteness.

Schwarz is careful to note that strong and weak definite articles do not exhaust the possibilities for expressing definiteness, and he mentions demonstratives as one example (see, in particular, Schwarz 2009:34–37). Since on Jenks’s account the Mandarin demonstrative is the strong definite, we believe it is worth understanding what, if anything, differentiates a strong definite from a demonstrative. We can do so with reference to English, where the two are lexically distinct.

The larger-situation use and part-whole bridging contexts mentioned in section 1 provide a good starting point. In both, the set denoted by the common-noun head is uniquely instantiated: world knowledge tells us that there is only one moon that we ordinarily talk about and that a given car can have only one license plate. While the English definite is acceptable in such cases, the English demonstrative is not.

On the ambiguity view of English definites, we can analyze the situation as in (7c). The weak definite survives; the strong definite and the demonstrative are ruled out. Note that the infelicity of Mandarin demonstratives in the corresponding cases seen in (3) is equally compatible with their being analyzed as garden-variety demonstratives as with their being analyzed as strong definites. These contexts therefore provide no reason to move from the null hypothesis, which is that Mandarin demonstratives are demonstratives rather than definites.

However, there is one other case discussed by Jenks that is meant to settle the question in favor of Mandarin demonstratives being ambiguous between demonstratives and strong definites. These are associative nouns that do not involve part-whole relationships of the kind seen in (3b) and (7b). Schwarz notes that several languages differentiate between these two types of associative uses and, while prohibiting strong definites from the part-whole cases, allow them in producer-product cases.⁷ Jenks provides the example in (8a) to establish that Mandarin allows the demonstrative to be used here, aligning it with languages that allow strong definites in these cases.

Sentence (8a) is meant to settle the case in favor of Mandarin demonstratives being strong definites since the English translation of the associated NP poet is acceptable with the definite determiner (and presumably not with a demonstrative). The explanation is as follows. The ordinary demonstrative meaning of na wei shiren ‘that Cl poet’ is ruled out analogously to the way that poet would be ruled out in English, but its strong definite meaning is acceptable analogously to the way the poet is acceptable. We believe this conclusion is too hasty.

Note that the antecedent noun in the main clause of (8a) also has a demonstrative, na shou shi ‘that Cl poem’. The canonical examples of bridging in the literature typically have indefinite antecedents. When we adjusted the data so that the sentence has an overt indefinite antecedent, as in (8b), the judgments changed and our consultants readily accepted the bare NP.⁸ We therefore do not see any basis for making the distinction between part-whole and producer-product subcases of associative anaphoric nouns in Mandarin. Consequently, we do not see sufficient empirical grounds to consider the Mandarin demonstrative anything other than what it appears to be: a regular demonstrative.

A theoretical issue with the way Index! relates to strong definites is also worth noting. Consider the sequence of English sentences in (9a–b).

The first mention of the president/the sun involves a weak definite, an instance of larger-situation uniqueness. It is not so clear how the second instance of these noun phrases should be analyzed. The phrases still satisfy uniqueness, so we could use the weak definite, but they also satisfy anaphoricity, so Index! might tell us to use the strong definite. Again, English does not help us in this regard because the ambiguity of the masks the distinction, but the Mandarin anaphoric noun is transparent.

Both instances of president/sun have to be bare; demonstratives in the second sentence are infelicitous. So, as in the other cases discussed here, the Mandarin demonstrative turns out to pattern with the English demonstrative, not with the English strong definite. We should point out that Jenks in a sense inherits this problem from Schwarz, but by proposing a principle like Index!, he makes a specific prediction for Mandarin that is not borne out.⁹

We will come back to the appropriate account of demonstratives in section 3. We turn now to the other side of the equation and test the generalization that nonsubject Mandarin bare nouns are infelicitous in anaphoric contexts.

Recall from (2) that Jenks (2018) claims that anaphoric bare nouns in direct and indirect object positions are infelicitous. However, we show on the basis of elicited and naturally occurring data that this is not always the case. We start with the elicited data. When the second sentence contains two bare nouns, one in subject position and one in object position, they can both refer anaphorically to the indefinite expressions in the first sentence.

Note that the bare nouns in (11b–c) are not subject to the exception Jenks makes for bare nouns in subject position. Crucially, while demonstratives are judged to be possible in subject and object positions in these examples, all the native speakers we consulted found two bare nouns to be acceptable, several even voicing a preference for two bare nouns over a demonstrative in one or both positions.

It has been suggested to us that having two bare nouns in the second sentence may create a contrastive environment and that Index! can be overridden and neutralized by contrastive focus, just as it can by continuing topics (Peter Jenks, pers. comm.). If focus can also introduce an index of its own, the occurrence of bare nouns in object position in (11) is to be expected under Index!. Although this possibility can help maintain Jenks’s (2018) proposal, we do not find it compelling, theoretically or empirically. Naturally occurring data establish that Mandarin bare nouns can be anaphoric in nonsubject positions in sentences that are not contrastive in any identifiable way.

In (10), we provide examples from the Beijing Language and Culture University Contemporary Chinese Corpus (BCC Corpus).¹⁰

The bare nouns nanhai ‘boy’ and gou ‘dog’ appear as objects in the second clause, referring anaphorically to indefinite expressions in the antecedent clause. Note that the subjects in the second clause in (12a) and (12b) are a proper name and a pronoun, respectively.

Example (13) illustrates that bare nouns in indirect object position can also refer anaphorically to the indefinite in the antecedent clause.

It is worth emphasizing that in the examples in (12) and (13), demonstrative expressions can also appear where the anaphoric bare nouns appear. Index! incorrectly predicts a competition between these two forms, with bare nouns losing out to demonstratives.

To make this point further, (14) illustrates that anaphoric bare nouns can appear multiple times in object (and subject) position in second and subsequent clauses.

In (14a), the anaphoric bare noun nühai ‘girl’ first appears in object position in the second clause and then appears in subject position in the third clause and object position in the fourth clause. In (14b), the anaphoric bare noun mao ‘cat’ first appears in the ba-construction in the second clause, leading to SOV (as opposed to SVO) word order, and appears again in object position in the third clause.

We have provided evidence based on elicitation from six native speakers of Mandarin and from naturally occurring sentences from a major corpus against the empirical generalizations in Jenks 2018. We have more such examples, which for reasons of space we do not include here. This suggests to us that Index! cannot explain the distribution of demonstratives and bare nouns in Mandarin and that an alternative is worth exploring. We show next what an account with the potential to capture the full range of facts might look like.

In this section, we suggest a way of analyzing Mandarin demonstratives that aligns them with demonstratives crosslinguistically, while delivering the apparent strong definite behavior noted in Jenks 2018. We also speculate on an alternative way of capturing the crucial contrasts in the anaphoric behavior of Mandarin bare nouns—that is, between the examples on which Jenks bases his claims for their nonanaphoricity and the examples we have provided that do show their anaphoricity.

Let us step back a bit from the specifics of Mandarin and ask what is known about demonstratives generally. One observation in the literature is that most, perhaps all, languages have demonstratives, while a large number of languages lack definite articles. Another is that demonstratives are the historical source for definite articles. There is a rich literature on the semantics of demonstratives, of course, but we will take our inspiration from Robinson 2005, as it bears most closely on the issue of interest here (see also Roberts 2002, Wolter 2006). Drawing on Löbner 1985, Robinson notes that demonstratives differ from definites in their ability to tolerate lack of uniqueness.

In fact, Robinson argues, a demonstrative not only tolerates nonuniqueness, it requires it.

Given these facts, she suggests that although demonstratives refer to a unique entity, they have a presupposition of nonuniqueness. We can represent it as in (17), and given that these facts hold crosslinguistically, we can take (17) to hold in Mandarin as well.

That is, each instance of a demonstrative picks out a unique entity that satisfies two properties, the common-noun property and the intended-referent property, here represented by the free variable y, in line with the convention in Schwarz 2009. The crucial difference lies in the presuppositional piece of the demonstrative’s meaning. It requires that the intended referent not be the sole member of the common-noun set. The demonstrative thus differs from the definite, where the commonnoun set bears a presupposition of uniqueness: |P| = 1.

Let us illustrate how this explains the contrasts in (15) and (16). Since |dog| = 2 in (15), each conjunct satisfies the nonuniqueness presupposition as well as the entailment of uniqueness: there are two dogs in the context, but the demonstratives in the two conjuncts have distinct indices, ensuring uniqueness calibrated to those indices. The problem with (16) lies with the presupposition of nonuniqueness. For (16a), normal world knowledge tells us that there is only one sun in our conceptual universe, and for (16b–c), it tells us that there is only one steering wheel/umpire per car or match.¹¹

Given what we have said so far, one may expect demonstratives and definites to be in complementary distribution, but this is not the case. Most relevant to present concerns are anaphoric contexts where both options seem possible in English.

Robinson (2005) notes, and we agree, that the version with the definite seems unmarked, while the version with the demonstrative suggests a slight sense of contrast. Setting that aside, how can we explain the difference between cases like (18b) and (16a–c), given the proposed semantics in (17)? To do so, we minimally modify the nonuniqueness presupposition in (17) as in (19a), to allow its satisfaction in a larger situation.¹²

We are importing the notion of widening, proposed by Kadmon and Landman (1993) to explain the polarity item any, for the satisfaction of the nonuniqueness condition in (19a). Domain widening has also been used by Dayal (2013) to explain the indefiniteness typically associated with bare plurals in episodic contexts. On the opposite side of the spectrum, D-linked expressions like each are thought to resist such widening (Kadmon and Landman 1993:378–379, Dayal 2016: 122–124). Here, we suggest that the initial situation invoked by the first sentence in (18b), with a unique woman and a unique man, is extended to include a larger situation that opens up the possibility of having other women in it, even if no such woman is salient. This is what licenses the demonstrative in the second sentence of (18b). The strong definite in (18a) is also felicitous, since its uniqueness presupposition is satisfied in the situation in which the first sentence is interpreted. For completeness, we include the weak definite. In this situation, it also happens to be satisfied since there is a unique element not only in the intersection of P and the set of individuals anchored to the indefinite in the first sentence, but in P itself. To anticipate, we propose that the Mandarin demonstrative is to be represented in terms of (19a), and the Mandarin bare noun in terms of (19b) and (19c).¹³ The choice between the strong and the weak definite is not trivial; we will return to it in the following sections but continue to focus on the demonstrative for now.

Our account of demonstratives rests on the view that domain widening in anaphoric cases such as (18b) involves a relatively smooth transition from a more restricted situation to a larger situation. One might ask why it is not similarly possible to license a demonstrative on a widened domain in larger-situation uses of the kind illustrated in (16a). The answer seems to be this. To allow the possibility of other suns, we would need to adjust our context to incorporate other solar systems. While such adjustments are certainly possible, it seems that speakers do not make the rather serious adjustments that would be required. And in the case of associative readings such as those in (16b–c), the demonstrative remains anchored to a single entity through a relationship defined on uniqueness. No matter how many other cars or matches there may be in the widened domain, there will still be only one steering wheel and one umpire anchored to the car and match referred to in the first sentence. To jump to the steering wheels of other possible cars or the umpires of other possible matches seems to be a bridge too far for the average speaker.¹⁴ However, note that immediate-situation uses involving ordinary common nouns, as in That boy is tall, would be acceptable under what would be described as a deictic use of the demonstrative.

Let us take stock. We have drawn on Robinson’s (2005) insight about the difference between definites and demonstratives to argue for the role of a nonuniqueness condition on the use of demonstratives and suggested a way in which their distribution can be differentiated from that of the definite determiner. Given that the behavior of demonstratives in core cases is relatively stable across English and Mandarin (and, as far as we know, most languages), we have suggested that Mandarin demonstratives are really just regular demonstratives. One might even argue that this is the null hypothesis, for, as we demonstrated in section 2.1, there is no empirical imperative to treat Mandarin demonstratives as ambiguous between demonstratives and strong definites.

While we presented ample evidence in section 2.2 against the generality of Jenks’s (2018) conclusion that Mandarin bare nouns cannot be anaphoric definites, we do not challenge his core data. We repeat a near-minimal pair as a reminder of the contrast at issue and to highlight the nature of the puzzle posed by anaphoric bare nouns in Mandarin.

There are two things to note about these examples. The first is that the syntactic position of the anaphoric definite does not determine its shape; there are nonsubject anaphoric nouns (nansheng ‘boy’) in both sentences and yet one is a bare noun while the other is a demonstrative. The second is that we are talking about preferences, not judgments of absolute (un)grammaticality. There is a preference for a demonstrative in cases like (20c), but the bare noun is not exactly ruled out for all speakers. The bare noun is fully acceptable for all speakers in (20b), but the demonstrative would not be deemed ungrammatical. One conclusion we can safely draw is that Index! does not provide the right level of granularity to capture the nuances of this paradigm. The challenge, then, is to find a different angle from which to approach the problem. Here we will build on a suggestion made by Gita Martohardjono (pers. comm.) to see if it can shed light on the observed contrast.

Let us start by asking how the context-setting sentence in (20a) would be interpreted. On a situation-semantics approach, we might posit something like (21a), taking s to be the minimal situation that has a boy and a girl in it.

Now let us consider the two follow-up sentences in (20b) and (20c), in isolation for the moment. Sentence (20b) has two bare nouns, nüsheng ‘girl’ and nansheng ‘boy’, each of which can be considered to be an immediate-situation unique definite. This is derivable via the iota type shift (e.g., Chierchia 1998, Cheng and Sybesma 1999, Yang 2001, Dayal 2004, Jiang 2012, 2020, Jenks 2018). For perspicuity, we include the presupposition on a separate line at the top. In (21b), we have the presupposition that there is a unique boy and a unique girl in the situation of evaluation and the assertion that the unique girl is sitting next to the unique boy. Interpreted in isolation, (20b) exemplifies immediate-situation definites in Hawkins’s (1978) classification and falls under Schwarz’s (2009) definition of weak-article definites. Note that this is how the bare noun tang ‘soup’ in (1a) is interpreted as a definite, on our account as in Jenks’s. The same explanation also covers larger-situation uses of bare nouns like yueliang ‘moon’ in (3a), where the situation of evaluation s′ would be identified with the world w_s.

In (20c), instead of a bare noun we have na-ge-nansheng ‘Dem Cl noun’. The truth-conditional contribution of the demonstrative is the same as that of a strong definite, but on the account sketched in section 3.1 it has a presupposition of nonuniqueness on the set of boys, satisfiable on a widened domain. Again, taking (20c) by itself, we predict it to be acceptable. And, indeed, it is. In immediate-situation uses, it would translate into a deictic use of the demonstrative (‘I met that boy yesterday’), a use that Jenks’s account recognizes, as would any account of demonstratives. Our account also has the advantage of indicating why demonstratives cannot function as larger-situation definites: yueliang ‘that Cl moon’ will not satisfy the nonuniqueness presupposition of demonstratives, as it is unique in the widest situation available, the world of the situation w_s. Note that the unacceptability of the demonstrative in larger-situation uses (e.g., (3a)) is categorical, not just a preference for the bare noun. This is expected since the unacceptability is due to presupposition failure.

With the basics in place, we return to anaphoric contexts. The contrast between (20b) and (20c) emerges when they are seen as continuations of (20a) in a narrative sequence. In (22), we illustrate what happens when (21b–c) are treated as follow-ups to (21a).

As the logical representations make explicit, both the bare noun and the demonstrative are defined in this context and end up referring to the same individual; that is, our account shows why both discourses are possible. However, we need to identify what could explain the preference for a bare noun in the first case and for a demonstrative in the second. The key, we believe, is in the relation between the initial situation s and the subsequent situation s′. In the version of situation semantics in which Schwarz couches his account of strong- and weak-article definites, there is no ontological difference between situations and individuals (Schwarz 2009:223). Given that, there is nothing substantive to the condition s ⩽ s′ in (22a) but there is in (22b): s is defined on two individuals and although the same two individuals define s′ in (22a), the individuals in s are a proper part of the individuals in s′ in (22b). We can therefore rewrite the formulas as in (23).

One might argue that since the initial situation extends across the whole discourse in (23a), the simplest type of definite would be one that employs the simple type-shift iota, which in the case of Mandarin is the bare noun. And, given that there are two distinct situations in (23b), a deictic definite might have an advantage. While there is no reason to doubt that there is still only the unique boy in s′, we can see that there are now potentially two situations in play, a situation s defined by a boy and a girl, and a situation s′ that presumably expands the original situation to include another individual. Though there is no indication that there are other boys in s′, the demonstrative can refer unambiguously to the unique boy in the initial situation s. We might represent this schematically, as in (24). While we have linguistic evidence only for the distinction in (24a) and (24b), we speculate that once the possibility of a distinct s′ arises, it brings with it the possibility of an extension such as (24c): namely, one in which iota (the operation that the definite bare noun taps into) is infelicitous but the demonstrative remains felicitous.

Looking at the contrast in terms of situations and the entities that constitute them provides us with a fresh angle from which to address this puzzle. We no longer predict complementary distribution, but we do allow for the possibility of a preference. If speakers feel confident that the initial situation remains unchanged, they have a choice between two felicitous options and they choose the simpler one—namely, the bare noun encoding the simple type-shift iota. Once the original situation is extended, however, speakers may play it safe and choose the demonstrative (which would remain felicitous even if the extension was drastic) over the definite (which could become infelicitous).

If this line of thinking has any merit, it predicts that anaphoric bare nouns should improve in (25), where the first sentence itself sets up a minimal situation with all three individuals, a boy, a girl, and (a woman) Mali. This prediction is indeed borne out, as the bare noun in (25b) is judged acceptable.

Interestingly, it is apparently even possible to have anaphoric bare nouns when the initial context is expanded incrementally to include additional participants.¹⁵

We speculate, then, that there is a crucial difference between the introduction of an individual in a way that implicitly shifts the initial context (as in (20c)) and the controlled introduction of an individual into a given context (as in (26)). In the first case, two distinct situations are brought into play; in the second case, a situation that satisfies uniqueness is minimally expanded to include another individual before the anaphoric noun is processed. In other words, the situation in which the last conjunct is interpreted in (26) is the same as the immediately preceding situation—namely, a situation with a boy, a girl, and the kid who then asks the unique boy in that same situation to go out.

We want to emphasize that we by no means think this is necessarily the last word on this issue; rather, we believe an explanation in these terms has the right contours to capture the gradient nature of the phenomenon. As far as we can tell, it covers the cases in Jenks 2018 showing the reduced acceptability of anaphoric bare nouns and the ones we have provided in section 2.2 showing their full acceptability.¹⁶Index!, we believe, is not the right tool for the job. Like Maximize Presupposition (Heim 1991), on which it is based, it leads to categorical distinctions in (un)acceptability and complementary distribution that do not characterize the anaphoric demonstratives vs. bare nouns in Mandarin.¹⁷

Let us consider the picture of definiteness that has emerged, expanding on Jenks’s (2018) distinctions (see again table 1). There are three lexical exponents that are major players crosslinguistically: demonstratives, strong definites, and weak definites, with a few languages (e.g., German) reflecting this three-way distinction lexically, while many (e.g., English and Mandarin) show only a two-way distinction. By introducing demonstratives into the picture, we also introduce a new diagnostic environment into the calculus, the deictic use of definites (see table 2), and we draw on context sentences like (15a) to determine whether a particular definite qualifies as a deictic definite. Note that we have included the German strong article among deictic definites. This is based on Schwarz’s (2009:34–37) discussion, which includes the following example:

Schwarz emphasizes that the strong article is not synonymous with the demonstrative, because (27) is only acceptable with prosodic focus on the two definites. His discussion nevertheless underscores the point we have tried to highlight in our discussion of Mandarin, that determining whether a particular form should be aligned with a strong-article definite or with a demonstrative is not straightforward.

Here, we should also point to yet another distinction that needs to be included in the picture of crosslinguistic variation given in table 2. Languages can also use structural options to express definiteness. Cantonese, closely related to Mandarin, includes bare classifier phrases in its inventory of definite structures. The following example, where superscripts indicate tones, shows one type of definite use:¹⁸

On the basis of examples like (28), Jenks (2018) proposes that Cantonese bare classifier phrases, rather than bare nouns, function canonically as anaphoric definites. However, Jenks’s view of Cantonese bare nouns is not uncontroversial (e.g., Sybesma and Sio 2008, Simpson, Soh, and Nomoto 2011, Cheng and Sybesma 2012). While we cannot go further into the facts here, the point that such sequences need to be factored into mapping out the full crosslinguistic typology of definiteness is well-taken.¹⁹

Returning to table 2, we note that it raises the issue of deciding how to map a two-way lexical distinction to a three-way conceptual distinction. Highlighting the importance of this issue is part of what we hope to have achieved in our discussion of Mandarin. A second question, and one that applies to all languages, is how to capture the distribution of different types of definites within a language. If lexically distinct items were in complementary distribution, it would be clear what research strategy one should pursue. The requirement would be to define the meanings of individual items such that only one of the meanings would survive in any given context. However, the reality is that there are overlaps, so any account must include ancillary assumptions to distinguish between the contexts that allow only one definite and those that allow more than one.²⁰

Focusing on Mandarin, we can safely take bare NPs to function as unique definites and demonstratives to function as deictic expressions. The question that remains is the status of anaphoric definites. We have shown that both types of NPs can function in canonical anaphoric contexts. We have suggested looking at situations in terms of the individuals that define them as a way of capturing overlaps in distribution. Once we recognize overlaps, however, the question of preference enters the picture. Focusing on demonstratives vs. strong-article definites, do speakers systematically accept one particular form over the other or is there individual variation on this score? Do all languages show the same preference for demonstratives vs. definites or does the preference depend on the nature of the definite—a lexical definite as in English vs. a bare NP as in Mandarin? At this point, not enough is known about the preferences related to demonstratives to answer this question responsibly.

To sum up, we see Jenks’s (2018) discussion of the choice between Mandarin demonstratives and bare nouns as a much-needed push to expand the study of definiteness to include demonstratives and bare nouns.²¹ We believe more fine-grained and sophisticated tests conducted on a statistically significant sample size are needed to empirically define the problem before a theory of competition between demonstratives and other exponents of definiteness can be clearly formulated.

We are extremely grateful to two anonymous reviewers for their feedback. We have also benefited from discussions with Dorothy Ahn, Vicki Carstens, Ming Chen, Yi-Hsun Chen, Kamil Deen, Shin Fukuda, Peter Jenks, Jess Law, Mingming Liu, Gita Martohardjono, Yimei Xiang, and Beibei Xu. All remaining errors and omissions are our responsibility.