1. Introduction
The restrictive/appositive distinction is by now familiar from the voluminous literature that touches on relative clauses (Stockwell et al. Reference Stockwell, Schachter and Partee1973, Jackendoff Reference Jackendoff1977, McCawley Reference McCawley1988, among many others). The distinction is worth our attention largely because, from a descriptive viewpoint at least, restrictive relatives and appositive relatives each have their unique grammatical properties. In languages like English, for instance, both the complementiser that and the zero complementiser may introduce a restrictive but not an appositive. Semantically, whereas restrictives restrict the set of individuals denoted by the relativised Head, appositives simply contribute subsidiary information about the referent of that Head. Prosodically speaking, appositives, but not restrictives, are typically separated by intonational breaks.
In this paper, we contribute to the unsettled debate about whether Mandarin relative clauses (RCs) exhibit the restrictive/appositive distinction, and if so, how the two types of relative can be effectively and systematically distinguished in the language. We demonstrate that the existing literature has overlooked the confound that demonstratives have distinct uses in language (Kaplan Reference Kaplan, Almog, Perry and Wettstein1989, Diessel Reference Diessel1999, Roberts Reference Roberts2003, Wolter Reference Wolter2004, Oshima & McCready Reference Oshima and McCready2017), and those uses in fact interact with the restrictive/appositive distinction in Mandarin RCs. Once this previously unnoticed factor is taken into proper account, the new generalisation that emerges is that there is no strict one-to-one correlation between restrictive/appositive status and the structural position of an RC, a conclusion that goes against the widely held consensus in the literature (cf. Chao Reference Chao1968, Huang Reference Huang1982, Lin Reference Lin2003, Constant Reference Constant2011, Lin & Tsai Reference Lin, Tsai, Li, Simpson and Tsai2015, among others). Instead, a more adequate way of characterising the restrictive/appositive distinction in Mandarin, we argue, should make reference to whether the co-occurring demonstrative is used deictically or anaphorically. Based on the new empirical findings, a formal explanation will then be offered which – for the first time – makes sense of the external distribution of the two types of Mandarin RC.
This paper is structured as follows. Section 2 outlines some necessary background on Mandarin RCs, while noting the highly inconclusive findings about the restrictive/appositive distinction from previous studies. Section 3 demonstrates how the distinct uses of demonstratives bear on the restrictive/appositive status of a Mandarin RC, and offers diagnostics for teasing apart restrictives and appositives (drawing on Constant Reference Constant2011, Lin & Tsai Reference Lin, Tsai, Li, Simpson and Tsai2015) that remain valid in the face of various confounds. In Section 4, we present an account of a curious ‘gap’ that we observe with the distribution of restrictive relatives. Section 5 summarises the main findings and highlights the broader implications of the present study.
2. Background on Mandarin RCs and the restrictive/appositive issue
Despite being a VO language, Mandarin has head-external prenominal RCs. Moreover, Mandarin RCs contain no relative pronounFootnote 1 but the invariant modificational marker de, and are always prosodically integrated with the relativised Head (see Chao Reference Chao1968, Li & Thompson Reference Li and Thompson1981 for classic descriptions). A canonical example of a Mandarin RC is given in (2):
Mandarin RCs may also co-occur with demonstratives, which is extremely common in the language. In such cases, they either immediately precede the demonstrative (as in (3)) or immediately follow the classifier (as in (4)). For expository ease, we will refer to the two types of RC as the ‘pre-Dem RC’ and the ‘post-Dem RC’, respectively.
Speaking of the restrictive/appositive distinction, the received wisdom in the literature is that the distinction directly correlates with the structural position of the RC in Mandarin (Chao Reference Chao1968, Huang Reference Huang1982, Lin Reference Lin2003, Constant Reference Constant2011, Lin & Tsai Reference Lin, Tsai, Li, Simpson and Tsai2015, among many others), although authors disagree greatly on what that correlation should be. More specifically, one view holds that pre-Dem RCs are restrictive and post-Dem RCs are appositive (Chao Reference Chao1968, Huang Reference Huang1982), whereas another view holds the exact opposite, i.e. that pre-Dem RCs are appositive and post-Dem RCs are restrictive (Tsai Reference Tsai1994, Lü Reference Lü and Shengshu1999). Some authors even hold that all Mandarin RCs are restrictive (Zhang Reference Zhang2001a, Shi Reference Shi2010), although Lin (Reference Lin2003) holds that this is true unless the relativised Head is a proper name, in which case the RC is necessarily appositive.
While the general absence of relative pronouns and the unavailability of prosodic cues in Mandarin certainly both play a role in explaining the mixed nature of the existing findings, there is, in fact, an equally important reason as well. Cinque (Reference Cinque2006, Reference Cinque2008) points out that there are, in fact, two types of appositive: integrated and non-integrated.
As it turns out, many of the standard tests for identifying appositive RCs (as in Jackendoff Reference Jackendoff1977, Potts Reference Potts2005) only single out a sub-type of appositives. To illustrate, consider the following test:
The intended logic behind the test, clearly, is that non-nominal phrases may be modified by appositives but not restrictives. In this respect, the following Mandarin example is sharply ungrammatical:
It would be premature to conclude from (8) that Mandarin lacks appositives, however. As (9a) shows, integrated appositives of the che/cui type do not permit non-DP anchors either:
Hence, the standard tests are useful to the extent that they distinguish between restrictives and non-integrated appositives (see also Constant Reference Constant2011, Lin & Tsai Reference Lin, Tsai, Li, Simpson and Tsai2015 for relevant discussion). Against this background, Del Gobbo (Reference Del Gobbo2010, Reference Del Gobbo2017) argues that Mandarin appositives are all of the integrated type, similar to che/cui-appositives in Italian. Because Mandarin appositives are integrated, it follows that they will pass or fail such tests exactly as restrictives do, which potentially explains why some studies (e.g. Zhang Reference Zhang2001a, Shi Reference Shi2010) would conclude that Mandarin RCs lack the restrictive/appositive distinction.
In order to ascertain whether the restrictive/appositive distinction truly exists in Mandarin, it is clear that tests that tap more into the semantic properties of the two types of RC will be more helpful for such a purpose. To foreshadow our results, the restrictive/appositive distinction does reveal itself once the right tests are applied. Before we attempt such a demonstration, however, let us first point out a previously unnoticed confound that would impact the results of the relevant diagnostics.
3. Diagnosing restrictives and appositives
3.1. Two uses of demonstrative
The potential confound, which has not yet been documented in the literature, concerns the fact that there are at least two uses of demonstrative in language (Diessel Reference Diessel1999, Roberts Reference Roberts2003, Wolter Reference Wolter2004, Oshima & McCready Reference Oshima and McCready2017). In its deictic (or exophoric) use, the demonstrative is accompanied by extralinguistic demonstration in physical space in some discourse situation. When uttering (10), for instance, the speaker points to an individual (as represented by the symbol ‘
’) close to or distant from them, and the demonstrative-containing nominal expression refers to that particular individual.
In its anaphoric use, on the other hand, the demonstrative-containing nominal expression refers back to a nominal expression in the preceding linguistic discourse. Following standard practice, we indicate coreferential relations with coindexation:
Importantly, the two uses of demonstrative interact with relative types in Mandarin, as we will now demonstrate in the next subsection.
3.2. The interaction of demonstratives with RCs
Among the various tests proposed in the RC literature, we apply three specific tests that we believe are effective for distinguishing between restrictives and appositives in the case of Mandarin, drawing on Constant (Reference Constant2011) and Lin and Tsai (Reference Lin, Tsai, Li, Simpson and Tsai2015). We should note that the novelty here does not lie in the tests themselves, but in the demonstration that different uses of demonstrative potentially affect the outcomes of such tests. To the extent that we ultimately arrive at a coherent picture, of course, the current results will not only constitute evidence for the restrictive/appositive distinction in Mandarin but also further establish the validity of the tests in question.Footnote 2
3.2.1. Diagnosing restrictive RCs
The first test is called undeniability. This test capitalises on the fact that appositives encode not-at-issue content, which usually cannot be suspended with so-called ‘epistemic riders’, i.e. intensional operators that may affect the speaker’s commitment towards the truth of the prejacent (Potts Reference Potts2005, Constant Reference Constant2011). For instance, the content of an appositive in the then-consequent usually cannot be embedded in the if-antecedent:
This is because the speaker should be committed to the appositive RC inference at the global level, the certainty of which, however, is exactly what the if-antecedents in (12) are trying to question.
By contrast, a restrictive relative, together with the definite article in the relativised Head, contributes the presupposition that there exists a uniquely identifiable entity of which the property denoted by the restrictive holds (Hooper & Thompson Reference Hooper and Thompson1973, Comrie Reference Comrie1989), and this presupposition can be filtered by the if-antecedent:
Hence, if an RC passes the test (i.e. if its content can be suspended with epistemic riders), it follows that it can be restrictive. If an RC does not pass the test, then it follows that it can only be appositive but not restrictive.
Before applying the current test to Mandarin, we should address one potential concern raised by a JL referee. The referee points out that appositive content is not always scopeless relative to an intensional operator (see also Amaral et al. Reference Amaral, Roberts and Smith2007, Oshima Reference Oshima2016), and there exist examples where the speaker is not committed to appositive content:
It is indeed too strong to claim that appositive content is always speaker-oriented and scopeless relative to intensional operators. However, Harris & Potts (Reference Harris and Potts2009) rightly note that when there is no extra context that facilitates perspective shifting, appositive content is overwhelmingly speaker-oriented. Importantly, Mandarin examples like (15), which involve appositive RCs as they modify proper names (see Lin Reference Lin2003, Constant Reference Constant2011), indeed sound degraded, if not completely odd:
Thus, the undeniability test should not be considered ineffective in Mandarin. As just shown in (15), it is typically difficult to suspend appositive content with epistemic riders.
Let us now apply the current undeniability test to Mandarin. First, we consider RCs with a deictic demonstrative as in (16). In order to force the deictic use, each utterance containing a demonstrative is accompanied with a pointing gesture directed at an individual within the speaker’s proximal physical space. Without presupposing that the RC in question is a restrictive or an appositive, we offer in (16a) and (16b) what the relevant RC inference should be under each relative type.
Interestingly, when the RC content is questioned with a if-antecedent as in (17), only the then-consequent with a post-Dem RC is felicitous, but the one with a pre-Dem RC is not:
One might wonder whether the if-antecedent should be of the form ‘if a student indeed won’ as in English examples with restrictives like (13) above. Such a form is not adopted because a deictic demonstrative identifies the relevant individual directly with a pointing gesture, which is thus incompatible with the use of an indefinite. Moreover, the current form ‘if this student indeed won’ as in (17) likewise allows for questioning the RC content – it is clear that the same student is discussed in both cases, so that if there exists a uniquely identifiable student being pointed at who won the game called
$ m $
, then an if-antecedent of the form ‘if the student being pointed at indeed won’ is indeed questioning whether the property expressed by the RC holds of
$ m $
. The following English contrast further confirms the validity of our current adaptation:
Now turning to the anaphoric use of demonstrative, both continuations in (19a) and (19b) turn out to be felicitous:
The following test – which forms an even more minimal pair with (17) – ensures that the different ways of introducing the referent of which the RC property holds in the antecedent do not constitute a confound:
The results in (17) and (19)–(20) thus indicate that when the demonstrative is used deictically, a pre-Dem RC can only be appositive as it is not deniable (thus failing the test), while a post-Dem RC can be restrictive. In contrast, when the demonstrative is used anaphorically, both pre-Dem and post-Dem RCs can be restrictive.
The second test is called anti-backgrounding. Given that appositives (and parentheticals more generally) typically introduce new information (Potts Reference Potts2005), redundancy will obtain if their content is backgrounded. By contrast, restrictives are not expected to exhibit any anti-backgrounding effect, as their content is presupposed.Footnote 4 The following pair illustrates:
Now, results from the anti-backgrounding test are consistent with the previous results from the undeniability test:
Again, the following test is included to ensure that the way of introducing the antecedent is not a confound (recall (20)):
We thus have converging evidence that in Mandarin, post-Dem but not pre-Dem RCs can be restrictive under the deictic use of demonstrative, whereas both types of RC can be restrictive under the anaphoric use. To put it in another way, pre-Dem and post-Dem RCs can both be restrictive, unless when the co-occurring demonstrative is used deictically, in which case the pre-Dem RC can only be appositive.
A JL referee notes that the contrast between (17a) and (19a), as well as that between (22a) and (23a), may instead be attributed to the requirement that the content of an appositive RC help the addressee identify the referent in a non-redundant manner. Accordingly, the referee suggests that pre-Dem RCs are necessarily appositive, and (17a) and (22a) are infelicitous because the pointing gesture makes the referent clearly identifiable from the addressee’s viewpoint, such that the use of appositive RCs is redundant. In contrast, (19a) and (23a) are felicitous because the use of appositive RCs can emphasise that the student referred to in the antecedent clause is identical to the one mentioned in the consequent clause, and hence its use is not redundant.
Interesting as the current suggestion is, it does not explain why the use of appositive RCs in (25) should lead to degradation, for they do potentially help identify the referent in a non-redundant manner.
By contrast, when restrictive RCs are involved, the above examples then become perfectly natural:
For this reason, we consider the current tests to be effective, and hence the pre-Dem RCs in (19a) and (23a) are restrictive rather than appositive.
3.2.2. Diagnosing appositive RCs
Insofar as the reasoning goes, however, positive results from the two tests above only indicate the availability of a restrictive parse and constitute no evidence that an appositive parse is also available (see Constant Reference Constant2011 for a similar point). As a response to this issue, we introduce and perform the final test based on compatibility with root-level adverbs (Emonds Reference Emonds1979). As (27) exemplifies, appositives but not restrictives may host root-level adverbs like frankly and fortunately:
As the test results in (28)–(29) show, both pre-Dem and post-Dem RCs can be appositive regardless of demonstrative use:
One might wonder, however, whether Mandarin root-level adverbs like laoshishuo can simply appear in any type of RC. Although it is not clear to us why one should depart from the null hypothesis, there is good reason to assume that they are indeed incompatible with restrictive RCs. In contrastive contexts that force RCs to be restrictive, for instance, the occurrence of such adverbs is impossible:
This reaffirms the validity of the root-level adverb test.
3.3. Interim summary
To summarise the foregoing results, a Mandarin RC that co-occurs with a demonstrative can always be appositive. But this in no way implies that the restrictive/appositive distinction does not exist: a Mandarin RC in similar configurations can also be restrictive, although restrictive relatives cannot precede a demonstrative when it is deictic. We thus arrive at a novel typology of Mandarin RCs, as illustrated in Table 1.Footnote 5
A formal typology of Mandarin RCs
4. Towards an explanation
The current typology clearly raises the question of why the distribution of Mandarin restrictive relatives should be restricted the way it is. In this section, we develop a formal account of such a restriction, building on ideas from existing works on demonstratives (Elbourne Reference Elbourne2005, Schwarz Reference Schwarz2009, Jenks Reference Jenks2018). Ultimately, the restriction will follow from the grammatical options available for semantic composition under the deictic and anaphoric uses of demonstratives as well as the different structural positions of the RC.
4.1. Background assumptions
Before presenting our account, let us first make explicit our assumptions about the syntax and semantics of Mandarin nominals and relatives, taking the following example as an illustration:
Setting aside the RC for now, we adopt the standard view that the demonstrative, numeral, classifier, and noun each heads its own projection arranged in a specific order (Tang Reference Tang1990, Li Reference Li1998, Reference Li1999, Huang, Li & Li Reference Huang, Li and Li2009), such that the basic structure of the noun phrase in (31) is as follows:
Following Trinh (Reference Trinh and Reich2011), Jenks (Reference Jenks2018), and others, we assume a semantic model whose domain includes both atomic and plural individuals (Link Reference Link and Bäuerle1983; Schwarzschild Reference Schwarzschild1996), as well as kind-level atomic and plural individuals (Dayal Reference Dayal2004). Accordingly, nouns denote a set of atomic and plural individuals, of token- and kind-level types, and both the token-level classifier
$ {\mathrm{Cl}}_{\mathrm{token}} $
(e.g. wei) and the kind-level classifier
$ {\mathrm{Cl}}_{\mathrm{kind}} $
(e.g. zhong) further restrict that set into a set of atomic token-level individuals and atomic kind-level individuals, respectively, as (33) illustrates:
Notice that the numeral yi ‘one’ in (31) is optional:
For the sake of brevity, we will omit the numeral yi as well as ignore its semantic contribution in what follows, as this does not affect the gist of our account to be presented below. Strictly speaking, such a numeral meaning remains present in examples like (34), which means that Mandarin has a covert Num0 yi (as argued in Zhang Reference Zhang2019) with the denotation shown in (33e) that we omit from our subsequent representations.Footnote 6
Turning to the RC, we follow Aoun and Li (Reference Aoun and Li2003) in assuming that Mandarin relativisation does not involve a complementation structure, and adopt for concreteness the so-called ‘matching’ analysis of relatives (Chomsky Reference Chomsky, Culicover, Wasow and Akmajian1977) involving null-operator movement.Footnote 7 Semantically, the relative CP denotes a property and composes with the relativised Head via Predicate Modification:
4.2. The core account
We can now proceed to the core of our account. Following Elbourne (Reference Elbourne2005), Schwarz (Reference Schwarz2009), Jenks (Reference Jenks2018), and others, we assume that demonstratives differ from the (uniqueness-based) definite article in being strong definites. Semantically, this means that the demonstrative will take an additional (semantic) argument as its domain restriction. Furthermore, we follow Hanink (Reference Hanink2021a) in assuming that both the proximal and the distal demonstratives introduce the presupposition that the referent picked out by the strong definite (under the deictic use) or the linguistic antecedent (under the anaphoric use) is proximal or distal (see also Roberts Reference Roberts, van Deemter and Kibble2002 and Elbourne Reference Elbourne2008), as shown belowFootnote 8:
Under the deictic use, the choice between proximal and distal demonstratives reflects the relative physical distance between the speaker and the referent. By contrast, under the anaphoric use, the exact distinction between ‘proximal’ and ‘distal’ is more complicated and we tentatively suggest that the proximal zhe implies speaker-privileged familiarity with the referent whereas the distal na implies shared familiarity between speaker and hearer (following Jenks Reference Jenks2018:509, fn. 8).Footnote 9
Structurally speaking, the first argument slot is saturated by ClP, whereas the second one is saturated by the specifier of DP:
We propose that an important distinction underlies the deictic use and anaphoric use of demonstratives. In the former case, where the demonstrative is used with a pointing gesture
$ \delta $
, the gesture obligatorily contributes an index argument in Spec-DP; see (39). Specifically, the gesture contributes a referential index
$ i $
, which picks out a referent
$ g(i) $
via the assignment function. This type
$ e $
denotation then type-shifts to an
$ \left\langle e,t\right\rangle $
property ‘
$ \lambda x.x=g(i) $
’ via IDENT, so as to compose with the semantics of D′.
By contrast, in the latter case where the demonstrative is used anaphorically, Spec-DP can be occupied by either a referential index (which is the same index as its linguistic antecedent and thus contributes an indexical property via IDENT) or phrases like adjectives or relatives as in (40), as long as the properties they denote can satisfy the presupposition associated with the
$ \iota $
-closure as shown in (37).Footnote 10 As pointed out by a JL referee, the idea that relatives and the index may compete for the same structural position has been proposed in Simonenko (Reference Simonenko2015), based on the fact that restrictive relatives systematically render DPs with demonstratives scope-sensitive (see also Simonenko Reference Simonenko2014). Jenks (Reference Jenks2018) also notes that high modifiers can fulfill the semantic function of a referential index (which can be taken to be associated with a null pronominal) for anaphoric definites in his account.
The current proposal thus has the following main consequence for pre-Dem RCs in Mandarin: when the demonstrative is used deictically, the pre-Dem RC can only be introduced via adjunction but not specification (as Spec-DP is already occupied by
$ {\delta}_i $
); when the demonstrative is used anaphorically, the pre-Dem RC can be introduced via adjunction or specification:
4.2.1. Accounting for the properties of pre-Dem RCs
Under our account, the phrasal status of a pre-Dem RC will directly determine its semantic status. When an RC adjoins to a type
$ e $
DP, the relative (being type
$ et $
) cannot further restrict its sister, and thus the two expressions cannot be composed via the regular Function Application.
Instead, we propose that the pre-Dem RC in this case is a kind of appositive that contributes a conventional implicature (CI). Following Potts (Reference Potts2005), we analyse CIs that arise from appositives, parentheticals, and expressives under a multidimensional semantic framework called ‘a logic for conventional implicatures’ (
$ {\mathrm{L}}_{\mathrm{CI}} $
; see also Karttunen and Peters Reference Karttunen, Peters, Oh and Dinneen1979). We distinguish between at-issue types and CI types by superscripts on types, which are defined as follows:
The type-driven composition rules are adjusted accordingly in Potts’ system. In particular, two expressions can be composed via the standard At-issue (Function) Application as in (44) or CI Application as in (45). In the latter case, an expression
$ \alpha $
with a CI type composes with
$ \beta $
’s denotation at a dimension independent of the at-issue dimension, while at the at-issue dimension,
$ \beta $
’s denotation is simply passed on to the mother node. The at-issue application and CI application are separated by a dot between them.
We propose that the pre-Dem RC in (46) is semantically a modifier of the CI type
$ \left\langle {e}^a,{t}^c\right\rangle $
, as the following derivation illustratesFootnote 11:
Because the content of the RC in this case cannot contribute to the at-issue dimension, as shown above, it follows that pre-Dem RCs that co-occur with a deictic demonstrative can only be appositive.
Turning to the anaphoric use of demonstratives, when the domain restriction argument of the demonstrative is saturated by a referential index, the pre-Dem RC must likewise adjoin to DP and will, for the exact same reasons, be appositive:
On the other hand, a pre-Dem RC may also be introduced via specification, as long as the property denoted by the RC can uniquely pick out a student with this property, thus satisfying the presupposition imposed by the ι-operator in the semantics of the demonstrative.Footnote 12 When that happens, the RC can saturate the domain restriction argument of the demonstrative and thus contributes to the at-issue dimension, as (48) illustrates. Hence, the availability of this option explains why pre-Dem RCs that co-occur with an anaphoric demonstrative can also be restrictive.
Our proposal therefore successfully derives the fact that pre-Dem RCs that co-occur with a deictic demonstrative can only be appositive, whereas pre-Dem RCs that co-occur with an anaphoric demonstrative can be appositive or restrictive.
4.2.2. Accounting for the properties of post-Dem RCs
The current proposal also straightforwardly explains the typology observed with post-Dem RCs as well. Unlike pre-Dem RCs, post-Dem RCs are necessarily introduced in a lower structural position, and therefore cannot saturate the extra argument slot of the demonstrative. Moreover, the deictic and anaphoric uses of demonstratives – in our account – differ in how that argument slot is semantically saturated, at which point the post-Dem RC will have been interpreted and whose semantic status, as such, cannot possibly be affected by such a late step of composition:
The absence of asymmetry between restrictives and appositives in the case of post-Dem RCs under the two uses of the demonstrative thus follows, and the full typology of Mandarin RCs is thereby derived.
Before we end this section, let us address the additional question of how restrictive and appositive post-Dem RCs can be derived. In the case of restrictives, the post-Dem RC adjoins to NP as a modifier, and the properties denoted by the RC and the NP are interpreted intersectively. Because ClP will saturate the first semantic argument of the demonstrative, and ClP contains the RC, the semantics of the relative will necessarily contribute to the at-issue dimension:
The derivation in (50) applies to both the deictic and anaphoric uses of a demonstrative (which is why the exact content of Spec-DP is left underspecified). Under the deictic use, Spec-DP is occupied by an index contributed by the gesture, whereas under the anaphoric use, Spec-DP can be occupied by either a referential index or another high modifier.Footnote 13
In the case of appositives, we follow Constant (Reference Constant2011) in assuming that appositives can be propositional, with a contextual variable in their semantics, such that CI application is triggered. The contextual variable takes the closest referent, i.e. the individual denoted by the DP. The closed proposition resulted then composes with the nominal anchor via ‘the rule of isolated CIs’ (Potts Reference Potts2005:66, ex. (3.41)), in which the CI material remains completely separate from the at-issue content. This way, post-Dem appositive RCs are derived as well:
Either way, it should be clear that both types of relative are compatible with whatever the three dots (‘…’) in the two derivations above will turn out to be. In other words, our account leads to the expectation that a post-Dem RC can be interpreted as a restrictive or an appositive regardless of whether the demonstrative is used deictically or anaphorically – exactly as desired.
5. Conclusion
In this paper, we revisited the classic issue of whether Mandarin relatives exhibit the restrictive/appositive distinction, and if so, how the two types of relative can be clearly distinguished – an issue that remained controversial despite more than 50 years of inquiry (since Chao Reference Chao1968). Our findings indicate that the distinction exists and that different uses of the demonstrative constitute a non-trivial factor in establishing the restrictive/appositive status of a Mandarin RC that appears in different structural positions, a factor that had not been previously noticed. Contrary to the common view that the restrictive/appositive status of a Mandarin RC directly correlates with its structural position, we noted that there is no strict one-to-one correlation once the potential confound of demonstratives is taken into account. In particular, the pre-/post-Dem position matters only if the demonstrative is used deictically, which blocks restrictive RCs in pre-Dem position; when the demonstrative is used anaphorically, the pre-/post-Dem position does not correlate with the restrictive/appositive distinction in any way. Differently put, there is a structural position that is exclusively appositive, but not one that is exclusively restrictive, for Mandarin RCs. The discovery of the current confound potentially explains why conflicting claims about the correlation have been reported in the vast body of existing literature. Moreover, we proposed a formal explanation for the typology of the two types of Mandarin RCs. By adapting the treatment of strong definites (Elbourne Reference Elbourne2005, Reference Elbourne2008, Schwarz Reference Schwarz2009, Jenks Reference Jenks2018, Hanink Reference Hanink2021a) to Mandarin demonstratives, we derive the observed typology directly from the different options available for saturating the domain restriction argument of the demonstrative depending on whether the RC appears in a pre-Dem or a post-Dem position, as well as whether the demonstrative is used deictically or anaphorically.
As a JL referee correctly notes, the present paper clarifies the restrictive/appositive distinction in Mandarin insofar as RCs that modify a demonstrative-containing nominal are concerned. In fact, Mandarin RCs may also occur in different positions in nominals without a demonstrative:
It may be worth further investigating the nature of the restrictive/appositive distinction in such cases as well.
The current study has at least two further implications that are worth highlighting. The first implication is that the present conclusion furnishes an argument against relating pre-Dem and post-Dem RCs derivationally (as has been previously proposed or suggested for Mandarin; see Cinque Reference Cinque2009, Constant Reference Constant2011, Zhang Reference Zhang2015), as schematised below:
In this derivation, semantic composition is successful, and the derivation converges. Hence, if pre-Dem RCs and post-Dem RCs were derivationally related, we would then expect a pre-Dem RC to allow for a restrictive reading when a deictic demonstrative is involved, just as in (53). Accordingly, we would fail to derive the typology shown in section 3.3.Footnote 14
Another implication is that the proposal extends previous analyses of strong definites and demonstrates that extralinguistic devices like pointing gestures actively interact with linguistic structure in Mandarin, a result that should bear on future studies on demonstratives more broadly. While the existing studies often analyse deictic and anaphoric demonstratives on a par such that they both involve some kind of index argument in the structure (Elbourne Reference Elbourne2008, Hanink Reference Hanink2021a, among others), the present study showed that the kind of index contributed by a gesture under the deictic use takes precedence over restricting modifiers like relative clauses as the specifier of a strong definite. In contrast, the referential index involved in the anaphoric use does not have such a privileged status and competes with restricting modifiers – a result that is in line with Simonenko (Reference Simonenko2014, Reference Simonenko2015). The proposed account thus contributes to our understanding of the linguistic properties of non-linguistic components in communication in general (Schlenker Reference Schlenker2018, Reference Schlenker2023, Esipova Reference Esipova2019, Reference Esipova2021).
Acknowledgements
We thank the editors and the three anonymous referees for their constructive comments. The usual disclaimers apply.
Open access
