Next Article in Journal
Phonation Variation as a Function of Checked Syllables and Prosodic Boundaries
Next Article in Special Issue
Parallel Corpus Research and Target Language Representativeness: The Contrastive, Typological, and Translation Mining Traditions
Previous Article in Journal
Preposition Allomorphy in Calabrian Greek (Greko) and Standard Modern Greek and Its Theoretical Implications
Previous Article in Special Issue
Perfect-Perfective Variation across Spanish Dialects: A Parallel-Corpus Study
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Time Reference in Mandarin Relative Clauses

1
Department of Chinese Studies, Université de Picardie Jules Verne, CERCLL UR 4283, 80000 Amiens, France
2
Department of Linguistics, Nantes Université/CNRS, LLING UMR 6310, 44000 Nantes, France
*
Author to whom correspondence should be addressed.
Languages 2022, 7(3), 170; https://doi.org/10.3390/languages7030170
Submission received: 2 October 2021 / Revised: 8 June 2022 / Accepted: 12 June 2022 / Published: 5 July 2022
(This article belongs to the Special Issue Tense and Aspect Across Languages)

Abstract

:
In this paper, we investigate constraints on the time reference of embedded clauses in Mandarin. We show that while English past-tensed embedded clauses disallow later-than-matrix readings in intensional contexts on a de dicto construal, Mandarin relative clauses with bare predicates yield temporally free readings across the board. We argue that the contrast between the temporal interpretations of bare embedded clauses in Mandarin vs. past-tensed embedded clauses in English is not due to a putative contrast between ‘tenseless’ languages (as Mandarin is traditionally assumed to be) and ‘tensed’ languages such as English. Mandarin is indeed not tenseless, but rather has a covert Non-Future tense, restricting the reference time of bare sentences to non-future times. Moreover, Mandarin superficially tenseless embedded clauses with overt—be it perfect, perfective, durative/progressive—aspectual marking do not allow later-than-matrix readings on a de dicto construal, just like tensed embedded clauses in English. We conclude that the freedom of interpretation of bare embedded clauses in Mandarin cannot be imputed to null semantically underspecified tense, but rather to null semantically underspecified aspect. Our analysis provides, to our knowledge, the first arguments for Non-Future tense in embedded contexts.

1. Introduction

In this paper, we examine the temporal interpretation of embedded clauses in Mandarin, a language with no overt grammatical tense (Li and Thompson 1981; Klein et al. 2000; Lin 2006 among many others). We put forth the following generalizations:
  • Mandarin bare Relative Clauses (RCs), that is, RCs with aspectually unmarked predicates, yield temporally free and, in particular, later-than-matrix readings1 even in intensional contexts.
  • Mandarin RCs with overt aspectual marking—be it perfect, perfective, or progressive/durative aspect—do not allow free (later-than-matrix) readings in intensional contexts on a de dicto construal, just like tensed embedded clauses in English.
The generalization in (ii), but not that in (i), conforms to the descriptive constraint known as the Upper Limit Constraint (ULC, Abusch 1994), according to which the tense of the embedding clause is an upper bound on the tense of an embedded clause.
As first pointed out by Abusch (1988), the temporal construal of RCs in intensional contexts correlates with whether the DP containing the RC is interpreted de re or de dicto: the later-than-matrix reading of a past-tensed RC is possible if the DP is read de re, but banned when the DP is read de dicto (Section 2), and the ULC is thus a constraint on intensional contexts:
(1)‘The local evaluation time is the upper limit for reference of tenses’.
The question, then, is what is the source of the striking contrast between the temporal interpretations of Mandarin bare RCs and English (past)-tensed embedded clauses in intensional contexts: why are later-than-matrix readings available for Mandarin bare embedded clauses, but not for English past-tensed embedded clauses? We argue that this contrast is not due to a contrast between grammatically ‘tensed’ vs. ‘tenseless’ languages, because Mandarin embedded clauses with overt aspect—be it perfect(ive) or imperfective/durative aspect—behave just like English past/present tensed RCs: they disallow later-than-matrix readings on a de dicto construal.2 Indeed, we refute the assumption that Mandarin is a tenseless language, a view defended by Li and Thompson (1981), Gōng (1991), Klein et al. (2000), Mei (2002), and Lin (2006), among many others. We show that Mandarin is only superficially tenseless (as also argued by Sybesma 2007; Tsai 2008; Lin 2015). That is, Mandarin has a silent Non-Future tense restricting the reference time of bare sentences to non-future times, as first proposed in Sun (2014) for independent clauses (see also Huang 2015; Chen and Husband 2018). We extend Sun’s hypothesis to embedded clauses which thus by hypothesis contain a null Non-Future tense selecting for times that precede or overlap the local evaluation time.
We then argue that the contrast between aspectually marked RCs (which do not allow temporally free readings in intensional contexts) and aspectually bare RCs which do, follows straightforwardly on the assumption that the latter contain a null semantically underspecified aspect, as a consequence of which the temporal ordering of the matrix and embedded event times is left undetermined. This is ultimately the source of later-than-matrix construals.
This paper is organized as follows. Section 2 presents the temporal construals of English relative clauses and Abusch’s Upper Limit Constraint. Section 3 and Section 4 discuss the time reference of Mandarin relative clauses without vs. with overt aspectual marking, and gives an overview of the availability of later-than-matrix readings in Mandarin vs. English RCs. Section 5 presents the Non-Future tense hypothesis defended by Sun (2014) for Mandarin root clauses, extending the analysis to Mandarin embedded clauses in Section 6. We show how the temporal interpretation of aspectually marked vs. unmarked RCs is derived on the Non-Future Tense hypothesis, together with the assumption that Mandarin has a silent underspecified aspect. Section 7 assesses the arguments (the first, to the best of our knowledge) provided here for Non-Future tense in embedded contexts, while offering two further arguments.

2. English Relative Clauses

The temporal interpretation of English relative clauses has been extensively investigated (Abusch 1988; Ladusaw 1977; Ogihara 1996; Stowell 1993, 2007 and references therein). In extensional contexts, English RCs yield temporally free construals (Enç 1986). In (2), the Chronos speech event is temporally ordered relative to the Utterance Time (UT) by the past tense in the RC, but remains unordered relative to the matrix past meeting event: the past speech event can either be backward-shifted ((3a)) or forward-shifted ((3b)) with regard to the past meeting event.
(2)John met a professor who gave a speech at Chronos.
(3)a.---[SPEECH---]---[MEETING---]---[UT--]--->earlier-than-matrix
b.---[MEETING---]---[SPEECH---]---[UT--]--->later-than-matrix
However, in intensional contexts, the later-than-matrix reading is available only if the DP (containing the RC) is interpreted de re (Abusch 1988). Consider (4). If the DP headed by ‘a professor’ is interpreted de re, the sentence conveys that there is a specific professor, say, Prof. Smith, that John was looking for, and Prof. Smith may have given a speech at Chronos before ((5a)) or after ((5b)) John’s looking for her. If the DP is interpreted de dicto, (4) conveys that John was looking for any professor who had given a speech at Chronos. The speech time must be earlier than the matrix searching event, so only the past-shifted reading in (5a) is available.
(4)John was looking for a professor who gave a speech at Chronos.
(5)a.---[SPEECH---]---[LOOKING FOR---]---[UT--]--->earlier-than-matrix
b.---[LOOKING FOR---]---[SPEECH---]---[UT--]--->*later-than-matrix
Why is the later-than-matrix reading not available for past-tensed RCs on a de dicto construal? This is where Abusch’s (1994) Upper Limit Constraint comes in. This constraint states that the tense of the embedding clause is an upper bound on the tense of an embedded clause ((1)) and, as such, rules out later-than-matrix readings of a RC, unless the relativized DP raises out of the scope of the matrix tense. The de re vs. de dicto contrast observed in the interpretation of RCs has thus been taken to provide evidence for a scopal analysis of free interpretations of RCs in English. Later-than-matrix readings are allowed on a de re construal because the DP containing the RC scopes above the matrix past tense. Forward-shifted readings are banned on a de dicto construal because the DP containing the RC remains in situ, and Abusch’s ULC then excludes the forward-shifted reading of the embedded past relative to the matrix past tense. This proposal correctly correlates the distribution of temporally independent readings with noun phrase interpretation (see Stowell 1993; Ogihara 1996; Kusumoto 2005 for further discussion).

3. Mandarin Bare Relative Clauses

We now turn to the temporal interpretation of Mandarin relative clauses, a language with no overt grammatical tense markers, focusing in this section on bare RCs, that is, RCs which also lack overt aspectual marking. We consider in turn two classes of bare RCs, depending on whether the main predicate of the RC denotes an event or a state since, as we shall see, the temporal construal of states appears to be more constrained in the absence of temporal adverbs setting the context.

3.1. Relative Clauses with Bare Eventives

Sun (2015) shows that Mandarin RCs with bare eventives allow temporally free readings in both extensional and intensional contexts, even on a de dicto construal. In an extensional context, such as (6), the dancing event described by the RC is temporally free: it can either precede, coincide with, or follow, be it either the matrix event time or the UT.
(6)Lùlupāishè-guo[RC yí-getiàobāléiwǔdenǚháir].
Lùlufilm-PFTone-CldanceballetDEgirl.
‘Lùlu filmed a girl who danced/is dancing/will dance ballet.’
In an intensional context, such as (7), the RC allows temporal construals where the dancing event either precedes ((8a)), coincides with ((8b)), or follows ((8c)) the searching event of the matrix clause. Suppose that the ballet show started at 8 p.m. and ended at 9:30 the night before the day where (7) is uttered. Now, when Kǎi meets Lùlu at 8:30 at the show, Lùlu was looking for a girl who at some point performed in the show. Lùlu could be looking for any girl who had already performed ((8a)), or who was dancing when Lùlu was looking for her ((8b)), or who was supposed to dance later in the show ((8c)).
(7)Context: Kǎi met Lùlu last night during a ballet show. He is now telling Méi what Lùlu was doing when he met her.
Dāngshí,Lùluzàizhǎo[RC yí-getiàobāléiwǔdenǚháir.]
at.that.timeLùluPROGlook-forone-CldanceballetDEgirl.
‘At that moment, Lùlu was looking for a girl who had danced/was dancing/would dance ballet.3
(Adapted from Sun 2015)
(8)a.---[DANCING---]---[LOOKING FOR---]---[UT--]--->earlier-than-matrix
b.---[DANCING-----[LOOKING FOR---]--]---[UT--]--->simultaneous4
c.--[LOOKING FOR--]—([UT--])--[DANCING--]--([UT--])-->later-than-matrix
Example (9), likewise, allows the temporally free construals in (10) and, in particular, the later-than-matrix construal in (10b) where, say, last year in May, Lǐsì wanted to marry any woman who would win the marathon in July, suggesting that there is no constraint on the temporal ordering of the matrix event time relative to the embedded event time even on a de dicto construal.
(9)Qùnián,Lǐsìxiǎng[RC yí-geyíngmǎlāsōngdenǚzǐ].
last.yearLǐsìwantmarryone-ClwinmarathonDEwoman
‘Last year, Lǐsì wanted to marry a woman who had won/would win the marathon.’
(10)a.---[WINNING--]---[WANTING---]--[UT--]--> earlier-than-matrix
b.---[WANTING--]--([UT--])-[WINNING--]--([UT--])--> later-than-matrix
We have established that bare eventive RCs allow later-than-matrix readings in both extensional and intensional contexts, and even on a de dicto reading. We now show that this generalization extends to the temporal construal of stative bare RCs.

3.2. Relative Clauses with Bare Statives

In extensional contexts and in the absence of time adverbs, bare stative predicates in RCs do not show the freedom of temporal interpretation that bare eventive predicates show (Sun 2015): the event described by the embedded clause in, say, (11), is understood to coincide with either the utterance time or the matrix event time. Example (11) thus conveys that Lùlu interviewed a painter who lives in Beijing at the utterance time ((12a)), or who had lived in Beijing at the moment of the interview ((12b)).
(11)Lùlucǎifǎng-guoyí-gezàiBěijīngshēnghuódehuàjiā.
Lùluinterview-PFTone-ClatBeijingliveDEpainter
‘Lùlu interviewed a painter who lives/lived in Beijing.’
(Adapted from Sun 2015)
(12)a.---[INTERVIEW---]----[LIVING----[UT--]---]--->earlier-than-matrix
b.---[LIVING---[INTERVIEW---]--(])--[UT--]--(])--> simultaneous
We are not warranted, however, to conclude that bare statives in RCs only allow simultaneous construals, where the event time of the RC coincides with either the utterance time (12a) or the matrix event time (12b). We can indeed show that later-than-matrix readings are in fact available by adding a temporal adverb to the sentence, which serves to temporally anchor the time of the event described by the bare stative embedded in the RC. The state time of the RC will thus coincide with the time denoted by the adverb, whether this time precedes, follows, or overlaps the matrix event time. Crucially, if the time denoted by the adverb is itself a time in the future relative to the matrix event time, a later-than-matrix reading arises, be it in an extensional or intensional context.
To illustrate, consider the intensional context in (13). Example (13) allows temporal construals where the state described by dāi zài bālí ‘stay in Paris’ either precedes ((14a)) or follows ((14b)) the event of Lùlu’s looking for a girl even on a de dicto construal. In other words, the temporal adverb qiūtiān ‘fall’ can refer either to the fall preceding the matrix event time or to the fall following the matrix event time. Thus, suppose that (13) is uttered in August 2020. The time adverbial qiūtiān ‘fall’ can refer to the fall of either 2019 or 2020.5,6
(13)Shànggeyuè,Lùluzài zhǎo [RC yí-geqiūtiāndāizài
last.month LùluPROGlook.forone-Clfallstayat
bālídenǚháir].
ParisDEgirl
‘Last month, Lùlu was looking for a girl who stayed/would stay in Paris in the fall.’
✓fall 2019, ✓fall 2020
(14)a.---[STAYING---]---[LOOKING FOR---]---[UT--]---> earlier-than-matrix
b.---[LOOKING FOR--]—([UT--])--[STAYING--]--([UT--])--> later-than-matrix
We conclude that bare stative and eventive RCs pattern alike in that they both allow temporally free readings relative to the matrix event time—be it in extensional or intensional contexts. It is only that it is harder to show with statives because they need to be temporally anchored to a reference time. This reference time, however, need not be the matrix event time, but rather can be provided by a temporal adverbial.7

3.3. Mandarin vs. English Relative Clauses

Let us recapitulate our findings. While the ULC holds for English past-tensed RCs that are read de dicto, it does not hold for Mandarin RCs with bare (be it eventive or stative) predicates on a de dicto construal: the later-than-matrix reading is available for Mandarin RCs with bare predicates.
What, then, is the source of this striking contrast in temporal interpretation between Mandarin and English RCs? Why can the former but not the latter appear to violate the ULC (Abusch 1994)? Could we impute this contrast to the fact that English is a tensed language, whereas Mandarin lacks overt tense? We contend that this cannot be the case, since, as we shall now show, the ULC is enforced in Mandarin intensional contexts when the RC is overtly marked with aspect (be it perfect(ive) or imperfective). We will conclude that the source of the difference in interpretation of Mandarin vs. English RCs is not surface tenselessness, but rather surface aspectlessness.

4. Mandarin Relative Clauses with Overt Aspect

We discuss below the temporal construals of aspectually marked RCs, distinguishing in turn for types of aspectual marking in the RC: perfect, perfective, progressive, and durative.8

4.1. Relative Clauses with Overt Perfect Aspect

Recall our example (7), repeated below as (15a). We saw that the bare eventive predicate in the RC allows the temporally free readings in (8). Now, if we add the overt perfect marker guo9 to the predicate tiào bāléiwǔ ‘dance ballet’ in the RC, as in (15b), the later-than-matrix construal which was available for (15a) on a de dicto construal of the relativized DP is no longer available: the dancing event must have occurred prior to the matrix searching event, as schematized in (16a). In other words, (15b) does not allow a later-than-matrix reading ((16c)), nor a simultaneous reading of the RC relative to the matrix past, where the searching and the dancing temporally overlap ((16b)).
(15)a.Shàngzhōu,Lùluzàizhǎo[RC yí-getiàobāléiwǔdenǚháir.]
last.weekLùluPROGlook-forone-CldanceballetDEgirl.
‘Last week, Lùlu was looking for a girl who had danced/was dancing/would dance ballet.’
b.Shàngzhōu,Lùluzàizhǎo[RC yí-getiào-guobāléiwǔ
last.weekLùluPROGlook-forone-Cldance-PFTballet
denǚháir.]
DEgirl.
‘Last week, Lùlu was looking for a girl who had danced ballet.’
(16)a.[DANCING--]--[LOOKING FOR--]--[UT--]--> earlier-than-matrix
b.[DANCING--[LOOKING FOR--]--]--[UT--]-->*simultaneous
c.[LOOKING FOR--]--[DANCING--]--> *later-than-matrix
Mandarin stative RCs with overt perfect aspect, just like Mandarin eventive RCs, only allow earlier-than-matrix readings on a de dicto construal. Example (17) below differs from (13) in two respects10: morphologically, the stative predicate in (17) is marked with perfect aspect guo, while in (13), it is bare; semantically, the event described by the RC in (17) is not temporally free, unlike (13), which allows either of the temporal readings in (14), repeated in (18). Example (17) conveys that Lùlu was looking for any girl who had stayed in Paris, but her stay in Paris is understood to precede Lùlu’s searching time. Thus, (18a), but not (18b), illustrates correctly the only temporal ordering conveyed by (17). Suppose that (17) is uttered in August 2020; the adverb qiūtiān ‘fall’ can only be interpreted as a fall earlier than August 2020 and, therefore, cannot refer to the fall of 2020.
(17)Shànggeyuè,Lùluzàizhǎo[RC yí-ge(qiūtiān)zàibālí
last.monthLùluPROGlook.forone-ClfallatParis
dāi-guodenǚháir.
stay-PFTDEgirl.
‘Last month, Lùlu was looking for a girl who had stayed in Paris (in the fall).’
✓fall 2019, ✗ fall 2020
(18)a.---[STAYING---]---[LOOKING FOR---]---[UT--]---> earlier-than-matrix
b.---[LOOKING FOR--]—([UT--])--[STAYING--]--([UT--])--> *later-than-matrix
In sum, on a de dicto reading, Mandarin RCs overtly marked by perfect aspect only allow earlier-than-matrix readings. Temporally free readings are no longer available.

4.2. Relative Clauses with Overt Perfective Aspect

Generally, to convey a past reading of an eventive predicate, a bare predicate is the most commonly used form in Mandarin RCs. Perfective aspect le is used in the RC when the speaker seeks to emphasize the fact that the described event has already taken place.11
Imagine that the latest Harry Potter movie was released last Wednesday, and for her TV report, Lùlu, a journalist, interviews some movie fans in a cinema hall. Suppose Lùlu is interested in both the expectations of those who have not seen it yet and the impressions of those who already have. The sentence in (19a) with the bare predicate kàn hālǐ bōtè ‘watch Harry Potter’ perfectly fits this scenario, and can thus be used to convey that Lùlu was looking for any Harry Potter movie fan, where the time of watching the movie could be earlier or later than the searching time.
In a similar scenario, if Lùlu is only interested in the impressions of the spectators, one can utter (19b) with perfective le to emphasize that Lùlu is looking for a fan who would have just seen the movie, and whose recollection of the movie is therefore still fresh. Now, in the very same scenario, should the speaker not wish to emphasize this requirement, she would use (19a) without le.
(19)Context: A Harry Potter movie was released last Wednesday. Lùlu was looking for a spectator to interview.
a.(Shàng-zhōusān,)Lùluzàizhǎo[RC yí-gekàn
last-WednesdayLùluPROGlook-forone-Clwatch
hālǐ bōtèdeyǐng-mí.]
HarryPotterDEmovie-fan.
‘(Last Wednesday,) Lùlu was looking for a movie fan who had watched/was watching/would watch Harry Potter.’
b.(Shàng-zhōusān,)Lùluzàizhǎo[RC yí-gekàn-le
last-WednesdayLùluPROGlook-forone-Clwatch-PFV
hālǐbōtèdeyǐng-mí.]
HarryPotterDEmovie-fan.
‘(Last Wednesday,) Lùlu was looking for a movie fan who had seen Harry Potter.’
If perfective le in (19b) is replaced by the perfect marker guo, as in (20), the watching event could have taken place in a relatively distant past with regard to Lùlu’s searching time. Thus, (20) could be used to convey that Lùlu was looking for a movie fan who had already seen, say, the first Harry Potter movie years earlier.
(20)Shàng-zhōusān,Lùluzàizhǎo[RC yí-gekàn-guo
last-WednesdayLùluPROGlook-forone-Clwatch-PFT
hālǐbōtèdeyǐng-mí.]
HarryPotterDEmovie-fan.
‘Last Wednesday, Lùlu was looking for a movie fan who had seen Harry Potter.’
It is important to point out that RCs with either le (e.g., (19b)) or guo ((20)) require the embedded (here watching) event to be completed before the matrix (searching) event, which itself serves as the evaluation time of the RC (since we are dealing here with an intensional context). (This anteriority requirement relative to the evaluation time (that is, the matrix searching time) will be accounted for by the semantics given for guo ((49)) and le ((53)) in Section 6.)
A few comments on the semantics of these particles are thus in order here, since a precedence relation is typically associated with perfect aspect, but not with perfective aspect (Kratzer 1998). Many authors explicitly state that le in Mandarin encodes anteriority to some extent. Both Lin (2003, 2006) and Sun (2014) explicitly write a precedence relation into the semantics of le, requiring either the topic time (Klein 1994) (‘t2’ in (21)) or the running time of the event (‘τ(e)’ in (22)) to be earlier than the evaluation time.
(21)le〛 = λPλt2λe∃e′[P(e) ∧ P(e′) ∧ e′ ≤E e ∧ τ(e′) ⊆ t2 ∧ t2 < τ(epro)]
12 (Lin 2003)
(22)leg,c = λP<v,t>. λt’.λt.∃e[P(e) = 1 ∧ t’ ⊇ τ(e) ∧ τ(e) < t](Sun 2014)
Now, although both the perfect marker guo in (20) and the perfective marker le in (19b) shift the time reference of their RC to the past of the matrix event time, they do not yield the same range of temporal interpretations when combined with the same predicate. While guo always marks anteriority, le can give rise to a continuous reading (Lin 2000, 2003; Jin 2002), as shown in (23b).13
(23)a.Mòmoyǎng-guo yì-zhīmāo
Mòmoraise-PFTone-Clcat
‘Mòmo raised a cat.’
b.Mòmoyǎng-le yì-zhīmāo
Mòmoraise-PFVone-Clcat
‘Mòmo is raising a cat.’
In other words, although Mandarin le is commonly referred to as a perfective marker, it cannot be defined as encoding a simple inclusion relation between the running time of an eventuality and the topic time (τ(e) ⊆ tTop), in the sense of Klein (1994) and Kratzer (1996), since le encodes not only perfectivity/completion but also imperfectivity/continuity (Lin 2003). According to Lin’s (2003) definition, in (21), le requires the inclusion of a subpart of the event time within the topic time, itself in turn constrained to precede the evaluation time. In other words, the event denoted by the embedded clause is required to have been at least partially realized at the evaluation time. This explains why some RCs with le yield simultaneous reading with regard to the matrix event time, as is the case in (24b), where the event of raising a cat is understood as having begun at a time earlier than the beginning of the searching event, and is still ongoing at the searching time (as illustrated in (25b)). In contrast, with perfect guo, the raising event described by the RC in (24a) must be located entirely in the past of the event time of searching, as illustrated in (25a).
(24)a.Shàngzhōu,Lùluzàizhǎo[RC yí-geyǎng-guoyì-zhī
last.weekLùluPROGlook-forone-Clraise-PFTone-Cl
māodenǚháir.]
catDEgirl
‘Last week, Lùlu was looking for a girl who had raised a cat.’
b.Shàngzhōu,Lùluzàizhǎo[RC yí-geyǎng-leyì-zhī
last.weekLùluPROGlook-forone-Clraise-PFVone-Cl
māodenǚháir.]
catDEgirl
‘Last week, Lùlu was looking for a girl who was raising a cat.’
(25)a.---[RAISING---]---[LOOKING FOR---]---[UT--]---> earlier-than-matrix
b.---[RAISING---[LOOKING FOR---]---(])---[UT--]-(])--> simultaneous
We do not take a stand here on whether Lin’s analysis of le is the most accurate one, since the semantics of le is a far-reaching and controversial issue. What is of importance for the purpose of this paper is to elucidate the interpretational contrasts between bare and aspectual marked RCs. The crucial generalization so far is that RCs with overt either perfect guo or perfective le do not allow than later-than-matrix readings on a de dicto construal.
Summarizing, there are two correlated differences between the RCs in (15a) vs. (15b) on the one hand, and (19a) vs. (19b) on the other: (i) morphologically, the verbs in the RC in (15b) and (19b) are overtly marked by perfect or perfective aspect; (ii) semantically, the RC in (15b) and (19b) no longer allow temporally free construals when read de dicto—the dancing/watching must have occurred prior to the matrix searching event, thus excluding forward-shifted readings of the RC relative to the matrix event time. Note that when the RC containing le receives a continuous reading, such as in (24b), the event time denoted by the RC must crucially have begun before the matrix event time and a later-than-matrix reading is thus also excluded. In other words, whether le encodes completion or continuity, a later-than-matrix reading where the event time of the RC is entirely in the future of the event time of the matrix is unavailable.

4.3. Eventive Relative Clauses with Overt Progressive Aspect

We now turn to eventive predicates with the overt progressive marker zài. RCs marked by progressive zài yield simultaneous readings, and take as topic time a time that is salient (either a time denoted by an adverb or another contextually salient time) (Sun 2015). On a de dicto reading, (26b) conveys that Lùlu was looking for any girl who, at the moment of searching, was dancing ballet. That is, the time of the dancing event denoted by the RC with a progressive verb form must coincide with the matrix event time, as shown in (27a). The later-than-matrix reading where the dancing event occurs after the searching event ((27c)) is thus not available.
(26)Context: When Kǎi met Lùlu at 9 last night during a ballet show, Lùlu uttered (a). At the present time of speech, Kǎi utters (b) to tell Méi what Lùlu was doing when he met her last night.
a.[Lùlu:] zàizhǎo[RC yí-gezàitiàobāléiwǔdenǚháir.]
1SGPROGlook-forone-ClPROGdanceballetDEgirl.
‘I’m looking for a girl who is dancing ballet.’
b.(Dāngshí),Lùluzàizhǎo[RC yí-gezàitiàobāléiwǔdenǚháir.]
at.that.timeLùluPROGlook-forone-ClPROGdanceballetDEgirl.
‘(At that moment,) Lùlu was looking for a girl who was dancing ballet.’
*’(At that moment,) Lùlu was looking for a girl who would be dancing ballet.’
(Adapted from Sun 2015)
(27)a.–[DANCING---[LOOKING FOR--]--]---[UT--]--> simultaneous
b.--[DANCING--]---[LOOKING FOR--]--[UT--]--> earlier-than-matrix
c.–[LOOKING FOR--]--[DANCING--]--> later-than-matrix
Note that in the absence of an overt time adverbial in (26b), the event time of the RC is understood as coinciding with the matrix event time of searching. If a past time adverb such as liǎng-ge xiǎoshí zhīqián ‘two hours earlier’ is added to the RC as in (28b), the time denoted by the adverb will serve as the topic time for the progressive RC. The dancing event will then be understood as in progress/ongoing two hours prior to the searching event. Example (28b) thus allows the earlier-than-matrix reading schematized in (27b).
(28)a.[Lùlu: ]zàizhǎo[RC yí-geliǎng-gexiǎoshí zhīqián/
1SGPROGlook-forone-Cltwo-Clhour before
qī-diǎn zàitiàobāléiwǔ denǚháir.]
seven-o’clockPROGdance balletDEgirl.
‘I’m looking for a girl who was dancing ballet two hours ago/at seven.’
b.[Kǎi: ](Dāngshí),Lùluzàizhǎo[RC yí-geliǎng-gexiǎoshí
at.that.timeLùluPROGlook-forone-Cltwo-Clhour
zhīqián zàitiàobāléiwǔ denǚháir.]
before PROGdanceballetDEgirl.
‘(At that moment,) Lùlu was looking for a girl who was dancing ballet two hours earlier.’
Crucially, when a progressive RC is read de dicto, future time adverbs fail to shift the time reference to a time later than the matrix event time. Example (29), for instance, does not mean that Lùlu, at 9 p.m., was looking for any girl who would be dancing ballet at 11 p.m. Example (29) only allows a de re reading, namely, where there is a girl who was dancing ballet at 11 p.m., and Lùlu was looking for her at 9 p.m.
(29)[Kǎi: ](Dāngshí),Lùluzài zhǎo[RC yí-geliǎng-gexiǎoshí
at.that.timeLùluPROGlook-forone-Cltwo-Clhour
zhīhòu zàitiàobāléiwǔdenǚháir.]
after PROGdanceballetDEgirl.
‘(At that moment,) Lùlu was looking for a girl who was dancing ballet two hours later.’
In sum, eventive RCs in Mandarin marked with progressive aspect, just like RCs marked with either perfect or perfective aspect, do not allow later-than-matrix readings when the RC is read de dicto in intensional contexts.

4.4. Relative Clauses with Overt Durative Aspect

The finding for RCs with the progressive aspect marker zài carries over to the durative marker zhe, which is another imperfective marker (Yeh 1993; Smith 1991; Pan 1996; Djamouri and Paul 2018; a.o.). Generally, zhe modifies atelic predicates (Lin 2006). Smith (1991) argues that zhe differs from zài in that while the former has a static meaning, the latter has a dynamic meaning. Example (30a) describes a state of Mòmo wearing blue glasses. Example (30b) focuses on the state of the window being closed, as opposed to the closing event, and (30c) is almost equivalent to an existential clause even if it contains the eventive verb guà ‘hang’.14
(30)a.Mòmodàizheyí-fùlánsèdeyǎnjìng.
MòmowearDURone-pairblueDEglass
‘Mòmo is wearing a pair of blue glasses.’
b.Chuānghùguānzhe.
windowcloseDUR
‘The window is closed.’
c.Qiángshàngguàzheyí-fúhuà’er.
wallonhangDURone-Clpainting
‘There is a painting on the wall.’
When zhe is present in a RC read de dicto, such as in (31b), it yields a simultaneous stative reading: the state of wearing blue glasses must overlap the time of searching ((32a)). The later-than-matrix reading illustrated in (32c) is not available for (31b). The corresponding bare RC in (31a) allows earlier ((32b)) and later ((32c)) than the matrix readings, in addition to the simultaneous (and most salient) reading ((32a)). Suppose that Mòmo and Tíngting are actresses who do not wear glasses in their daily life, but will have to wear special blue glasses to perform in a play. Before the play begins, Lùlu is looking for any of the actresses who will wear these blue glasses in the play. One can use (31a) but not (31b) to report this scenario. In other words, bare RCs but not RCs with durative zhe allow later-than-matrix readings.
(31)a.(Dāngshí),Lùluzàizhǎo[RC yí-gedài lánsèyǎnjìng
at.that.timeLùluPROGlook.forone-Clwearblueglasses
denǚháir.
DEgirl.
‘At that moment, Lùlu was looking for a girl who was wearing/wore/would wear/wears blue glasses.’
b.(Dāngshí),Lùluzàizhǎo[RC yí-gedàizhelánsè
at.that.timeLùluPROGlook.forone-ClwearDURblue
yǎnjìngdenǚháir.
glasses DEgirl.
‘At that moment, Lùlu was looking for a girl who was wearing blue glasses.’
(32)a.–-[WEARING---[LOOKING FOR--]--]---[UT--]--> simultaneous
b.--[WEARING–]--[LOOKING FOR--]----[UT--]-->earlier-than-matrix
c.--[LOOKING FOR--]--([UT-])--[WEARING--]--([UT-])-->later-than-matrix
To summarize, in the absence of overt aspect, Mandarin RCs allow temporally free (earlier-than-matrix, simultaneous, and later-than-matrix) readings with regard to the matrix event time, and this is the case even on a de dicto construal in intensional contexts. In contrast, in the presence of overt aspect, they have restricted temporal interpretations depending on the semantics of the aspectual marker: (i) while RCs with perfect guo only allow earlier-than-matrix readings, (ii) RCs with perfective le, progressive zài, or durative zhe allow either earlier-than-matrix or simultaneous readings.
We conclude that Mandarin RCs with overt aspect, unlike bare RCs, but just like English tensed RCs, do not allow later-than-matrix readings on a de dicto construal, in keeping with the ULC.

4.5. Recapitulating

Table 1 below recapitulates the availability of later-than-matrix readings in Mandarin vs. English embedded clauses. While in extensional contexts, both English and Mandarin embedded clauses allow later-than-matrix readings, in intensional contexts, the later-than-matrix reading is not available in the presence of an overt past tense (English) or of an overt aspect (Mandarin), on a de dicto construal. Embedded clauses in Mandarin with neither overt tense or aspect marking allow later-than-matrix readings, in violation of the ULC.
The obvious question is then why, in intensional contexts, do Mandarin RCs with bare predicates allow later-than-matrix readings on a de dicto reading, unlike Mandarin RCs with aspectually marked predicates.
One might seek to ascribe this freedom of temporal interpretation to the status of Mandarin as presumably a tenseless language, as claimed by Li and Thompson (1981), Klein et al. (2000), Smith and Erbaugh (2005), and Lin (2003, 2006). This hypothesis is not tenable since, as we shall now show, Mandarin does indeed have tense.

5. Non-Future Tense in Mandarin

Following Kratzer (1998 a.o.), we assume that tenses are pronouns that denote time intervals. Like pronouns, they are variables that receive their value from a contextually determined assignment function. Tense morphemes thus introduce a time variable which, following Klein (1994), we call the topic time (henceforth, TopT)—that is, the time for which an assertion is made, together with a restriction on its possible values. In particular, past tense introduces the condition that the TopT precede the evaluation time, and present tense introduces the condition that it coincide with the evaluation time. While the evaluation time of a root clause uttered out of the blue is the utterance time, in subordinate contexts, it may also be the matrix event or attitude time.
Sun (2014) argues for syntactic and semantic tense in Mandarin. Her specific claim is that Mandarin has a T(ense) projection hosting a covert Non-Future tense (NF in (33)), which restricts the TopT of bare sentences to the non-future times, as shown in (33). This hypothesis is largely inspired by Matthewson’s (2006) silent TENSE hypothesis for St’át’imcets.15
(33)〚 NF 〛g,c<i, ii> = λr. λt: tr. t(where r is the evaluation time and t the TopT)
(Adapted from Sun 2014)
Sun’s Non-Future tense hypothesis explains why future time adverbs fail to fix the temporal reference of Mandarin bare sentences, as illustrated in (34) with the bare predicate hěn jǔsàng ‘very frustrated’.16 Examples (34a) and (34b) are well formed, yielding a past and a present reading, respectively, because both dāngshí ‘at that time’ in (34a) and xiànzài ‘now’ in (34b) denote non-future intervals that constrain the time reference of their clause to (respectively) precede or overlap the utterance time, thus falling in the domain of Non-Future tense. Crucially, (34c), on the other hand, is not felicitous because the time interval denoted by míngtiān-shídiǎn ‘tomorrow at 10 o’clock’ lies entirely after the utterance time and, as such, does not fall in the domain of Non-Future tense.
(34)a.Context: Kǎi met Lùlu yesterday and he was very surprised because usually she is very calm, but…
DāngshíLùluhěnjǔsàng.
at.that.timeLùluveryfrustrated
‘At that moment, Lùlu was very frustrated.’
b.XiànzàiLùluhěnjǔsàng.
nowLùluveryfrustrated
‘Now, Lùlu is very frustrated.’
c.Context: Lùlu always gets very frustrated when she receives her transcript with her grades. The transcript of the semester will be delivered tomorrow morning.
*Míngtiānshí-diǎnLùluhěnjǔsàng.
tomorrowten-o’clockLùluveryfrustrated
Intended: ‘Tomorrow at 10 o’clock, Lùlu will be very frustrated.’
In (35) and (36) below, we give a simplified derivation of the semantics of (34a) (full derivations will be given in Section 6, once we introduce aspect). We assume that temporal frame adverbs such as dāngshí ‘at that time’, xiànzài ‘now’, and míngtiān-shídiǎn ‘tomorrow’ denote properties of times (Kamp and Reyle 1993; Demirdache and Uribe-Etxebarria 2004) and, as such, can be base-generated adjoined to the AdjP in (35), also a predicate of times, with which the adverb thus combines via Predicate Modification (Heim and Kratzer 1998).17 Example (34a) is well-formed on this derivation since the TopT coincides with a past time which in the context is a subinterval of the day before the utterance time (tu in (35) below) at which Kǎi met Lùlu, and as such, falls in the domain of Non-Future tense.
(35)
Languages 07 00170 i001
(36)
a.〚 NF 〛g,c = λr. λt: tr. t
b.dāngshíg,c = λt. tThat Moment (where K met L)
c.AdjPLùlu hěn jǔsàngg,c = λt. FRUSTRATED (L)(t)
d.dāngshí Lùlu hěn jǔsàngg,c = λt.tThat Moment & FRUSTRATED (L)(t)
e. ti NF tug,c = g(i); defined only if g(i) ≤ tu.
f.〚 TP 〛g,c = 1 iff g(i)That Moment & FRUSTRATED (L)(g(i));
defined only if g(i) ≤ tu.
Further evidence for Non-Future tense comes from bare sentences such as (37), from Sun (2014, p. 205), inspired by Matthewson (2006). If uttered in 2021, (37) conveys that Gǔlóng, a Chinese novelist who passed away in 1985, used to smoke, and Mòyán, a writer who is alive in 2021, is a smoker. The smoking habits of these two individuals are true for different time intervals. The one and only main predicate of the sentence (and, by hypothesis, the one and only syntactic tense (TP) projection) in (37) thus yields past and present readings simultaneously. Example (37) can only be translated to English with a biclausal sentence ((37a)), because the temporal information encoded in (37) cannot be conveyed with a single tense in English. Only a tense that can select simultaneously for both past and present intervals could fit the truth conditions of the covert tense heading TP, and this is exactly how Non-Future is defined (cf. (33)).
(37)GǔlóngMòyándōuchōuyān.
Gǔlóng andMòyán DOUsmoke
a.‘Gǔlóng used to smoke and Mòyán smokes.’
b.*‘Gǔlóng used to smoke/smokes and Mòyán will smoke.’
Importantly, (37) uttered in 2021 cannot be used to convey simultaneously a past state of Gǔlóng’s smoking and a future state of Mòyán’s smoking ((37b)). Now, suppose (37) were uttered in 1970, at a time when Gǔlóng was still alive. Example (37) could not be used either to convey a present ongoing state of Gǔlóng’s smoking and a future state of Mòyán’s smoking. The unavailability of the readings in (37b) is correctly predicted under the Non-Future Tense Hypothesis since Non-Future cannot range over an interval that falls after the utterance time.
Consider now the contrast between (38) vs. (37) on the one hand, as well as (38) vs. (39) on the other hand. Note that (38), if uttered, say, in March 2021, can be used to convey simultaneously a past state of Lùlu staying in Paris in the summer of 2020, and a future-oriented state of Lùlu staying in Paris in the summer of 2021. Why does this construal involving past and future-oriented eventualities, unavailable in (37), become available in (38)? Because the single bare predicate dāi zài bālí ‘stay in Paris’ in (40) can be used to describe simultaneously a past state/stay in Paris and a future state/stay in Paris that is being planned at the utterance time—that is, (38) can be used to report that Lùlu stayed in Paris last summer and is planning to stay in Paris next summer (of this year). Non-Future is thus ranging over a past state and a present plan for a future state. In contrast, its English counterpart in (39) with a single tensed predicate (be it past or future) is not grammatical.
(38)QùniánxiàtiānjīnniánxiàtiānLùlu(dōu)dāizàibālí.
last.yearsummerandthis.yearsummerLùluDOUstayatParis
‘Lùlu stayed in Paris last summer and will stay in Paris this summer.’
(39)*Lùlu stayed/will stay in Paris last summer and next summer.
We turn now to the contrast between (38) vs. (37) again, vs. this time that between (38) vs. (40) below, where a future-oriented reading is here unavailable. The generalization is that a Mandarin bare predicate can simultaneously describe past and future-oriented states only if the forthcoming state can be understood as a futurate—that is, as a future eventuality that is planned at the utterance time (Copley 2008; Sun 2014; Thomas 2015). As Copley (2002, 2008) emphasizes, bare states yield futurate readings only if the described eventuality is plannable or under the control of some natural force. Since neither Mòyán’s smoking ((37)) nor Lùlu’s being nervous during the exam ((40)) can readily be construed as planned eventualities, simultaneous past and futurate readings are unavailable in (37) and (40), even when the future time adverb míngtiān ‘tomorrow’ is present.
(40)*ZuótiānmíngtiānkǎoshìshíLùlu(dōu)hěnjǐnzhāng.
yesterdayandtomorrowexamtimeLùluDOUverynervous
Intended: ‘Lùlu was nervous yesterday during the exam, and will be nervous tomorrow during the exam.’
We close this section by pointing out that the Non-Future tense hypothesis has also been defended for Mandarin by Huang (2015), who argues for a syntactic Non-Future tense on the basis of the distribution of the morpheme jiāng, a future ‘tense’ morpheme according to Huang, as well as by Chen and Husband (2018), who provide experimental evidence for a Non-Future tense in Mandarin.

6. Deriving Embedded Time Reference: The Distribution of Later-Than-Matrix Readings

Let us now go back to the question of why later-than-matrix readings are available for Mandarin bare RCs, but not for English (past-)tensed RCs. This contrast between English and Mandarin cannot be imputed to a putative difference between grammatically ‘tensed’ vs. ‘tenseless’ languages since, as we have just established, Mandarin is not a tenseless language: Mandarin sentences that appear on the surface to not have items dedicated to expressing tense are in fact underlyingly tensed since Mandarin has a silent tense and, moreover, this tense, although more underspecified than tense in, say, English, is not semantically underspecified since it restricts the TopT to non-future times with regard to the local evaluation time.

6.1. (Covert) Aspect

Now, although the freedom of interpretation of bare embedded clauses in Mandarin cannot be imputed to a null semantically underspecified tense, since Mandarin has a covert Non-Future tense, it can, however, be imputed to a null semantically underspecified aspect. We thus contend that the contrast between aspectually marked RCs which do not allow temporally free readings in intensional contexts in Mandarin and aspectually bare RCs which do, follows straightforwardly on the assumption that the latter contains a null semantically underspecified aspect, as a consequence of which the temporal ordering of the matrix and embedded event times will ultimately be left undetermined, as we shall see in Section 6.3.
We illustrate in (41) below our assumptions about how tense and aspect combine compositionally with the main predicate (VP) of their sentence. Both tense and aspect each combine with two temporal arguments (as can be seen in the lexical entries given for Non-Future tense in (33) above, and (41) for Mandarin (im)perfect(ive) aspects in (43) below), both of which are projected in the syntax, as shown in (41).18
(41)
Languages 07 00170 i002
Tense imposes a condition on the relation between the TopT and the evaluation time, which is the utterance time by default in an independent clause, though in subordinate contexts, it may be bound by the matrix event or attitude time (as we shall see in (48), (52), (56), and (61) below). The tense node takes as sister an aspect projection (AspP), which denotes a property of times, <i, st>. Following Kratzer (1998), aspect crucially serves to map its sister node VP, which itself denotes a property of eventualities, <v, st>, onto a property of times so that the VP can ultimately combine with tense, as shown in (41). On Kratzer’s proposal, the aspect head mediates between events and times by introducing a running time function, and in so doing can impose conditions on the relation between the event time and the TopT. These restrictions are illustrated with the denotations given below for the perfect, which constrains the event to be over by the TopT, and the imperfective, which constrains the TopT to be included in the (running) time of the event (Kratzer 1998, p. 107).
(42)a.Perfect:λP<v,<s t>>.λti.λws.∃e1 (time(e) < t & P(e)(w) = 1)
b.Imperfective:λP<v,<s t>>.λti.λws.∃e1 (t ⊆ time(e) & P(e)(w) = 1)
We break with the assumption encoded in (42) that the event time is existentially bound in the lexical entry of the aspect heads, but rather assume that aspect (like tense) combines with two temporal arguments (as shown in (43) with the denotation of perfect aspect guo and progressive zài), both of which are projected in the syntax. The event time will be existentially bound in the syntax (as shown in the tree in, e.g., (48), where the event time is t5).
(43)a.Perfect: 〚guog,c = λp<v,st>. λti. λui: t < u. λws. ∃e[p(e)(w) & τ(e) = t]
b.Progressive: 〚zàig,c = λp<v,st>. λti. λui: ut. λws. ∃e[p(e)(w) & τ(e) = t]
On this proposal, aspect, just like tense, combines with two temporal arguments projected in the syntax. While tense imposes a condition on the relation of the TopT relative to the evaluation time, aspect imposes a condition on the relation of the TopT relative to the running time of the event (τ(e) in (43)). This specific implementation thus captures Klein’s seminal proposal that aspect relates the event time to the TopT, which tense in turn relates to another time (that is, the local evaluation time).
Given these assumptions, AspP is always projected—that is, is always present, even in sentences with no overt aspectual marker appearing on the surface—since it serves to mediate between eventualities and times by introducing, via a running time function, the eventuality time argument. This means that alongside the overt aspectual heads in (43), Mandarin must also have a covert aspectual head whose denotation is given in (44).
(44)Underspecified covert aspect
a.Asp°∅〛g,c = λp<v,st>. λti. λui: tu or t > u. λws. ∃e[p(e)(w) & τ(e) = t]
b.Asp°∅〛g,c = λp<v,st>. λti. λui. λws. ∃e[p(e)(w) & τ(e) = t]
(where t is the event time and u the TopT)
The silent aspectual morpheme in (44), just like the overt aspectual morphemes in (43), combines with two temporal arguments, the TopT itself introduced by tense and the event time introduced via a running time function, and, as such, serves to map properties of eventualities onto properties of times. Crucially, however, the silent aspectual morpheme in (44), unlike the overt aspectual morphemes in (43), imposes no restriction on the relation between the TopT and the running time of the event. We give two different implementations of this hypothesis in (44). According to the lexical entry given in (44a), the relation holding between these two time intervals can be interpreted as either precedence, overlap, or subsequence, while according to the lexical entry in (44b), there is simply no condition holding on the temporal ordering between these two times.19 As we shall see, the hypothesis that the temporal ordering of the event time relative to the TopT is free, undetermined, is ultimately the source of the later-than-matrix construals that RCs allow when they are aspectually unmarked/bare on the surface.
On this proposal, both tense and aspect are always projected in Mandarin since Mandarin has a silent Non-Future tense present in all finite clauses, as well as a silent unspecified aspect present whenever there is no overt aspectual head. In what follows, we show how these two assumptions derive the interpretative differences in the temporal construals of Mandarin bare vs. aspectually marked RCs, without having to stipulate a constraint such as the ULC. We conclude that the source of these interpretative contrasts lies ultimately not in surface tenselessness or underspecified/undetermined tense since Mandarin has a silent Non-Future tense, but rather in surface aspectlessness and underspecified/undetermined aspect since Mandarin has a silent radically underspecified aspect.

6.2. Intensional Contexts with Overt Aspect: Lack of Later-Than-Matrix Readings

6.2.1. RCs with Overt Perfect Aspect guo

In an intensional context, when the DP containing a perfect RC is embedded in a matrix clause describing a past event, as is the case in (45), the time of the event described by the RC is constrained to precede the matrix event time on a de dicto construal. That is, Lùlu in (45) was looking for a girl who had already danced ballet before the searching time. The later-than-matrix reading where the dancing event happens after the searching time is not available. We now show how the temporal readings of (45), and the lack of the later-than-matrix readings, automatically fall out from the assumptions put forth in Section 6.1.
(45)Context: Kǎi met Lùlu last week. Kǎi is now telling Méi what Lùlu was doing when he met her.
Dāngshí, Lùluzài zhǎo yí-ge(liǎng-zhōu qián)
at.that.time LùluPROGlook-for one-Cl(two-week before)
tiào-guo bāléiwǔdenǚháir.
dance-PFTballet DEgirl
‘At that moment, Lùlu was looking for a girl who had danced ballet (two weeks ago/earlier).’
In an intensional context, on a de dicto reading, the RC is anchored into the matrix via the binding of the embedded evaluation time (tRC evaluation in (46)/the lower t2 in (48)) by the matrix event time (tmatrix event/the higher t2 in (48)). As shown in (48a), Non-Future tense in the RC restricts its TopT (tRC top in (46)/t3 in (48)) to times that are non-future relative to the evaluation time, which itself coincides with the matrix event time (via binding). As shown in (46b), perfect aspect (guo), in turn, restricts the event time of the RC (tRC event in (46)/t5 in (48)) to times earlier than the TopT in the RC. Consequently, tRC event is constrained to always precede tmatrix event, as illustrated in (49a,b), and a later-than-matrix reading is therefore excluded. Mandarin perfect aspect thus plays a role similar to the English past tense in embedded clauses.
(46)a.NF tense in the RC:tRC top ≤ tRC evaluation (= tmatrix event)
b.PFT aspect:tRC event < tRC top
(47)----[tRC event]------[tmatrix event]--->→ ✗ later-than-matrix
(48)
Languages 07 00170 i003
(49)Relative clause with guo:
a.Asp° guog,c = λp<v,st>. λti. λui: t < u. λws. ∃e[p(e)(w) & τ(e) = t]
b.VP tiào bāléiwǔg,c = λxe. λev. λws. DANCE BALLET(x)(e)(w)
c. nǚháirg,c = λxe. λti. λws. GIRL(x)(t)(w)
d. t3 NF t2g,c = g(3); defined only if g(3) ≤ g(2)
e.〚∃t〛g,c = λp<i,st>. λws. there is an interval u such that p(u) = 1
f.AspP t5 guo [x4 tiào bāléiwǔ]g,c = λui: g(5) < u. λws. ∃e[DANCE BALLET(g(4))(e)(w) & τ(e) = g(5)]
g.〚TP〛g,c = λws. ∃e[DANCE BALLET(g(4))(e)(w) & τ(e) = g(5)]; defined only if g(5) < g(3), and g(3) ≤ g(2)
h.〚CP〛g,c = λxe. λti: tg(2). λws. ∃e[DANCE BALLET(x)(e)(w)] & ∃u[ u < t & τ(e) = u]
i.〚NP〛g,c = λxe. λti: tg(2). λws. GIRL(x)(t)(w) & ∃e[DANCE BALLET(x)(e)(w) & ∃u[u < t & τ(e) = u]]
The syntactic structure of (45) is given in (48), and its semantic derivation in (49) and (50). As (49g) shows, perfect aspect guo places the time of dancing g(5) in the past of the TopT g(3) (that is, g(5) < g(3)), and NF checks whether the TopT g(3) is a non-future interval with regard to the evaluation time g(2) (that is, g(3) ≤ g(2)). The evaluation time g(2) is itself identified with the matrix event time (hosted in the specifier of the matrix AspP in (48)) via binding. It thus follows that the RC dancing time g(5) is indirectly ordered in the past relative to matrix searching time g(2) (since g(5) < g(3) ≤ g(2)), and this is why RCs with perfect aspect fail to yield later-than-matrix readings. The temporal interpretation of (45) is thus correctly predicted.
(50)Matrix clause:
a.zàig,c = λp<v,st>. λti. λui: ut. λws. ∃e[p(e)(w) & τ(e) = t]
b.dāngshíg,c= λti. tThat Moment (when K met L)
c. zhǎog,c = λp<e,i,st>. λye. λev. λws. ∀〈t’, w’〉 [REACH ONE’S GOAL(y)(e)(w)(t’)(w’) → ∃ze[p(z)(t’)(w’) = 1 & FIND (z)(y)(t’)(w’)
d.〚(45)〛g,c = λws. g(1) ⊆ That Moment & ∃tiev [ τ(e) = t & g(1) ⊆ t] & ∀〈t’, w’〉 [REACH ONE’S GOAL(L)(e)(w)(t’)(w’) → ∃zefv [GIRL(z)(t’)(w’) & DANCE BALLET(z)(f)(w’) & τ(f) < t’] & FIND (z)(L)(t’)(w’)]; defined only if g(1) ≤ tu
A few words are in order on how the meaning of the matrix clause is composed. The semantic value of zhǎo ‘look for’, given in (50c), involves a future-oriented finding event. It should be read as ‘for all 〈t’, w’〉 such that, at t’ in w’, y reaches the goal of the event e in w, there is some z such that z has the property p at t’ in w’, and y finds z at t’ in w’. According to (50d), (45) is defined only if the matrix TopT g(1) precedes or coincides with the utterance time tu. Where defined, (45) is true in w if g(1) is included in That Moment. Example (50d) reads as ‘there is an interval t and an event e, such that t is the running time of e, and t includes g(1); for all 〈t’, w’〉 such that, at t’ in w’, L reaches her goal for the event e in w, there is some z such that z is a girl at t’ in w’, and there is an event f of z dancing ballet in w’, such that the running time of f precedes t’, and L finds z at t’ in w’.

6.2.2. RCs with Overt Perfective Aspect le

We now turn to the RC in (51) with the overt perfective marker le (repeated from (19b)). We give in (53) and (54) below the semantic value of le and the semantic derivation of (51) on a de dicto construal.
(51)(Shàng-zhōusān,)Lùluzàizhǎo[RC yí-gekàn-le
last-WednesdayLùluPROGlook-forone-Clwatch-PFV
hālǐ bōtèdeyǐng-mí.]
Harry PotterDEmovie-fan.
‘(Last Wednesday,) Lùlu was looking for a movie fan who had seen Harry Potter.’
Or ‘ Lùlu is looking for a movie fan who has seen Harry Potter.’
The derivation of (51) differs from that of (45) only in that with perfective aspect (le), the RC event time g(5) can either precede or coincide with the TopT g(3) (that is, g(5) ≤ g(3)). Tense then checks whether the TopT g(3) is a non-future interval with regard to the evaluation time g(2) (that is, g(3) ≤ g(2)). Once again, since the evaluation time g(2) is itself identified with the matrix event time via binding, the RC searching time g(5) is constrained to either precede or coincide with the matrix searching time, which itself is required to precede or coincide with the time of utterance. Putting all this together yields the ordering g(5) ≤ g(3) ≤ g(2) ≤ tu, which ensures that no subinterval of the dancing time g(5) can lie in the future of the searching time g(2). This is why/how RCs with perfective aspect fail to yield later-than-matrix readings.
(52)
Languages 07 00170 i004
(53)Relative clause with le:
a.Asp° leg,c = λp<v,st>. λti. λui: tu. λws. ∃e[p(e)(w) & τ(e) = t]20
b.AspP t5 le [x4 kàn hālǐ bōtè]g,c = λui: g(5) ≤ u. λws. ∃e[WATCH HARRY POTTER(g(4))(e)(w) & τ(e) = g(5)]
c.〚TP〛g,c = λws. ∃e[WATCH HARRY POTTER(g(4))(e)(w) & τ(e) = g(5)]; defined only if g(5) ≤ g(3), and g(3) ≤ g(2)
d.〚NP〛g,c = λxe. λti: tg(2). λws. MOVIE-FAN(x)(t)(w) & ∃e[WATCH HARRY POTTER(x)(e)(w)] & ∃u[u < t & τ(e) = u]
(54)〚(51)〛g,c = λws. g(1) ⊆ Wednesday before the day of tu & ∃tiev[ τ(e) = t & g(1) ⊆ t] & ∀〈t’, w’〉 [REACH ONE’S GOAL(L)(e)(w)(t’)(w’) → ∃zefv [MOVIE-FAN(z)(t’)(w’) & WATCH HARRY POTTER(z)(f)(w’) & τ(f) ≤ t’] & FIND (z)(L)(t’)(w’)]; defined only if g(1) ≤ tu

6.2.3. RCs with Overt Progressive Aspect zài

Recall that RCs containing progressive marker zài allow simultaneous readings (Section 4.3). Example (26b), repeated below as (55), has the syntax in (56), similar in all relevant respects (except for the meaning of the aspectual head) to (45) with perfect guo. Its semantic derivation is given in (57) and (58).
(55)(Dāngshí),Lùluzàizhǎo[RC yí-gezàitiàobāléiwǔdenǚháir.]
at.that.timeLùluPROGlook-forone-ClPROGdanceballetDEgirl.
‘(At that moment,) Lùlu was looking for a girl who was dancing ballet.’
*’(At that moment,) Lùlu was looking for a girl who would be dancing ballet.’
Here, progressive aspect zài constrains the RC event time g(5) to coincide with the TopT g(3) (g(5) ⊆ g(3)), while Non-Future constrains the TopT g(3) to precede or coincide with the RC evaluation time (g(3) ≤ g(2)). Once again, since the RC evaluation time and the matrix event time are assigned the same value g(2) (via binding), the RC dancing time g(5) is required to coincide with the matrix searching time g(2), itself constrained to either precede or coincide with utterance time. Putting all this together yields the ordering g(5) ⊆ g(3) ≤ g(2) ≤ tu, which ensures that no subinterval of the dancing time g(5) can lie in the future of the searching time g(2). This is why/how RCs with progressive aspect fail to yield later-than-matrix readings.
(56)
Languages 07 00170 i005
(57)Relative clause with zài:
a.zàig,c = λp<v,st>. λti. λui: ut. λws. ∃e[p(e)(w) & τ(e) = t]
b.AspPt5 zài [x4 tiào bāléiwǔ]g,c = λui: ug(5). λws. ∃e[DANCE BALLET(g(4))(e)(w) & τ(e) = g(5)]
c.〚TP〛g,c = λws. ∃e[DANCE BALLET(g(4))(e)(w) & τ(e) = g(5)]; defined only if g(3) ⊆ g(5), and g(3) ≤ g(2)
d.〚NP〛g,c = λxe. λti: tg(2). λws. GIRL(x)(t)(w) & ∃e[DANCE BALLET(x)(e)(w)] & ∃u[tu & τ(e) = u]
(58)〚(55)〛g,c = λws. g(1) ⊆ That Moment & ∃tiev [τ(e) = t & g(1) ⊆ t] & ∀〈t’, w’〉 [REACH ONE’S GOAL(L)(e)(w)(t’)(w’) → ∃zefv [GIRL(z)(t’)(w’) & DANCE BALLET(z)(f)(w’) & t’ ⊆ τ(f)] & FIND (z)(L)(t’)(w’)]; defined only if g(1) ≤ tu
At this stage, we have derived the temporal interpretations of embedded RCs with overt—be it perfect(ive) or imperfective—aspect (as identified in Section 4 above), without having to stipulate the ULC. The lack of later-than-matrix readings followed automatically from how the meaning of the RC and the matrix clause compose, on the assumption that Mandarin has a covert Non-Future tense. We now turn to the interpretations of aspectually bare RCs.

6.3. Intensional Contexts without Overt Aspect

Recall example (7), repeated below as (59), which showed that Mandarin RCs without overt aspect allow freely ordered temporal construals, since the time of dancing in the RC can either precede, follow, or coincide with the matrix searching time.
(59)Dāngshí, Lùluzàizhǎo [RC yí-getiàobāléiwǔdenǚháir.]
at.that.timeLùluPROGlook-for one-Cldance balletDEgirl.
‘At that moment, Lùlu was looking for a girl who had danced/was dancing/would dance ballet.’
In an intensional context, on a de dicto reading, the RC in (59) (just like the RCs in (45), (51), or (55) above) is anchored into the matrix via binding of the RC evaluation time by the matrix event time. Now, the only difference between (59) and (45) is that there is no overt aspect in the RC in (59). Recall the assumptions defended in Section 6.1: AspP is always projected since Asp° serves to mediate between eventualities and times by introducing, via a running time function, the eventuality time argument. Aspect is thus always present, even in sentences with no overt marker appearing on the surface. This means that alongside overt aspectual heads (e.g., guo, le, zài, or zhe), Mandarin also has the covert aspectual head, whose denotation was given in (44) above. This covert aspectual head is semantically underspecified: it imposes no restriction on the temporal ordering of its two time arguments. Consequently, the ordering between the RC event time and the RC TopT can be freely construed as precedence, coincidence, or subsequence, as illustrated in (60c,d). We go through the semantic derivation of (59) in (60) and (61) step by step to see how the availability of later-than-matrix reading automatically falls out.
(60)
a.NF tense in the RC: tRC toptRC evaluation (= tmatrix event)
b.∅Asp:tRC eventtRC top OR tRC top < tRC event
→ Ordering undetermined
c.tRC eventtRC top→ ✗later-than-matrix
i.---[tRC top]---[DANCING--]-----[LOOKING FOR]---->
ii.---[DANCING--]---[tRC top LOOKING FOR]------>
d.tRC top < tRC event → ✓later-than-matrix
i.---[tRC top]---[DANCING--]-----[LOOKING FOR]---->
ii.---[tRC top]---[LOOKING FOR]---[DANCING---]----->
iii.---[tRC top LOOKING FOR]---[DANCING---]----->
(61) Languages 07 00170 i006
(62)Relative clause with covert aspect:
a.Asp°∅〛g,c = λp<v,st>. λti. λui: tu or u < t. λws. ∃e[p(e)(w) & τ(e) = t]
b.〚TP〛g,c = λws. ∃e[DANCE BALLET(g(4))(e)(w) & τ(e) = g(5)]; defined only if g(3) ≤ g(2), and g(5) ≤ g(3) or g(5) > g(3)
c.〚NP〛g,c = λxe. λti: tg(2). λws. GIRL(x)(t)(w) & ∃e[DANCE BALLET(x)(e)(w)] & ∃u[ut or t < u & τ(e) = u]
(63)〚(59)〛g,c = λws. g(1) ⊆ The Moment & ∃tiev [τ(e) = t & g(1) ⊆ t] & ∀〈t’, w’〉 [REACH ONE’S GOAL(L)(e)(w)(t’)(w’) → ∃zefv [GIRL(z)(t’)(w’) & DANCE BALLET(z)(f) (w’)] & FIND (z)(L)(t’)(w’)]; defined only if g(1) ≤ tu
Compare the derivation of (59) in (62) and (63) with that of the RC in (45), where perfect guo constrains the time of dancing g(5) to precede the TopT g(3) (g(5) < g(3)). Now, silent aspect in (61)–(62a) places no constraint on the ordering of these two times: g(5) can precede g(3), as was the case with perfect guo in (48) and (49); overlap g(3), as was the case with progressive zài in (56) and (57); or follow g(3). Suppose then that g(3) follows g(5). Tense then constrains the TopT g(3) to precede or coincide with the RC evaluation time (g(3) ≤ g(2)), but since g(3) is free to fall after g(5) (g(5) < g(3)), the ordering (g(3) ≤ g(2)) < g(5)) can freely be generated. Since, moreover, the RC evaluation time and the matrix event time are assigned the same value g(2) (via binding), then on this ordering, the RC dancing time g(5) is future-shifted relative to the matrix searching time g(2). This is why/how RCs with bare predicates allow later-than-matrix readings.
Recapitulating, why are later-than-matrix readings possible in Mandarin bare RCs, but not in RCs with overtly marked aspect, or English tensed RCs? Mandarin bare RCs are tensed (have a silent Non-Future) and, in intensional contexts, their evaluation time is bound by the matrix event time (g(2) in all of the derivations in Section 6.2 and Section 6.3). When aspect is silent, however, the (indirect) ordering of the matrix event time g(2) and the embedded event time g(5) is undetermined and, as such, can be freely construed as precedence, yielding an earlier-than-matrix construal, as coincidence yielding a simultaneous construal, or subsequence yielding the later-than-matrix construal proscribed by the ULC. The underspecification of silent aspect thus ultimately explains why Mandarin embedded clauses with bare predicates allow later-than-matrix readings, even in intensional contexts.

7. Conclusions: Evidence for Non-Future Tense in Embedded Contexts

We first sought in this paper to establish that Mandarin bare embedded clauses in intensional contexts allow later-than-matrix readings. We then argued that this freedom of temporal interpretation could not be imputed to the lack of tense: Mandarin clauses are only superficially tenseless, since they carry a silent Non-Future tense (Sun 2014).
We have shown that the intricate distribution of the temporal readings of Mandarin relative clauses with or without aspectual markers nicely follows the Non-Future tense hypothesis first defended by Sun (2014) for Mandarin root clauses, together with the assumption that Mandarin has a silent underspecified aspect. The source of temporally free readings of bare embedded clauses in Mandarin does not lie in surface tenselessness or undetermined tense since Mandarin has a silent Non-Future tense, but rather in surface aspectlessness and undetermined aspect.
We close this paper by assessing the arguments provided here (the first to the best of knowledge) for Non-Future tense in embedded contexts, as well as offering two further arguments.
One of the core arguments for Non-Future tense in independent clauses (Matthewson 2006; Sun 2014) comes from the absence of future readings for bare sentences. As discussed at length in Section 5, Mandarin bare sentences cannot be used to describe future eventualities even when a future adverbial is added to the sentence. Future (though not futurate) readings require a future-oriented modal. This restriction follows straightforwardly the assumption that Mandarin has a covert Non-future tense restricting the reference of the TopT to non-future times.
Turning to embedded contexts, we have seen that RCs overtly marked with aspect do not allow readings in intensional contexts where the RC eventuality lies in the future relative to utterance time, if the matrix eventuality is itself in the past or the present. On the assumption that Mandarin has a covert Non-Future tense, this restriction follows straightforwardly how the meaning of the RC and the matrix clause compose, without stipulating the ULC—but why not appeal to the ULC itself to rule out later-than-matrix readings without positing silent Non-Future tense in the embedded clause? (This is indeed the position that Bochnak (2016) takes for Washo, a language with morphologically tensed and tenseless clauses.) Well, first, Non-Future in embedded contexts is the null hypothesis once we have established its existence in matrix contexts. Second, the ULC is clearly too strong since it does not hold in Mandarin aspectually unmarked RCs in intensional contexts, as established in Section 4. Thirdly, the other major argument provided in the literature for Non-Future tense in independent clauses carries over to embedded contexts. Thus, consider the RC in (64b):
(64)a.Preceding discourse context:
GǔlóngshēngXiānggǎng,MòyánshēngShāndōng.
Gǔlóngbe.bornatHong Kong Mòyánbe.bornatShandong
‘Gǔlóng was born in Hong Kong, and Mòyán was born in Shandong.
b.Zhèliǎng-ge[RC chōuyānhěnxiōngde]zuòjiā dōu
DEMtwo-ClsmokeveryterribleDEwriterDOU
hěnyǒumíng.
veryfamous
‘These two writers who smoked a lot and smokes a lot (respectively) are both very famous.’
The argument provided in (64) parallels that given in (37) above for independent clauses. As the preceding context in (64a) makes clear, the two writers under discussion in (64b) are Gǔlóng, a novelist who passed away in 1985, and Mòyán, a writer (alive in 2021 when the sentence is uttered). Example (64b) thus conveys that Gǔlóng, who used to be a heavy smoker, and Mòyán, who is heavy smoker, are both very famous. Their smoking habits as depicted by the RC are true for different time intervals. The one and only main predicate in the RC (and, by hypothesis, the one and only syntactic tense (TP) projection in the RC) thus yields past and present readings simultaneously. Example (64b) can only be translated to English with two occurrences of the predicate ‘smoke’ inside the RCs because the temporal information encoded in the RC in (64b) cannot be conveyed with a single tense in English. Only a tense that can select simultaneously for both past and present intervals could fit the truth conditions of the covert tense heading the TP in the RC, and this is exactly how Non-Future is defined.
Finally, consider (65), which provides yet another argument for Non-Future tense in embedded context from the temporal reference of adverbs. The predicate mǎi cǎipiào ‘buy lottery tickets’ in the RC marked by guo can only be understood as describing a past-shifted event relative to the matrix search time. That is, Lùlu was/is looking (at the past time under discussion) for a customer who had already bought tickets at the search time. The time adverb shíèryuè ‘December’ can in principle refer to either a past or a future December. If Lùlu’s search time is in, say, June 2021, shíèryuè in (65) can felicitously refer to the previous December—that is, December 2020—but not the following December, that is, December 2021. Now, this constraint on the reference of the adverb will follow the Non-Future tense hypothesis, but not if the guo marked RC in (65) is tenseless. To see why, consider the syntax of the RC as sketched in (66) (all irrelevant information omitted).
(65)(Dāngshí,)Lùluzhǎoguo[RC yí-gezàishíèryuèyǐqián
at.that.timeLùlulook-for PFTone-Clat Decemberbefore
mǎiguocǎipiàodegùkè].
buyPFT lottery.ticket DEcustomer.
‘At that moment, Lùlu had looked for a customer who had bought lottery tickets before December.’
(66)TopT modification by a temporal adverb (in the RC):
Languages 07 00170 i007
Perfect guo in the RC places the RC event time in the past of the RC TopT. Since we are dealing with an intensional context, the embedded evaluation time will be bound by the matrix event time. The RC TopT is thus constrained to be a time before December that does not fall after the matrix search time. Perfect guo in the matrix places the search time in the past of the past topic time under discussion. These constraints yield the temporal ordering schematized below:
(67)BUYING TICKETS < [TopT-RC BEFORE DECEMBER] ≤ SEARCH < [TopT-Matrix AT THAT PAST MOMENT]
It thus follows that if Lùlu’s search time is in, say, June of 2021, the adverb shíèryuè in (65) can felicitously refer to the preceding December (of 2020), but not to the upcoming December (of 2021). Suppose, however, that tense in Mandarin imposes no restriction on the embedded TopT relative to the embedded evaluation time—that is, relative to the matrix event time (since in intensional contexts the former binds the latter). The only constraints we would then have are thus those in (68a) and (68b) imposed respectively by perfect aspect in the RC (68a), and perfect aspect in the matrix (68b). So what, then, prohibits either of the orderings in (68c–d) which satisfy both these constraints, but are nonetheless unavailable?
(68)a.BUYING TICKETS < [TopT-RC BEFORE DECEMBER] (via perfect guo in the RC)
b.SEARCH < [TopT-Matrix AT THAT PAST MOMENT] (via perfect guo in the matrix)
c.✗ later-than-matrix
SEARCH < [TopT-Matrix AT THAT PAST MOMENT] < BUYING TICKETS < [TopT-RC BEFORE DECEMBER] < UT.
d.✗ earlier-than-matrix
BUYING TICKETS < SEARCH < [TopT-Matrix AT THAT PAST MOMENT] < [TopT-RC BEFORE DECEMBER] < UT.
(69)Dāngshí,Lùluzhǎoguo[RC yí-gejiānghuìzàishíèryuè
at.that.timeLùlulook-forguoone-ClWOLLatDecember
yǐqiánmǎiguocǎipiàodegùkè].
beforebuyPFTlottery.ticketDEcustomer.
‘At that moment, Lùlu had looked for a customer who would have bought lottery tickets before December.’
Note importantly that on the ordering in (68d), only the past reference time of the adverb is future-shifted relative to the matrix search event. That is, the ticket buying event itself remains past-shifted relative to the matrix search event, so this unavailable construal is not a later-than-matrix reading, but rather an earlier-than-matrix reading. In order to convey the forward-shifted readings schematized in (68c,d), a modal is required, as shown in (69). The Non-Future tense hypothesis straightforwardly accounts for the unavailability of such construals, whether the embedded event is itself past- or future-shifted relative to the matrix event, by ensuring that (any subinterval of) the TopT of the RC does not lie in the future of the matrix search time—that is, of the embedded evaluation time in intensional contexts.

Author Contributions

The contribution of each author is equal. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the Laboratoire de Linguistique de Nantes (LLING UMR 6310, Nantes Université/CNRS), the Centre d’Études des Relations et Contacts Linguistiques et Littéraires (CERCLL UR 4283) and Université de Picardie Jules Verne.

Acknowledgments

We are particularly indebted for their extremely insightful comments to Orin Percus, Tim Stowell, and the three reviewers of Languages (addressing the latter’s perspicacious questions greatly improved the manuscript). The current paper also benefited from discussion with Oana Lungu, Hadil Karawani, Rint Sybesma, María Arche, Lisa Matthewson, and Ruoying Zhao. Special thanks go to the audiences of the “Tenselessness” workshop in Greenwich, London (2017); the “Marking Aspect and Tense” (MAT) workshop in Amiens (2021); and the “Looking for Non-Future Tense” (NFT) workshop in Nantes for discussion (2022).

Conflicts of Interest

The authors declare no conflict of interest.

Abbreviations

AdjP: adjective phrase, Cl: classifier, DEM: demonstrative, DP: determiner phrase, DUR: durative, EvlT: evaluation time, LF: logical form, NF: non-future, PFT: perfect, PFV: perfective, PP: prepositional phrase, PROG: progressive, RC: relative clause, Top/top: topic, TopT: topic time, ULC: upper limit constraint, UT: utterance time.

Notes

1
To our knowledge, the term ‘later-than-matrix reading’ is first used in Kusumoto (1999), referring to the temporal ordering of the event time of the embedded clause with regard to the matrix event time, on construals where the former follows the latter.
2
We restrict our attention here to RCs, the issues that the time reference of complement clauses raises going well beyond the scope of this paper. Suffice to say, however, that the generalizations given in (i) and (ii) in the text above carry over to complement clauses in Mandarin (see Demirdache and Sun 2017; Sun and Demirdache 2018).
3
Note that the eventive predicate tiào bāléiwǔ ‘dance ballet’ in the RC can also receive a habitual reading, in which case the RC describes a girl who was/is a ‘ballet dancer’. For the purpose of this paper, we focus exclusively on episodic readings and leave aside the generic/habitual readings of eventive predicates, but see Sun (2014) for extensive discussion.
4
The bracketing in (8b) and (12b) is intended to illustrate a simultaneous reading where the matrix search/interview time coincides with (that is, starts at or during) the past dancing/living time of the RC. The reader might wonder why the RC event time can extend up to/after the UT in (12b) but not in (8b); this is merely because of the presence of the adverb dāngshí ‘that time’ in (7), which (given the context provided for (7)) restricts the search time to a time located last night.
5
That stative predicates receive simultaneous readings with regard to the utterance time or the matrix event time—in the absence of an overt time adverb in the RC—leads Sun (2015) to incorrectly conclude that stative RCs, unlike eventive RCs, do not allow temporally free readings. Her generalization is roughly as follows. Out of the blue, the default anchor time for the state eventuality in the RC in, say, (i) is either the utterance time or the matrix event time, thus yielding a simultaneous interpretation where the state described by the RC (the time of staying in Paris) overlaps either the utterance time, or the matrix event search time. When, however, an adverb is added to temporally anchor the state eventuality described by the RC, the bare stative RC can be construed as either past- or future-shifted relative to the matrix event time.
(i)Shànggeyuè,Lùluzàizhǎo[RC yí-gedāizàibālídenǚháir].
last.monthLùluPROGlook.forone-CLstayatParisDEgirl
‘Last month, Lùlu was looking for a girl who is/was staying in Paris.’
Now, this construal is also available in the absence of a temporal adverb as long as the discourse context is properly set up. Suppose, for instance, that Lùlu wants to interview people about the COVID-19 shutdown in 2020. Example (i) could be felicitously uttered, in June of 2021 on a past-shifted reading where Lùlu is looking for a girl to interview who lived in Paris during the 2020 shutdown. Or, suppose that Lùlu is preparing a TV program about Paris scheduled to be filmed the following fall. In this context, (i) could be felicitously uttered in July of 2021 on a future-shifted reading where Lùlu was looking for a girl who will be living in Paris at the future time of filming.
6
Note, moreover, that since there is no temporal specification of the relevant fall in (13) (e.g., whether it is last or next fall), the RC can also receive a habitual reading (cf. footnote 3), where Lùlu was looking for a girl who generally stays in Paris in the fall. As mentioned previously (footnote 3), however, we leave aside here habitual and generic construals, restricting ourselves to episodic readings.
7
For further discussion of the temporal interpretation of bare stative vs. eventive predicates, see Caudal and Bednall (2022) (in this Special Issue), who convincingly argue against the view found in the literature that aktionsart largely determines the temporal anchoring of sentences with bare predicates (e.g., Smith and Erbaugh 2005), concluding that biases towards a particular temporal anchoring can always be overruled by additional contextual temporal information, whatever the event structure type of the predicate.
8
For an extensive discussion of Mandarin perfective aspect, see Zhao (2022) (in this Special Issue).
9
Often referred to as an ‘experiential’ marker, guo indicates that ‘the event has been experienced at some indefinite time, often in the past’ (Klein et al. 2000). It differs from the English perfect in that it has a ‘discontinuity’ effect (Chao 1965; Smith 1991; Lin 2006). See also discussions in Zhao (2022) (in this Special Issue) and Bertrand et al. (2022) (in this Special Issue) for a cross-linguistic perspective.
10
Note that (17) also differs from (13) in that the locative PP zài bālí ‘in Paris’ is in a preverbal position. This is because a post-verbal locative PP would be infelicitous with guo.
11
The analysis of le is extensively debated among scholars, many of which distinguish verbal le, commonly analyzed as encoding perfectivity, from sentential le, analyzed as an inchoative marker (Chao 1965; Li and Thompson 1981; Sybesma 1999; Paul 2015). Only verbal le is relevant for the discussion in this paper.
12
In (21), e′ ≤E e means e’ is a subpart of the event e; epro is a pronoun-like free variable.
13
What follows owes a lot to a reviewer who inquired about later-than-matrix readings of le on its continuous meaning, and brought to our attention verbs such as yǎng ‘raise’, which yield continuous rather than past-shifted readings when modified by le.
14
We set aside cases where zhe modifies a verb other than the main verb of the sentence, such as (i), where zhe marks simultaneity of the event denoted by the verb kāi ‘drive’ with the event denoted by the main verb tīng yīnyuè ‘listen to music’. In this case, the verb modified by zhe does not encode a result state.
(i)jīngchángkāizhechētīngyīnyuè
3SGoftendriveDURcarlistenmusic
‘He often listens to music while driving his car.’
15
i in (33) is the type of time intervals.
16
The paradigms in (34) and (37) in the text are adapted from St’át’imcets (Matthewson 2006) to Mandarin. The reader is also referred to Chen and Husband (2018), who provide interesting experimental evidence for the contrasts illustrated in (37).
17
In Section 6, we introduce aspect, which serves to map predicates of eventualities onto predicates of times, following Kratzer (1998), and in this way can impose conditions on the relation between event time and TopT. Temporal frame adverbs can modify either of these two times (Hornstein 1990; Kamp and Reyle 1993; Demirdache and Uribe-Etxebarria 2000, 2004). Example (66) in Section 7 illustrates the syntax and semantic composition of modification of the TopT once we introduce AspP.
18
v in (41) is the type of events.
19
We leave these two options open here, as we see at this stage no empirical differences in the predictions they respectively make.
20
We give this definition of le for simplicity. The exhaustive semantics of le remains controversial in the literature due to the complexity of its uses and as such is well beyond the scope of this paper. Suffice to say that capturing the full range of temporal construals of le when applied to different types of predicates (e.g., whether or not a result state is inferred) would require reformulating (53a) in order to make reference to subevent structure, in the spirit of Lin (2003, 2006).

References

  1. Abusch, Dorit. 1988. Sequence of Tense, Intensionality and Scope. In The Proceedings of the 7th West Coast Conference on Formal Linguistics. Edited by H. Borer. Stanford: CSLI Publications, pp. 1–14. [Google Scholar]
  2. Abusch, Dorit. 1994. Sequence of Tense Revisited: Two Semantic Account of Tense in Intensional Contexts. In Ellipsis, Tense and Questions. Edited by Hans Kamp. Amsterdam: University of Amsterdam, pp. 87–139. [Google Scholar]
  3. Bertrand, Anne, Yurika Aonuki, Sihwei Chen, Joash Gambarage, Laura Griffin, Marianne Huijsmans, Lisa Matthewson, Daniel Reisinger, Hotze Rullmann, and Raiane Salles. 2022. Nobody’s Perfect. In Languages, Special Issue "Tense and Aspect across Languages". Edited by Henriëtte de Swart and Bert Le Bruyn. Basel: MDPI. [Google Scholar]
  4. Bochnak, Ryan. 2016. Past time reference in a language with optional tense. Linguistics and Philosophy 39: 1–48. [Google Scholar] [CrossRef] [Green Version]
  5. Caudal, Patrick, and James Bednall. 2022. Aspectuo-temporal underspecification in Anindilyakwa: Descriptive, theoretical, typological and quantitative issues. In Languages, Special Issue "Tense and Aspect across Languages". Edited by Henriëtte de Swart and Bert Le Bruyn. Basel: MDPI. [Google Scholar]
  6. Chao, Yuen Ren. 1965. A Grammar of Spoken Chinese. Berkeley and Los Angeles: University of California Press. [Google Scholar]
  7. Chen, Sherry Young, and E. Matthew Husband. 2018. Contradictory (Forward) Lifetime Effects and the Non-Future Tense in Mandarin Chinese. Proceedings of the LSA Annual Meeting 3: 1–14. [Google Scholar] [CrossRef] [Green Version]
  8. Copley, Bridget. 2002. The Semantics of the Future. Ph.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, USA. [Google Scholar]
  9. Copley, Bridget. 2008. The plan’s the Thing: Deconstructing Futurate Meanings. Linguistic Inquiry 39: 261–74. [Google Scholar] [CrossRef]
  10. Demirdache, Hamida, and Hongyuan Sun. 2017. Constraints on embedded time reference in Mandarin: Zero vs. overt aspect. Paper presented at the Tenselessness Workshop, University of Greenwich, London, UK, 5–6 October 2017. [Google Scholar]
  11. Demirdache, Hamida, and Myriam Uribe-Etxebarria. 2000. The primitives of temporal relations. In Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. Edited by Roger Martin, David Michaels and Juan Uriagerreka. Cambridge: MIT Press, pp. 157–86. [Google Scholar]
  12. Demirdache, Hamida, and Myriam Uribe-Etxebarria. 2004. The syntax of time adverbs. In The Syntax of Time. Edited by Jacqueline Guéron and Jacqueline Lecarme. Cambridge: The MIT Press, pp. 143–79. [Google Scholar]
  13. Djamouri, Redouane, and Waltraud Paul. 2018. A new approach to -zhe in Mandarin Chinese. In Studies in Japanese and Korean Historical and Theoretical Linguistics and Beyond. Edited by William McClure and Alexander Vovin. Leiden: Brill, pp. 110–23. [Google Scholar]
  14. Enç, Mürvet. 1986. Towards a Referential Analysis of Temporal Expressions. Linguistics and Philosophy 9: 405–26. [Google Scholar] [CrossRef]
  15. Gōng, Qiānyán. 1991. Tán xiàndài Hànyǔ de shízhì biǎoshì hé shítài biǎodá xìtǒng [About the system of expression of tense and time in Modern Chinese]. Zhōngguó Yŭwén 1991: 251–61. [Google Scholar]
  16. Heim, Irene, and Angelike Kratzer. 1998. Semantics in Generative Grammar. Oxford: Blackwell Publishers. [Google Scholar]
  17. Hornstein, Norbert. 1990. As Time Goes By. Cambridge: The MIT Press. [Google Scholar]
  18. Huang, Nick. 2015. On syntactic tense in Mandarin Chinese. In Proceedings of the 27th North American Conference on Chinese Linguistics. 2. Edited by Hongyin Tao. Los Angeles: UCLA, pp. 406–23. [Google Scholar]
  19. Jin, Lixin. 2002. Ciwei ‘Le’ De Shiti Yiyi Ji Qi Jufa Tiaojian [The Tense/Aspectual Meaning of the Sentence-final Le and its Syntactic Conditions]. Shijie Hanyu Jiaoxue 1: 34–43. [Google Scholar]
  20. Kamp, Hans, and Uwe Reyle. 1993. From Discourse to Logic. Dordrecht: Kluwer. [Google Scholar]
  21. Klein, Wolfgang. 1994. Time in Language. London: Routledge. [Google Scholar]
  22. Klein, Wolfgang, Ping Li, and Henriëtte Hendriks. 2000. Aspect and assertion in Mandarin Chinese. Natural Language and Linguistic Theory 18: 723–70. [Google Scholar] [CrossRef]
  23. Kratzer, Angelika. 1996. Severing the External Argument from its Verb. In Phrase Structure and the Lexicon. Edited by Johan Rooryck and Laurie Zaring. Dordrecht: Kluwer. [Google Scholar]
  24. Kratzer, Angelika. 1998. More Structural Analogies between Pronouns and Tenses. In Proceedings from Semantics and Linguistic Theory VIII (SALT 8). Edited by Devon Strolovitch and Aaron Lawson. Ithaca: Cornell University, pp. 92–109. [Google Scholar]
  25. Kusumoto, Kiyomi. 1999. Tense in Embedded Contexts. Ph.D. Dissertation, University of Massachusetts at Amherst, Amherst, MA, USA. [Google Scholar]
  26. Kusumoto, Kiyomi. 2005. On the Quantification Over Times in Natural Language. Natural Language Semantics 13: 317–57. [Google Scholar] [CrossRef]
  27. Ladusaw, William. 1977. Some Problem with Tense in PTQ. In Texas Linguistic Forum. Austin: University of Texas, vol. 6, pp. 89–102. [Google Scholar]
  28. Li, Charles N., and Sandra A. Thompson. 1981. Mandarin Chinese: A Functional Reference Grammar. Berkeley: University of California Press. [Google Scholar]
  29. Lin, Jo-wang. 2000. On the Temporal Meaning of the Verbal -le in Chinese. Language and Linguistics 1: 109–33. [Google Scholar]
  30. Lin, Jo-wang. 2003. Temporal Reference in Chinese. Journal of East Asian Linguistics 12: 259–311. [Google Scholar] [CrossRef]
  31. Lin, Jo-wang. 2006. Time in a language Without Tense: The case of Chinese. Journal of Semantics 23: 1–53. [Google Scholar] [CrossRef]
  32. Lin, Tzong-Hong Jonah. 2015. Tense in Mandarin Chinese Sentences. Syntax 18: 320–42. [Google Scholar] [CrossRef]
  33. Matthewson, Lisa. 2006. Temporal semantics in a superficially tenseless language. Linguistics and Philosophy 29: 673–713. [Google Scholar] [CrossRef]
  34. Mei, Kuang. 2002. How languages express time differently. IIAS Newsletter 28: 46. [Google Scholar]
  35. Ogihara, Toshiyuki. 1996. Tense Attitudes and Scope. Dordrecht: Kluwer Academic Publishers. [Google Scholar]
  36. Pan, Haihua. 1996. Imperfective Aspect Zhe, Agent Deletion, and Locative Inversion in Mandarin Chinese. Natural Language & Linguistic Theory 14: 409–32. [Google Scholar]
  37. Paul, Waltraud. 2015. New Perspectives on Chinese Syntax, Series: Trends in Linguistics. Studies and Monographs. Berlin: De Gruyter Mouton. [Google Scholar]
  38. Smith, Carlota. 1991. The Parameter of Aspect. Dordrecht: Kluwer. [Google Scholar]
  39. Smith, Carlota, and Mary Erbaugh. 2005. Temporal interpretation in Mandarin. Linguistics 43: 713–56. [Google Scholar] [CrossRef]
  40. Stowell, Tim. 1993. Syntax of Tense. Los Angeles: University of California. [Google Scholar]
  41. Stowell, Tim. 2007. The Syntactic Expression of Tense. Lingua 117: 437–63. [Google Scholar] [CrossRef]
  42. Sun, Hongyuan. 2014. Temporal Construals of Bare Predicates in Mandarin Chinese. Ph.D. Thesis, Université de Nantes/Leiden University. [Google Scholar]
  43. Sun, Hongyuan. 2015. Evidence against a scope analysis of temporally free readings of relative clauses in Mandarin. In Cahier Chronos: Taming the TAME Systems. Edited by Emmanuelle Labeau and Qiaochao Zhang. Amsterdam and New York: Rodopi, pp. 63–81. [Google Scholar]
  44. Sun, Hongyuan, and Hamida Demirdache. 2018. Later-than-Matrix Temporal Construals in Mandarin. Nantes: Université de Nantes and Université de Picardie Jules-Verne. [Google Scholar]
  45. Sybesma, Rint. 1999. The Mandarin VP. Dordrecht: Kluwer. [Google Scholar]
  46. Sybesma, Rint. 2007. Whether we Tense-agree overtly or not. Linguistic Inquiry 38: 580–87. [Google Scholar] [CrossRef]
  47. Thomas, Guillaume. 2015. The present tense is not vacuous. Journal of Semantics 32: 685–747. [Google Scholar] [CrossRef]
  48. Tsai, Wei-Tien Dylan. 2008. Tense anchoring in Chinese. Lingua 118: 675–86. [Google Scholar] [CrossRef]
  49. Yeh, Meng. 1993. The Stative Situation and the Imperfective ZHE in Mandarin. Journal of the Chinese Language Teachers Association 28: 69–98. [Google Scholar]
  50. Zhao, Ruoying. 2022. Decomposing perfect readings. In Languages, Special Issue “Tense and Aspect across Languages”. Edited by Henriëtte de Swart and Bert Le Bruyn. Basel: MDPI. [Google Scholar]
Table 1. Availability of later-than-matrix readings.
Table 1. Availability of later-than-matrix readings.
Extensional/Intensional de reIntensional de dicto
BarePast/PFT/PFV
PROG/DUR
BarePast/PFT/PFV
PROG/DUR
English
Mandarin
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Sun, H.; Demirdache, H. Time Reference in Mandarin Relative Clauses. Languages 2022, 7, 170. https://doi.org/10.3390/languages7030170

AMA Style

Sun H, Demirdache H. Time Reference in Mandarin Relative Clauses. Languages. 2022; 7(3):170. https://doi.org/10.3390/languages7030170

Chicago/Turabian Style

Sun, Hongyuan, and Hamida Demirdache. 2022. "Time Reference in Mandarin Relative Clauses" Languages 7, no. 3: 170. https://doi.org/10.3390/languages7030170

Article Metrics

Back to TopTop