There are two major groups of analyses in the literature on the English past tense; existential semantics with domain restriction, and the pronominal (i.e. anaphoric) analysis. In this paper, I discuss a group of data difficult for both of these analyses: the obligatory use of the English past in some contexts without any salient antecedent, such as asking
Two major groups of analyses of the past tense in English and across languages are: (i) existential semantics with appropriate domain restriction, and (ii) pronominal analysis. In particular, the literature has extensively explored the idea that English has a pronominal tense (
(1)
(Pointing at a church:)
Who built/#has built this church?
Borromini built/#has built this church.
(
(2)
Mary, my colleague, was born/#has been born in New Zealand.
To explain these data, previous authors have proposed two major groups of analyses: that the English past is lexically ambiguous between a pronominal past and an existential past, or ambiguous between a pronominal past and a sort of present perfect reading. In any case, these data, and the vast amount of other data suggesting the the English past is pronominal, create a dilemma for a unified semantic analysis. In addition, there is no satisfactory analysis that accounts for the distribution of the pronominal and the non-pronominal readings.
In this paper, I propose that the distribution of these ‘problematic’ instances of the English past tense can be predicted: they suggest that the English past tense is lexically ambiguous between a pronominal tense, and a tense that presupposes existence and uniqueness of a past reference time with respect to a certain property. The English past tense also displays similar presupposition projection patterns observed with anaphoric and unique definite DPs in the nominal domain.
The literature on definite descriptions has an intensive discussion on whether definiteness is better characterized by uniqueness (
This paper is organized as follows: In Section 2, I discuss data that are difficult to explain under a pure pronominal analysis of the English past and the previous attempts to account for them. In Section 3, I argue that these data can be accounted for if we allow the English past tense to also carry existence and uniqueness presuppositions, parallel with similar patterns in definite DPs. In Section 4, I will discuss one possible way to formalize this idea. Following the recent analyses of the two kinds of definites, I propose that the English past tense is lexically ambiguous between the two readings, with the anaphoric past tense having an additional index argument, which can be dynamically bound.
The data in this paper are obtained from three speakers mostly: one speaker from New Zealand, one from the UK, and one Australian. The speakers are given the context provided in the examples, and two different tasks: (i) produce sentences under the intended reading and context; (ii) compare the past tense and present perfect in a given sentence. In particular, the first task is used to determine the distribution of the past tense and whether it is the most natural choice for speakers in contexts with and without a salient past reference time (e.g. we are talking about what Mary did at yesterday’s party, and the speaker is invited to describe the situation; we are looking at an open window with no contextually salient past reference time, and the speaker asks about that opening event). It is motivated by the observation in Kratzer (
While I will not provide an analysis of the present perfect in the scope of this paper, this asymmetry between the past tense and the present perfect is too strong and consistent to be ignored. I will assume the more ‘relaxed’ version of the Extended Now analysis of the present perfect as in Grønn & Von Stechow (
Partee (
(3)
I didn’t turn off the stove.
Regardless of the scoping choices, treating the past tense as the simple existential quantifier ∃ derives the wrong readings: either ∃
Although Partee (
(4)
Yesterday, John played catch/#has played catch with his dog in the park.
(5)
(Mary and Sue are planning to go to the gym, but Sue has to cancel last minute. Mary goes alone and comes back after an hour.)
Sue: So, how was your workout/#has your workout been?
These observations motivate the analysis that the English past tense behaves like a pronoun. It denotes a contextually salient time interval, in a way just like pronouns denote contextually salient individuals.
However, Kratzer (
(6)
(Pointing at a church. There is no contextually salient past time when the following question comes up:)
Who built this church? Borromini built this church.
In contrast, the German simple past tense behaves as expected, being unacceptable in this context.
(7)
#Wer
Who
baute
built
diese
this
Kirche?
church
Borromini
Borromini
baute
built
diese
this
Kirche.
church
(8)
Wer
Who
hat
has
diese
this
Kirche
church
gebaut?
build
Borromini
Borromini
hat
has
diese
this
Kirche
church
gebaut.
built
‘(Lit.) Who has built this church? Borromini has built this church.’
Note that we cannot analyze the English data as some sort of accommodation of the antecedent. Assuming that the pragmatics of accommodation works the same across languages given the same context, we would not expect this contrast. This suggests that the contrast between English and German illustrates the different grammatical properties of the past tenses.
Kratzer also refutes the idea that maybe this contrast is a matter of terminology. She points out that the German past tense is strictly anaphoric, quoting the following contrast:
(9)
a.
We will answer every letter that we got.
✔
✔
b.
Wir
we
werden
will
jeden
every
Brief
letter
beantworten,
answer
den
that
wir
we
bekamen.
got
‘We will answer every letter that we got.’
✔
In (9a), the English past tense can be used without previously mentioned or contextually salient past times. This contrasts with its German counterparts: the German simple past can only be used if there is a contextually salient past time. If this time is not available, the present perfect is obligatory.
In order to account for the English data, Kratzer proposes that the English simple past sentences without salient antecedent times actually have the semantics of a perfect aspect (10) with present reference time. In other words, the English simple past morphology is lexically ambiguous between the pronominal past reading and a present perfect reading.
(10)
a.
⟦Perfect⟧ =
b.
⟦Borromini built this church⟧ =
After saturating
Note that the present perfect semantics here is not to be confused with the present perfect construction (in the morphological sense). Kratzer (
However, Matthewson et al. (
The idea of restricting the domain of an existentially quantified past tense has also been previously explored in the literature. The central idea is that Partee’s stove sentence (11a) only shows that the English past tense needs salient contextual domain restriction. Authors who adopt this view (
(11)
a.
I didn’t turn off the stove.
b.
¬∃
such that[I-turn-off-stove(
This idea of domain restriction has been applied to those uses of the English past problematic for the pronominal analysis. In particular, Matthewson et al. (
However, assuming a simple existential analysis runs into the obvious problem that the existential reading of the simple past is not always available in contexts without a salient past time. This leads to the question of why domain restriction works in some cases but not others. This is illustrated in the following contrast noted by Matthewson et al. (
(12)
(I am curious which of my friends has read
#Who read
(13)
(There has been confusion about what our book club’s chosen book was this month. Some of us read
Who read
In (12), there is no contextually salient past time, hence the past tense is infelicitous. This contrasts with (13), where there is a salient reference time in the context. Matthewson et. al point out that this contrast would follow if the English past were purely pronominal, but this would leave the church example (6) unexplained. In other words, there needs to be some kind of restriction of the ‘existential’, i.e. the ‘out-of-the-blue’ use of the English past tense.
In order to reconcile the existential past analysis for (6) with the many other sentences where the past tense is infelicitous without salient reference times, Matthewson et al. (
The resulting reading is similar to Ogihara’s (
(14)
∃
Matthewson et al. (
Matthewson et al. (
(15)
(I bought a brand new copy of
(16)
Who littered?
However, requiring the event to be specific seems to both over- and undergenerate this use of the past tense, since Matthewson et al. (
(17)
(Uttered without any contextually salient past time:)
??Mary lost a chess game against Sue.
For states, too, we cannot tell the difference between (18a) and (18b). The former is about a specific state of Sheldon being a child prodigy, and the latter about a specific state of Sheldon Cooper being sick.
(18)
(Uttered without any contextually salient past time:)
a.
Sheldon Cooper, the physicist over there, was a child prodigy.
b.
#Sheldon Cooper, the physicist over there, was sick.
In general, for these examples, it seems that what’s needed is not the specificity of the event, but rather, the (specific) event must be known or can be inferred by both participants.
A remaining problem is whether a domain restricted existential past can be distinguished from a pure pronominal past.
In Section 3, I will argue that those instances that Matthewson et al. (
A previous account based on the notion of definiteness is Grønn & Von Stechow (
In particular, the indefinite reading of the English past is meant to cover the data considered problematic for a pure pronominal tense approach. This is achieved by adding a covert indefinite article (and a covert definite articles for the typical pronominal past). The definite an indefinite articles presuppose familiarity/novelty of the discourse referent for tense, respectively.
(19)
(
This analysis is similar to Matthewson et al.’s (
Michaelis (
(20)
#Mary has cleaned her room yesterday.
However, if the present perfect is specifically marked for non-anaphoricity, then in the absence of any explicit or implicit antecedent, most theories of markedness or competition would predict that the present perfect should be the most appropriate form to use. Using the past tense in this case would give rise to pragmatic inferences due to not using the most appropriate present perfect. This prediction is not borne out. The actual observation is that, for those ‘existential’ uses of the past tense, the present perfect is prohibited in these contexts (21).
(21)
(Pointing at a church. There is no contextually salient past time when the following question comes up:)
#Who has built this church? Borromini has built this church.
In addition, it has been noted that in such contexts, using the present perfect would give rise to inferences such as ‘hot news’ or some kind of repeatability (
(22)
a.
(There is no contextually salient past time when the following question comes up:)
Gutenberg discovered/#has discovered the art of printing.
b.
(There is a demon orchestrating important discoveries in human history, observing the humans, he says:)
Now that Gutenberg has discovered printing, it’s time to lead these humans to do the next thing… (
The repeatability inference is illustrated in (23). Given that Mary is already an adult (being the speaker’s colleague), it cannot be that the speaker is announcing her recent birth. To the extent that the sentence can be interpreted, the only interpretation seems to be that Mary has been born several times, each time at a different place, and now she is born in New Zealand, which is of course impossible.
(23)
(Uttered without any contextually salient past time:)
??Mary, my colleague, has been born in New Zealand.
However, Michaelis (
Hence, I conclude that Michaelis’ (
As the preceding section shows, there is a group of data where the English past tense can be used without any contextually salient antecedent, which poses a challenge for the pure pronominal analysis. Previous analyses can be roughly grouped into three categories: (i) the English past tense is lexically ambiguous between the anaphoric reading and a ‘present perfect’ reading (
In addition, there are two worth noting points here: (i) the past tense is not just allowed to appear in those out-of-the-blue contexts, but obligatory; (ii) these data have in common in that the reference time seems to be presupposed rather than presented as new information.
As pointed out by Matthewson et al. (
However, there is one observation regarding the present perfect I would like to point out. In (24), if we want to assert that there does not exist an event of Borromini building this church, the negated present perfect sentence should simply be true. However, even in this case, the past tense is obligatory:
(24)
(Pointing at a church:)
a.
#Borromini hasn’t built this church./#This church hasn’t been built by Borromini.
b.
Borromini didn’t built this church./This church wasn’t built by Borromini. Michelangelo did./It was built by Michelangelo.
It seems that the building event of this church is presupposed, and the assertion is not about its existence, but rather who the agent is. In Section 3.1, I will elaborate on this idea and propose that these ‘out-of-the-blue’ uses of the past tense have in common that the reference time is presupposed to exist and be unique. In other words, the English past tense is lexically ambiguous between an anaphoric past and a unique past. Examples like (24) reflect one way of satisfying the uniqueness of the reference time, namely, letting it be the time span of the unique change-of-state event giving rise to the current state.
There are two other ways in which this uniqueness presupposition can be satisfied: (i) when the event is assumed to have taken place in the past; (ii) when the reference time is taken to be the (past) lifetime or time span of a person or entity under discussion. I will illustrate these in Sections 3.2 and 3.3.
In the rest of this section, I will show that the uniqueness of the reference time is indeed a presupposition, and compare it with patterns observed with anaphoric and unique definites in the nominal domain. I will also discuss some evidence supporting analyzing uniqueness as a separate presupposition of the English past tense.
Let us first consider a group of data that always license the past tense without a contextually salient reference time.
(25)
(Talking about my colleague Mary, without any contextually salient past time:)
Mary was born in New Zealand./#Mary has been born in New Zealand.
(26)
(Mary is looking at a new book on Sue’s desk. There is no contextually salient past time:)
a.
Where did you get it?/#Where have you gotten it?
b.
Who wrote this book?/#Who has written this book?
(27)
(Pointing at a church, without any contextually salient past time:)
Who built this church?/#Who has built this church?
Borromini built this church./#Borromini has built this church.
(28)
(Uttering the sentence knowing that we already have the technology of printing. There is not contextually salient past time.)
The art of printing was discovered by Gutenberg./#has been discovered
(29)
(Mary enters the office. She opened the window an hour ago, but now it is closed. She asks:)
Who closed the window?/#Who has closed the window?
(30)
(Mary is looking at a huge mess in her kitchen. She asks her roommates angrily:)
Who did this?/#Who has done it?
In each of these examples, it is true that the Common Ground entails a state which must have followed from a previous change-of-state event (
In other words, the reference time in these examples is the time span of the unique change-of-state event that gives rise to a state in the Common Ground. Since time is dense, there can be infinitely many large intervals in which the unique change-of-state event is located. In order to guarantee uniqueness and be formally precise, I adopt von Fintel et al.’s (
(31)
For a temporal property
a.
b.
For all other
{
i.e.
Here, we can define
(32)
where
In other words, the first group of contexts where the presupposition of the unique past tense can be summarized as:
(33)
If the context satisfies:
a.
There is a unique change-of-state event
b.
The event under discussion is that change-of-state event;
then the reference time is the (unique) maximally informative time interval with respect to: ∃
The other group of data where the past tense is obligatory even without a previously mentioned past reference time is shown in (34). Unlike the examples in the previous subsection, these examples do not involve a perceived change-of-state event.
(34)
(Uttered without contextually salient past times:)
a.
(We know Sheldon is an adult.)
Sheldon was raised in Texas.
b.
(Leonard is currently in his late twenties and we know that he has a degree:)
Leonard went to Princeton.
c.
(Penny is currently in her late twenties:)
Penny didn’t go to college.
d.
(Bill is currently in his late twenties:)
Bill didn’t finish high school.
(
These examples suggest that we need a different mechanism to derive the reference time in these sentences. Let us consider what they have in common. In (34a), we know that there must have been such a time during which Sheldon was a child. In (34b), the speaker is assuming that Leonard indeed went to university, and the reference time is the interval corresponding to his university years. In (34c), on the other hand, the reference time is the interval in which Penny could have gone to university, except that she did not actually go, unlike in (34b). The same applies to (34d), where Bill is assumed to have been in high school and he was supposed to have graduated. The assertion is that he did not actually graduate during that time.
Unlike in the existential past analyses, the past reference time is not asserted as new information, but rather as a background assumption. I argue that, like the examples in the previous subsection, the past reference time is also presupposed to exist in these examples, and they also satisfy uniqueness. In particular, the context entails that there is a unique past interval in which a particular event is expected or assumed to have taken place, but it may or may not actually did.
The following example further confirms that the reference time should be presupposed rather than asserted:
(35)
Mary is 12. #She didn’t go to university.
In (35), if the past tense here is simply existential, the sentence should be just true: there does not exist a past interval in which Mary goes to university. However, native speakers judge (35) to be infelicitous instead of simply true or redundant. Following the reasoning presented above for (34), (35) is infelicitous because Mary is only 12, and our world knowledge tells us that we probably cannot find a past interval in which she was supposed to be in university. (37) shows that if the context entails information that licenses this assumption, the past tense improves.
(36)
(Mary is a twelve-year-old child prodigy. Her school usually recommends children like her to attend university classes by the end of the fourth grade.)
But Mary didn’t go.
We can compare the past tense with the present perfect for this reading. The present perfect version of (35) here is fine, without the oddness of (35):
(37)
Mary is 12. (Of course) she hasn’t been to university.
The contrast between (35) and (37) further suggests that the present perfect here has the simple existential reading (cf. the existential perfect in Iatridou et al. (
Unlike the change-of-state examples in the previous subsection, the reference time for these expected and assumed events can be accommodated more easily. Whether accommodation takes place affects the felicity of the present perfect. (38) illustrates the subtle difference between the present perfect and the past version of the same sentence.
(38)
(I’m introducing my friend Alex to another friend. There is no contextually salient past time:)
a.
This is Alex. She didn’t go to university.
b.
This is Alex. She hasn’t been to university.
(38a) sounds a bit more condescending, because it has an assumption that there is a past time in which Alex could/should have gone to university, but she did not for some reason. (38b) does not rely on this assumption. In the former case, the listeners also need to accommodate this assumption upon hearing the past tense. This judgment is further confirmed by the following example, where the past tense is strongly preferred:
(39)
(Bill, who is 25, comes from a family that greatly values education. All his siblings have a degree.)
But Bill didn’t go to/hasn’t been to university.
Since Bill is already 25 years old, given what we know about his family, he was probably expected to enroll in college when he was about 18. The sentence then asserts that he did not actually go during that time. The present perfect sentence, on the other hand, does not make an assertion about that past interval, and simply asserts that Bill hasn’t been to university (yet).
Having concluded that these examples indeed involve a presupposed past reference time, let us now consider how this interval is unique. Like before, we can adopt maximal informativeness. In plain words, the reference time in these examples should be the unique maximally informative interval with respect to a temporal property like
(40)
a.
The ordering source is a function that assigns to any evaluation world
b.
For a set of worlds
∀
c.
For any set of worlds
Even if the evaluation world is the actual world
Having defined the set of best worlds according to this standard, we can say that if the speaker believes that a
(41)
an interval
The unique maximally informative interval with respect to this temporal property will be the unique minimal time interval that would envelop all the possible
Let us illustrate it with an example.
(42)
Let
a.
Suppose the best possible worlds according to the speakers’ expectations about Penny’s education in
(We actually don’t know if she actually went to university in
b.
The unique maximally informative
In addition, it is important that the
(43)
In this case we would not be able to find a unique maximally informative time interval with respect to this property. For instance, if we do not know if the state of Sheldon being a boy has ended or not, and suppose in some world in the Context Set
I believe this is a good prediction because in all of the examples with the expected event/state, the event or state is assumed to have taken place and completed/terminated.
For sentences in (34), this analysis will give us the unique time intervals such as Sheldon’s childhood span, the time in which Leonard/Penny was at the appropriate age for university, and the time in which Bill was supposed to be in high school. I argue that these intervals are the reference times for these sentences.
We can also test this theory by constructing some new contexts, such as:
(44)
(It is required by law that every child receives the polio vaccine before the age of 7. There is no contextually salient past time when the following is uttered:)
Mary (walking with her friend, passing a hospital): I was/#have been vaccinated for polio at this Hospital.
As expected, in (44), the past tense can be used without any contextually salient antecedent time. In addition, using the present perfect seems to suggest that it is not required by law that everyone must take the polio vaccine.
We can summarize the second case of uniqueness as:
(45)
If the context satisfies:
a.
There is a set of stereotypical expectations and Common Ground information, according to which some event is expected to have taken place in the past;
b.
The event under discussion is such an event;
then the reference time is the (unique) maximally informative time interval with respect to (41) above.
The English past tense is also obligatorily used without an antecedent when talking about a dead person (46a), or an entity that no longer exists (46b).
(46)
(Uttered without a contextually salient past time:)
a.
Madam Curie was a physicist.
b.
The Northern Wei was an imperial dynasty that ruled over Northern China.
When the speakers are ignorant about whether the person/entity under discussion still exists, using the past tense without an antecedent has an inference that the the person/entity no longer exists. This inference is especially strong with individual-level predicates.
(47)
(Uttered without a contextually salient past time:)
Mary was very talented.
It is tempting to analyze the inference in (47) as a result of failing to use the present tense to describe a living person (
(48)
a.
#Madam Curie has been a physicist.
b.
#The Northern Wei has been an imperial dynasty that ruled over Northern China.
There is nothing obvious in the assertive semantics of the present perfect that prohibits these examples: it is clear that in some time span leading up to the speech time, there is a bounded state of Marie Curie being a physicist and the Northern Wei being an imperial dynasty. Since we are describing past states, these examples cannot be covered by the competing present tense analysis. We will also need to derive the infelicity of the present perfect. In addition, as Matthewson et al. (
(49)
(Talking about Einstein. There is no contextually salient past time.)
a.
Einstein visited Princeton.
b.
#Einstein has visited Princeton.
(50)
(Talking about Princeton. There is no contextually salient past time.)
a.
#Princeton was visited by Einstein.
b.
Princeton has been visited by Einstein.
(49)-(50) are a set of famous examples illustrating how the topic can affect the felicity of the tenses. In both examples, there is no contextually salient past time that would license the anaphoric use of the past tense. However, in (49), given the common knowledge that Einstein is no longer alive, the past tense is chosen over the present perfect. In (50) where the topic is Princeton (which still exists), we observe the opposite pattern. There are various proposals in the literature on the present perfect that try to derive this pattern, such as Inoue (
I argue that the past tense is obligatory in all the examples above because the uniqueness presupposition is satisfied. The reference time is taken to be the lifetime or the time span of the person/entity under discussion, which is a unique interval.
In this case, the relevant temporal property is:
(51)
The unique maximally informative interval
We can summarize the third case of uniqueness as:
(52)
If the context satisfies:
a.
The person or entity under discussion no longer exists;
then the reference time is the (unique) maximally informative time interval with respect to (51) above.
There is a group of data worth mentioning. It has been noted by Heny (
(53)
How did Cicero die?
He was executed by Marcus Antonius.
(54)
Shakespeare said ‘In many’s looks the false heart’s history is writ.’
Partee (
Under my analysis, both (53) and (54) would be instances of the uniqueness reading. Both the reference times are unique (the lifetime of Cicero or Shakespeare), and the asserted events are located in that time.
There seems to be an implicit assumption in these previous accounts that if the reference time cannot be identified, then it is not an anaphoric reading. However, it seems that identification is not necessary in general, even for the anaphoric reading. The classic stove example in Partee (
(55)
(At the office, Bill suddenly tells his friend:)
I didn’t turn off the stove!
The sentence is naturally interpreted as about the ‘20 minutes before Bill left the house this morning’. The speakers do not need to first have that interval in mind (or as a topic), since this interval should be easily accommodated given the speakers’ world knowledge. The speakers also do not need to know when exactly Bill left the house.
Similarly, a sentence like (56) sounds perfectly natural, despite the absence of an explicit past time.
(56)
(Two friends meet up after a long time.)
How are you?
I bought a new car!
The reference time of the past tense sentence is naturally interpreted as about ‘the past interval since we last saw each other’. Again, the speakers do not need to identify the exact date they last met. This example does not seem so different from the stove example above in this sense. In the stove example, there are (infinitely) many intervals in the past during which the speaker did not turn off the stove. In (56), likewise, the speaker could have bought a car many times
There is one group of data that cannot be subsumed under the uniqueness analysis, however. It involves the use of
(57)
(There is no contextually salient past time:)
a.
I already told you, I’m not interested!
(
b.
Can you clean the kitchen?
I already did. (…and I mowed the lawn too.)
The use of the past tense in these examples seems indistinguishable from the resultative use of the present perfect. Unlike the earlier examples such as (55) and (56), in (57b), it is difficult to come up with an appropriate past interval that would serve as the antecedent for the past tense. The reference time does not seem to satisfy any kind of uniqueness either.
Since these examples with
The use of Maximal Informativeness to derive uniqueness and the relevant smallest and largest intervals is similar to the proposal of Von Fintel et al. (
(58)
a.
⟦the
When defined, the reference of
b.
For all
(
If
(59)
a.
the height of this tree
b.
(59a) refers to the maximally informative measure with respect to
On the other hand, if
(60)
a.
the number of people sufficient to lift this piano
b.
We can see that (60a) refers to the smallest possible number of people who can lift the piano in
This result is similar to our analysis in the previous subsections. In Section 3.1 and 3.2, the unique past tense denotes the smallest possible interval
We can also show that the uniqueness of the past tense is a presupposition by standard presupposition projection tests. Two typical environments where presuppositions project out are negation and yes-no questions. We can see that the existence and the uniqueness inference survives all of these environments, for both change-of-state and non-change-of-state events.
(61)
(There is no contextually salient past time:)
a.
This church wasn’t built by Borromini./It is not the case that this church was built by Borromini.
b.
Was this church built by Borromini?
(62)
(There is no contextually salient past time:)
a.
Bill didn’t finish high school.
b.
Did Bill finish high school?
(63)
Marie Curie wasn’t a mathematician.
In addition, we also other reasons to believe that this kind of past tense cannot be existential. Chen et al. (
Applying these tests to the English past tense, we can see that the unique past tense does not fit the criteria for the existential tense, in that it does have the bound use (see next section) and it is also scopeless, like the pronominal past. For example:
(64)
This church wasn’t built by an Italian.
The interpretation of the past tense in (64) is independent of the negation, just like in
In the discussion above, we tried to determine the reading of the past tense using mostly one criterion: whether the past tense is felicitous without a previously mentioned antecedent. Sometimes the judgement can be subtle, especially for expected events. In this subsection, I present a group of data unnoticed in the previous literature, which systematically determines which kind of presupposition the English past tense has.
In particular, the two readings of the English past tense correspond to the two readings of the definite article. One way to distinguish the anaphoric and unique definites is the presupposition projection pattern in quantified sentences.
Briefly, in a sentence of the form
(65)
Every student saw it1.
(66)
Every student who has a1 cat pets it1.
The same applies to an anaphoric definite. For example, the German strong article (as opposed to the weak article) has been shown to be strictly anaphoric (
(67)
In
in
jeder
every
Bibliothek,
library
die
that
ein
a
Buch
book
über
about
Topinambur
topinambur
hat,
has
sehe
look
ich
I
#im/
in-theweak/
in
in
dem
thestrong
Buch
book
nach,
ob
whether
man
man
Topinambur
topinambur
grillen
grill
kann.
can
‘In every library that has a book about topinambur, I check in the book whether one can grill topinambur.’
(
On the other hand, if
(68)
In every car, the steering wheel is on the left.
If the unique definite is a larger situation definite, whose uniqueness is evaluated globally, then the sentence does not have a covarying reading.
(69)
Everyone in this company hates the boss.
Here,
Before we proceed, I will first address a theoretical issue: can this contrast justify the separation of the two definite readings?
Under a familiarity-only analysis of definites, such as Heim (
(i) For a sentence of the form
(ii) The sentence (66) with an anaphoric pronoun under the covarying reading does not have a presupposition any more, while (68) with the unique definite still has a presupposition that each
(iii) The antecedent-anaphor relationship between an indefinite and an anaphoric element seems to be special, and cannot be simply captured by uniqueness-only accounts with domain restriction. and it is related to the problem of the formal link in the literature (
(70)
(Assuming all couples here are in a cisgender relationship.)
a.
Every married woman is sitting next to him.
# under the covarying reading.
b.
Every married woman is sitting next to the man.
# under the covarying reading.
c.
Every woman married to a man is sitting next to him.
In (70a), the anaphoric pronoun
(71)
a.
(Assuming all couples here are in a cisgender relationship.)
Every married woman is sitting next to the husband.
b.
(Assuming that each book here has a unique author.)
Every book is signed by the author.
Interestingly, relational nouns require the anaphoric-only strong article in German (
For the purpose of this paper, the crucial observation is that under the uniqueness reading of the English past tense, the presupposition projection patterns with the covarying reading of unique definites in (68), which would be unexpected if it were strictly anaphoric.
To test this, first observe that the English present perfect can provide an antecedent for the anaphoric past tense.
(72)
a.
Mary has met Sue already. She met her at a party.
b.
Sue has been to Egypt. She saw some pyramids.
Without getting into the details of the present perfect, I will simply use the present perfect as an indefinite in the testing the antecedent-anaphor relationship.
The past tense in the following sentences can be uttered without a previously mentioned antecedent, illustrated in the (a) sentences below. This is the case for perceived change-of-state events (73a), expected events (74a), and lifetime (75a). The (b) sentences show that when we put the past tense under
(73)
(Looking at churches in a town. There is no contextually salient past time.)
a.
This church was built by an Italian.
b.
Every church in this town was built by an Italian.
(Satisfied.)
(74)
(At a gathering of alumni from several American Universities. There is no contextually salient past time.)
a.
This woman here graduated from an American university.
b.
Every person here graduated from an American university.
(Satisfied.)
(75)
(Knowing that Einstein and Marie Curie are no longer alive. There is no contextually salient past time.)
a.
Marie Curie won the Novel Prize in physics.
b.
Both Albert Einstein and Marie Curie won the Nobel Prize in physics.
(Satisfied.)
(76)
(Talking about dead famous people from different eras–Marie Curie (20th century), Michael Faraday (19th century), Antoine-Laurent de Lavoisier (18th century); Archimedes (ancient Greece). There is no contextually salient past time.)
a.
Marie Curie made several important scientific discoveries.
b.
Each of these people made important scientific discoveries.
(Satisfied.)
To compare, under the strictly anaphoric interpretation of the past tense, we will not be be able to get the covarying reading. This is illustrated in (77). (77) on its non-habitual interpretation, strongly suggests a non-covarying, anaphoric reading that every boy danced during the same salient past interval (e.g. at the party last Friday).
(77)
Every boy danced with Mary.
As I mentioned above in (72), it is possible to use the present perfect as a sort of indefinite past. We can introduce an antecedent to the anaphoric past tense in the nuclear scope with the present perfect in the restrictor of
(78)
a.
Every girl who has been to this club danced.
b.
Every boy who has visited Paris also went to Marseilles.
In addition, the previous accounts reviewed in Section 2 cannot account for this contrast satisfactorily. One major issue is that they cannot control for the distribution of the ‘out-of-the-blue’ (i.e. unique past for us) use of the past tense. In addition, analyzing the past tense as existential fails to account for the presupposition projection patterns observed.
In the previous section, we established that there are data which cannot be satisfactorily explained with a pure pronominal (i.e. anaphoric) analysis of the English past tense, and we also found evidence that these instances of the past tense are actually similar to unique definites. In other words, the English past tense (with the relevant temporal property) behaves more like the definite article
This novel observation opens up a possibility of analyzing the English past tense on par with the analysis of definites in the recent literature, where familiarity and uniqueness are separated as sub-concepts of definiteness, and that languages may have different lexical items for each definite reading (
(79)
a.
b.
Schwarz (2009: (
For the English past tense, since both the anaphoric and the uniqueness readings are expressed with the same morphology, we may suggest a lexical ambiguity analysis. However, there is a concern: in general, even if other languages may morphologically distinguish the two readings, there is no good reason to believe that English is lexically ambiguous, in both the nominal (
Some commonly used tests for lexical ambiguity include: conjunction, ellipsis, and contradiction, illustrated below (
(80)
a.
The colours are light.
b.
The feathers are light.
c.
??The colours and feathers are light.
(81)
I saw his duck and swallow under the table and I saw hers too.
(82)
The bank isn’t a bank.
(80) tests the two senses of the adjective
If a word is underspecified, however, these tests would give us the opposite results: the conjunction would be allowed, the elided part may have the other reading, and we would observe a contradiction. For example, in English,
(83)
a.
John has a son.
b.
Mary has a daughter.
c.
Both John and Mary have a child.
(84)
Mary likes her child, and John does too.
(85)
a.
#John’s child isn’t a child.
b.
#John has a boy, and he doesn’t have a child.
Similarly, NPs in classifier-languages, such as Mandarin Chinese, is underspecified for number. We have the same test results:
(86)
a.
Zhangsan
Zhangsan
you
has
yi
one
zhi
mao.
cat
‘Zhangsan has a cat.’
b.
Lisi
Lisi
you
has
liang
two
zhi
mao.
cat
‘Lisi has two cats.’
c.
Zhangsan
Zhangsan
he
and
Lisi
Lisi
dou
all
you
have
mao.
cat
‘Both Zhangsan and Lisi have cats.’
d.
Zhangsan
Zhangsan
xihuan
likes
ziji
self
de
mao.
cat
Lisi
Lisi
ye
two
shi.
‘Zhangsan likes his single cat, and Lisi does too. (Lisi likes his own multiple “cats”)’
(87)
a.
#Zhangsan
Zhangsan
de
mao
cat
bu
shi
mao.
cat
‘(Intended:) Zhangsan’s cat is not multiple “cats”.’
b.
#Lisi
Lisi
you
has
liang
two
zhi
mao,
cat
suoyi
therefore
ta
he
bu
shi
you
has
mao.
cat
‘(Intended:) Lisi has two cats, therefore it’s not the case that he has a single cat.’
The general patterns of lexical ambiguity and underspecificity with respect to the tests are summarized below:
# | ✔ | |
same interpretation only | different interpretations allowed | |
✔ | # |
Testing the English past tense for the two readings can be tricky, because even if we have conjoined NPs as an argument of the VP, there is still one tense in the sentence. However, we can still construct some examples, such as the following. Imagine that we are in Florence, and the speaker is pointing at the famous dome of the Florence Cathedral. The background information is that both Brunelleschi and Ghiberti were involved in the project in 1418, but Ghiberti was only briefly involved and Brunelleschi eventually got all the credit for the construction of the dome:
(88)
(Pointing at the dome of the Florence Cathedreal:)
Who built this dome? And what did Ghiberti do in 1418?
??Ghiberti and Brunelleschi built this dome.
(89)
What did Penny do Monday night? And what about Leonard?
They both hung out with Sheldon (on Monday night).
Here, it is impossible for the past tense to both be unique (as in
Consider the following example. Here the background information is that Marie Curie and Pierre Curie were both involved in the discovery of radium in 1898.
(90)
Who discovered radium? And what did Pierre Curie achieve in 1898?
?Marie Curie and Pierre Curie discovered radium in 1898.
Here, the sentence is again odd despite the fact that there is nothing wrong with the statement itself. Again, I take it to be a clash between the two senses of the past tense, in particular, their different presuppositions. From the discussion in the previous sections, we conclude that sentences with the unique past tense in general presupposes the existence and the uniqueness of past interval (but crucially not which interval it exactly is).
This presupposition is satisfied when the speakers talk about the discovery of a substance known today. Hence, (90) with the unique past tense has the discovery event (and the unique reference time) as old information, and both
A related point here is that even without the conjunction, a sentence like
(91)
a.
Who discovered radium?
MARIE CURIE discovered radium.
b.
What did Marie Curie achieve in 1898?
Marie Curie discovered RADIUM.
If we take away the temporal adverbial
A reviewer points out that in (92a), by letting both answers have subject-oriented focus, the sentence becomes acceptable. However, since it involves a past temporal adverbial, the past tense is already automatically anaphoric to that time, regardless of the fact that the first question involves a unique past tense. Let us compare it with (92b) without the adverbial. Unlike (92a), (92b) sounds more like the speaker is only answering the first question (i.e. with a unique past tense). Likewise, (92c) with elided TP(s) also seems to suggest that the speaker is only answering on of the questions. Both pattern with the lexical ambiguity ellipsis examples (i.e. same interpretation only).
(92)
Who discovered radium? And who was involved in a significant achievement in 1898?
a.
Marie Curie and Pierre Curie discovered radium in 1898.
b.
Marie Curie and Pierre Curie discovered radium.
c.
Marie Curie and Pierre Curie.
Now let us consider lifetime as reference time. Assume that Sheldon is a physicist who is still alive, and he got a grant in 2010 to work on quantum mechanics. Max Planck, who is no longer alive, was one of the founding fathers of quantum mechanics.
(93)
What did Sheldon do in 2010? And who was Max Planck?
?They both worked on quantum mechanics.
Again, the utterance in (93) seems odd, and I suggest it follows from the clash of the two interpretations of the past tense.
We may also try to conduct the test with expected events. For example:
(94)
(At a gathering of alumni from several American university:)
Everyone here went to college in the US.
Here, the context is rich enough to admit the presupposition that for each person, we can find a unique past interval during which they went to university. This reference time is the old information, and covaries with each person. The new information seems to be the adjunct
Without such contextual information,
(95)
Last fall, Mary’s younger sister went to college.
Given the contextual information above, we may try the conjunction test again:
(96)
(Introducing Mary and her younger sister.)
?Mary’s younger sister started college last year. She and Mary have something in common: they both went to college in the US.
(96) seems a little odd. There may be multiple reasons behind this judgement: (i) given that if Mary’s younger sister only started college last year, she should still habitually goes there and we will not be able to find a suitable unique interval that is in the past; (ii) this means that
To sum up, the conjunction test results suggest an ambiguity analysis of the two readings of the English past tense.
To apply the ellipsis test, we need TP ellipsis where one of the tenses is elided. It seems that the interpretation of the explicit tense determines the interpretation of the elided tense, just as what we expect with a lexical ambiguity analysis.
Let us consider the following scenario with perceivable change-of-state events. We are looking at two windows, one of them is John’s and the other Bill’s. We know that every night, they both keep their windows closed. John’s window is half open now, which means he must have opened it at some point. We also know that once Bill gets up, every half an hour he opens his window completely for five minutes and closes it after getting a good ventilation.
(97)
(Looking at John’s half open window and Bill’s now closed window.)
??John opened his window narrower than Bill.
(97) is odd in this situation because as we see, John’s window is half open while Bill’s window is currently closed, and the sentence would only be appropriate and true if the elided tense were anaphoric to the last time Bill opened his window. The oddness suggests that this reading is not available. Trying to interpret the elided tense as unique is odd because given what we see and what we know, we cannot find a unique maximally informative past interval that would serve as the reference time for the elided TP.
Alternatively, suppose we know that both John’s and Bill’s windows were closed at night and both are open now. The following utterance will be felicitous event without a contextually salient past time.
(98)
(Looking at both open windows.)
John opened his window narrower than Bill.
The contrast between (97) and (98) suggests a pattern parallel to that of lexical ambiguity.
The contradiction examples are easy to construct. For example, we may consider a context that satisfies the uniqueness presupposition only, such as:
(99)
(Everyone in the department has a PhD.)
a.
Professor Snape got his PhD from Harvard.
b.
Professor Snape got his PhD from Harvard, but he did not get it last year.
In (99a), we do not need a contextually salient past time to use the past tense, because the context satisfies the uniqueness presupposition. (99b) shows that if we have an anaphoric past in the second conjunct, the overall sentence is not contradictory, which patterns with typical lexical ambiguities.
For perceivable change-of-state events, we may construct the following sentence:
(100)
(Looking at a sand castle. There is no contextually salient past time:)
Mary built this sand castle. However, she didn’t build it yesterday.
Again, the first past tense has the uniqueness reading, since it does not require a contextually salient antecedent. The second past tense is anaphoric to
We can also let the first conjunct have an anaphoric past tense by adding a past adverbial.
(101)
(Looking at a new book.)
Leonard did not buy this book yesterday, but he was the person who bought it.
In (101), the past tense in the second conjunct is not anaphoric to any contextually salient time, but they are still felicitous because this context satisfies the uniqueness presupposition: there is a new book, so it must have been bought by someone in a uniquely identifiable past buying event (and the corresponding maximally informative past interval).
The conclusion of this section is that the two readings of the English past tense pattern with lexically ambiguous words regarding the three tests. I will conclude that the English past tense is indeed lexically ambiguous between an anaphoric past tense and a unique past tense, with different presuppositions.
In this section, I will present a possible analysis of the English past tense, assuming that it is indeed lexically ambiguous between the anaphoric and the uniqueness readings. In particular, I follow the literature on definites and assume that the anaphoric reading comes from dynamic binding of an index argument (
(102)
The PASTANA has the following definition:
(103)
where
I propose that the temporal property here is always a trivial one, namely,
The index
The unique past tense, on the other hand, does not have the index argument.
(104)
The PASTUNI is defined as:
(105)
PASTUNI
where
MI(
I also propose that the temporal property in (104) may take three possible values, corresponding to the three ways uniqueness can be satisfied. In particular, we have:
(106)
where
(107)
a.
b.
c.
an interval
(108)
The unique past tense can yield three possible values, depending on the temporal property with respect to with uniqueness is evaluated: (i)
In conclusion, we have evidence that the English past tense is not just anaphoric, but also has a uniqueness presupposition. It behaves more like the definite article
Sharvit (
Although the past tense morphology may not denote a past time if embedded under a future operator, such as
In this paper, there are many examples illustrating the obligatory use of the past tense, the sentences are presented with the present perfect as comparison. This follows from the general observation that the present perfect does not need a contextually salient reference time to be felicitous (cf. Kratzer’s (
It has been pointed out by a reviewer that ‘without a contextually salient antecedent’ is not the same as ‘discourse initial’. For example,
Kratzer (
It is important to distinguish a pronominal tense and a quantificational tense with contextually salient domain restriction. Pronominal tenses, like pronouns, are type
In addition, assuming that the temporal adverbial provides domain restriction gives rise to readings that are too weak, and results in unwanted entailments. This is known as the Bäuerle—von Stechow problem in the literature (
(i) a. John coughed exactly twice yesterday.
b. John coughed exactly once yesterday.
c. ∃
d. ∃
If
On the other hand, if we take
(ii) ∃
Grønn & Von Stechow (
Note that (34b) is ambiguous between a ‘going to Princeton for his degree’ reading and a ‘physically going to Princeton’ reading. Under the latter reading, (34b) is infelicitous unless there is a salient past reference time. This is likely due to how easy it is to assume the time in which this event is expected. See the discussion below.
Note that for
Usually, this is a situation exemplifying
There are several possible analyses of the present perfect that captures this observation. For example, Grønn & Von Stechow (
Of course, if the context contains enough information for the speaker to assume that each boy was expected to have danced with someone, then (77) can also get the covarying reading, with the past tense licensed by uniqueness.
One exception to the conjunction test is a construction known as the zeugma, where the speaker intentionally conjoins two senses of the same word, such as in
In fact, having
See also the discussion of the antecedent-anaphor relationship in Section 3.7.1.
For a list of standard abbreviations, refer to the Leipzig glossing rules.
I would like to thank the reviewers and editors of the ‘Definiteness and referentiality’ special issue of Glossa for their helpful and detailed comments, especially Radek Šimík for the super detailed reviews and help along the way (especially for his patience when I was caught in between covid anxiety and thesis stress). The paper wouldn’t have been finalized without your help! I would also like to thank the audience at the workshop on definiteness at DGfS 2019, where I presented an early version of this paper. The paper went through a lot of changes and improvements since then, and many linguists have given me guidance, valuable discussions, and suggestions. I would like to thank (in alphabetical order): Carla Bombi, Richard Breheny, Berit Gehrke, Henriëtte de Swart, Peter Jenks, Nathan Klinedinst, Lisa Matthewson, Takanobu Nakamura, Radek Šimík, Yasutada Sudo, and Gregor Williamson.
The author has no competing interests to declare.