Reference Time and English Tenses

Autor: Meyer-Viol, W.P.M. & Jones, J.S.

Linguist and Philos, 34
2011, págs. 223-256
ISSN: 1573-0549

Esta obra está protegida por el derecho de autor y su reproducción y comunicación pública, en la
modalidad puesta a disposición, se ha realizado en virtud del artículo 32.4 de la Ley de
Propiedad Intelectual. Queda prohibida su posterior reproducción, distribución, transformación
y comunicación pública en cualquier medio y de cualquier forma.
Linguist and Philos (2011) 34:223–256
DOI 10.1007/s10988-011-9100-y

RESEARCH ARTICLE

Reference time and the English past tenses

W. P. M. Meyer-Viol Æ H. S. Jones

Published online: 11 December 2011

Ó Springer Science+Business Media B.V. 2011

Abstract We offer a formal account of the English past tenses. We see the
perfect as having reference time at speech time and the preterite as having
reference time at event time. We formalize four constraints on reference time,
which we bundle together under the term ‘perspective’. Once these constraints
are satisfied at the different reference times of the perfect and preterite,
the contrasting functions of these tenses are explained. Thus we can account
formally for the ‘definiteness effect’ and the ‘lifetime effect’ of the perfect, for
the fact that the perfect seems to ‘explain’ something about the present, and that
the perfect cannot presuppose a past time point. We explain why perfect and
preterite can sometimes be interchangeable, and we offer a solution to the
‘present perfect puzzle’. We explain the unacceptability of notorious examples of
the perfect such as *Gutenberg has discovered the art of printing. We give
greater definition to the familiar notions of ‘current relevance’ and ‘extended
now’.

Keywords Past tenses  Perfect  Preterite  Reference time  Lifetime effect 

Presuppositions  Update

W. P. M. Meyer-Viol (&)
King’s College, London, London, UK
e-mail: [email protected]

H. S. Jones
Oxford University, Oxford, UK
e-mail: [email protected]

123
224 W. P. M. Meyer-Viol, H. S. Jones

1 Introduction

This paper proposes a formal account of the English perfect and preterite tenses,
which is centred around the notion of ‘reference time’.1 Like Reichenbach we see
the perfect as having its reference time at ‘speech time’ and the preterite as having
its reference time at ‘event time’. Reichenbach saw reference time as a time point or
interval but otherwise left the notion undefined. We view reference time as a
‘perspective’ at which certain constraints must be satisfied. In this respect our notion
has something in common with Klein’s Topic Time (1992), which is the time
interval for which a claim is made on a particular occasion. In our account, refer-
ence time brings with it four constraints. Once these constraints are satisfied in the
perfect at speech time and in the preterite at event time, the differences and overlaps
between these two tenses can be explained. The first constraint is that there is
always some matter to be settled (or ‘choice to be decided’, as we describe it) at
reference time: the ‘current relevance’ of the perfect can be accounted for by the
fact that, in the perfect, reference time coincides with speech time. The second
constraint relates to presuppositions. In both tenses speech time and reference time
are presupposed; this means that event time is not presupposed in the perfect, but it
is in the preterite. As a result, the perfect is unacceptable when predicating a past
event whose time point is already given. This explains why, for example, *John has
died in Paris. His death has been witnessed by his landlady sounds awkward since
‘died’ and ‘witnessed’ refer to the same time point, while John has died in Paris.
His death was witnessed by his landlady, does not. Third, positional adverbials
denoting time points must be either functions from speech time to reference time
(e.g. today, yesterday) or predicates on reference time (e.g. at six). This rules out
past time adverbials with the perfect, whose reference time coincides with speech
time, but licenses them with the preterite, whose reference time coincides with a
(past) event time. Fourth, both perfect and preterite require a referent at reference
time. This explains the ‘perfect lifetime effect’, i.e. the fact that the perfect is
unacceptable if it does not have a referent at speech time, although we argue that
‘lifetime effect’ is too narrow a term, as a referent need not have a lifetime.
These four effects, which together we call ‘perspective’, interact with the other
tense features of time reference and aspect. In terms of the time reference of the
event predicated, the perfect and preterite are both past tenses. We model two
aspects, a marked ‘continuative’ (i.e. imperfective) aspect and an unmarked aspect.
Together, the features of perspective and aspect define each (past) tense. Thus he
spoke has preterite-perspective and non-continuative aspect; he was speaking
has preterite-perspective and continuative aspect; she has spoken has perfect-
perspective and non-continuative aspect; and she has been speaking has perfect-
perspective and continuative aspect.
With our model of ‘perspective’ we can account formally for individual features
of the perfect and preterite such as the ‘definiteness effect’ and the ‘lifetime effect’,
and explain the unacceptability of notorious examples of the perfect such as

1
We do not offer a cross-linguistic account of these tenses although elements of our formalization may
be used in such an account.

123
Reference time and the English past tenses 225

*Gutenberg has discovered the art of printing. We also shed light on the notions of
‘current relevance’ and ‘extended now’ which have been used to characterise the
perfect. The interaction between ‘perspective’ and aspect allows us to account for
‘up-to-now’ perfects such as she has been living in Paris for ten years, and also to
model the ‘imperfective paradox’ (e.g. he has been/was climbing Everest), although
this effect is not confined to the perfect and preterite. We offer an explanation as to
why the English perfect and preterite are sometimes interchangeable. Finally we
account for the fact that in the pluperfect, but not the perfect, event time may be
modified by temporal adverbs.
This paper is organized as follows. The next section summarises the various
effects of the perfect and preterite described in previous treatments, and outlines our
own approach in informal terms. Section 3 sets out a formal model for the perfect
and preterite. In Sect. 4 we discuss in detail how the model accounts for all the
effects described in Sect. 2. Section 5 concludes.

2 Previous treatments

The English perfect and preterite tenses both predicate situations occurring before,
and evaluated in, the speaker’s present. The temporal truth conditions of the perfect
(she has come) and of the preterite (she came) are satisfied if the coming occurred
before speech time.2
The functions of the perfect and preterite clearly differ, however. Since
Reichenbach (1947), and even before,3 linguists have identified in the perfect tense
some connection between, on the one hand, what is predicated of the past and, on
the other, the time of speech. Reichenbach (1947, pp. 296–297) distinguishes the
preterite from the perfect in terms of ‘reference time’: in the preterite, reference
time coincides with ‘event time’, while in the perfect ‘reference time’ coincides
with ‘speech time’. Reichenbach does not define ‘reference time’ precisely, but
subsequent studies have developed the notion in various ways (see Rothstein 2008,
pp. 6–13). Reichenbach’s notion that the preterite is anchored to the past while the
perfect is anchored to the present has led to two broad interpretations of the perfect,
which may be summarized as ‘current relevance‘ and ‘extended now’. Thus Comrie
(1976, p. 52) defines the perfect tense as indicating ‘the continuing present rele-
vance of a past situation’ (‘current relevance’), while McCoard (1978, p. 18) sees it
as expressing a ‘past event within a time span which is continuous with the present’
(‘extended now’). The preterite, by contrast, while it may allow a similar link with
the present, does not mark it as the perfect does.
Both the ‘current relevance’ and ‘extended now’ approaches have been made
more precise since their original formulations. For example, Portner (2003) argues
that current relevance lies in the fact that the perfect presupposes that there is a
discourse topic at speech time for which the predicate in the perfect tense provides
an answer. A number of recent studies have developed the concept of ‘extended

2
Dowty (1979, p. 331): ‘the past and perfect have exactly the same truth conditions’.
3
e.g. Bryan (1936).

123
226 W. P. M. Meyer-Viol, H. S. Jones

now’, focusing in particular on the relationship between the interval in question and
speech time. This has led to the use of the term ‘perfect time span’ (e.g. Pancheva
and Stechow 2004; Rothstein 2008), which does not necessarily include speech
time.4
One difficulty faced by these approaches is to reconcile them with other, specific,
effects of the perfect that have been identified in the literature. Before setting out
our own model, we summarize these specific effects below.

2.1 The ‘definiteness effect’

According to this effect, the perfect predicates a situation in an indefinite past time,
whereas the preterite predicates a situation in definite past time. This effect is most
clearly shown in that the perfect typically may not occur with a ‘definite’ past-time
expression.5 We see this in the different acceptability of the following sentences:
(1) *I have seen him yesterday

(2) I have seen him today

(3) *I have seen him this morning [said in the afternoon]

(4) I have seen him this morning [said that morning]6

This observation is not new. Pickbourn (1789, pp. 30–32)7 defines the perfect tense
as ‘perfect indefinite’, and gives the following examples to illustrate the use of the
perfect and preterite with time expressions: Philosophers have made great dis-
coveries in the present century but Philosophers made great discoveries in the last
century.8
This effect will be accounted for in our model by the fact that first, as in
Reichenbach (1947), the reference time of the perfect is speech time and the ref-
erence time of the preterite is event time, and that second, all temporal expressions9
are interpreted relative to reference time. This explains the different acceptability of
sentences (1)–(4). However, ‘definiteness’ does not seem appropriate names for this
effect. What makes (1) and (3) unacceptable is just that they exclude speech time,
while (2) and (4) include it.

4
For a summary of the literature, see Portner (2011).
5
Binnick (1991, pp. 38–39): ‘We call the preterite (simple past tense) definite because it fixes the event
at some definite time in the past. [. . .] We call the perfect, on the other hand, indefinite, because it does
not assume a particular time.’
6
‘Definite’ time expressions are acceptable if they are, as McCoard (1978, pp. 53–54) dubs them,
‘iterative experientials’, e.g. Whenever I’ve seen him on Sunday morning he has looked like death.
7
James Pickbourn, in his Dissertation on the English verb (1789), described a number of the features of
the perfect which still preoccupy linguists. We shall refer in footnotes to some of his other observations.
8
Along similar lines Higginbotham (2009, p. 170) argues that the adverbials ‘in the past’, ‘before’,
‘recently’ are compatible with the perfect because they are ‘closed under temporal succession’, that is,
they do not exclude speech time.
9
Except durational adverbials, e.g. for ten years, or since 2009. See under the ‘Up-to-now perfect’
below.

123
Reference time and the English past tenses 227

2.2 Presupposition of past time

Michaelis (1994, p. 143) notes that, in the following narrative, the perfect is
anomalous after the first sentence.
(5) Hayward police have arrested the prime suspect in last week’s string of laun-
dromat robberies. # Two off-duty officers have confronted the suspect as he left
a local 7–11. # A back-up unit has been called in to assist in the arrest.
In the first sentence the perfect is acceptable (assuming it is the first sentence in
the narrative) because the sentence does not presuppose a past time-reference.
However, if the second and third sentences refer to the same events as the first, they
capitalize on, or presuppose, its time-reference, and the perfect is anomalous. The
preterite would be acceptable in both cases, because the preterite can presuppose a
past time reference. Michaelis’s analysis leads her (1994, p. 125) to characterise the
relationship between the perfect and preterite tenses in terms of markedness, with
the preterite unmarked for (temporal) ‘anaphoricity’ and the perfect, at least in some
of its uses, marked for its absence.10
Our formalization accounts for this effect by making a distinction between the
‘given’ and ‘new’ information in an utterance. The perfect introduces a past time-
point as ‘new’ information, but it may not presuppose one as ‘given’, while with the
preterite a past time-point is ‘given’.
In (5) the presuppositional effect is inter-sentential. However, the same effect
may apply intra-sententially. Kiparsky (2002, p. 4) points out that the following
sentence is unacceptable:
(6) *John has been born in Paris
If the speaker and listener both know John, there is always a time-point determined
at which he was born. As John’s birth must be presupposed at some past time-point
the perfect is, according to our account, correctly ruled out.

2.3 ‘Lifetime effects’

According to one definition of the ‘lifetime effect’, a sentence in the perfect must
refer to something which still exists at speech time, although there is no consensus
about what this ‘something’ is. Smith (1997, p. 108) argues that it is the gram-
matical subject, which would explain why, as Chomsky notes (1970, p. 85), the first
of the following sentences is unacceptable but the second acceptable (in both cases
when uttered after Einstein’s death):
(7) *Einstein has visited Princeton

(8) Princeton has been visited by Einstein

10
As Partee shows (1973, pp. 602–603), the definite interval need not be specified linguistically when the
preterite is used. As an example, I didn’t turn off the stove, when uttered out-of-the-blue half-way down
the turnpike, refers to ‘a definite interval whose identity is generally clear from the extralinguistic
context’.

123
228 W. P. M. Meyer-Viol, H. S. Jones

Others argue that it is the topic, or more generally, what is ‘referred to’,11 which
must still exist at the time of utterance. Following McCawley’s suggestion (1971,
p. 106), we might argue that *Einstein has visited Prínceton is unacceptable if the
topic of discourse is Einstein (with the sentence stress falling on Princeton), but
Eínstein has visited Princeton is acceptable if the topic is Princeton (with the
sentence stress falling on Einstein). Without reference to intonation, Inoue (1979,
p. 576) points out that this sentence would be acceptable if Einstein were repre-
sentative of a set of subjects which did exist in the present, as in the following:
(9) Einstein has visited Princeton, Yukawa has, Friedman has. . .
Inoue wrote this in 1979, after Einstein’s death but while Yukawa and Friedman
were still alive. However, even after the deaths of Yukawa and Friedman the sen-
tence seems acceptable, if we take the three men as creating a reference to a set (say
of Nobel Prize winners or great thinkers), which does still exist at speech time.
The following sentences, from McCawley (1971, p. 106), shed further light on
what can constitute a ‘referent’ of a verb in the perfect:
(10) Frege has contributed a lot to my thinking

(11) Frege has been denounced by many people

(12) *Frege has been frightened by many people

As a physical person capable of being frightened, Frege does not exist and therefore
cannot be the referent, but as a thinker through his work, he does, and the perfect is
in those cases acceptable.12
Our account requires that there is a referent at reference time, which coincides
with speech time when the perfect is used; this accounts for the effects above.
Examples (10) and (11) suggest that ‘lifetime effect’ is a misleading term for this
phenomenon, as do these two sentences:
(13) John has died

(14) The bomb has exploded

because they show that an entity may still be a referent after its lifetime.

11
Curme (1935, p. 321): ‘the person or thing referred to must be living or still existing, thus related to the
present’, cited by Inoue (1979, p. 561).
12
The ‘lifetime effect’ was noted by Pickbourn (1789, p. 34), who uses the following pair to illustrate the
constraint: (i) Priests have (in all ages) claimed great power and (ii) *The Druid priests have claimed
great power. ‘We may say [the first sentence]’, Pickbourn explains, ‘because the general order of the
priesthood still subsists. But if we speak of the Druids, or any particular order of priests, which does not
still exist, we cannot use [the perfect].’ The difficulty of identifying what in the sentence must still exist
for the perfect to be acceptable is also noted by Pickbourn (1789, pp. 33–34), ‘We may say,‘‘Cicero has
written orations’’; but we cannot say ‘‘Cicero has written poems’’. ½. . . We suppose Cicero, as it were,
still existing, and speaking to us in his orations; but as his poems are lost, we cannot mention them in the
same manner’. As these examples show, the acceptability of a sentence depends on the speaker’s and
listener’s knowledge, in this case their knowledge of the extant works of Cicero. Pickbourn, the keeper of
a boarding-house in Hackney, London, seems to assume that this was common knowledge at the time.

123
Reference time and the English past tenses 229

‘Lifetime effect’ is also a term that has been used to describe how, when the
preterite occurs with (some) stative predicates, the inference is that the referent no
longer exists. Musan (1997, p. 271) gives the following examples:

(15) Gregory was from America

(16) Gregory had blue eyes

(17) Gregory resembled Jörg Bieberstein

According to our account, reference time lies before speech time in the preterite.
Since these predicates denote a permanent quality, the use of such tenses forces the
conclusion that the referent no longer exists: in these cases this would typically
mean that the referent is dead. We shall call this lifetime effect the ‘preterite lifetime
effect’ as opposed to the ‘perfect lifetime effect’ mentioned before. However, we
also show that it is possible to find contexts where sentences such as those above do
not force the conclusion that the referent is dead. For example, the sentence Mrs
Thatcher was a great prime minister, and she was a woman can be uttered when
Mrs Thatcher is alive. The reason for this, we shall argue, is that the referent here is
Mrs Thatcher as prime minister, which is not a permanent quality, so that the
preterite is not ruled out during the referent’s lifetime.

2.4 The perfect as explanation

Two categories of perfect which have featured persistently in the literature are the
‘existential perfect’ and the ‘perfect of result’. With the ‘existential’ (or ‘experien-
tial’) perfect, according to Comrie (1976, p. 58), ‘a given situation has held at least
once during some time in the past leading up to the present’. Thus, when we say
(18) John has lived in Paris13
we are stating that at some point(s) in the past it is true that John lived in Paris.
However, this is an understatement of the function of the perfect. For, as long as the
‘definiteness’, ‘presuppositional’ and ‘lifetime’ constraints mentioned above are
satisfied, this definition merely restates the temporal condition of past tenses in
general. Moreover, sentences such as this are typically not uttered in isolation, and
there is always some point at issue in the present which the verb in the perfect
relates to. In (18), it might be to show that John is a man of the world, for example.
In the ‘perfect of result’, a past situation predicated by the perfect implicates a
result at the time of utterance.14 The result of
(19) John has arrived
could, according to Comrie (1976, p. 56), be that John is here. It is widely accepted
that the present result is an implicature which is pragmatically determined and is not

13
Example from Declerck (1991, p. 322).
14
Portner (2003, p. 471) defines the result in such cases as ‘a state which results from the past event
under the scope of have’.

123
230 W. P. M. Meyer-Viol, H. S. Jones

the only possible result in the circumstances.15 Thus, Declerck (1991, p. 325)
suggests that the sentence
(20) They’ve fallen into the river
could have various resultative interpretations, such as: ‘Please help them’, ‘That’s
why they are soaking wet’, ‘That’s how they got pneumonia’, ‘That’s why they are
late’. The problem with this analysis is that it is easy to think of implicatures of
these sentences which are not results in the speaker’s present. The implicature of
(19) could be that John left in good time, that he will catch his flight, or that he is a
reliable person, and the implicature of (20) could be that they took the wrong path,
that they will be late, or that they are careless. So although these sentences are
evaluated at the time of utterance, the implicature need not have present time
reference and need not even be a result.
Inoue (1979, p. 585) addresses this point when she argues that the perfect has an
‘explanatory’ sense. As she puts it, ‘a speaker of English uses a sentence in the
present perfect when the information he is giving appropriately exemplifies or
explains the topic of discourse’.16 Portner (2003, p. 464) argues along similar lines
when he refers to the type of perfect which ‘provides evidence for something, not
that it indicates any results’.
Our formalization accounts for this ‘explanatory’ effect by requiring that a
predicate always settles some matter (or ‘decides some choice’ as we call it in the
formalization) at reference time which, in the perfect, coincides with speech time.
The matter being settled may be explicit or implicit in context. This part of the
formalization can be seen as an extension of Portner’s (2003) modelling of the
perfect as a question and answer, in which the perfect operator provides ‘a complete
or partial answer to the discourse topic’ at speech time.17

2.5 The ‘up-to-now’ perfect

The perfect tense can be used with an adverbial of duration to bring the situation up
to the speaker’s present.18 Two examples from previous treatments are:
(21) Jack Norbert has taught at MIT since 196919
15
See for example Moens and Steedman (1988, p. 19).
16
Inoue also writes (1979, p. 577): ‘the relationship between the proposition in the present perfect and
the one in the topic of discourse is that of entailment in a broad sense, supplemented by the so-called
pragmatic presuppositions, i.e. the speaker’s assumptions in uttering a given sentence.’ She considers an
example (1979, pp. 585–586) in which the conversation is about whether Mrs Gandhi is publicity-shy and
inaccessible. The speaker thinks not, and says, I don’t know if she is inaccessible. John Chancellor of
NBC has met with her, and Walter Cronkite of CBNS has met with her, too, you know. In this case, neither
is Mrs Gandhi’s accessibility (or otherwise) a result of her meeting these people, nor is it only in the
speaker’s present. Rather, the situation predicated by the perfect provides evidence for a quality which is
not limited in time at all.
17
Portner: ‘A sentence S of the form PERFECTð/Þ presupposes: 9q½ANSðqÞ & Pðp; qÞ, where ANS is
true of any proposition which is a complete or partial answer to the discourse topic at the time S is uttered’
(2003, p. 501). ‘The operator P is similar to an epistemic must’ (2003, p. 499).
18
This use of the perfect is sometimes called ‘universal’.
19
Inoue (1979, p. 566).

123
Reference time and the English past tenses 231

(22) I have lived here for ten years20

These cases also allow an interpretation which does not bring the situation up to the
present:
(23) Jack Norbert has taught at MIT since 1969 (as well as before)
(24) I have lived here for ten years (and therefore have a right to residency)
An up-to-now reading is preferred, without an adverbial of duration, when the
perfect is ‘imperfective’, e.g.
(25) She has been living in Paris
When the perfect is both imperfective and combined with a durational adverb, an
up-to-now reading seems obligatory, e.g.
(26) She has been living in Paris for ten years
Our formalization accommodates these effects. When the perfect is not imperfec-
tive, as in (21)–(24), the durational adverbial can be interpreted as lasting until
speech time [(21)–(22)], or ending beforehand [(23)–(24)]. When the perfect is
imperfective, as in (25) and (26), an up-to-now reading is obligatory.

2.6 Interchangeability of perfect and preterite

As scholars have noted, the perfect and preterite are, in some cases, interchangeable.
Klein (1992, p. 531) points out that the preterite can have current relevance, as in:
(27) Why is Chris so cheerful these days?
Well, he won a million in the lottery
The perfect has won would also be acceptable in the second sentence. However, if it
is qualified by a definite past time expression, or if it presupposes a past time point,
only the preterite could be used, in line with Sects. 2.1. and 2.2. In our formalization
we model all tenses as deciding some choice at reference time. In the preterite this
coincides with event time. However, if some choice is explicit (as in the example
above) or implicit at speech time, it can be decided by implicature by the choice
decided at reference time (¼ event time). In this case the choice at reference time is
‘Did Chris win the lottery or not?’ and the choice at speech time is ‘Is Chris cheerful
these days or not?’ Indeed, in all tenses, the choice decided at reference time may
implicate the deciding of some choice at another time; for example, the choice
decided at reference time (¼ speech time) in the perfect may itself decide a choice
in the future or in universal time.

2.7 The pluperfect and the ‘present perfect puzzle’

The English perfect is unacceptable with definite time expressions which qualify the
event predicated (see Sect. 2.1). However, in the pluperfect, definite past time

20
McCoard (1978, p. 46).

123
232 W. P. M. Meyer-Viol, H. S. Jones

expressions are acceptable. If these time expressions qualify a Reichenbachian

reference time which falls between event time and speech time, the pluperfect may
be analysed as a perfect-in-the-past. However, time expressions qualifying the event
are also acceptable with the pluperfect. In the sentence
(28) Chris had left at six
we may interpret ‘at six’ as referring either to the time of Chris’s arrival (event-time
modification) or to some later time by which he had arrived (reference time mod-
ification). The fact that event time may be qualified in the pluperfect but not in the
perfect is part of the ‘present perfect puzzle’ (see Klein 1992). In line with previous
treatments, e.g. Comrie (1976, p. 56), our formalization accounts for this effect by
analysing the pluperfect as either a perfect-in-the-past Pret(Perf) or a preterite-in-
the-past Pret(Pret).21

3 Formalization

The formalization will center around event time, speech time and reference time,
with the latter as a perspective point.
In a natural representation we have some linearly ordered number of time points
(at most three), one of which is annotated with some (tense-less) proposition. The
last point in the ordering we identify with the speech time (as we don’t deal with
future tenses); the annotated point we identify with the event time. The reference
time is not identified by structural means, but is interpreted as a perspective point
that may coincide with any of the temporal points. In our formalization the logical
form of a perfect like John has walked to school and a preterite like John walked to
school will be rendered as Perf ðJwtsÞ and PretðJwtsÞ respectively, where Jwts
represents the a-temporal ‘john-walking-to-school’ event, and ‘Perf ’ and ‘Pret’ are
the perfect- and preterite-operator respectively.
These forms are not interpreted directly. We associate them with a temporal
variable n, denoting the reference time, and translate them to representational
structures that can be depicted as follows:

In the structures there are elements of the following kind.

 Above the line a temporal variable may occur. Such a variable will be interpreted
as new, (alternatively: the variable is declared or introduced). All variables
occurring in the structure that do not occur above the line are called given
(alternative formulation: they are presupposed).
21
This puzzle includes the fact that the future perfect (Chris will have left at six) and the perfect
infinitive (Chris seems to have left at six) can also have event time modification. As we are discussing the
past tenses we confine ourselves to the way the ‘present perfect puzzle’ relates to the pluperfect, but our
analysis is applicable mutatis mutandis to these verb forms as well.

123
Reference time and the English past tenses 233

So, in the translation of the perfect the variable ‘m’ above the line is new, and placed
in the past of a given variable (i.e. n) to be qualified by a non-temporal formula (i.e.
Jwts). In the translation of the preterite the variables ‘m’ and ‘n’ occur only below
the line, so both are given, and the past point is qualified by the content.
 Below the line we find a temporal ordering of a number of variables, x  y,
inclusion statements n 2 i, and a propositional formula associated with one of the
temporal variables.22
The representational structures will be interpreted in, or ‘talk about’, a branching
time structure ðT; <Þ consisting of a set T of time points structured by the binary
relation < of ‘precedence’ as a directed order branching towards the future.
Within this structure we will locate (the interpretation of the) propositional
formulas that represent events as a collection of run-times. A run-time of an event in
this structure will be a right closed interval over T, that is, a <-linear and <-convex
subset of T including an <-last element.
The temporal variables are interpreted as <-unconnected sets of time points, that
is, as possibly simultaneous time points, and a formula like Jwtsx is interpreted as
true if all time points in the interpretation of the temporal variable x lie within a
Jwts-run time.
But in our analysis of the perfect-preterite contrast, the temporal structure of the
events is only relevant up to a point. In this contrast notions of presupposition and
relevance will fulfill a prominent role and their analysis will be of a general nature.
For both notions a dynamic interpretation of utterances seems appropriate, and, in
this sense, our analysis resembles Heim’s File Change Semantics (1982) and Kamp
and Reyle’s Discourse Representation Theory (1993): an utterance is considered to
be an update of an interpretation (of the language). That is, an utterance is asso-
ciated with an input interpretation and an output interpretation. ‘Presuppositions’
then constrain the domain of the utterance, the input interpretations, and an utter-
ance has ‘relevance’ if the output interpretation settles some issue, decides some
choice, at the input interpretation. Section 3.3 will introduce this semantics.

3.1 Propositional formulas

We introduce a logical representation language of tensed utterances for which we

will define a semantics. The language will be introduced in stages. In this section we
will define the language L1 in which (tenseless) content is expressed. In Sects. 3.2
and 3.4 we add to this language to get languages L2 and L, respectively.
The language L1 is built from atomic variables A1 ; A2 ; . . .. An L1 / will have the
shape
/ ::¼ Ai j :/ j ð/ _ :/Þ:
In order to interpret propositional variables A we will introduce the notion of an
event as a set of right-closed intervals over the branching time structure ðT; <Þ

22
The propositional variables could be distinguished by their ‘Aktionsart’, but in this paper we will not
attempt this (see Sect. 3.8).

123
234 W. P. M. Meyer-Viol, H. S. Jones

where < is the (binary) relation of temporal precedence. Time branches towards the
future, representing indeterminacy.23 A set i  T is a right closed interval if

1. 9w 2 i such that 8w0 2 i, if w 6¼ w0 then w0 < w

2. w < w0 < w00 and w; w00 2 i implies w0 2 i
So right closed intervals are <-linear and convex sets including a last element. We
collect these intervals in the set C.
Interpretation functions I 2 INT map a a finite number of atomic variables A to
sets IðAÞ  C of right closed intervals in the branching structure ðT; <Þ, the ‘run-
times’ of event A.
Within a set of intervals IðAÞ we will identify the set BndðIðAÞÞ of right bounds:
BndðIðAÞÞ ¼ fw 2 i j i 2 IðAÞ; w is the <-last element of ig
Clauses will be interpreted by atomic propositions, but we will also need to for-
malize the notion of a choice and for this we will introduce (a kind of) negation and
a choice construction: An L1 -formula / _ :/ will have the interpretation:
1. Ið/ _ :/Þ ¼ Ið/Þ [ Ið:/Þ,
and to interpret ‘negation’ we will introduce a function  2 PðCÞPðCÞ , that is a
function that maps subsets of C to subsets of C such that for X  C:
 X \ X ¼ ; (non-contradiction)
 X ¼ X (double negation)
This function can be interpreted as giving a specific alternative X to X
 BndðXÞ ¼ BndðXÞ
and we set for the L1 -formula /:
2. Ið:/Þ ¼ Ið/Þ:
Note that BndðIð/ _ :/ÞÞ ¼ BndðIð/ÞÞ [ BndðIð:/ÞÞ.

3.2 Temporal structure

In order to address temporal structure we will add to the language means to refer to
(speech, event, or reference) times in the form of a set of temporal variables and a
set of interval variables.

3.2.1 Temporal variables

Temporal variables are not interpreted as referring to time points: they are intended
to be part of a representational structure as either speech time, event time, or
reference time. These Reichenbachian time points cannot be identified with actual
time points or intervals (e.g. John has been killed in Paris. His body was found this
morning by . . . : In the first sentence the reference time coincides with the speech
23
This also allows a simple solution to the so-called ‘imperfective paradox’.

123
Reference time and the English past tenses 235

Fig. 1 k; l; m; n are unconnected sets in the branching time structure. According to Definition 1.4 we have
k  m, but k 6 l; k 6 n

time. In the second it lies at event time; the temporal structure stays the same, but
the perspective changes).
It may be worth emphasizing that the speech time is just as representational as the
event time and reference time; the speech time is not the actual time at which an
utterance takes place. A speech time may last minutes, that is, a sequence of
utterances may pile up information about a speech time and such a sequence takes
time. Moreover, a novel may start with ‘‘Churchill has been dead for two
years.. . .’’, locating the speech time in the past.
We will interpret the temporal variables as sets of time points which we will call
evaluation points. Evaluating at evaluation points allows for ‘partial’ interpreta-
tions, as truth at an evaluation point is then naturally interpreted as ‘true at all time
points in the set’, and falsehood at an evaluation point as ‘false at all time points in
the set’, and neither situation need obtain.
For a set of time points to be an evaluation point it has to be an unconnected set u  T:
that is, for all w; v 2 u, neither w < v nor v < w, no temporal precedence relations hold
between elements of an unconnected set. An unconnected set selects time points from
different paths in the branching time structure and, because there are no temporal
relations between them, can be seen as a collection of possibly simultaneous time points
across branches of the structure. Fig. 1 represents the evaluation points.
In the language we introduce temporal variables, n; n0 ; n00 ; m; m0 ; . . . collected in
the set EV to range over unconnected sets.

3.2.2 Interval variables

Our formalization of continuative (or imperfective) tenses will require reference to

intervals over the branching structure. We first define the notion of a convex set: a
set X  T is convex if w < w0 < w00 and w; w00 2 X imply w0 2 X. So a convex set
consists of a selection of intervals in the branching structure.
In the language we introduce interval variables, i; i0 ; . . . collected in the set IV to
range over convex sets. Every temporal structure ðT; <Þ then determines a qua-
druple ðT; <; U; CÞ where U is the collection of non-empty unconnected sets,
evaluation points, and C the collection of non-empty convex sets.

123
236 W. P. M. Meyer-Viol, H. S. Jones

The temporal and interval variables are used to construct (atomic) L2 -formulas of
the following three forms:
 Content formulas. /x is an atomic L2 -formula, for / an L1 -formula and
x 2 EV [ IV.
 Ordering statements. m  n is an atomic L2 -formula, for n; m 2 EV.
 Inclusion statements. n 2 i is an atomic L2 -formula, for n 2 EV and i 2 IV.
The atomic formulas can be combined using the propositional connectives ^
(conjunction), _ (disjunction), and  (material implication). So this language can
express temporal ordering, membership of an interval, and propositional content at a
temporal point or interval. We extend the interpretation functions I to interpret the
EV variables on unconnected sets: IðnÞ 2 U for all EV-variables n such that n 6¼ m
implies IðnÞ 6¼ IðmÞ, and IðiÞ 2 C for all i 2 IV.
We now define the notion
I  ... ‘‘ . . . ’’ is true under interpretation I;
for the formulas of the language L2 (see Fig. 2 for a representation).
Definition 1 For content formula /n the notion ‘I  /n ’ (formula / is true at
evaluation point n), is defined by
1. / is true at time n: I  /n () IðnÞ  BndðIð/ÞÞ
2. / is true at interval i : I  /i () IðiÞ  Ið/Þ S
3. evaluation point n lies in interval i : I  n 2 i () IðnÞ  IðiÞ
4. evaluation point n precedes point m : I  n  m () 8w 2 IðnÞ9w0 2 IðmÞ :
w < w0 and 8w0 2 IðmÞ9w 2 IðnÞ : w < w0 .
3.3 The relevance part

The relevance of an utterance will consist in the fact that it settles some issue, or
decides some choice. Every utterance will be relevant in a strict sense to be
explained, but only in case of the perfect does this give rise to a present choice being
decided by an event in the past.
The formalization consists of two parts. First we fix what we mean by a ‘choice’
and then we formulate what it is to decide a choice.

Fig. 2 Temporal structure. I  Am ; I 6 Al ; I 6 :Al ; I 6 ðA _ :AÞl ; I 6 An ; I 6 :An ; I  ðA _ :AÞn

123
Reference time and the English past tenses 237

Fig. 3 Example of a choice being decided. m  n and I  ðA _ :AÞn . After updating m with B, n reduces
to n0 (Definition 1.4) and I  An0

By the semantics, the following cases can arise for L2 -formulas / and n 2 EV:
1. I  ð/ _ :/Þn , because I  /n or I  :/n
2. I  ð/ _ :/Þn ; but I 6 /n and I 6 :/n
3. I 6 ð/ _ :/Þn .
In case the second holds, we will say that ð/ _ :/Þ is a choice at n, and we let
CHI ðnÞ be the set of all choices at n (according to I): at every time point t 2 IðnÞ
either / holds or :/, but neither holds across IðnÞ. (3) above highlights that choices
do not hold at all evaluation points: generally CHI ðnÞ 6¼ CHI ðn0 Þ in case n 6¼ n0 . So
it makes sense to talk about the choices at n 2 U (according to I).
We introduce a choice predicate Chð/Þ in the language and set
I  Chð/Þn () ð/ _ :/Þ 2 CHI ðnÞ:24
The interpretation of the relevance of an utterance / at n is that it decides some
choice w _ :w at n, and this suggests a dynamic interpretation: utterance / maps an
interpretation function I, such that I  ChðwÞn , to an interpretation function I 0 that
extends I and such that I 0  wn , that is, the choice is decided, i.e. the disjunction is
split. Fig. 3 represents this situation.
Let’s define the components of this interpretation. The relation of extension is
defined as follows
Definition 2 Interpretation I 0 extends interpretation I, notation, I I 0 (or I 0
I), if
1. DomðIÞ  DomðI 0 Þ.
I 0 is defined over I’s domain

2. for all v 2 DomðIÞ : IðvÞ  I 0 ðvÞ, where ‘’ is some partial order on the range
of I.25
Now, based on the notion of one interpretation extending another, we can give
the dynamic semantics of L2 -formulas.
24
Note I  Chð/Þ () I  Chð:/).
25
The relation ‘’ on the range of the interpretation functions I may be interpreted as ‘identity’,
‘inclusion’, ‘embedding’, ‘injection’, etc., but for our purposes we will set: if I I 0 then for temporal
variables IðnÞ  I 0 ðnÞ () I 0 ðnÞ  IðnÞ, for all other vocabulary elements v; IðvÞ  I 0 ðvÞ ()
IðvÞ ¼ I 0 ðvÞ.

123
238 W. P. M. Meyer-Viol, H. S. Jones

Definition 3 We associate with an L2 -formula X a set ½½X of pairs (I, I0 ) of

interpretations, such that
ðI; I 0 Þ 2 ½½X
implies
1. I is defined over all variables in X,
2. I I 0 ,
so I 0 is defined over the domain of I,
3. I 0  X,
so the target I 0 makes X true,
4. if I I 00 I 0 and I 00  X then I 00 ¼ I 0 ,
so, I 0 is a -smallest interpretation above I satisfying (3).
If ðI; I 0 Þ 2 ½½X, then we call I an input- and I 0 an output-interpretation of utterance
X. In case X ¼ /n , that is a content formula, we add the following conditions on
input and output:
5. if n 2 EV then for all interval variables i 2 DomðIÞ : IðnÞ  IðiÞ,
the evaluation point is included in all intervals identified by I,
6. I 0  wn for some w such that I  ChðwÞn ,
/ decides some choice at n.
As I 0 ðnÞ  IðnÞ in case ðI; I 0 Þ 2 ½½/n , the transition from I to I 0 ‘zooms in’ on a
subset I 0 ðnÞ of IðnÞ and the formulas that hold at IðnÞ will be amongst the ones
holding at I 0 ðnÞ but new facts may have been added (the update), in particular, an
I-question w _ :w at n may have been decided by I 0 .
By Definition 3.6 there is always some choice decided at n. In our representation
of the past tenses n will identify the reference time. So, in the perfect an update of
the (past) event time always decides a choice at the speech time, in the preterite
always at event time.
Information about the past decides a present choice as follows. Let n; m 2
DomðIÞ; I  n  m, ðI; I 0 Þ 2 ½½An  (we update the past of m with A) and
ðw _ :wÞ 2 CHI ðmÞ (ðw _ :wÞ is a choice at m). If now I 6 An , then I 0 ðnÞ  IðnÞ
and so, by Definition 1.4, I 0 ðmÞ  IðmÞ: by updating the past (n) of m, we update m
itself. So, I 0  wm is possible for ðw _ :wÞ 2 CHI ðmÞ.26
Note that, unlike the perfect, the preterite is not required to decide a choice at
speech time, but it may do so, that is, it may, or may not, have current relevance,
depending on the pragmatic situation. For instance, if a question has been asked
introducing a choice, then Grice’s maxims alone guarantee that the subsequent
utterance decides (or contributes to the decision of) that choice. So, if the question
refers to the present and the answer is in the preterite then, by Grice’s maxims alone,
the answer has current relevance. The same can also be true if some choice is
merely implicit at speech time.

26
Note that there need be no structural relation between the formulas that occur in the choice (w; :w)
and the formula deciding it (A). In that sense the ‘relevance’ of an utterance at n (relevant at n iff it
decides some choice at n) is independent of its structure.

123
Reference time and the English past tenses 239

This requires of our formalization that it be able to represent consecutive

utterances. In our dynamic interpretation there is a natural way to represent this. For
X and Y L2 formulas we introduce the binary connective ‘:’ where X : Y represents
that utterance X is followed by utterance Y, with the interpretation:
½½X : Y ¼ fðI; I 0 Þ j 9I 00 : ðI; I 00 Þ 2 ½½X; ðI 00 ; I 0 Þ 2 ½½Yg:
This formal definition will be essential for our analysis of utterances in the preterite
as answers to questions (e.g. Klein’s example in (27)), but it will also accommodate
Michaelis’s example in (5).
Finally, it is worth mentioning that the following situation may occur
½½ChðwÞn : /n   ½½wn ; but ½½/n  6 ½½wn :
Updating input interpretation I with a choice w _ :w and subsequently with /, at n
yields output interpretations I 0 under which w holds at n (the choice is decided), but
updating I with / alone will not output such interpretations: w ‘follows from’ / at n
only if there is a choice w _ :w to decide at n.

3.4 The presuppositional part

In our analysis the perfect and preterite will differ in their presuppositional struc-
ture. The perfect introduces a fresh past evaluation point, whereas a preterite pre-
supposes such a point. Consequently, in the perfect all information about the past
time point must be new.
We introduce in the vocabulary of a formula a division between given and new.
For n an evaluation variable, we call the new syntactic construct fng a declaration
and we prefix an L2 -formula / with a declaration fng giving a prefixed formula
fng : /. A vocabulary element in / that occurs in a declaration is called new (or
bound) in the prefixed formula fng : /, and all the remaining vocabulary elements
of / are called given (or free) in fng : /. In our motivating examples at the
beginning of the section this division is depicted as occurring above the line (new)
and occurring only below the line (given).
We extend the mapping ½½  over declarations by stipulating ðI; I 0 Þ 2 ½½fmg if
1. I I 0 ,
2. DomðI 0 Þ DomðIÞ ¼ fmg.
That is, m does not lie in the domain of I and the domain of the extension I 0 is that of
I plus the variable m.
Now we consider the representational language L for / an L2 -formula and fng a
declaration:
w ::¼ / j fng j w : w;

where ‘:’ was defined at the end of the previous section.

123
240 W. P. M. Meyer-Viol, H. S. Jones

3.5 Formalization of the tenses

Finally, putting the parts together, we can define our formalization of ‘perfect’,
‘preterite’, and ‘continuative’.
Definition 4 We represent utterances in the perfect, preterite and continuative, as
the abbreviations of the following L formulas:
1. Perf ð/Þn ¼df fmg : m  n : /m
2. Pretð/Þn ¼df fg : n  m : /n
3. Contð/Þn ¼df n 2 i : /i :
For clarity, we can represent the tensed formula as boxes with declarations above
the line and the rest listed below the line. For the perfect and the preterite the
translations are respectively:

In a tensed formula /n we will identify Reichenbach’s time points as follows: in /n

the reference time is (identified by) n, the speech time is (identified by ) the -last
element in the structure, and the event time is (identified by) the unique temporal
variable that is updated with the non-temporal formula. This gives Reichenbach’s
attribution of reference time, event time, and speech time for the perfect and
preterite.
This formalization addresses the presuppositional and relevance differences
between preterite and perfect:
 The preterite Pretð/Þn updates the reference time n in the past of the speech time,
whereas the perfect Perf ð/Þn arbitrarily chooses an evaluation point m that lies in
the past of the speech time n and updates it. In the perfect (where m is the fresh
variable) a past time point is declared (m) about which the interpretation function
I 0 in ðI; I 0 Þ can have no previous knowledge because it does not lie in the domain
of I.
 Every utterance decides a choice at reference time. In the preterite the reference
time and event time coincide, so the non-temporal formula / updating the event
time may decide the choice question / _ :/ at the reference time (as this time is
updated), but in the perfect this is not possible, because the event time that is
updated by / lies in the past of the reference time.27
The representation structure for the present continuative has the following
translation

27
Although we don’t claim that we have an exhaustive analysis of the so-called ‘perfect of recent past’
(see Comrie 1976, p. 60), the formalization has a natural explanation for recency, as the closer the event
time is to the speech time, the more choices are decided.

123
Reference time and the English past tenses 241

where i is a variable ranging over intervals and n 2 i if IðnÞ \ IðiÞ 6¼ ;. Nesting of

the operators Pret; Perf , with Cont is restricted by the formula syntax. Combina-
tions ContðContð/ÞÞ; ContðPerf ð/ÞÞ; ContðPretð/ÞÞ don’t occur because i 2 i0 ,
n  i, and i  n are not well-formed conditions. On the other hand, nesting of boxes
occurs, for instance, for the perfect continuative:

The condition that the reference time is included in every interval occurring in the
formula (Definition 3.5) has the effect that the speech time is included in i, thus in
the formalization of John has been living in Paris John still lives there at speech
time.
The pluperfect can be represented by the following nesting:

which will locate the reference time on the middle variable (k  n  m).
Alternatively the pluperfect can be represented by

where the variable ‘k’ may be identical to ‘m’. By the basic constraint on inter-
pretations I all temporal variables in its domain are mapped to a strict linearly order
sequence. So, if the variables k and m are interpreted distinctly, then n  k  m or

123
242 W. P. M. Meyer-Viol, H. S. Jones

n  m  k where the reference time is located on the first in both cases.28 This
ambiguity in the representation of the pluperfect (Pret(Perf) versus Pret(Pret))
where, in the latter, event time and reference time coincide, allows for specification
of either event time or reference time by temporal adverbs.

3.6 Temporal adverbs

Temporal adverbs like ‘yesterday’, ‘today’, etc, are represented as functions from
speech time to reference time, while those like ‘at six’ are interpreted as predicates
on the reference time where IðYesterdayðnÞÞ 2 fIðmÞ 2 U j I  m  ng (yester-
day maps the speech time to a past point) and IðTodayðxÞÞ [ IðxÞ is an <-uncon-
nected set. So yesterday is acceptable as a qualification of the event time in the
preterite (as this coincides with the reference time):

but unacceptable as a qualification in the perfect:

3.7 ‘Lifetime effects’

In our account of the ‘lifetime effects’, the reference time will again play the central
role. The lifetime effects in the perfect and the preterite point to the fact that the
denotations of names interact with the tense system: names must have a referent at
reference time.
This means that in the perfect names have a referent at speech time, while in the
preterite they have a referent at event time, and it explains the unacceptability of
utterances like *Einstein has visited Princeton and *Mrs. Thatcher was a woman
without any context (see Sect. 2.3).
To formalize this we assume that our propositional formulas are of the form
Pa1 ; . . . ; an , that is, a predicate symbol ‘P’ followed by a sequence of names
‘a1 ; . . . ; an ’ and /ðaÞ will stand for a (possibly complex) formula in which that

28
The fact that this representation does not fix either ‘m’ or ‘k’ as the speech time does not constitute a
problem as both are ‘given’ and thus interpreted already by any ‘input’ interpretation I. It is this
interpretation that decides which of ‘m’ or ‘k’ is the speech time.

123
Reference time and the English past tenses 243

name a occurs. We then introduce a predicate R on names (i.e. RðaÞ is a formula)

such that for a name a; IðRðaÞÞ  T, that is IðRðaÞÞ is a set of time points, and
I  RðaÞn () IðnÞ  IðRðaÞÞ. Here ‘I  RðaÞn ’ has the intended meaning ‘a has a
referent at time n according to I’. The following representations then represent the
reference constraints for the two past tenses:

where the constant ‘a’ stands for ‘Einstein’ and the formula RðaÞn is contributed by
the ‘visiting Princeton’ predicate and expresses the requirement that the name a has
a referent at n, the reference time.
But both Einstein has visited Princeton and Mrs. Thatcher was a woman can be
made acceptable by context. For the first, suppose that the remains of famous
scientists are doing the rounds of prestigious scientific institutions. Then Einstein
has visited Princeton is perfectly acceptable, the referent of ‘Einstein’ being his
remains, and these do exist at speech time.29 For the second, suppose somebody
claims that Great Britain never had a female prime minister. Then one can counter
with Thatcher was a woman (emphasis on ‘Thatcher’). In this case the referent of
‘Thatcher’ is a past prime minister.
What these examples of contexts have in common is that they combine names
with (definite) descriptions: we consider ‘Einstein as his remains’, and ‘Mrs.
Thatcher as prime minister’. ‘Einstein as his remains’ has a referent at speech time
(the remains exist at speech time) explaining the acceptability of Einstein has visited
Princeton, and ‘Mrs. Thatcher as prime minister’ does not exist at speech time,
explaining the acceptability of Mrs Thatcher was a woman.
We can introduce names a/ðxÞ for ‘a-as-a-/-er’. These have a referent if they
satisfy the description, that is:
I  Rða/ðxÞ Þn () I  /ðaÞn :

So,‘Thatcher-as a prime-minister’ has a referent as long as Thatcher is prime min-

ister. The use of these names-under-a-description in utterances is then defined by
½½/ðawðxÞ Þn  ¼ fðI; I 0 Þ 2 ½½/ðaÞn  j I  wðaÞn g:

The assumption is then that all names a are used as under some description. But that
obliges us to find a description for the denotational use of a name. The obvious
candidate description is the reference predicate RðxÞ, for, by the above definitions:
½½/ðaRðxÞ Þn  ¼ fðI; I 0 Þ 2 ½½/ðaÞn  j I  RðaÞn g:

So, ½½/ðaRðxÞ Þn   ½½/ðaÞn : RðaÞn g.

29
We thank Alexander Davies for this example.

123
244 W. P. M. Meyer-Viol, H. S. Jones

The reference constraints for the perfect and preterite in their formal represen-
tation are as follows:

where the reference constraint RðawðxÞ Þn comes down to wðaÞn . In the perfect, where
n is the speech time, awðxÞ has a referent if a has property wðxÞ (e.g. ‘wðxÞ’ equals
‘these are the remains of x’ and ‘a’ equals ‘Einstein’) at n whether a on its own has a
referent at n or not. In the preterite, if wðawðxÞ Þ holds at event time (e.g. wðxÞ equals
‘x is the prime minister’ and ‘a’ equals ‘Mrs. Thatcher’), then awðxÞ need not have a
referent at speech time, even if a on its own does.
Independent of the names and their use as descriptions, the predicates themselves
may impose constraints on what the referent can be. In the Einstein example just
mentioned, the predicate ‘visiting Princeton’ requires a physical referent at speech
time. However in Frege has contributed a lot to my thinking, using the predicate
‘contributing a lot to my thinking’ the referent of the argument need not be a
physical entity, but only an influence. That is, it does not contribute a referent
constraint RðaÞn to the logical form.

3.8 Aktionsart categories

In Sect. 3.1 we model one type of propositional formula, which we call an ‘event’.
Events are collections of intervals which include an end-point. These end-points
may be culmination-points, as in the Vendlerian categories of accomplishment and
achievement (e.g. climb Everest and reach the summit respectively), or the final
bound of a Vendlerian activity (e.g. walk);30 they may also be the final bound of a
temporary ‘Davidsonian’ state such as sit, stand, lie, sleep, wait and gleam. Our
formalization does not model ‘Kimian’ states such as weigh, know, resemble, be ill,
be Dutch, be a woman. These states may be temporary (e.g. be ill) or permanent
(e.g. be a woman), but they do not include an end-point.31 The only claim that we
make about such states is that they do not consist of collections of intervals. For this
reason such states are, in all tenses, incompatible with the function Contð/Þ, which
can only take a set of intervals as its argument (see Definition 4). This incompat-
ibility explains why sentences like *John will be/is/was/has been being Dutch or
*John will be/is/was/has been being ill are unacceptable. Although we have not
modelled a propositional formula for Kimian states, we see no reason why they
cannot be accommodated in our formalization, reflecting perfects such as she has
been ill.

30
See Vendler (1967).
31
See Rothmayr (2009) for a discussion of Davidsonian and Kimian statives.

123
Reference time and the English past tenses 245

4 Discussion

We review, in light of our model, the various effects of the perfect and preterite
mentioned in Sect. 2. We also discuss the ‘imperfective paradox’.

4.1 The ‘definiteness effect’

In our formalization a temporal adverbial like today or yesterday has only the
reference time to work with. Since in the perfect the reference time lies at speech
time, a past-time adverbial is ruled out. In the preterite the reference time lies in the
past, so a past-time adverbial is acceptable. The acceptability of a sentence with
the perfect or preterite therefore derives not from a notion of definiteness, but from
(i) the fact that any temporal expressions must qualify reference time, and (ii) that
reference time is located at different points in the structure in the perfect and
preterite. The four cases below (repeated from Sect. 2.1) illustrate this for the perfect.
(1) *I have seen him yesterday

(2) I have seen him today

(3) *I have seen him this morning [said in the afternoon]

(4) I have seen him this morning [said that morning]

(1) and (3) are ruled out because the temporal expressions exclude reference time
(which coincides with speech time) while (2) and (4) are acceptable because they
include reference time. In the terms of the formalization, the temporal adverbials
yesterday and today are functions from speech time to reference time. In the perfect
these points coincide, and speech time is mapped to itself, making adverbials which
include speech time acceptable. Adverbials which precede speech time are correctly
applied only if reference time precedes speech time, which is not the case in the
perfect, but is the case for the preterite (Definitions 4.1 and 4.2). The next four cases
show the preterite in equivalent contexts.
(29) I saw him yesterday

(30) I saw him today

(31) I saw him this morning [said in the afternoon]

(32) I saw him this morning [said in the morning]

(29)–(32) are acceptable because the temporal expressions can coincide with a
reference time which, in the preterite, precedes speech time.32

32
The terms ‘definiteness’ does not fit the facts well in any case. For example, the time expression in his
lifetime is indefinite but is compatible with the preterite, as in he saw wonders in his lifetime, while the
expression in the last ten seconds is definite but is compatible with the perfect in I have seen him in the
last ten seconds.

123
246 W. P. M. Meyer-Viol, H. S. Jones

The constraint that temporal expressions must qualify reference time also
explains why adverbials of present time which exclude the past are compatible with
the perfect but not with the preterite, as in the pair:
(33) It has been two years now since Mary went to America

(34) *It was two years now since Mary went to America.

4.2 Presupposition of past time

In the formalization a given past evaluation point may have been specified before,
whereas a newly introduced past evaluation point may not. Any information that
forces the identification of the new past point with a previous given one is excluded
from the perfect. This explains the inter- and intra-sentential effects of (5) and (6),
and of the following example:
(35) John has died in Paris. His death was witnessed by his landlady
rendered in the formalization as

fmg : ðm  n ^ JdiPm Þ; fg : ðm  n ^ Jdwblm Þ33

Let’s call the composition of these two L formulas M. The first formula of M
introduces a new temporal variable m as lying in the past of a given n and updates it
with JdiP. This m is given in the second formula of M and updated with Jdwbl.
The unacceptability of *John has died in Paris. His death has been witnessed by
his landlady, rendered as
fmg : ðm  n ^ JdiPm Þ; fkg : ðk  n ^ Jdwblk Þ;
follows from the fact that the propositions are interpreted at two (disjoint) evalu-
ation points. In the latter (but not the former!), the interpretation I that is the output
of the first utterance (with reference time n) must be undefined on the event time, so
k and m cannot share information.
The different effects of the perfect and preterite here are analogous to those of
names and pronouns respectively in Discourse Representation Theory: a name
introduces a referent, whereas a pronoun presupposes a referent.
(6) *John has been born in Paris
In (6) the perfect is ruled out because, once an interpretation of ‘John’ is given, there
is always some time-point determined at which he was born, so *John has been born
in Paris would have to be located at a given past time-point. The acceptability of
(36) #John has died in Paris
depends on whether John’s death is given or new information. If John’s death is
known to speaker and hearer, (36) is unacceptable even if ‘in Paris’ is new, because
a past time-point is already given. However, if John’s death is not known, (36) is
33
This is equivalent to fmg : ðm  n ^ JdiPm Þ : Jdwblm .

123
Reference time and the English past tenses 247

acceptable because the perfect is introducing a new past time-point, not presup-
posing a given one.
The unacceptability of (37) below has proved hard to explain ever since it was
cited by Dietrich (1955, p. 195):

(37) *Gutenberg has discovered the art of printing

The sentence in this form is ruled out by the ‘lifetime effect’ if Gutenberg is taken as
the referent, because as the discoverer of anything he cannot be the referent at
reference time which, in the perfect, falls at speech time. However, the sentence is
still unacceptable if it is passivized along the lines of the Einstein example in (8), as
follows:

(38) *The art of printing has been discovered by Gutenberg

The perfect is ruled out here for the same reason that *John has been born in Paris
is ruled out. So long as the art of printing is known to speaker and listener, its
discovery would have to be located at a given past time-point, and this is un-
acceptable in the perfect.

4.3 ‘Lifetime effects’

The ‘perfect lifetime effect’, as defined in the Sect. 2.3, is captured in our account
by the fact that the perfect and preterite must each have a referent at reference time
and that they have different reference times. In (9), repeated below, the perfect is
acceptable because the sentence creates a reference to a set which can still be a
referent at reference time (which coincides with speech time in the perfect), for
example the set of Nobel Prize winners or great thinkers:
(9) Einstein has visited Princeton, Yukawa has, Friedman has. . .
In the preterite, by contrast, the referent coincides with a reference time which in
this tense lies in the past, and no claim is made about any referent at speech time:
(39) Einstein visited Princeton, Yukawa did, Friedman did. . .
In (10)–(12), repeated below, although Frege is the referent in each case, his quality
as a referent varies. After his death he cannot be a referent as being capable of fear,
so (12) is ruled out, but if ‘Frege’ is used here as a name for ‘Frege’s thinking’ or
‘Frege’s works’, he can be a referent, which makes (10) and (11) acceptable.
(10) Frege has contributed a lot to my thinking

(11) Frege has been denounced by many people

(12) *Frege has been frightened by many people

The preterite is acceptable in the equivalent sentences because Frege can be a
referent at a past reference time both as a physical person and as a name for his
thoughts and works.

123
248 W. P. M. Meyer-Viol, H. S. Jones

Examples (10) and (11) suggest that ‘lifetime effect’ is an inappropriate term, as
do (13) and (14), repeated here:
(13) John has died

(14) The bomb has exploded

In (13) let us assume that John’s death is new, not given information, so that the
sentence does not fall foul of the presuppositional constraint explained above. As
for the ‘lifetime effect’, John’s death does not seem to prevent his being the referent
at reference time, showing that an object may be a referent even as a physical entity
after its own end. We see this also in the use of the present tense to refer to dead
people, as in John is dead. (14) is acceptable because the bomb exists in its after-
effects at speech time.
We showed in Sects. 2.3 and 3.7 that stative predicates denoting permanent
qualities, when used in the preterite, do mean that the referent no longer exists but
that the referent may have a narrower description in context than out of context: we
can say Mrs Thatcher was a woman if we are denoting Mrs Thatcher as prime
minister, rather than Mrs Thatcher as a human being. However, without the
restriction, such predicates used in the preterite do invite the reading that the ref-
erent no longer exists, as these examples, repeated from Sect. 2.3, show:
(15) Gregory was from America

(16) Gregory had blue eyes

(17) Gregory resembled Jörg Bieberstein

In the perfect the same sentences would be at best infelicitous, but this time owing
to the ‘perfect lifetime effect’. For example, suppose we assume that, according to
the lifetime effect, Gregory is the referent at speech time in *Gregory has had blue
eyes. We would then be predicating of Gregory a permanent quality in the past
which still holds in the present. Similarly *John has been Dutch is unacceptable
because being Dutch is a permanent quality, whereas John has been ill is acceptable
because being ill is not necessarily so.

4.4 The perfect as explanation

In our formalization, every utterance decides some choice at reference time, which
in the perfect coincides with speech time. This choice is decided by the past event
predicated in the perfect. The choice being decided is always context-dependent and
may be implicit. So, in (18) and (19), repeated below, the past events of ‘John’s
living in Paris’ and ‘John’s arrival’ decide some choice at reference time, which
coincides with speech time.
(18) John has lived in Paris

(19) John has arrived

123
Reference time and the English past tenses 249

The formalization of (18) as Perf ðJliPÞn (where ‘JliP’ is the ‘john-living-in-Paris’

proposition) has the form fmg : m  n : JliPm where ðI; I 0 Þ 2 ½½Perf ðJliPÞn  only
if there is some choice ðw _ :wÞ 2 CHi ðnÞ (that is, at reference time ¼ speech time),
such that I 0  wn (Definition 3.6). However, depending on the pragmatic situation,
the choice at speech time may, in addition, implicate a choice at some other time
point. For example, the choice being decided at speech time by (18) could be ‘Does
John know Paris?’ and that could implicate the deciding of the choice ‘Who will
show us around Paris?’ where the time reference is future; or it could implicate the
deciding of the choice ‘Is John a man of the world?’ where the time reference is
universal. Likewise in (19), the choice being decided at speech time might be ‘Is John
here?’ and that could implicate the deciding of the choice ‘Will John catch his
plane?’ (future time reference) or ‘Is John reliable?’ (universal time reference).

4.5 The ‘up-to-now’ perfect

The formalization deals with ‘up-to-now’ perfects by interpreting the continuative

perfect (i.e. when it is Contð/Þ) in an interval which includes reference time ¼
speech time (Definition 3.5). Accordingly, sentences (25) and (26), repeated here,
require ‘up-to-now’ interpretations:
(25) She has been living in Paris

(26) She has been living in Paris for ten years

Without Contð/Þ, there are only temporal variables involved in the interpretation,
so that sentences (21) and (22), repeated below, permit, but do not require, an
‘up-to-now’ reading:
(21) Jack Norbert has taught at MIT since 1969

(22) I have lived here for ten years

4.6 The ‘imperfective paradox’

When the aspect is Contð/Þ, there is an interval i in which the proposition / is true,
and an evaluation point n lies somewhere in the interval i. If / includes a culmi-
nation as part of its meaning (i.e. if it is what is often called a ‘telic’ proposition),
the evaluation point does not necessarily include the end-point of /. Consider, for
instance, the following, which represents the so-called ‘imperfect’ tense:

123
250 W. P. M. Meyer-Viol, H. S. Jones

The speech time m lies in the future of n, an element of interval i which may end in a
cumulation point c. We have n  c and n  m, but, because of the branching
structure (see Sect. 3.2.2) it is possible that neither m  c nor c  m. This explains
why, for example, John will be/is/was/has been climbing Everest can be true even if
John does not reach the summit.34

4.7 Interchangeability of perfect and preterite

When there is an explicit question under discussion, as in the Klein example (27),
perfect and preterite can be interchangeable. Consider
(270 ) Is Chris cheerful these days?
Well, he won a million in the lottery
where a yes-no question is employed instead of Klein’s why-question.35 An explicit
question at speech time allows an answer in the preterite as well as in the perfect. If
it is an issue at speech time whether Chris is cheerful, then that choice is decided if
at some past time point Chris won the lottery, that is,
½½m  n : ChðCctdÞn : Cwmlm   ½½Cctdn ; 
where Cwml stands for Chris winning a million in the lottery and Cctd for Chris
being cheerful these days. Then both perfect and preterite have ‘current relevance’.
First the perfect as an answer to the question:

The event Cwml lies in the past m of n, it has to decide some choice at reference
time, n, and in this case it answers the question.
In the preterite as an answer to the question:

Cwml has to decide an (implicit) choice at m (by Definition 3.6) and there is no
requirement for it to decide some choice at speech time. But the question introduces
an explicit choice at speech time which, in this case, is decided by updating m (the
past).

34
See Dowty (1977, 1979) on the ‘imperfective paradox’, and Mittwoch (1988, p. 236) for a formal-
ization of the perfect progressive along the lines of Dowty’s formalization of the imperfect.
35
A yes-no question can be differentiated from a why question by its set of acceptable answers. Chð/Þn
represents a yes-no questions if the set of answers consists of fw j ½½Chð/Þn : w  ½½/n  or
½½Chð/Þn : w  ½½:/n g whereas Chð/Þ represents a why-question if the answer set consists of
fw j ½½Chð/Þn : w  ½½/n g alone.

123
Reference time and the English past tenses 251

The analysis above could also be applied to a similar example in Portner (2011),
in which the Gutenberg example is used in the preterite with current relevance
which is made explicit by a question: What contributions have Germans made
which are still important to the German economy today?—Gutenberg discovered
printing, which is the basis of our world-renowned academic handbook industry.
There need not be an explicit question for the preterite to decide a choice at
speech time: the choice may be implicit. In example [19], the sentence with the
perfect John has arrived could decide the choice ‘Is John here’ at speech time. Now
suppose the sentence John arrived five minutes ago is uttered in the same context: it
can be read as deciding the choice ‘Did John arrived five minutes ago?’ but also, by
implicature, the choice ‘Is John here?’

4.8 The pluperfect and the ‘present perfect puzzle’

In our formalization the pluperfect is defined as either Pret(Perf) or Pret(Pret). The

fact that the pluperfect can function as either a perfect-in-the-past or a preterite-in
the-past is illustrated in the following sets of examples. In each set, the first two
examples show a contrast between the preterite and the perfect that is highlighted by
our analysis of those tenses; in the third example we show that, allowing for the shift
back in time, the pluperfect can correspond to either of the first two. First we look at
presuppositional effects.
(40) I don’t know if John worked in advertising during his time in Paris but he
worked in marketing.

(41) I don’t know if John worked in advertising during his time in Paris but he has
worked in marketing.

(42) I didn’t know if John had worked in advertising during his time in Paris but he
had worked in marketing.
In (40), a past time point is given by the first ‘worked’, and the second ‘worked’
identifies with this, so they have the same time reference, and John‘s work in
marketing must have occurred when he was in Paris. In the perfect [(41)] this is
excluded, and the time reference of ‘has worked’ cannot be identified with that of
the first verb ‘worked’; therefore his work in marketing may or may not have been
during that time. In the pluperfect, either reading is possible, and we would need
further information to disambiguate them. For example, ‘at that time’ added to the
end of (42) would force a Pret(Pret) reading, while ‘at some point’ would force a
Pret(Perf) reading.
If we know from context that the time reference of the pluperfect has to be
identified with a given time point, we must interpret the pluperfect as Pret(Pret).
This would be the case in a pluperfect version of example (5), in which only
the first verb could be read as Pret(Perf) since, thereafter, the time reference is
presupposed.
The pluperfect can be similarly ambiguous in terms of lifetime effects, as the
following examples show.

123
252 W. P. M. Meyer-Viol, H. S. Jones

(43) John lived a good life

(44) John has lived a good life

(45) John had lived a good life

In (43), owing to the ‘preterite lifetime effect’ the implicature is that John is dead,
because ‘having a life’ is a permanent quality, and the preterite would be used only
after John ceased to exist. On the other hand (44) can be used only if John is alive,
because the perfect needs a referent at speech time, which in this case must be
John as a living being (see Sects. 2.3 and 3.7). However, the pluperfect allows
either reading. For example, (45) could refer to a situation at John’s graveside
(Pret(Pret)), or at his bedside (Pret(Perf)).
Finally, the pluperfect can be ambiguous when it is continuous, i.e. (Cont) in our
formalization.
(46) She was living in Paris

(47) She has been living in Paris (repeated from (25))

(48) She had been living in Paris

In (46), which we analyse as Pret(Cont), the past time reference of her living in
Paris must be given by the context; for example, the sentence might continue, when
the Germans arrived. On the other hand (47), which we analyse as Perf(Cont) has
an ‘up-to-now’ sense, and the sentence might go on, and she hasn’t even got an
identity card. The pluperfect ([48]) could be used in either context: She had been
living in Paris when the Germans had arrived and She had been living in Paris but
didn’t even have an identity card.
As the above examples show, the pluperfect can function as either a Pret(Perf)
or a Pret(Pret). The acceptability of event time modification in the pluperfect
reflects the Pret(Pret) analysis. Thus, Chris had left at six can be represented as
follows under our formalization (where Clas stands for the atemporal chris-
leaving-at-six):

This will locate the reference time on the middle variable (k  n  m).

123
Reference time and the English past tenses 253

Alternatively the pluperfect can be represented by

which identifies the reference time and the event time (n  k  m or n  m  k).
The analysis of the pluperfect as a Pret(Pret) does not prevent it from having an
explanatory sense (the equivalent of ‘current relevance’) at the intermediate time
point. As we argued in Sects. 2.6 and 4.7, the preterite can have ‘current relevance’
if there is some choice to be decided at speech time; indeed, when the current
relevance arises from a past event with a definite time reference, only the preterite
may be used. Correspondingly both a Pret(Perf) and a Pret(Pret) pluperfect can
decide a choice at the intermediate (or any other) time point.

5 Conclusion

This paper offers a unified account of the English perfect and preterite by formal-
izing the notion of reference time. We identify a bundle of constraints which
must all be satisfied at reference time, and show that the contrasting functions of
the perfect and preterite reflect their distinct reference times. This difference in
‘perspective’ allows us to account for well-known effects of the perfect: the ‘def-
initeness’ and ‘lifetime’ effects, the fact that the perfect seems to ‘explain’ some-
thing about the present, and the way in which the perfect cannot presuppose a past
time point. Moreover, the interaction between ‘perspective’ and aspect explains the
‘up-to-now’ perfect. This interaction also accounts for the ‘imperfective paradox’,
which is a feature of all English tenses, not only the perfect and preterite. Our
account enables us to explain cases where the unacceptability of the perfect is
especially puzzling, such as *Gutenberg has discovered the art of printing. We
account for the overlap in function between perfect and preterite and we offer a
solution to the ‘present perfect puzzle’.
The well-established notions of ‘current relevance’ and ‘extended now’ clearly
capture something important about the function of the perfect. In our formalization
both notions are reflected in the connections which are created in the perfect
between event time and speech time: first, the perfect predicates a past event which
settles some matter, or ‘decides some choice’, at speech time (the perfect as
explanation); second, the perfect predicates something in the past about an entity
which is still a referent at speech time (the ‘perfect lifetime effect’); and third, in the
‘up-to-now’ perfect, the event spans the interval between event time and speech
time. However, neither ‘current relevance’ nor ‘extended now’ accommodates the
more specific effects of the perfect: we would have to stretch these notions unduly if
we wished to include the ‘definiteness’ and presuppositional effects within them.

123
254 W. P. M. Meyer-Viol, H. S. Jones

Our formalization, which models a bundle of features which we term ‘perspective’,

offers an account of all the specific features of the perfect listed in Sect. 2, including
the relationship between the perfect, preterite and pluperfect.
This paper opens up at least two avenues for further research. First, in our
model we formalize propositions as ‘events’, which include all Aktionsart cate-
gories except Kimian states. The only constraint we place on Kimian states is
that they do not represent collections of intervals; this explains why they are
incompatible with continuative aspect, which demands such a collection. A
model of Kimian states would allow us to formalize the ‘preterite lifetime effect’,
whereby Kimian states denoting permanent qualities, when used in the preterite,
can implicate that the referent no longer exists (e.g. John was Dutch) provided
the name ‘John’ does not stand for some (definite) description which is confined
to the past (see Sect. 3.7).
Second, the model developed in this paper accounts for the perfect and preterite
tenses in English, where distinct forms correspond to the different functions that we
have described. In other languages the constraints on the verb forms labelled
‘perfect’ and ‘preterite’ often differ from those on their English counterparts. For
example, in German both the perfect and the preterite can be used with past time
adverbials, or when a past time point is presupposed, or without a referent at speech
time. Thus in the following sentences, where the preterite would be acceptable in
both languages, the perfect is also acceptable in German but not in English (based
on Klein 2000, p. 359):
(49) Einstein hat Princeton besucht
*Einstein has visited Princeton
(50) Ich habe den Brief gestern um zehn abgeschickt
*I have sent off the letter yesterday at ten
The distribution between perfect and preterite in German is also regionally
determined, with the preterite ‘more or less extinct’ in some parts of the German-
speaking world (Klein 2000.). However, where the two tenses are both used, there
are some contexts which do demand the perfect, such as here (Klein 2000, p. 360):
(51) Schau, der Stuhl ist umgekippt
Look, the chair has toppled over
Here the preterite in German would sound odd but would be acceptable in English.
A number of recent accounts of the German perfect have been offered, some lan-
guage-specific,36 some as part of cross-linguistic studies.37 The model we propose in
the present paper could be tested on German and other languages to see if our notion
of perspective is a cross-linguistic category, or just a bundle of effects seen in
modern English.

Acknowledgements The authors are grateful for comments and suggestions by Alexander Davies,
Kerstin Hoge, Will Randall and an anonymous reviewer at Linguistics and Philosophy.

36
e.g. Stechow (1999), Klein (2000), and Musan (2002).
37
e.g. Iatridou at al. (2001), Pancheva and Stechow (2004), Rothstein (2008), and Schaden (2009).

123
Reference time and the English past tenses 255

References

Binnick, R. (1991). Time and the verb: A guide to tense and aspect. Oxford: Oxford University Press.
Bryan, W. (1936). The preterite and the perfect tense in present-day English. Journal of English and
Germanic Philology, 35, 363–338.
Chomsky, N. (1970). Deep structure, surface structure, and semantic interpretation. In R. Jakobson &
S. Kawamoto (Eds.), Studies in general and oriental linguistics (pp. 52–91). Tokyo: TEC Corpo-
ration.
Comrie, B. (1976). Aspect. Cambridge: Cambridge University Press.
Curme, G. O. (1935). A grammar of the English language, volume 2: Parts of Speech and Accidence.
New York: D.C. Heath.
Declerck, R. (1991). Tense in English: Its structure and use in discourse. London: Routledge.
Dietrich, W. (1955). Erweiterte form, Präteritum und Perfektum im Englischen. Eine Aspekt- und
Tempusstudie. Munich: Hueber.
Dowty, D. (1977). Toward a semantic analysis of verb aspect and the English continuative progressive.
Linguistics and Philosophy, 1, 45–77.
Dowty, D. (1979). Word meaning and montague grammar: The semantics of verbs and times in gener-
ative semantics and in Montague’s PTQ. London: Reidel.
Heim, I. (1982). The semantics of definite and indefinite noun phrases in English. PhD Thesis, University
of Massachusetts, Amherst.
Higginbotham, J. (2009). Tense, aspect and indexicality. Oxford: Oxford University Press.
Iatridou, S., Anagnostopoulou, E., & Pancheva, R. (2001). Observations about the form and meaning of
the perfect. In M. Kenstowicz (Ed.), Ken Hale: A life in language (pp. 189–238). Cambridge: MIT.
Inoue, K. (1979). An analysis of the English present perfect. Linguistics, an Interdisciplinary Journal of
the Language Sciences, 17, 561–589.
Kamp, H., & Reyle, U. (1993). From discourse to logic. Dordrecht: Kluwer.
Kiparsky, P. (2002). Event structure and the perfect. In D. I. Beaver, L. D. C. Martinez, B. Z. Clark, &
S. Kaufmann (Eds.), The construction of meaning. Stanford: CSLI Publications.
Klein, W. (1992). The present perfect puzzle. Language, 68(3), 525–552.
Klein, W. (2000). An analysis of the German Perfekt. Language, 76(2), 358–382.
McCawley, J. D. (1971). Tense and time reference in English. In C. J. Fillmore & D. Terence Langendoen
(Eds.), Studies in linguistic semantics (pp. 96–113). New York: Holt, Rinehart and Winston.
McCoard, R. W. (1978). The English perfect: Tense-choice and pragmatic inferences. Amsterdam:
North-Holland.
Michaelis, L. (1994). The ambiguity of the English present perfect. Journal of Linguistics, 30, 111–157.
Mittwoch, A. (1988). Aspects of English aspect: On the interaction of perfect, progressive and durational
phrases. Linguistics and Philosophy, 11, 203–254.
Moens, M., & Steedman. M. (1988). Temporal ontology and temporal reference. Computational
Linguistics, 14(2), 15–28.
Musan, R. (1997). Tense, predicates, and lifetime effects. Natural Language Semantics, 5, 271–301.
Musan, R. (2002). The German perfect. Its semantic composition and its interaction with temporal
adverbials. Dordrecht: Kluwer.
Pancheva, R., & von Stechow, A. (2004). On the present perfect puzzle. In K. Moulton & M. Wolf (Eds.),
Proceeding of the North East Linguistics Society (=NELS) 34. Amherst: GLSA.
Partee, B. (1973). Some structural analogies between tenses and pronouns in English. The Journal of
Philosophy, 70(18), 601–609.
Pickbourn, J. (1789). A dissertation on the English verb (Facsimile edition, 1968). Menston: Scolar Press.
Portner, P. (2003). The (temporal) and (modal) pragmatics of the perfect. Linguistics and Philosophy, 26,
459–510.
Portner, P. (2011). Perfect and progressive. In C. Maienborn, K. von Heusinger, & P. Portner (Eds.), Semantics:
An international handbook of natural language meaning. Berlin: Mouton de Gruyter. Available at http://
semanticsarchive.net/Archive/jlmMTE0M/Portner-perfect-progressive-HSKdraft.pdf. Accessed 23
Nov 2011.
Reichenbach, H. (1947). Elements of symbolic logic. London: Macmillan. (Repr. Dover, New York,
1980.)
Rothmayr, A. (2009). The structure of stative verbs. Amsterdam: Benjamins.
Rothstein, B. (2008). The perfect time span: On the present perfect in German, Swedish, and English.
Amsterdam: John Benjamins.
Smith, C. (1997). The parameter of aspect (2nd ed.). Dordrecht: Kluwer.

123
256 W. P. M. Meyer-Viol, H. S. Jones

Schaden, G. (2009). Present perfects compete. Linguistics and Philosophy, 32(2), 115–141.
Vendler, Z. (1967). Linguistics in philosophy. New York: Cornell University Press.
von Stechow, A. (1999). Eine erweiterte Extended Now Theorie für Perfekt und Futur. Zeitschrift für
Literatur und Linguistik, 113, 86–118.

123