The Expressive Power of the English Temporal Preposition System

Ian Pratt and David Bree

1 Introduction

Arti cial Intelligence Group

Department of Computer Science,
University of Manchester
Manchester M13 9PL
United Kingdom

Sentences involving temporal prepositions number

among our simplest and most natural expressions of
temporal location and extent:
(1) Sue worked on the letter until 10 o'clock
(2) Bill will post the parcels by 10 o'clock
(3) Charles has sent me email within the last
10 minutes.
Altogether, there are eighteen or so temporal prepositions in English, depending on exactly how you count
them. The temporal prepositions considered here are
before, after, since, until, by, at, for, in, within, on,
during, through, throughout, over, from : : : to, between
: : : and and ago. The last of these, ago, is normally
classi ed as an adverb rather than a preposition, but,
as we will see, ts neatly into the temporal prepositional system. There are a number of related prepositions such as about and around which can have a
temporal function, but which we ignore here.
One of the noteworthy features of temporal prepositions is that, in general, they are not embedded one
within the other. Thus
(4) *Sue wasn't working on the letter by 10
o'clock until 10 o'clock
(5) *Charles had sent me electronic mail
within the last 5 minutes until 10 o'clock
count, for most competent speakers of English, as
nonsense. Examples where two temporal prepositions
can work together, as in
(6) In ve minutes, David will have been working in his o ce for 6 hours
are the exception rather than the rule. In the Brown
corpus of American English 10], which has a total of
52,126 sentence, about 10% (4,894) include at least
one preposition used temporally. Only 366 (7.5%) of
these sentences, i.e. less than 1% of the corpus, have
two or more prepositions used temporally. Only two
sentences have more than two temporal prepositions:1
(f32:020) Many years later (on August 3, 1915),
Lucy Upton wrote Winslow's daughter soon
1
code numbers refer to the sub-corpus and the sequential number of the sentence within that sub-corpus

to be graduated from Smith College: \While I

love botany which, after dabbling in for years,
I studied according to the methods of that day
exactly forty years ago in a summer school, it
must be fascinating to take up zoology in the
way you are doing.

(p09:067) And the hours were six-thirty in the

morning until eleven at night on Saturdays
and during sales, ...

In f32:020 we see the most common situation: the

temporally used prepositions (in italics) are not in
the same clause. In p09:067 we see how temporal
prepositions, when there is more than one within a
clause, have a hierarchical structure rather than a
direct interaction: the in and the at qualify the two
clock times, until builds a cyclic period between these
times, which in turn is restricted by the on and the
during.
This general lack of iterative use, combined with
the limited number of temporal prepositions available to start with, suggests that the expressive power
of the temporal preposition system is limited. That
is, within the space of logically possible assertions
concerning the temporal distribution of states and
events, prepositions give us access to only a very limited subspace. The main burden of this paper is to
identify this subspace and to explain how, exactly,
the English temporal propositions span it.
The approach does not make use of the well-known
temporal relationships proposed by James Allen 1].
His system provided a means of representing all the
possible temporal orderings between two events in
time, in particular when they overlapped in some
way, e.g. one starting and nishing before the other.
In English this relationship is not indicated by the
prepositions but rather by explicit references to the
beginning/start of and to the end/ nish of such periods. In English temporal prepositions serve not only
to order two events but also to indicate their absolute time duration and their actual time of occurrence relative to the time of reference or utterance of
the discourse, neither of which are included in Allen's
system.

2 Temporal functions

This section illustrates the avour of the analysis we

propose by means of simple examples.
Consider the sentences:
(7) Bill posted the parcel between 9 o'clock and
10 o'clock.
(8) Sue worked on the letter from 9 o'clock to
10 o'clock.
Sentence (7) reports an event|namely, Bill's posting the parcel|and locates that event as having occurred sometime within the interval 9:00,10:00]. Sentence (8) also reports an event|namely, Sue's working on the letter|but locates that event as having
occurred throughout that interval.
Following common practice in semantics (e.g.
Richards 16]), we take (7) and (8) to have, as signi cant constituents, the tenseless sentences
(9) Bill post the parcel
(10) Sue work on the letter
respectively, which receive truth-values relative to intervals of time. Thus (9) is true over an interval J
just in case Bill posts the parcel over that interval
likewise, (10) is true over an interval J just in case
Sue works on the letter over that interval, and so on.
We will have more to say presently on what it means
for a tenseless sentence to be true over an interval
but for the present, let us take this notion as given.
On this approach, it is natural to give the truthconditions of (7) and (8) as, respectively:
(11) There exists an interval J wholly contained within the interval 9:00, 10:00],
such that (Bill post the parcel) is true over

(12) For all intervals J wholly contained within

the interval 9:00, 10:00], (Sue work on the
letter) is true over J .
And we might write these truth-conditions symbolically as
(13) 9J (J 9:00, 10:00] & (Bill post the
parcel)(J ))
(14) 8J (J 9:00, 10:00] ! (Sue work on the
letter)(J ))
thus viewing tenseless sentences as one-place predicates whose arguments range over time intervals. To
be sure, there are alternative approaches to the semantics of tensed sentences to the one taken here.2
Some writers, for example, have taken tensed sentences to quantify over events (e.g. Parsons 14]), or over
events and time-points, depending on the type of sentence
(Herweg 7]). And among those who adopt the intervalapproach, there is sometimes disagreement on which intervals, exactly, given tenseless sentences should be evaluated at (see Lascarides 11], where the discussion hinges
on just this question).
2

However, the basic framework adopted here should

not necessarily be understood as a rejection of these
alternatives. It is merely that we have to work within
some framework or other, and ours, as we shall see,
has the virtue of simplicity for the topic at hand.
But we expect that the conclusions presented below
can probably be translated|with varying degrees of
e ort|into the frameworks favoured by a number
of approaches to natural language semantics. So,
those who think they have an argument against the
interval-based framework adopted here should be able
to reformulate what they nd in this paper in their
preferred terms.
Generalizing our analysis of (7) and (8), the temporal
prepositional
constructions
between : : : and and from : : : to can be seen as functioning so as to map a tenseless sentence onto the
propositions:
(15) 9J (J t1 t2] & (J ))
(16) 8J (J t1 t2] ! (J ))
respectively, where t1 and t2 are the prepositional arguments. We shall call forms (15) and (16) temporal
functions|that is: functions by which an underlying
tenseless sentence is mapped to a proposition. Our
goal in this paper is precisely to specify the range
of temporal functions which the prepositional system
makes available.
The temporal functions (15) and (16) express restricted existential and universal quanti cation over
intervals. That is, they have the general forms:
(17) 9J (J I & (J ))
(18) 8J (J I ! (J ))
where I is a time interval determined (in these cases)
by the arguments to the prepositions. We call temporal functions having form (17) existential temporal
functions, and those having form (18) universal temporal functions. The interval I , over which the quanti cation ranges, we call the reference interval of the
temporal function.
In giving the semantics of temporal prepositions
in terms of operations they perform on underlying
tenseless sentences, we need to distinguish between
tenseless sentences that report a completed event from
those that report an ongoing process. Such aspectual
distinctions are familiar and widely accepted in semantics, following Vendler 17], who di erentiated between the four aspectual classes of activities, achievements, accomplishments and states. Less widely accepted are Vendler's speci c proposals as to the number and nature of the di erent aspectual classes. (See,
for example, Herweg 7], and Lascarides 11], for their
own treatments.) But for the purposes of accounting
for the semantics of temporal prepositions, the following minimal distinction su ces.
We suppose that some tenseless sentences, which
describe `ongoing' process- or state-types, have the

property that whenever they are true over an interval I , they are true over every subinterval of I .
We call these non-completion sentences. We suppose in addition that other tenseless sentences, which
describe `completed' event-types, have the property
that whenever they are true over an interval I , they
are false over some subinterval of I . We call these
completion sentences. Note that the aspectual distinctions made here are the minimal ones required
for presenting a semantics of temporal prepositions.
We do not mean to argue that there might be no other
reason to de ne additional aspectual classes satisfying stronger requirements.

3 The group-I temporal functions

In the previous section, we proposed that the temporal prepositions between : : : and and from : : : to should
be seen as having temporal functions of the forms:
(17) 9J (J I & (J ))
(18) 8J J I ! (J ))
respectively, where I , the so-called reference interval,
is xed directly by the arguments to those prepositions. In this section, we examine a group of temporal
prepositions which also express temporal functions of
this form, but where the reference interval is xed
indexically or anaphorically.
The prepositions since, by, until, within, in and for,
when used with appropriate verb-tense and -aspect,
can express temporal functions of forms (17) and (18)
with any of the eight reference intervals:
t, TOU], TOU-t, TOU],
t, TOR], TOR-t, TOR],
TOU, t], TOU, TOU + t],
TOR, t], TOR, TOR + t],
where t is the argument to the temporal preposition,
TOU the time of utterance, and TOR the reference
time of the discourse. These reference intervals can
be identi ed by their position on three dimensions
of variation: indexical/anaphorical, forward-looking/
backward-looking and absolute/relative.
Indexicals have one end of the reference interval
xed at the time of utterance. The tense in this
case cannot be past. Anaphoricals have one end of
their reference interval xed by the time of reference
of the discourse, so their tense cannot be present.
Further tense restrictions are placed by the existential/universal feature of the function. We do not go
into these in detail.
The absolute use occurs when the argument to the
preposition is a time on the time axis, e.g. 6 o'clock.
The relative use is when this argument is a temporal duration, e.g. one hour. This duration is used
as an o set from the time of utterance/reference in
determining the reference interval, I .
Whether an existential or a universal function is intended is indicated by an aspect of the matrix clause.
If it is progressive, then the function is universal.

The direction is indicated by another aspect of the

matrix clause. If it is perfect, then the function is
backward-looking if not, it is forward looking.
The reference intervals, I , for the backwardlooking functions have their upper bound xed at the
time of utterance (TOU) or reference (TOR). Their
lower bound is determined by the argument to the
preposition, t, either absolutely, or by an o set relative to the upper bound. For the reference intervals
for forward-looking functions this is reversed: their
lower bound is xed by the time of utterance or reference and their upper bound is xed by t, either absolutely or relatively.
Example sentences of each function, i.e. Absolute or Relative, Backwards or Forwards, indexical
or anaphorical and Universal or Existential, are:
(ABiU) Sue has been working since 6 o'clock.
(ABiE) Sue has cried once since 6 o'clock.
(ABaU) Sue had/(will have) been working
since 6 o'clock.
(ABaE) Sue had cried once since 6 o'clock.
(AFiU) Sue is/(will be) working until 6 o'clock.
(AFiE) Sue will call by 6 o'clock.
(AFaU) Sue worked/(would be working) until
6 o'clock.
(AFaE) Sue called/(would call) by 6 o'clock.
(RBiU) Sue has been working for a week.
(RBiE) Sue has the best score in 3 days.
(RBaU) Sue had/(will have) been working for
a week.
(RBaE) Sue had the best score in 3 days.
(RFiU) Sue is/(will be) working for a week.
(RFiE) Sue will call within an hour.
(RFaU) Sue would be working for a week.
(RFaE) Sue would call within an hour.
As can be seen, backward-looking indexicals are
not happily classi ed as existentials. They are generally used, e.g. in the Brown corpus, to give a frequency of occurrence, as in ABiE and ABaE, or to
pick out an extreme value, as in RBiE and RBaE, in
a time interval.
The rules for the interpretation of English sentences involving propositions in this usage are summarized in table 1. The permitted combination of
tense and aspect, both perfective and progressive,
are taken from the above examples. Note that the
past+future corresponds to the use of would, as in
RFaU and RFaE above. By way of illustration, the
mapping of these functions onto the time axis is depicted in gure 1.

at

by

since

?
6
t

since

?
6

: : : ago/
: : : before

at

-?
6

until

in
for

within
for

-6

TOR/TOU
t
TOR/TOU
The absolute functions
The relative functions
Figure 1: The absolute and relative functions in groups I and II

relative/ forward/ indexical/

absolute backward anaphorical
absolute backward indexical

reference
interval I
t TOU]

anaphorical
forward
relative

backward
forward

existential/
universal on J
universal
existential
universal
existential
universal
existential
universal
existential
universal
existential
universal
existential
universal
existential
universal
existential

TOR]

indexical

TOU t]

anaphorical

TOR t]

indexical

(TOU ; t) TOU]

anaphorical

(TOR ; t) TOR]

indexical

TOU (TOU + t)]

anaphorical

TOR (TOR + t)]

preposition
since
since
until
by
until
by
for
in
for
in
for
within
for
within

tense
present
past/fut
past
pres/fut
future
past/
future
present
past/fut
past
pres/fut
future
past+
future

in/
after

?
-

perprofective gressive
Y
Y
N
Y
Y
N
N
Y
N
N
Y
N
Y
Y
N
Y
Y
N
N
Y
N
N
Y
N

Table 1: The group-I temporal functions: (9=8)J (J I & (J )).

TOU is the time of utterance
TOR is the time of reference

t
t

98
8
9
( ; + )]
8
9
(TOU- - , TOU- + )]
8
9
(TOR- - , TOR- + )]
8
9
(TOU+ - , TOU+ + )] 8
9
(TOR+ - , TOR+ + )] 8
9

relative/ forward/ indexical/

absolute backward anaphorical
absolute backward

reference
interval I
(t ; " t + ")]

forward
relative

backward

indexical
anaphorical

forward

is the argument to the preposition phrase when it is a time

is the argument to the preposition phrase when it is a duration.

indexical
anaphorical

" t

on J

"

t
t

at

t "

"

ago

t "

"

before

t "

"

in

t "

"

after

Table 2: The group-II temporal functions: (9=8)J (J I & (J )).

TOU is the time of utterance
TOR is the time of reference
and " is a small duration.

preposition
at

tense

perprofective gressive
past
N
Y
N/Y
N
present/
N
Y
future
N
past
N
Y
N
past
Y
Y
N
future
N
Y
pres/fut
N
past+
N
Y
future
N

is the argument to the preposition phrase when it is a time

is the argument to the preposition phrase when is a duration

4 The group-II temporal functions

Various temporal prepositions, including at, in, after,

before and ago (usually classed as an adverb) can be
used to express temporal functions of the by now familiar forms (17) and (18) but where the reference
interval I is of negligible temporal extension. These
temporal functions are thus used to locate events and
states in time rather than to report on their duration.
The reference interval centres around a point which
is given either absolutely or relatively. In the latter
case, the function may be determined relative to the
TOU (indexical) or the TOR (anaphorical), and may
be xed by adding (forward-looking) or subtracting
(backward-looking) the o set.
Even though the reference interval is negligible, it
is possible for there to be universal quanti cation.
Again, universal quanti cation is indicated by the
progressive aspect.
Only one of the prepositions, at, has an absolute temporal function, either existential or universal.
When the argument of the preposition is earlier than
the time of utterance, then the past tense is used,
when later, a non-past tense. The other prepositions
cover times relative to the time of utterance or reference. As for group-I functions, they can be either
backward or forward-looking. However, unlike groupI functions, a di erent preposition is used with a time
of utterance (ago, in) than with a time of reference
(before, after).
Examples of the group-II functions, either Absolute or Relative, Backward or Forward-looking, indexical or anaphorical and Universal or Existential
are:
(AB.U) We were working at 2 o'clock.
(AB.E) We (had) nished at 2 o'clock.
(AF.U) We are/(will be) working at 2 o'clock.
(AF.E) We (will) nish at 2 o'clock.
(RBiU) We were laughing three hours ago.
(RBiE) We stopped three hours ago.
(RBaU) We had been laughing three days before.
(RBaE) They had stopped three days before.
(RFiU) We will be laughing in three hours.
(RFiE) They (will) stop in three hours.
(RFaU) The trains would be running after
three days.
(RFaE) They would stop after three days.
Note that the perfect aspect is used with the
backward-looking anaphoricals (BaE and BaU) but,
surprisingly, not necessarily with the indexicals (BiE
and BiU).
The results are summarized in table 2 and again
depicted in gure 1.

before

before

after

after

Figure 2: The group-III temporal functions

5 The group-III temporal functions

Under this group we understand existential and universal temporal functions with reference intervals
START, t] or t, END], where t is the time supplied
by the prepositional argument and START and END
are contextually supplied parameters indicating the
start- and end-times of the temporal universe of discourse. That is, the group-III temporal functions are:
(19) 9J (J START t] & (J ))
(20) 8J (J START t] ! (J ))
(21) 9J (J t END] & (J ))
(22) 8J (J t END] ! (J )).
We have already met certain uses of until, by and
at with the perfective aspect, in which we took the
reference interval to be START, t]. However, the
main prepositions expressing the group-III temporal
functions are before and after. Both can be used existentially and universally:
(23) Tim died before/after the explosion.
(24) Sue worked before/after her marriage.
The mapping of these prepositions onto the time
axis is depicted in gure 2.

6 The group-IV temporal functions

This group contains just two temporal functions of

the general forms (17) and (18), where the reference
interval I is given directly by a pair of arguments
to the prepositions. That is, the group-IV temporal
functions are:
(25) 9J (J t1 t2] & (J )).
(26) 8J (J t1 t2] ! (J )).
We have already met prepositions expressing these
temporal functions: between : : : and and from : : : to,
since they were used in section 2 to introduce the
general approach taken here.
(27) Tim died between 5 and 6 o'clock.
(28) Sue worked from dawn to dusk.

7 The group-V temporal functions

This group also contains just two temporal functions

of the form (17) and (18), where the reference interval I is given directly as the argument to the preposition. Several temporal prepositions can be used to

8 The oating temporal functions

between ... and

from ... to

-6

t1

t2

Figure 3: The group-IV temporal functions

Group
III
IV
V

reference 9=8
interval I on J
START, t] 8=9
t, END]
8=9
t1 t2 ]
8

9
8
9

preposition
before
after
from: : : to
between: : : and
throughout/for/
through/over
in/on/during

Table 3: The groups III, IV and V temporal functions: (9=8)J (J I & (J )).
t

is the argument to the preposition phrase

and t2 are the arguments to a double preposition.

t1

express the group-V universal function: throughout,

for, through, over.
(29) Judy was working throughout the day.
(30) Judy is at home for Christmas.
(31) Judy survived through the winter.
(32) Judy worked over the summer.
In, on and during express the group-V existential
function:
(33) Judy worked best in the morning.
(34) Judy worked on Sunday.
(35) Judy worked hardestduring examinations.
The di erences in use encountered in the Brown
corpus between throughout, for, through and over on
the one hand and in, on and during on the other are
discussed in Bree et al. 4].
in/on/during
through/over
throughout/for

-6
-

Figure 4: The group V temporal functions

Leaving aside some doubtful cases (particularly concerning the preposition over), we have so far considered temporal functions belonging to one of two categories: the existential functions, which allow one to
say that something is true over some subinterval J of
a given reference-interval I , and the universal functions, which allow one to say that something is true
over all subintervals J of a given reference-interval
I . We now turn to a group of temporal functions for
which only the duration of the subinterval is xed, not
its location in time: the oating functions. Prepositions expressing functions in this group include for,
within, in and over.
The oating functions also fall into the two categories existential and universal. We begin with an
oft-noted ambiguity in for-sentences. The sentence
(36) Sue has worked for one hour
can be used to say that Sue was working throughout
the interval stretching backward one hour from the
TOU. In that case, we give (36) truth-conditions:
(37) 8K (K TOU TOU ; 1 hr] ! (Sue be
working on the letter)(K )).
But (36) can also be used to say that Sue spent a
period of one hour sometime in the past working on
the letter. In which case, plausible truth-conditions
would be, by analogy with (37),
(38) 9K (K START TOU ] & dur(K ) = 1hr
& 8J (J K ! (Sue be working on the
letter)(J )))
where dur(K ) is a function that returns the duration
of interval K . Thus, for, on this reading, has a temporal function of the form:
(39) 9K (K I & dur(K ) = t
& 8J (J K ! (J )))
where I is, in this case, the interval START, TOU].
It is as if the reference interval in form (18) has been
allowed to oat.3 So we might say that temporal
functions having form (39) are oating universal functions.
The sentence
(40) Charles will solve the problem within one
hour
is also ambiguous. It can be used in a discourse to say
that Charles will solve the problem sometime in the
interval stretching forward one hour from the TOU.
In that case, we give (40) truth-conditions:
3 Actually, this use of for is complicated by the fact that
the intervals over which the event in question is said to
have been taking place can sometimes be summed: that
is, (36) can be taken to be true if, e.g. Sue worked on
the letter over three periods of twenty minutes each. We
ignore this feature of for, noting it as a phenomenon which
does not t neatly into the framework developed here.

(41) 9K (K TOU TOU + 1 hr] & (Charles

solve the problem)(K )).
But (40) can also be used to say that Charles will
solve the problem sometime in the future, taking not
more than one hour to do so. In which case, plausible
truth-conditions would be
(42) 9K (K TOU END] & dur(K )=1 hr
& 9J (J K & (Charles solve the
problem)(J )))
Thus, within, on this reading, has a temporal function
of the form:
(43) 9K (K I & dur(K ) = t
& 9J (J K & (J )))
where I is, in this case, the interval TOU, END].
We say that temporal functions having form (43) are
oating existential functions.
Examples of the oating use that are Backwardor Forward-looking, Universal or Existential and indexical (or anaphoric), are:
(BUi) Judy (has) worked for one hour yesterday.
(BUi) The premium was/(has been) paid over
3 years.
(BEi) Judy (has) solved the problem in one
hour.
(BEi) Judy (has) solved the problem within one
hour.
(FUi) Judy will work for one hour tomorrow.
(FUi) The premium will be paid over 3 years.
(FEi) Judy will solve the problem in one hour.
(FEi) Judy will solve the problem within one
hour.
(.Ui) Judy works each day for 10 hours.
(BUa) Judy had worked for one hour already.
We see that for and over indicate a universal oating
function whereas in and within indicate an existential
oating function. The di erences between these pairs
is discussed in Bree et al. 4].
The backward/forward-looking distinction is carried by the tense or perfect aspect as is usual: past
tense or perfect aspect indicating backward-looking,
future tense indicating forward-looking. The present
tense indicates a universal time interval, i.e. I is
START, END]. The tense also indicates whether the
interval I is bounded by the time of utterance, as in
most of the above examples, or by the time of reference, as in the last example BUa. The di erence
between indexical and anaphoric use is again carried
by the tense and aspect as normally, so we do not
consider it further here.
The rules for the oating temporal function are
given in table 4 and given some depiction in gure 5.

over
(with)in

-for
?

t
Figure 5: Floating temporal functions
reference
9=8
interval I
on J
START, TOR]/ 8
START TOU]
TOR, START]/
TOU START]

9
8
9

prepos- tense
perition
fective
for/
present
Y
over
past
Y/N
in
present
Y
within
past
Y/N
for/ present/
N
over
future
Y/N
in/
present
Y
within
future
Y/N

Table 4: The oating temporal functions: 9K (K

I & dur(K ) = t & (9=8)J (J K & (J )))
TOR is the time of reference
TOU is the time of utterance
t is the argument to the preposition phrase

9 Conclusion

In this paper, we chart the expressive power of

the English temporal preposition system by determining the forms of the temporal functions|
functions mapping underlying tenseless sentences
onto propositions|which those prepositions can express. We distinguish just four such forms, namely:
(44) 9J (J I & (J ))
(45) 8J (J I ! (J ))
(46) 9K (K I & dur(K ) = t
& 9J (J K & (J )))
(47) 9K (K I & dur(K ) = t
& 8J (J K ! (J ))).
We have shown how the parameters I and t are
xed in a number of ways, using, as well as the prepositional arguments, indexical, anaphorical and other
contextual parameters. We distinguish ve groups
of temporal functions having forms (44) and (45).
These are: 16 group-I functions, where one of the
bounds on the reference interval I is xed at either
the TOR or the TOU 12 group-II functions in which
the reference interval I is some sort of interval straddling a time point speci ed either absolutely or relative to the TOU or TOR 4 group-III functions in
which one bound of the reference interval is either
the start or the end of the temporal universe of discourse 2 group-IV functions in which the endpoints
of the reference interval are given directly by a pair
of arguments and 2 group-V functions in which the

reference interval is given by a single argument. In total, then, we have 36 temporal functions of forms (44)
and (45)|that is: 2 temporal forms, each with 18
ways of specifying the reference interval I .
We have seen that not all these 36 cases are equally
easy to specify in particular backward-looking existentials in group-I are awkward. Also there is more
than one preposition for the group-V functions, indicating that some of these carry additional nontemporal information.
In addition, we have distinguished a nal group of
8 oating functions having forms (46) and (47)|each
with a 4 ways of specifying I .
We have investigated how temporal prepositions,
by combining with tense and aspect, span this logical
space and we have explained how the well-known restrictions on the use of temporal prepositions with
underlying tenseless sentences of various aspectual
classes arise.

References

1] Allen, James \Towards a general theory of action

and time" AI Journal, 23(2), 1984.
2] Bree, D. S. \Words for Time" in Macar, Pouthas
and Friedman 12]
3] Bree, D. S., R. A. Smit and J.P. van Werkhoven
\Translating Temporal Prepositions between
Dutch and English", Journal of Semantics, 7,
pp. 1{51.
4] Bree, D. S., Allel Feddag and Ian Pratt \Towards
a formalisation of the semantics of some temporal prepositions" Time and Society, 2(2), 1993,
pp. 219{240.
5] Dowty, David, R. \The e ects of Aspectual Class
on the Temporal Structure of Discourse: Semantics or Pragmatics?", Linguistics and Philosophy, 9(1), 1989, pp. 37{61.
6] Van Eynde, Frank \Towards a Dynamic and
Compositional Treatment of temporal Expressions" in Proceedings, Eighth Amsterdam Colloquium
7] Herweg, Michael \A Critical Examination of
Two Classical Approaches to Aspect" Journal of
Semantics, 8, 1991, pp. 363{402.
8] Hinrichs, Erhard \Temporal Anaphora in Discourses of English", Linguistics and Philosophy,
9(1), 1989, pp. 63{82.
9] Hornstein, Norbert As Time Goes by: Tense
and Universal Grammar, Cambridge, MA: MIT
Press 1990.
10] Kucera, H., and W.N. Francis Computational
analysis of present-day American English, Providence, RI: Brown University Press, 1967.

11] Lascarides, Alex \The progressive and the imperfective paradox", Synthese 87(3), 1991, pp.
401{448.
12] Macar, F., V. Pouthas and W. J. Friedman
Time, Action and Cognition: Towards Bridging
the Gap, NATO ASI Series D, vol. 66, Dordrecht:
Kluwer, 1992.
13] Mittwoch, Anita \Aspects of English aspect: on
the interaction of perfect, progressive and durational phrases", Linguistics and Philosophy,
11(2), 1988, pp. 203{254.
14] Parsons, Terence \The progressive in English",
Linguistics and Philosophy, 12(2), 1989, pp. 213{
241.
15] Rice, Sally, A: \Polysemy and Lexical Representation: the case of three English Prepositions",
Proceedings of the fourteenth annual conference
of the cognitive science society, Hillsdale, NJ:
Lawrence Earlbaum, 1992.
16] Richards, Barry, Inge Bethke, Jaap van der Does
and John Oberlander Temporal representation
and inference, London: Academic Press, 1989.
17] Vendler, Zeno Linguistics and Philosophy,
Ithaca: Cornell University Press, 1967