TEMPORAL ONTOLOGY AND TEMPORAL REFERENCE 
Marc Moens 
Centre for Cognitive Science 
University of Edinburgh 
Edinburgh, Scotland EH8 9LW 
Mark Steedman 
Department of AI 
Centre for Cognitive Science 
University of Edinburgh 
and 
Department of Computer and Information Science 
University of Pennsylvania 
Philadelphia, PA 19104-6389 
"Two weeks later, Bonadea had already been his lover 
for a fortnight." 
--Robert Musil, Der Mann ohne Eigenschaften. 
A semantics of temporal categories in language and a theory of their use in defining the temporal 
relations between events both require a more complex structure on the domain underlying the meaning 
representations than is commonly assumed. This paper proposes an ontology based on such notions as 
causation and consequence, rather than on purely temporal primitives. A central notion in the ontology 
is that of an elementary event-complex called a "nucleus." A nucleus can be thought of as an association 
of a goal event, or "culmination," with a "preparatory process" by which it is accomplished, and a 
"consequent state," which ensues. Natural-language categories like aspects, futurates, adverbials, and 
when-clauses are argued to change the temporal/aspectual category of propositions under the control of 
such a nucleic knowledge representation structure. The same concept of a nucleus plays a central role 
in a theory of temporal reference, and of the semantics of tense, which we follow McCawley, Partee, and 
Isard in regarding as an anaphoric category. We claim that any manageable formalism for natural- 
language temporal descriptions will have to embody such an ontology, as will any usable temporal 
database for knowledge about events which is to be interrogated using natural language. 
1 INTRODUCTION 
It is often assumed that the semantics of temporal 
expressions is directly related to the linear time concept 
familiar from high-school physics--that is, to a model 
based on the number line. However, there are good 
reasons for suspecting that such a conception is not the 
one that our linguistic categories are most directly 
related to. When-clauses provide an example of the 
mismatch between linguistic temporal categories and a 
semantics based on such an assumption. Consider the 
following examples, suggested by Ritchie 1979: 
1. When they built the 39th Street bridge... 
a .... a local architect drew up the plans. 
b .... they used the best materials. 
c .... they solved most of their traffic problems. 
To map the temporal relations expressed in these ex- 
amples onto linear time, and to try to express the 
semantics of when in terms of points or intervals 
(possibly associated with events), would appear to 
imply either that when is multiply ambiguous, allowing 
these points or intervals to be temporally related in at 
least three different ways, or that the relation expressed 
Copyright 1988 by the Association for Computational Linguistics. Permission to copy without fee all or part of this material is granted provided 
that the copies are not made for direct commercial advantage and the CL reference and this copyright notice are included on the first page. To 
copy otherwise, or to republish, requires a fee and/or specific permission. 
0362-613X/88/010015-28503.00 
Computational Linguistics, Volume 14, Number 2, June 1988 15 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
between main and when-clauses is one of approximate 
coincidence. However, neither of these tactics explains 
the peculiarity of utterances like the following: 
2. #When my car broke down, the sun set. 
The unusual character of this statement seems to arise 
because the when-clause predicates something more 
than mere temporal coincidence, that is, some contin- 
gent relation such as a causal link or an enablement 
relation between the two events. Our knowledge of the 
world does not easily support such a link for (2), at least 
if we don't indulge in the fiction that the natural 
universe is conspiring against the speaker. Nor is the 
relation predicated between the two events by when the 
one that we normally think of as scientifically causal, 
for when seems to predicate an intransitive relation. 
Consider: 
3. a. When John left, Sue cried. 
b. When Sue cried, her mother got upset. 
c. When John left, Sue's mother got upset. 
From (3a) and (b) it would be unwarranted to conclude 
the state of affairs that is described in (c). And this 
causal aspect of the sentence's meaning must stem from 
the sense-meaning of when, because parallel utterances 
using while, just after, at approximately the same time 
as, and the like, which predicate purely temporal coin- 
cidence, are perfectly felicitous. 
We shall claim that the different temporal relations 
conveyed in examples (1) and (2) do not arise fi'om any 
sense-ambiguity of when, or from any "fuzziness" in 
the relation that it expresses between the times referred 
to in the clauses it conjoins, but from the fact that the 
meaning of when is not primarily temporal at all. Nor is 
it simply causal, as Example 3 shows. We will argue 
instead that when has a single sense-meaning, reflecting 
its role of establishing a temporal focus, which we 
follow Isard and Longuet-Higgins (1973) in relating to 
Reichenbach's reference time (cf. introduction to this 
collection). The apparent diversity of meanings arises 
from the nature of this referent and the organisation of 
events and states of affairs in episodic memory under a 
relation we shall call contingency, a term related, but not 
identical to a notion like causality, rather than mere 
temporal sequentiality. This contingent, nontemporal 
relation on the representation of events in episodic 
memory also determines the ontology of propositions 
associated with linguistic expressions denoting events 
and states. It is to these that we turn first. 
2 TEMPORAL AND ASPECTUAL CATEGORIES 
Propositions conveyed by English sentences uttered in 
context can, following Vendler, be classified into tem- 
poral or aspectual types, partly on the basis of the 
tenses, aspects, and adverbials with which they can 
co-occur (cf. Dowty 1979, and the introduction to the 
present collection). The term aspectual type refers to the 
relation that a speaker predicates of the particular 
happening that their utterance describes, relative to 
other happenings in the domain of the discourse. What 
the speaker says about those relations is of course quite 
distinct from what those relations objectively are. In 
partictdar, the speaker's predications about events will 
typically be coloured by the fact that those events are 
involved in sequences that are planned, predicted, 
intended, or otherwise governed by agencies of one 
kind or another. For want of some established term to 
cover this very general class of dependencies between 
events, we will use the term contingency. Thus an 
utterance of 
4. Harry reached the top 
is usually typical of what we will call a culmination-- 
informally, an event which the speaker views as punc- 
tual or instantaneous, and as accompanied by a transi- 
tion to a new state of the world. ~ This new state we will 
refer to as the consequent state of the event. It does not 
necessarily include all events that are objectively and in 
fact consequences. It rather includes only those conse- 
quences that the speaker views as contingently related 
to other events that are under discussion, say by 
causing them or by permitting them to occur. For 
reasons that are discussed in Section 3.2 below, expres- 
sions like these readily combine with the perfect, as in 
5. Harry has reached the top. 
The point may perhaps best be made by noting that 
there is another class of punctual expressions that is not 
normally associated with a consequent state. For exam- 
ple, 
6. John hiccupped 
is not usually viewed as leading to any relevant change 
in the state of the world. It typifies what we call a point 
expression. A point is an event (not necessarily an 
instantaneous one) that is viewed as an indivisible 
whole and whose consequences are not at issue in the 
discourse--which of course does not mean that de facto 
consequences do not exist. Such expressions are evi- 
dently not the same as culminations, for they are rather 
odd in combination with the perfect, as in 
7. #Harry has hiccupped. 
The reasons for this will also be discussed below. 
Sentences like 
8. Harry climbed 
typify a third aspectual category, which we will call for 
obvious reasons a process. Most utterances of such 
sentences describe an event as extended in time but not 
characterised by any particular conclusion or culmina- 
tion. As was pointed out by Vendler, expressions like 
these can be combined with a for-adverbial but not with 
an/n-adverbial: 
16 Computational Linguistics, Volume 14, Number 2, June 1988 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
9. Harry climbed for several hours. 
#Harry climbed in several hours. 
In contrast, 
10. Harry climbed to the top 
typically describes a state of affairs that also extends in 
time but that does have a particular culmination asso- 
ciated with it at which a change of state takes place. We 
classify most utterances of such sentences as a fourth 
aspectual type, called a culminated process. Culminated 
processes, in contrast to ordinary processes, combine 
readily with an/n-adverbial but not with a for-adverbial. 
11. Harry climbed all the way to the top in less than 45 
minutes. 
#Harry climbed all the way to the top for less than 
45 minutes. 
All of the above categories describe what common 
sense suggests we call events---that is, happenings with 
defined beginnings and ends. We distinguish these 
"hard-edged" categories from a class of indefinitely 
extending states of affairs, which, equally commonsen- 
sically, we call states. Example 12 typically describes 
one kind of state: 
12. Harry is at the top. 
Part of the appeal of Vendler's account, and such 
descendants as the present proposal, is that it suggests 
that part of the meaning of any utterance of a sentence 
is one of a small number of temporal/aspectual profiles 
distinguished on a small number of dimensions. In 
present terms, the event-types can be distinguished on 
just two dimensions, one concerned with the contrast 
between punctuality and temporal extension, the other 
with the association with a consequent state. This 
subcategorisation can be summarized as in Figure 1. 
+conseq 
-conseq 
EVENTS STATES 
atomic 
CULMINATION 
recognize, spot, 
win the race 
POINT 
hiccup, 
tap, wink 
extended 
CULMINATED 
PROCESS 
build a house, 
eat a sandwich 
PROCESS 
run, swim. walk, 
play the piano 
understand, 
love, know, 
resemble 
Figure 1. 
We have included in Figure 1 examples of verbs which 
typically yield propositions of the relevant types, and 
we shall assume that such verbs (or, strictly speaking, 
the associated uninstantiated propositions) are lexically 
specified as bearing that type. However, it cannot be 
stressed too often that these aspectual profiles are 
properties of sentences used in a context: sense-mean- 
ings of sentences or verbs in isolation are usually 
compatible with several (or even all possible) Vendle- 
rian profiles, as Dowty and Verkuyl have pointed out-- 
hence the frequent use of words like "typically" and 
"readily" above. The details of this taxonomy and the 
criteria according to which utterances can be catego- 
rised are less important than the observation that each 
primitive entity of a given type, such as the culmination 
event of Harry's reaching the top, carries intimations of 
other associated events and states, such as the process 
by which the culmination was achieved and the conse- 
quent state that followed. What linguistic devices like 
tenses, aspects, and temporal/aspectual adverbials ap- 
pear to do is to transform entities of one type into these 
other contingently related entities, or to turn them into 
composites with those related entities. 
For example, we shall argue below that the progres- 
sive auxiliary demands that its argument be a process, 
which it predicates as ongoing. If it is combined with an 
event type that isn't a process, say with a punctual 
event as in Harry was hiccupping, then it will cause that 
original event to be reinterpreted as a process, in this 
case the process of iteration or repetition of the basic 
event. Similarly, we shall argue that a perfect auxiliary 
demands a culmination, predicating of the time referred 
to that the associated consequent state holds. The 
notion of "time referred to" is related to Reichenbach's 
reference time in Section 4.1 below. If the perfect is 
combined with an event description for which world 
knowledge provides no obvious culmination, then the 
ensemble will tend to be anomalous. So, for example, 
Harry has reached the top is fine, but The clock has 
ticked, and Harry has hummed, to the extent that they 
are acceptable at all, seem to demand rather special 
scenarios in which the tick of the clock and the mere act 
of humming have a momentousness that they usually 
lack. 
The phenomenon of change in the aspectual type of a 
proposition under the influence of modifiers like tenses, 
temporal adverbials, and aspectual auxiliaries is of 
central importance to the present account. We shall talk 
of such modifiers as functions which "coerce" their 
inputs to the appropriate type, by a loose analogy with 
type-coercion in programming languages (cf. Ait-Kaci 
1984). Thus the effect on meaning of the combination of 
the progressive with an expression denoting an atomic 
punctual event as in Sandra was hiccupping occurs in 
two stages: first the point proposition is coerced into a 
process of iteration of that point. Only then can this 
process be defined as ongoing, and hence as a progres- 
sive state. These two stages might be represented as in 
the following diagram: 
Computational Linguistics, Volume 14, Number 2, June 1988 17 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
13. (point (Sandra hiccup)) 
(process (iteration (point (Sandra hiccup)))) 
(progressive (process (iteration (point (Sandra 
hiccup))))) 
The temporal/aspectual ontology that underlies the phe- 
nomenon of aspectual type coercion can be defined in 
terms of the transition network shown in Figure 2, in 
which each transition is associated with a change in the 
content and where, in addition, the felicity of any 
particular transition for a given proposition is condi- 
tional on support from knowledge and context. 
EVENTS 
atomic I extended 
~N/''~ +prep.proce,s TED~ "~" HABITUAL STATE CULMINA 
÷ conseq. CULMI. AT\[ON PROCESS / 7i 
21) . tnprogres~ STATE 
-conseq. POINT, ~ PROCESS iteration 
LEX\[CAL STATE 
STATES 
Figure 2. 
Rather than attempting to explain this diagram from first 
principles, we present below a number of examples of 
each transition. However, it is worth noting first that 
many of the permissible transitions between aspectual 
categories illustrated in Figure 2 appear to be related to 
a single elementary contingency-based event structure 
which we call a nucleus. A nucleus is defined as a 
structure comprising a culmination, an associated pre- 
paratory process, and a consequent state. 2 It can be 
represented pictorially as in Figure 3: 
preparatory process consequent state 
I//////////////////////I/////////////////// 
I 
culmination 
Figure 3. 
Any or all of these elements may be compound: for 
example, the preparation leading to the culmination of 
reaching the top of Mt. Everest may consist of a number 
of discrete steps of climbing, resting, having lunch, or 
whatever. The consequent state may also be compound; 
most importantly, it includes the further events, if any, 
that are in the same sequence of contingently related 
events as the culmination. SimilArly, the culmination 
itself may be a complex event. For example, we shall 
see below that the entire culminated process of climbing 
Mt. Everest can be treated as a culmination in its own 
right. \]in this case, the associated preparatory process 
and consequent state will be different ones to those 
internal to the culminated process itself. 
3 ASPECT 
3.1 THE PROGRESSIVE 
According to the present theory, progressive auxiliaries 
are fimctions that require their input to denote a proc- 
ess. Their result is a type of state that we shall call a 
progressive state, which describes the process as ongo- 
ing at the reference time. Thus the following sentence, 
among other meanings that we shall get to in a moment, 
can simply predicate of a present reference time that the 
process in question began at some earlier time and has 
not yet stopped: 
14. The president is speaking. 
If the input to a progressive is atomic then by definition 
it cannot be described as ongoing. However, as was 
noted in the introduction, it may be coerced into a 
process by being iterated, as in 
15. Harry is hiccupping. 
There is another route through the network in Figure 2, 
where the point is coerced into a culmination, i.e., as 
constituting an atomic event that does have conse- 
quences associated with it. In this case, the interpreta- 
tion for (15) parallels the one given for Harry was 
reaching the top, below. However, this particular ex- 
ample is deliberately chosen in order to make that 
interpretation unlikely. 
If a progressive combines with a culminated process, 
as in: 
16. Roger was running a mile 
--then the latter must also first be coerced to become a 
process. The most obvious way to do this is to strip off 
the culmination and leave the preparatory process be- 
hind. It is this process that is stated to be ongoing at the 
past reference time. Another possible coercion is to 
treat the entire culminated process as a point, and to 
iterate it. This interpretation appears to be the one that 
is forced by continuing (16) as in: 
17. Roger was running a mile last week. This week he is 
up to three. 
When a culmination expression like reach the top is 
used with a progressive, it must be coerced to become 
a process in a slightly more complicated way. The most 
18 Computational Linguistics, Volume 14, Number 2, June 1988 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
obvious path through the network in Figure 2 from the 
culmination node to the process node involves first 
adding a preparatory process to the culmination to make 
it a culminated process, then stripping off the culmina- 
tion point as before. Thus sentences like the following 
describe this preparatory process as ongoing at the past 
reference time: 
18. Harry was reaching the top. 
Again, an iterated reading is possible in principle, but 
pragmatically unlikely here. 
As a result of the coercions implicit in the last two 
examples, it is no longer asserted that the culminations 
in question ever in fact occurred, but only that the 
associated preparatory processes did. Thus there is no 
contradiction in continuations that explicitly deny the 
culmination, like: 
19. a. Harry was running a mile, but he gave up after 
two laps. 
b. Harry was reaching the top when he slipped and 
fell to the bottom. 
The fact that, according to the present theory, progres- 
sives coerce their input to be a process so that any 
associated culmination is stripped away and no longer 
contributes to truth conditions provides a resolution of 
the imperfective paradox (Dowty 1979), without appeal- 
ing to theory-external constructs like inertia worlds. 
3.2 TIlE PERFECT 
A perfect, as in 
20. Harry has reached the top 
is a function that requires its input category to be a 
culmination. Its result is the corresponding consequent 
state. The most obvious of these consequences for (20) 
is that Harry still be at the top, although as usual there 
are other possibilities. Informal evidence that this in- 
deed is the function of the perfect can be obtained by 
noticing that perfects are infelicitous if the salient 
consequences are not in force. Thus, when I'm on my 
way to get a cloth to clean up the coffee I accidentally 
spilled, I can say 
21. I have spilled my coffee. 
After cleaning up the mess, however, all the obvious 
consequences associated with this event seem to be 
over. In that context, it would be infelicitous to utter 
(21). 
If the input to a perfect is not a culmination, then the 
perfect will do its best to coerce it to be one, subject to 
the limitations imposed by contextual knowledge. If the 
hearer cannot identify any relevant consequences, as 
seems likely for the following example, then coercion 
may simply fail, in which case a perfect will be infelic- 
itous, as was noted earlier: 
22. #The star has twinkled. 
To be able to use a culminated process expression like 
climbing Mount Everest with a perfect auxiliary, it first 
has to be coerced into a culmination. Requiring such a 
transition might seem unnecessary since a culminated 
process already implies the existence of a culmination 
with consequences to which the perfect could refer. But 
consider Figure 4 as a possible rendering of the nucleus 
associated with climbing Mt. Everest: 
climbing the mountain being at the top 
t///lll//I/////////////f//I/I/I///////I/I/I 
reaching the summit 
of Mt. Everest 
Figure 4. 
Ifa perfect could be used to single out the consequences 
of a nucleus associated with a culminated process 
expression, then having climbed Mt. Everest could be 
used to refer to the state of having reached the/summit 
or being at the top. However, this does not seem to be 
the case. A reporter who has managed to establish radio 
contact with a mountaineer who has just reached the top 
of Mt. Everest is unlikely to ask 
23. Have you climbed Mt. Everest yet? 
The question rather seems to concern consequences of 
the culminated process as a whole. We capture this fact 
by making the perfect coerce the culminated process to 
become a culmination. The transition network allows 
this to happen if the entire event of climbing Mt. 
Everest is treated as a single unit by making it a point, 
so that it can become a culmination in its own right. The 
perfect then delivers a rather different kind of conse- 
quent state. 
A process like work in the garden can be coerced by 
a perfect auxiliary in essentially the same way: the 
process of working, possibly associated with a culmina- 
tion point, is treated as a single unit. This pointlike 
entity can then be used as the starting point for the 
construction of a new nucleus, by treating it as a 
culmination in its own right, provided that there are 
associated consequences. As a result, a question like 24 
can only be used felicitously if John's working in the 
garden was (for example) part of a prearranged plan, or 
a particular task John had to finish before something 
else could happen: 
24. #Has John worked in the garden? 
This account also explains the infelicity of a sentence 
like (25): 
25. #They have married yesterday. 
The sentence could only refer to the consequences of 
getting married yesterday as opposed to getting married 
Computational Linguistics, Volume 14, Number 2, June 1988 19 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
some other time. But most of what we think of as 
consequences of events are independent of the specific 
time at which the event occurred. (In this respect they 
are different from the preparatory processes, which are 
argued below to be implicated in certain futurates.) If a 
certain situation is a consequence of an event taking 
place at a particular time, then a perfect auxiliary may 
be used to describe that event. Thus a superstitious 
person believing that disastrous consequences are likely 
to result from actions performed on an unpropitious 
date can say: 
26. They have married on Friday the 13th! 
But even on Saturday the 14th, such a person still 
cannot use (25), for it would not provide the essential 
information about the date, thus flouting Grice's maxim 
of quantity. 
The account given here also explains the well-known 
contrast between the infelicitous (27a) and its felicitous 
counterpart, (b): 
27. a. #Einstein has visited Princeton. 
b. Princeton has been visited by Einstein. 
Whatever causal sequence of events and their conse- 
quences associated with the individual (Einstein) we 
take to be the one we are currently talking about, (a) 
cannot be used felicitously to refer to a part of that 
sequence since all such causal sequences seem to be to 
do with his enduring consciousness and are therefore by 
definition over. However, (b) can be uttered felicitously 
to refer to that same event because the relevant causal 
sequence must be one whose event and consequences 
apply to the institution of Princeton University (whose 
corporate consciousness endures) and many such con- 
sequences are still in train. 
The hypothesis we advance that the perfect has only 
one temporal meaning has a precedent in the work of 
Inoue 1979. Moens 1987 has extended the present 
analysis to show that the distinctions McCawley 1971, 
1981 and Comrie 1976 draw between different kinds of 
perfects (such as "perfect of current relevance," "hot 
news," "result," etc.) are nothing but different conse- 
quent states, depending on the nature of the verbal 
expression and the particular core event it expresses, 
and the specific kind of episodes in which our general 
knowledge tells us such core events typically occur. 
3.3 ADVERBIALS 
For-adverbials can only be used felicitously with proc- 
ess expressions: 
28. John worked in the garden for five hours. 
The resulting combination is a culminated-process 
expression. Evidence for this can be found in the ease 
with which an expression like (28) can be combined with 
a perfect, unlike its process counterpart: 
29. #John has worked in the garden. 
John has worked in the garden for five hours. 
An expression like playing the sonata can readily occur 
with a for-adverbial, suggesting that its basic category-- 
by which we mean the type assigned in the lexicon and 
inherited by the proposition in the absence of any 
coercion--is that of a process. As a result, (30) carries 
no implication that Sue finished playing the sonata: 
30. Sue played the sonata for a few minutes. 
Another' route through the network is possible in order 
to account for examples like (30): Sue's playing the 
sonata, like any other event, can be viewed as an 
unstructured point. A transition to turn it into a process 
then results in an iteration of occurrences at which Sue 
plays the sonata. This route through the network seems 
to be ruled out for (30) because it finds no support in our 
knowledge about sonatas and about how long they 
typically last. It does result, however, in a likely inter- 
pretation for a sentence like 
31. Sue played the sonata for about eight hours. 
A similar transition path is needed to make sense of 
examples like the following, in which a culmination is 
coerced to become a point, and then in turn coerced to 
become a process by being iterated: 
32. John arrived late at work for several days. 
The aspectual network would wrongly predict the ex- 
istence of a for-adverbial paradox, parallel to the imper- 
fective paradox, if for-adverbials were permitted to 
freely coerce culminated processes (and hence culmina- 
tions) to be (not necessarily completed) processes. The 
theory might seem to wrongly predict that (a) below 
would mean roughly the same as (b): 
33. a. #Red Rum won the race for the first few min- 
utes. 
b. Red Rum was winning the race. 
However, it is hard to find a context in which (a) means 
anything at all. The reason for this lies in the way 
English syntax and morphology control coercion in the 
aspectual transition network. The transition from cul- 
mination to consequent state, for example, demands the 
presence of a perfect. Similarly, the arc from process to 
progressive state may be traversed only if a progressive 
auxiliary is present in the sentence. For other transi- 
tions, such as the one resulting in an iterated process or 
an habitual state, English has no explicit markers and 
they can be made freely. 
The transition from culminated process to process is 
not one that can be made freely in English, but seems to 
require the presence of a progressive -ing-form. As a 
result, turning the culmination in (33a) into a process by 
first adding a preparatory process and then stripping off 
the culmination point is not allowed. It is allowed in (b), 
but only because the example contains the required 
progressive -ing-form. The only other transition path in 
the aspectual network that can account for the combi- 
nation of a culmination with a for-adverbial is the one 
20 Computational Linguistics, Volume 14, Number 2, June 1988 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
that turns the culmination into a point, and then iterates 
it to be a process. This interpretation is not felicitous for 
(33a), either, given our knowledge about what consti- 
tutes winning a race. However, as with (32), it is 
acceptable for 
34. Nikki Lauda won the Monaco Grand Prix for sev- 
eral years. 
Sometimes, a for-adverbial in combination with a cul- 
mination seems to describe a time period following the 
culmination rather than an iterated process: 
35. John left the room for a few minutes. 
This adverbial is of a different kind, however, express- 
ing intention rather than duration. It is merely by 
accident that English uses the same device to convey 
these different meanings. In French or German, for 
example, the two constructions are clearly distinct, as 
shown in the following translations of (35) and (32): 
36. Jean a quitt6 la chambre pour quelques minutes. 
Johann verliess fiir einige Minuten das Zimmer. 
37. Pendant des ann~es Jean est arriv6 en retard au 
travail. 
Jahrelang erschien Johann zu spat zur Arbeit. 
Not all aspectual/temporal adverbials expressing a time 
span have the same functional type./n-adverbials, for 
example, coerce their input to be a culminated process 
expression, as do related phrases like "it took me two 
days to .... " This means that combination with a 
culmination expression requires a transition to the cul- 
minated process node. According to the aspectual net- 
work in Figure 2 this transition is felicitous if the 
context allows a preparatory process to be associated 
with the culmination, as in (38): 
38. Laura reached the top in two hours. 
The/n-adverbial then defines the length of this prepa- 
ratory period. 
Since the arcs describe how one must be able to view 
the world for transitions to be made felicitously, it is 
obvious that there are expressions that will resist cer- 
tain changes. For example, it will be hard to find a 
context in which an/n-adverbial can be combined with 
a culmination expression like Harry accidentally spilled 
his coffee, since it is hard to imagine a context in which 
a preparatory process can be associated with an invol- 
untary act. Indeed, sentences like the following only 
seem to be made tolerable to the extent that it is 
possible to conjure up contexts in which the event only 
appears to be accidental: 
39. In fifteen minutes, Harry accidentally spilled his 
coffee. 
A similar problem arises in connection with the follow- 
ing example: 
40. John ran in a few minutes. 
The process expression John ran has to be changed into 
a culminated-process expression before combination 
with the/n-adverbial is possible. One way in which the 
network in Figure 2 will permit the change from a 
process to a culminated process is if the context allows 
a culmination point to be associated with the process 
itself. General world knowledge makes this rather hard 
for a sentence like John ran, except in the case where 
John habitually runs a particular distance, such as a 
measured mile. If the /n-adverbial had conveyed a 
specific duration, such as in four minutes, then the 
analysis would make sense, as Dowty has pointed out. 
However, the unspecific in a few minutes continues to 
resist this interpretation. 
However, another route is also possible for (40): the 
process of John running can be made into an atomic 
point, and thence into a culmination in its own right. 
This culmination can then acquire a preparatory process 
of its own--which we can think of as preparing to run-- 
to become the culminated process which the adverbial 
requires. This time, there is no conflict with the content 
of the adverbial, so this reading is the most accessible of 
the two. 
Since the transition network includes loops, it will 
allow us to define indefinitely complex temporal/aspect- 
ual categories, like the one evoked by the following 
sentence: 
41. It took me two days to play the "Minute Waltz" 
in less than sixty seconds for more than an hour. 
The process expression play the Minute Waltz is co- 
erced by the /n-adverbial into a culminated process, 
including a culmination of finishing playing the Minute 
Waltz. Combination with the for-adverbial requires this 
expression to be turned into a process--the only possi- 
ble route through the network being that through the 
point node and iterating. The resulting culminated- 
process expression describes the iterated process of 
playing the Minute Waltz in less than sixty seconds as 
lasting for more than an hour. The expression it took 
me .... finally, is like an /n-adverbial in that it is 
looking for a culminated-process expression to combine 
with. It would find one in the expression to play the 
Minute Waltz in less than sixty seconds for more than 
an hour, but combination is hampered by the fact that 
there is a conflict in the length of time the adverbials 
describe. In the case of (41), the whole culminated 
process is instead viewed as a culmination in its own 
right (via the path through the point node). Knowledge 
concerning such musical feats then supplies an appro- 
priate preparatory process that we can think of as 
practicing. The phrase it took me two days then defines 
the temporal extent of this preparatory process needed 
to reach the point at which repeatedly playing that piece 
of music so fast for such a considerable length of time 
became a newly acquired skill. We assume that the 
ordering of these successive coercions, like others 
Computational Linguistics, Volume 14, Number 2, June 1988 21 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
induced by the perfect and the progressive, are (not 
necessarily unambiguously) under the control of syntax. 
4 TENSE AND TEMPORAL FOCUS 
4.1 TENSE 
The aspects and temporal/aspectual adverbials consid- 
ered above all act to modify or change the aspectual 
class of the core proposition, subject to the limits 
imposed by the network in Figure 2, and by contextual 
knowledge. However, tenses and certain other varieties 
of adverbial adjuncts have a rather different character. 
Tense is widely regarded as an anaphoric category, 
requiring a previously established temporal referent. 
The referent for a present tense is usually the time of 
speech, but the referent for a past tense must be 
explicitly established. This is done by using a second 
type of "temporal" adjunct, such as once upon a time, 
at five o'clock last Saturday, while I was cleaning my 
teeth, or when I woke up this morning. 
Most accounts of the anaphoric nature of tense have 
invoked Reichenbach's (1947) trinity of underlying 
times and his concept of the positional use of the 
reference time. Under these accounts, temporal ad- 
juncts establish a referent to which the reference time of 
a main clause and subsequent same-tensed clauses may 
attach or refer, in much the same way that various 
species of full noun phrases establish referents for 
pronouns and definite anaphors (see foreword). 
Reichenbach's account is somewhat inexplicit as far 
as extended, noninstantaneous events go. In particular, 
he makes it look as though the reference time is always 
an instant. However, we believe that the following 
account is the obvious generalisation of his and proba- 
bly what he intended anyway. 
In Reichenbach's system a simple past tense of an 
atomic event is such that reference time (R) and event 
time (E) are identical, while progressives and perfects 
are such that R and E are not identical. 3 The only 
coherent generalisation of his scheme to durative events 
is to maintain this pattern and assume that R and E are 
coextensive for an utterance like: 
42. Harry ran a mile. 
It follows that R may be an extended period (cf. 
Steedman 1982). R may also be an extended period for 
a state such as a progressive, although in this case the 
corresponding event time is still quite separate, of 
course. 
What is the nature of this referent, and how is it 
established? The anaphoric quality of tense has often 
been specifically compared to pronominal anaphora (cf. 
McCawley 1971; Partee 1973; Isard 1974). However, in 
one respect, the past tense does not behave like a 
pronoun: use of a pronoun such as "she" does not 
change the referent to which a subsequent use of the 
same pronoun may refer, whereas using a past tense 
may. In the following example, the temporal reference 
point fi)r the successive conjoined main clauses seems 
to move on from the time originally established by the 
adjunct: 
43. At exactly five o'clock, Harry walked in, sat down, 
and took off his boots. 
Nor is this just a matter of pragmatic inference; other 
orders of the clauses are not allowed: 
44. #At exactly five o'clock, Harry took off his boots, 
sat down and walked in. 
This fact has caused theorists such as Dowty 1986, 
Hinrichs 1984, and Partee 1984 to stipulate that the 
reference time autonomously advances during a narra- 
tive. However, such a stipulation (besides creating 
problems for the theory vis-d-vis those narratives where 
reference time seems not to advance) seems to be 
unnecessary, since the amount by which the reference 
time advances still has to be determined by context. The 
concept of a nucleus that was invoked above to explain 
the varieties of aspectual categories offers us exactly 
what we need to explain both the fact that the reference 
time advances and by how much. We simply need to 
assume that a main-clause event such as Harry walked 
in is interpreted as an entire nucleus, complete with 
consequent state, for by definition the consequent state 
comprises whatever other events were contingent upon 
Harry walking in, including whatever he did next. 
Provided that the context (or the hearer's assumptions 
about the world) supports the idea that a subsequent 
main clause identifies this next contingent event, then it 
will provide the temporal referent for that main clause. 
If the context does not support this interpretation, then 
the temporal referent will be unchanged, as in: 
45. At five o'clock, my car started and the rain stopped. 
In its ability to refer to temporal entities that have not 
been explicitly mentioned, but whose existence has 
merely been implied by the presence of an entity that 
has been mentioned, tense appears more like a definite 
NP (e.g., the music in the following example) than like 
a pronoun, as Webber 1987 points out. 
46. I went to a party last night. The music was wonder- 
ful. 
4.2 WHEN-CLAUSES 
The definite nature of tense together with the notion of 
the nucleus as the knowledge structure that tensed 
expressions conjure up explain the apparent ambiguity 
of when-clauses with which this paper began. A when- 
clause behaves rather like one of those phrases that are 
used to explicitly change topic, such as and your father 
in the following example (cf. Isard 1975): 
47. And your father, how is he? 
22 Computational Linguistics, Volume 14, Number 2, June 1988 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
A when-clause does not require a previously established 
temporal focus, but rather brings into focus a novel 
temporal referent whose unique identifiability in the 
hearer's memory is presupposed. Again, the focused 
temporal referent is associated with an entire nucleus, 
and again an event main clause can refer to any part of 
this structure conditional on support from general or 
discourse specific knowledge. For example, consider 
again Example 1 with which we began (repeated here): 
48. When they built the 39th Street bridge... 
a .... a local architect drew up the plans. 
b .... they used the best materials. 
c .... they solved most of their traffic problems. 
Once the core event of the when-clause has been 
identified in memory, the hearer has two alternative 
routes to construct a complete nucleus: 
a) to decompose the core event into a nucleus and to 
make a transition to one of the components, such as 
the preparatory activity of building or to the conse- 
quent state of having built the bridge; or 
b) to treat the entire event as a single culmination and 
compose it into a nucleus with whatever preparation 
and consequences the context provides for the activ- 
ity of building a bridge, and to make the transition to 
either one of those~ 
Either way, once the nucleus is established, the refer- 
ence time of the main clause has to be situated some- 
where within it--the exact location being determined by 
knowledge of the entities involved and the episode in 
question. So in Example 48a, the entire culminated 
process of building the bridge tends to become a culmi- 
nation (via a path in Figure 2 that passes through the 
point node), which is associated in a nucleus with 
preparations for, and consequences of, the entire busi- 
ness, as in Figure 5: 
they prepare they have built 
to build the bridge 
Illllllllll/ll/l/llllllllllllllllllll 
I 
they build 
the bridge 
Figure 5. 
The drawing up of the plans is then, for reasons to do 
with knowledge of the world, situated in the preparatory 
phase. 
In Example b, in contrast, people tend to see the 
building of the bridge as decomposed into a quite 
different preparatory process of building, a quite differ- 
ent culmination of completing the bridge and some 
consequences that we take to be also subtly distinct 
from those in the previous case as was argued in Section 
3.2. The resulting nucleus is given in Figure 6. The use 
of the best materials is then, as in (a), situated in the 
preparatory process---but it is a different one this time. 
they build they have completed 
the bridge 
I////////////I///////////////////// 
J 
they complete 
the bridge 
Figure 6. 
Example c is like (a) in giving rise to the nucleus in 
Figure 5, but pragmatics seems to demand that the main 
clause be situated somewhere in the consequent state of 
building the bridge. 
Thus a main clause event can potentially be situated 
anywhere along this nucleus, subject to support from 
knowledge about the precise events involved. But Ex- 
ample 2, repeated here, is still strange, because it is so 
hard to think of any relation that is supported in this 
way: 
49. #When my car broke down, the sun set. 
The when-clause defines a nucleus, consisting of what- 
ever process we can think of as leading up to the car's 
breakdown, the breakdown itself, and its possible or 
actual consequences. It is not clear where along this 
nucleus the culmination of the sun set could be situated: 
it is not easy to imagine that it is a functional part of the 
preparatory process typically associated with a break- 
down, and it is similarly hard to imagine that it can be a 
part of the consequent state, so under most imaginable 
circumstances, the utterance remains bizarre. 
The constraints when places on possible interpreta- 
tions of the relation between subordinate and main 
clause are therefore quite strong. First, general and 
specific knowledge about the event described in the 
when-clause has to support the association of a com- 
plete nucleus with it. Secondly, world knowledge also 
has to support the contingency relation between the 
events in subordinate and main clauses. As a result, 
many constructed examples sound strange or are. con- 
sidered to be infelicitous, because too much context has 
to be imported to make sense of them. 
In all of the cases discussed so far, the main clause 
has been an event of some variety. With stative main 
clauses, as in the following examples, the interpretation 
strategy is somewhat different. Statives show no sign of 
being related under what we are calling contingency, 
presumably because contingency is by definition a 
relation over events. In particular, they do not enter in 
a causal or contingent relation with a when-clause the 
way corresponding sentences with events as main 
Computational Linguistics, Volume 14, Number 2, June 1988 23 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
clauses do. They therefore merely predicate that the 
state in question holds at the time of the culmination: 
50. When they built that bridge 
• . . I was still a young lad. 
• . . my grandfather had been dead for 
several years. 
• . . my aunt was having an affair with the 
milkman• 
• . . my father used to play squash. 
However, a stative main clause can be turned into an 
event expression; in that case, a contingency relation is 
predicated to exist between the two events• Thus the 
following example seems to involve an inceptive event, 
which begins the state of knowing: 
51. When Pete came in, I knew that something was 
wrong. 
Such changes of type are similar to others discussed 
above but are not treated in the present paper. 
5 REFERRING TO FUTURE EVENTS 
Bennett and Partee 1972, speaking of the difference 
between the present perfect and the simple past, remark 
that one might expect a similar distinction among future 
tenses• One could conceive of a construction parallel to 
the perfect, whose event time would be in the future and 
whose reference time would be the time of speech, 
conveying a notion of current relevance; and there 
could be a construction parallel to the simple past, with 
both reference and event times in the future. Bennett 
and Partee suggest that English is not as one would 
expect and follow Reichenbach in saying that these two 
functions are conflated in a single device, the modal 
future using will. Although it is true that the modal 
future shares features of both perfect and simple past, it 
is nevertheless also the case that there are two classes 
of futurate expressions, with properties parallel to each 
of the two past expressions. 
The candidate for the role parallel to the perfect is the 
so-called futurate progressive (Smith 1983): 
52. Robert was working on the speech project until he 
got a job offer from Sussex• 
As Dowty 1979, 1986 argues, examples like (52) can be 
both a past imperfective progressive (answering a ques- 
tion about Robert's past activities) and a past futurate 
progressive (answering a question about Robert's plans 
at some past time and meaning something like Robert 
was going to work on the speech project, but he didn't). 
However, the difference between the two interpreta- 
tions seems to be a matter of pragmatic world knowl- 
edge rather than sense-semantics, corresponding to the 
two different ways of constructing a nucleus (cf. Section 
4). The imperfective progressive decomposes the core 
event into a nucleus and makes a transition to the 
preparatory process, indicating that it is in progress at 
the time of reference• The futurate progressive, through 
the use of an adverbial signaling an event time posterior 
to the reference, forces the whole event to be treated as 
a single unit, which is then composed into a new 
nucleus. The progressive then indicates that the prepa- 
ration leading up to the event as a whole was in progress 
at the time of reference (as usual, without asserting that 
that event or even its onset was ever reached). The 
futurate progressive thus resembles the perfect in say- 
ing something about a (past or present) reference time 
that is entirely separate from the event time. 
The candidate for the role parallel to the simple past 
among the futurates is to be found in the simple, or 
non-modal future, sometimes (confusingly) called the 
tenseness future: 
53. He leaves on Tuesday. 
While the futurate progressive shares with the perfect 
the property of needing no nonpresent adverbial, the 
nonmodal future cannot be used in this way. For 
example, in response to a question about the current 
state of affairs as specific as Why are you being so rude 
to your boss these days? or as general as What's new?, 
one may respond with an unanchored progressive (54a), 
much as with a perfect (54b). But one may not reply 
with an unanchored nonmodal future (54c), although an 
anchored one (54d) is quite all right• 
54. a. I am leaving. 
b• I have handed in my notice. 
c. *I leave• 
d. I leave next month• 
In its requirement for an established non-present refer- 
ence time, the nonmodal future resembles the past 
tense• The resemblance (which was noted in Leech 
1971) is supported by the following further observa- 
tions. A when question concerning the past progressive 
is ambiguous, reflecting the separation of reference time 
and event time. By contrast, the nonmodal future does 
not really seem to occur in the past at all, except of 
course in reported or indirect speech; it just becomes 
indistinguishable from the simple past. It follows that 
(55) carl be answered with (a) or (b). But (56) can only be 
answered with (a), not with (b). 
55. When were you leaving? 
a• Last week (ambiguous)• 
b. Next week. 
56. When did you leave? 
a. Last week (unambiguous). 
b. *Next week. 
These similarities suggest the symmetry depicted infor- 
mally in Figure 7 between the perfect, the simple past, 
the futurate progressive, and the nonmodal future. The 
hatching again informally indicates the extent of the 
consequent state and the preparatory process associ- 
ated with the perfect and the futurate progressive, 
24 Comlmtational Linguistics, Volume 14, Number 2, June 1988 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
respectively. That is not to imply that the two are the 
same sort of entity: they are both states, but of a 
different kind. The perfect is a consequent state; the 
futurate progressive is a state derived from a prepara- 
tory process. This difference is indicated by the pres- 
ence of a defined upper bound on the latter. The 
Reichenbach diagram in Figure 7 for the nonmodal 
future is of course the one that is ascribed (traditionally 
and by Bennett and Partee) to the modal future, a 
construction to which we will return in a moment. 
Before doing so there are some problems remaining to 
be disposed of. 
If the futurate progressive is the true counterpart of 
the perfect,why is it not subject to the same restriction 
against nonpresent adverbials? 
57. a. John is leaving (tomorrow). 
b. John has left (*yesterday). 
The answer lies in the differences between preparatory 
processes and consequent states, rather than in the 
aspects themselves. In both cases the adverbial must 
associate with the core event of leaving rather than the 
present reference time. Thus (a) concerns the prepara- 
tions for leaving tomorrow (as opposed to some other 
time), while (b) concerns the consequences of leaving 
yesterday (as opposed to some other time). As was 
pointed out in Section 3.2, most of what we think of as 
consequences of events are independent of absolute 
time. This makes it hard to think of consequences 
associated with John's leaving yesterday as opposed to 
those associated with John's leaving generally. Prepa- 
ratory processes do not share this property: the prepa- 
ratory process associated with John's leaving tomorrow 
is conceivably very different from that associated with 
John's leaving next week. 
PAST 
E 
I/////////////////// 
I I I 
S,R E,R S 
(perfect) (simple past) 
John has left. John left. 
FUTURATE 
E 
IIIIIIIIIIIIIIIIIIIli 
I I I 
S,R S E,R 
(futurate progressive) (nonmodal future) 
John is leaving (tomorrow). John leaves tomorrow. 
Figure 7. 
One other difference between the futurate categories 
and the past categories should be mentioned. If the 
nonmodal future is the correlate of the simple past, it 
should be possible to have nonmodal futures of per- 
fects, just as with pasts of perfects. But Vetter 1973 has 
pointed out that the following is odd: 
58. The Dodgers have finished for the season next 
Sunday. 
Nevertheless, such futurates do appear in the context of 
futurate temporal adjuncts, as in the following example: 
59. Once the Dodgers play the Red Sox next Sunday, 
they have finished for the season. 
The other English futurate expressions also fit into the 
scheme of Figure 7. The "be going to" construction 
typified by 
60. I am going to buy a guitar. 
clearly belongs with the progressives, being distin- 
guished from them by the nature of the processes that it 
implicates (see Leech 1971; Palmer 1974; Wekker 1976, 
and references therein). The "be to" construction typ- 
ified by 
61. I am to be Queen of the May 
also seems to belong with the progressives, although its 
modal character has been remarked by Leech and 
Palmer. 
Finally, where does the modal future fit into this 
scheme? A full analysis of the modals would go beyond 
the scope of this paper, so the following remarks will be 
sketchy. The modal future clearly has a reference time 
not coincident with speech time, like the nonmodal 
future but unlike the futurate progressive. Neverthe- 
less, Bennett and Partee are quite right that the modal 
future says something about the present as well as the 
past. The source of its relevance to the time of speech 
must therefore have to do with the relation between 
modals and the time of speech. We make the following 
tentative suggestion about this relation. 
Palmer 1974 pointed out a systematic ambiguity 
within the epistemic modals as between a futurate and a 
strictly present meaning, and Steedman 1977 related 
this to the similar ambiguity of a present-tensed sen- 
tence. What needs to be added seems to be the idea that 
these (suspiciously untensed looking) modals define 
properties of the time of speech (as is implied by the 
speech-act theoretic analysis of Boyd and Thorne 1969) 
and do not of themselves have anything to do with 
reference time and event time, unlike the true tensed 
and aspectual auxiliaries. More specifically, will says of 
the time of speech that it leads the speaker to infer a 
proposition (possibly but not necessarily one concern- 
ing the future). Must says something very similar but 
seems to leave the speaker out of it and says that the 
proposition follows from the state of the world at speech 
time. May says that the proposition is permitted by the 
Computational Linguistics, Volume 14, Number 2, June 1988 25 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
state of the world at speech time. These senses are 
exhibited below. 
62. a. You will be my long-lost brother Willy. 
a'. You will marry a tall dark stranger. 
b. You must be my long-lost brother Willy. 
b'. You must marry a tall dark stranger. 
c. You may (or may not) be my long-lost brother, 
Willy. 
c'. You may (or may not) marry a tall dark 
stranger. 
But, as has often been suggested before, the future 
epistemic modals have nothing to do with future tense in 
the strict sense of the word. 4 
6 TOWARD A FORMAL REPRESENTATION 
We have argued in this paper that a principled and 
unified semantics of natural-language categories like 
tense, aspect, and aspectual/temporal adverbials re- 
quires an ontology based on contingency rather than 
temporality. The notion of nucleus plays a crucial role in 
this ontology. The process of temporal reference in- 
volves reference to the appropriate part of a nucleus, 
where appropriateness is a function of the inherent 
meaning of the core expression, of the coercive nature 
of co-occurring linguistic expressions, and of particular 
and general knowledge about the area of discourse. 
The identification of the correct ontology is also a 
vital preliminary to the construction and management of 
temporal databases. Effective exchange of information 
between people and machines is easier if the data- 
structures that are used to organise the information in 
the machine correspond in a natural way to the concep- 
tual structures people use to organise the same infor- 
mation. In fact, the penalties for a bad fit between 
data-structures and human concepts are usually crip- 
pling for any attempt to provide natural language inter- 
faces for database systems. Information extracted from 
natural-language text can only be stored to the extent 
that it fits the preconceived formats, usually resulting in 
loss of information. Conversely, such data-structures 
cannot easily be queried using natural language if there 
is a bad fit between the conceptual structure implicit in 
the query and the conceptual structure of the database. 
The contingency-based ontology that we are advo- 
cating here has a number of implications for the con- 
struction and management of such temporal databases. 
Rather than a homogeneous database of dated points or 
intervals, we should partition it into distinct sequences 
of causally or otherwise contingently related sequences 
of events, which we might call episodes, each leading to 
the satisfaction of a particular goal or intention. This 
partition will quite incidentally define a partial temporal 
ordering on the events, but the primary purpose of such 
sequences is more related to the notion of a plan of 
action or an explanation of an event's occurrence than 
to anything to do with time itself. It follows that only 
26 
events that are contingently related necessarily have 
well-defined temporal relations in memory. 
A first attempt to investigate this kind of system was 
reported in Steedman 1982, using a program that veri- 
fied queries against a database structured according to 
some of the principles outlined above; a more recent 
extension of this work was reported in Moens 1987. 
Even~s are stored as primitives in the database, possibly 
but not necessarily associated with a time point. Ex- 
tended events are represented in terms of a pair of 
punctual events, identifying their starting point as well 
as the point at which they end (in the case of processes) 
or culminate (in the case of culminated processes). 
Apart from the obvious accessibility relations of 
temporal precedence and simultaneity, events can also 
enter into the relation of contingency introduced above. 
It is significant that the relation used in the implemen- 
tation is identical to the notion of causality used by 
Lansky 1986 in an entirely different problem area. She 
developed a knowledge representation scheme for use 
in planners in which events are reified and modeled with 
an explicit representation of their temporal as well as 
causal relations. In this scheme, a mechanism is pro- 
vided for structuring events into so-called "locations of 
activity", the boundaries of which are boundaries of 
"cau~;al" access. As a result, two events with no causal 
relation between them cannot belong to the same loca- 
tion of activity--as in the episodes introduced above. 
Because we follow Lansky in making the contin- 
gency relation intransitive, we avoid certain notorious 
problems in the treatment of when-clauses and perfects, 
which arise because the search for possible conse- 
quences of an event has to be restricted to thefirst event 
on the chain of contingencies. Thus, when (3) is as- 
serted, repeated here as (63a) and (b), it would be wrong 
to infer (c): 
63. a. When John left, Sue cried. 
b. When Sue cried, her mother got upset. 
c. When John left, Sue's mother got upset. 
The reason is exactly the same as the reason that it 
would be wrong to infer that Sue's mother got upset 
because John left, and has nothing to do with the purely 
temporal relations of these events. It should also be 
noted that the notion of contingency used here (in line 
with Lansky's proposals) is weaker than the notion of 
causality used in other representation schemes (for 
example, that of McDermott 1982 or Allen 1984): if 
Event A stands in a contingent relation to Event B, then 
an occurrence of A will not automatically lead to an 
occmrence of B: John laying the foundations of the 
house is a prerequisite for or enables him to build the 
walls and roof, but does not cause it in the more 
traditional sense of the word and does not automatically 
or inevitably lead to him building the walls. 
The transitions in the network are implemented as 
inference procedures in the database. Answering a 
query involving the aspectual auxiliaries and adverbials 
Coml~utational Linguistics, Volume 14, Number 2, June 1988 
Marc Moens and Mark Steedman Temporal Ontology and Temporal Reference 
discussed before consists of finding a matching event 
description in the database and checking its aspectual 
type; if the event description is found not to have the 
required aspectual type, it can be changed by means of 
the inference procedures, provided such a change is 
supported by knowledge in the database about the event 
in question. 
7 CONCLUSION 
Many of the apparent anomalies and ambiguities that 
plague current semantic accounts of temporal expres- 
sions in natural language stem from the assumption that 
a linear model of time is the one that our linguistic 
categories are most directly related to. A more princi- 
pled semantics is possible on the assumption that the 
temporal categories of tense, aspect, aspectual adver- 
bials, and of propositions themselves refer to a mental 
representation of events that is structured on other than 
purely temporal principles, and to which the notion of a 
nucleus, or contingently related sequence of prepara- 
tory process, goal event, and consequent state, is 
central. 
ACKNOWLEDGMENTS 
We thank Jon Oberlander, Ethel Schuster, and Bonnie 
Lynn Webber for reading and commenting upon drafts. 
Parts of the research were supported by: an Edinburgh 
University Graduate Studentship; an ESPRIT grant 
(project 393) to CCS, Univ. Edinburgh; a Sloan Foun- 
dation grant to the Cognitive Science Program, Univ. 
Pennsylvania; and NSF grant IRI-10413 A02, ARO 
grant DAA6-29- 84K-006f, and DARPA grant N0014- 
85-K0018 to CIS, Univ. Pennsylvania. An earlier ver- 
sion of some parts of this paper was presented as Moens 
and Steedman 1987. 
NOTES 
1. Readers familiar with Vendler's work will realise that we have 
changed his terminology. We have done so both for notational 
convenience and to avoid the considerable confusion that has 
arisen concerning the precise meaning of the old terms. The new 
nomenclature is also intended to reflect the fact, also noted by 
Dowty (1979), that Vendler's "accomplishments," which we will 
refer to as "culminated processes," are composite events, consist- 
ing of a process which is associated with a particular culmination 
point. 
2. A similar tripartite event structure is proposed in Passonneau 
(1987, cf. this volume). 
3. In attributing this view to Reichenbach, we are assuming that there 
is an oversight or a misprint in his diagram for the past progressive, 
p. 290: the diagram seems to suggest that R and E are coextensive, 
whereas what is intended is that the punctual reference time is 
included in an extended event time, as in his diagram for the 
present progressive. We also ignore here one of his analyses of the 
modal future, which we regard as incorrect (cf. Section 5). 
4. It is ~.n implication of such an analysis that there should be no truly 
past w~rsion of epistemically modal propositions. Where past 
tense s of the epistemic modals do occur, they must, like the past 
nonntodal future, always be either counterfactual or indirect or 
repolted speech. This seems to be the case. Mary McCarthy 
(1974), speaking of David Halberstam's use in The Best and the 
Brigt~test of "what she could only describe as the Future Past," as 
in 
i. ~,1: a dinner party after the Bay of Pigs Bundy would tell 
lriends... 
and 
ii. The power and prestige that the McNamara years would 
bring... 
called it "that awful tense, seeming to endow the author with 
prophetic powers," signifying "a future already plangent when it 
has not yet happened." The source of that awful power (which also 
accntes to the past tenses of the nonmodal future and, as McCarthy 
also remarks, the modal-like "be to" construction), is of course the 
shifting of the speech or consciousness time into the past, rather 
than the reference time. 

REFERENCES 
Ait-Kaci, H. 1984 A Lattice Theoretic Approach to Computation 
Based on a Calculus of Partially Ordered Type Structures. Ph.D. 
thesis, University of Pennsylvania. 
Allen, J. F. 1984 Towards a General Theory of Action and Time. 
Artificial Intelligence 23:123-154. 
Bennett, M. and Partee, B. 1972 Toward the Logic of Tense and 
Aspect in English. Indiana University Linguistics Club, Blooming- 
ton, IN. 
Boyd, J. and Thorne, J. 1969 The Semantics of Modal Verbs. Journal 
of Linguistics 5:57-74. 
Comrie, B. 1976 Aspect: An Introduction to the Study of Verbal 
Aspect and Related Problems. Cambridge University Press, Cam- 
bridge, England. 
Dowty, D. 1979 Word Meaning and Montague Grammar. D. Reidel, 
Dordrecht, W. Germany. 
Dowty, D. 1986 The Effects of Aspectual Class on the Temporal 
Structure of Discourse: Semantics or Pragmatics?. Linguistics and 
Philosophy 9:37--61. 
Hinrichs, E. 1986 Temporal Anaphora in Discourses of English. 
Linguistics and Philosophy 9:63-82. 
Inoue, K. 1979 An Analysis of the English Present Perfect. Linguistics 
17:221-222, 561-590. 
Isard, S. D. 1974 What would you have done if... Theoretical 
Linguistics 1:233-255. 
Isard, S. D. 1975 Changing the Context. In Keenan, E. (ed.) Formal 
Semantics of Natural Language. Cambridge University Press, 
Cambridge, England. 
Isard, S. and Longuet-Higgins, C. 1973 Modal Tic-tac-toe. In Bogdan, 
R. J. and Niiniluoto, I. (eds.) Logic, Language and Probability. 
Reidel, Dordrecht, W. Germany:189-195. 
Lansky, A. L. 1986 A Representation of Parallel Activity Based on 
Events, Structure, and Causality. In Workshop on Planning and 
Reasoning about Action. Timberline Lodge, Mount Hood, OR: 
50-86. 
Leech, G. N. 1971 Meaning and the English Verb. Longman, Lon- 
don. 
McCarthy, M. 1974 Sons of the Morning. In The Seventeenth Degree. 
Harcourt Brace Jovanovich, New York, NY. 
McCawley, J. D. 1971 Tense and Time Reference in English. In 
Fillmore, C. and Langendoen, T. (eds.) Studies in Linguistic 
Semantics, Holt, Rinehart and Winston, New York, NY: 96-113. 
McCawley, J. D. 1981 Notes on the English Present Perfect. Austra- 
lian Journal of Linguistics 1:81-90. 
McDermott, D. 1982 A Temporal Logic for Reasoning about Pro- 
cesses and Plans. Cognitive Science 6:101-155. 
Moens, M. 1987 Tense, Aspect and Temporal Reference. Ph.D. 
thesis. Centre for Cognitive Science, University of Edinburgh, 
Edinburgh, Scotland. 
Moens, M. and Steedman, M. 1986 Temporal Information and Natural 
Language Processing. Research Paper/RP-2. Center for Cognitive 
Science, University of Edinburgh, Edinburgh, Scotland. 
Moens, M. and Steedman, M. 1987 Temporal Ontology in Natural 
Language. In Proceedings of the 25th Annual Meeting of the 
Association for Computational Linguistics. Stanford University, 
Stanford, CA: 1-7. 
Palmer, F. 1974 The English Verb. Longman, London. 
Partee, B. 1973 Some Structural Analogies Between Tenses and 
Pronouns in English Journal of Philosophy 70:601-609. 
Partee, B. 1984 Nominal and Temporal Anaphora. Linguistics and 
Philosophy 7:243-286. 
Passonneau, R. J. 1987 Situations and Intervals. In Proceedings of the 
25th Annual Meeting of the Association for Computational Lin- 
guistics, Stanford University, Stanford, CA: 16-24. 
Reichenbach, H. 1947 Elements of Symbolic Logic. Macmillan, 
London. 
Ritchie, G. D. 1979 Temporal Clauses in English. Theoretical Linguis- 
tics 6:87-115. 
Sag, I. 1973 On the State of Progress on Progressives and Statives. In 
Bailey, C. and Shuy, R. (eds.) New Ways of Analyzing Variation 
in English. Georgetown University Press, Washington, DC: 83-95. 
Smith, C. S. 1983 States and Dynamics. Language 59:47%502. 
Steedman, M. J. 1977 Verbs, Time and Modality. Cognitive Science 1: 
216-234. 
Steedman, M. J. 1982 Reference to Past Time. In Jarvella, R. and 
Klein, W. (eds.). Speech, Place andAction. John Wiley and Sons, 
New York, NY: 125-157. 
Vendler, Z. 1967 Verbs and Times. Linguistics in Philosophy. Cornell 
University Press, Ithaca, NY: chap. 4, 97-121. 
Verkuyl, H. J. 1972 On the Compositional Nature of the Aspects. D. 
Reidel, Dordrecht. 
Vetter, D. 1973 Someone Solves This Problem Tomorrow. Linguistic 
Inquiry 4:104-108. 
Webber, B. L. 1987 The Interpretation of Tense in Discourse. In 
Proceedings of the 25th Annual Meeting of the Association for 
Computational Linguistics. Stanford University, Palo Alto, CA: 
147-154. 
Wekker, H. 1976 The Expression of Future Time in Contemporary 
British English. North Holland, Amsterdam. 
