Towards a typology of discontinuous past marking

Abstract

The paper discusses verbal markers of the past tense with a meaning roughly charac-terizable as “past and not present” or “past with no present relevance”. This type of past time reference (labelled “discontinuous”) is opposed to standard past markers, which normally do not provide any information about the state of affairs in the present domain. Discontinuous past can be analyzed as a special cross-linguistically valid type of past tense marking. It occurs in a considerable amount of genetically unrelated languages of different areas (especially in Oceania and West Africa), though in cur-rent descriptions it may sometimes hide behind misleading terms or fall under the broad headings of “past” or “anterior”.

Full text

Vladimir A. Plungian (Moscow) & Johan van der Auwera (Antwerp)
Towards a typology of discontinuous past marking *
Abstract
The paper discusses verbal markers of the past tense with a meaning roughly characterizable as
“past and not present” or “past with no present relevance”. This type of past time reference (labelled
“discontinuous”) is opposed to standard past markers, which normally do not provide any information
about the state of affairs in the present domain. Discontinuous past can be analyzed as a special cross-
linguistically valid type of past tense marking. It occurs in a considerable amount of genetically unrelated
languages of different areas (especially in Oceania and West Africa), though in current descriptions it
may sometimes hide behind misleading terms or fall under the broad headings of “past” or “anterior”.
0. Introduction
This paper will focus on a type of verbal markers of the past tense with a meaning
roughly characterizable as “past and not present” or “past with no present relevance”.
These markers are different from “standard” past markers, which normally do not provide
any information (whether negative or positive) about the state of affairs in the present
domain of the time axis. We propose the label “discontinuous” for this non-standard type
of past time reference. A typological investigation shows that the category of discon-
tinuous past appears to be represented in many genetically unrelated languages of differ-
ent areas – though in current descriptions it may sometimes hide behind misleading terms
or fall under the broad headings of the “past” or “anterior”. Discontinuous past can thus
be analyzed as a special cross-linguistically valid type of past tense marking.
The paper is organized in the following way. Section 1 discusses the semantics of the
“standard” past markers, while section 2 draws attention to various verbal systems in
which the use of the tense markers is not consistent with what is required from standard
past markers. These cases call for explanation, and having introduced the notion of discon-
tinuous past for that purpose (section 3), we further concentrate on the place which discon-
tinuous past markers may occupy in a verbal system (section 4), on the other meanings
they may have (section 5), and on their morphological properties, which appear to reflect
iconically some important semantic properties of these markers (section 6). In the remainder
of this section, we briefly touch upon the possible diachronic origins and the evolution of
Sprachtypol. Univ. Forsch. (STUF), Berlin 59 (2006) 4, 317–349
*This paper was started in 2001/2002 in Brussels, when the authors worked on a joint project sup-
ported by the Royal Flemish Academy of Belgium for Science and the Arts. The work on Polynesian
languages goes back to Vladimir Plungian’s fellowship at the University of Kiel in 1999 (granted by
the Alexander von Humboldt Foundation). We are grateful to Sibylle Kriegel (Aix-en-Provence),
Ulrike Mosel (Kiel), Claire Moyse-Faurie (Paris), Pavel Petrukhin,Dmitry Sitchinava and Anna
Urmanchieva (Moscow) for their help.

discontinuous past markers, as well as the relation of discontinuous past to other TAM
values.
1. On defining past time reference
Among what are usually called “past tense markers” in grammars and descriptive
studies, one peculiar type of markers (referred to below as “idle”) deserves special
attention. They are worth discussing, first of all, because they seem to contrast strikingly
with common expectations about “average” past tense markers. Since these expectations
are seldom formulated explicitly, let us begin with some clarifications.
It would be not very surprising to say that past tense markers are expected to denote
past situations, i.e., situations with past time reference. The latter notion is commonly
taken as meaning “occurring before the moment of speech” (see, for example, Comrie
1985 or Michaelis 1998). However, statements like this do not make clear what it is that
has to occur before the moment of speech in order to warrant the status of a grammatical
marker of the past tense: is it necessarily the whole situation or would a portion do as well?
At first sight, if a situation occurs partially in the past and partially in the present (or
future), grammatical markers of past time reference cannot apply. Consider (1) or (2).
(1) I am just opening the tin.
(2) I always get up early in the morning.
In (1), the process of opening takes place at the moment of speech, though it has begun
in the past and a part of it undoubtedly belongs to the past domain of the time axis. In (2),
the situation referred to is a set of single occurrences, some (if not all) of which also belong
to the past domain. What triggers the use of the (simple) present or habitual marker is the
speaker’s belief that other occurrences of the sort will continue into the present and future
domain.
Based on examples like (1) or (2), one might want to improve the analysis and contend
that grammaticalized past time reference applies only to the situations which are entirely
included in the past domain (i.e. they entirely precede the moment of speech), so that no
part of them continues into present or future domains. This generalization, however,
appears somewhat hasty, because it is at odds with at least two types of data. First, exam-
ples like (3) or (4) are very problematic for the “entirely before” hypothesis.
(3) Sarah has painted the wall red.
(4) I have been thinking about it all day.
Both in (3) and (4), English has recourse to the special form of the “present perfect”,
intended to capture the fact that, although most of the situation belongs to the past
domain, some fragment of it is considered to be simultaneous with the moment of speech.
In (3), it is the consequent state (‘be [painted] red’), which is said to hold at present. In (4),
it is the endpoint of the situation which is said to coincide with the present moment – or
even perhaps the medial phase, because (4) does not necessarily mean that the speaker has
stopped thinking at the time of utterance.
Of course, one could object that the verbal forms in (3) and (4) are not past – they are
aspectually perfect (because they refer to the consequent phase), but temporally present
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
318

(because this phase is claimed to include the moment of utterance), and hence they would
not be true counterexamples. We do not want to go into the intricate discussion about
whether the English Present Perfect instantiates a past or a present tense (or none,
or perhaps both; all views have been advocated in the immense literature on the topic).
Notice, however, that the most widespread view seems to be that the Present Perfect is a
kind of past tense (cf., from different positions, Comrie 1985: 78, Maslov 1988, Hornstein
1990, Michaelis 1998, or Dahl &Hedin 2000). But even if we decide to consider special-
ized verbal forms like the English Present Perfect present (or, at least, not past), there
are many languages which lack specialized perfect forms, and where the reference to the
consequent state of a past situation obtaining at present is made with “ordinary” past
forms. In other words, these are languages with no morphological distinction between the
past and the perfect, but in which the past forms may have a perfect meaning (among
others). For these languages, semantic effects like that of (3) are undoubtedly part of their
past tense semantics. The verbal form under consideration is usually a kind of generalized
preterit, which describes the situation as completed before the moment of utterance,
leaving open the question of its consequent phase: it may or may not last up to the
present1.
Thus, the case of the perfect (if we consider it a past tense) or of the generalized preterit
with the resultative meaning prompts us to say that the existence of the result (or of some
more vaguely determined consequent state) at the moment of speech does not preclude
the choice of a past tense marker. This is the first amendment to the “entirely-before”
hypothesis.
The second set of facts – even more important – relates to another type of aspect-related
complications. This time, it is the durative aspect which may pose problems. Compare
examples (5) and (6).
(5) At ten, he fell asleep.
(6) At ten, he was already sleeping.
The verbal form in (5) is aspectually perfective, since it refers to the whole situation
(including its endpoint). The verbal form in (6) is durative: it refers to some medial
fragment of the situation. Both (5) and (6) are past. In the case of (5), there is a clear
understanding that the whole situation is prior to the moment of speech. On the other
hand, (6) is ambiguous in this respect. It reports only that at some moment in the past a
situation (a state) obtained. It does not report anything about the current state of affairs.
Moreover, according to Declerck 1991: 393–394 (where this example is borrowed from),
the most plausible pragmatic interpretation of (6) is that the state still obtains at the
moment of speech.
What then is the difference between the past construal illustrated in (6) and the present
one illustrated in (1)? Generally speaking, in both cases there are temporal fragments
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4
1Generalized preterits are a widespread phenomenon; in Europe, they are attested in Latin, (spoken)
French, some German and Italian dialects, Hungarian, East and West Slavic. Diachronically, all such
forms are late developments of original perfects, which come to supplant aoristic pasts (taking on
their functions, but preserving the primary resultative values as well); hence, the term “expanded
perfects” would be appropriate. For more details on this diachronic evolution see Bybee et al. 1994;
specifically on European languages, see Abraham 1999 and Lindstedt 2000 (Harris 1983 is an older,
but still very informative account of the situation in Romance area).
319

before as well as at the moment of speech. Obviously, the difference does not lie in the
temporal reference of the whole situations.
The problem of the temporal reference of durative situations has already received con-
siderable attention in the literature (see, e.g., Klein 1994 and Gosselin 1996). To account
for the difference between (1) and (6), it is commonly proposed that temporal reference
does not necessarily specify the relation between the moment of speech and the time of the
whole situation. The time reference of a situation is generally made not in a direct way, i.e.
not with regard to the time span occupied by the situation as a whole, but in an indirect
way, i.e. with regard to some portion of the time axis which is related to the situation. The
time span the reference is made to can be called “reference interval” (it corresponds to
Klein’s “topic time” and Gosselin’s “intervalle de référence”). Thus, the situations in
(1) and (6) are both durative, because the “reference interval” is included in the time of the
situation, but (1) is present and (6) is past, because the reference interval includes the
moment of speech in the first case and precedes it in the second. It means that, generally
speaking, the “standard” meaning of the past tense is characterized by the two following
properties:
(7a) The past tense marker, when applied to a verbal lexeme V, specifies the temporal
location of some reference interval RI only, not necessarily of the whole situation
SV: RI may be included in SV, coincide with SV, or include SV.
(7b) In the case of past time reference, the reference interval is located before the
moment of speech; the temporal location of the whole situation is not specified.
The most important consequence of (7a–b) is that the past tense is largely under-
specified as concerns the temporal interpretation of the whole situation. It makes a precise
statement about one fragment of the situation and it leaves unexpressed the temporal
reference of the other fragments. This can be recovered from a wider context, but it is not
part of the grammatical meaning of tenses.
2. A descriptive paradox: “idle” past markers
In the previous section, we have formulated one important property of the standard past
tense markers, namely, that they generally provide a temporal reference for a specific
reference interval and not for the whole situation.
A second important property of the standard past markers is that they are highly
grammaticalized, in the sense that they form part of a full-fledged grammatical category of
tense. It presupposes that in each instance when a verbal form is used, a tense marking
is obligatory. There may be exceptions, but these occur in precise and highly specific
conditions. The latter can be morphosyntactic (if, e.g., the expression of tense is blocked
with non-finite forms) or semantic (e.g., there may be no tense distinction in non-indicative
moods). However, if tense is grammaticalized, we may expect that in the overwhelming
majority of cases when (indicative finite) verbal forms are used, they will be marked
for tense.
Yet there exist languages which contradict the above expectations about standard inflec-
tional markers of tense (more particularly, of the past tense). Although these languages are
reported to have markers of the past tense, the behaviour of these alleged past markers is
very different. In the first place, the markers are distinguished by a conspicuously low
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
320

frequency – as compared to the frequency of all the past situations referred to in the texts.
It means that the bulk of the situations with past time reference in fact are not associated
with any explicit marking of the past tense, and only a minor subset of the situations with
past time reference are covered by these “idle” past markers. These data suggest that we
are possibly dealing with the markers of something different from the standard past tense –
the hypothesis which will indeed be argued for later.
In descriptive studies, however, “idle” past markers more often than not tend to be
assimilated to the standard past markers and are thus included under the general heading of
the past, possibly with a note on their slightly deviant behaviour (due to language-specific
settings). A characteristic example is the quotation from Dixon’s Fijian grammar, where it
is stated that, in Fijian, “the past tense marker [aa] may not be used at all (…), may seldom
or never recur in the narrative” (Dixon 1988: 72). In other words, rare as it may be, it is still
believed to be a past tense marker. The low frequency of such markers sometimes fails to
be made explicit, as, for example, in a grammar of (East) Futunan (where the situation
is not very different from Fijian). The list of verbal grammatical markers includes the
following comment:
La marque du passé na: La marque temporelle na sert à exprimer une action qui
s’est déroulée dans le passé, ou un état passé (Moyse-Faurie 1997: 92).
In a sense, this is true, but this claim must be completed with another claim, stating that,
though “the past marker na” may serve for past time reference, in most cases, it is absent
in Futunan texts relating past events. Let us take one small excerpt from a traditional
Futunan narrative to demonstrate this property:
(8) East Futunan narrative text [after Frimigacci et al. 1995: 180]
Ko le futi na tu’u i Kolotai, kau no’atia
fc sp banana.tree stand l(place) 1sg attach
se vaa ki ai. Ti ’au fafine faka-’ilo mai,
nsp liana lanand come women caus-know dir
koi tu’u ai lona sonaki. Kau Fano loa o nofo i
cont stand an poss:3grape 1sg Go then to stay l
SokisokiNaolau maasina kua lima,
(place) to count month pf five
ti ’o’ono, kua puli le futi la!
and Visit pf disappear sp banana.tree emph
The translation given in Frimigacci et al. 1995 is as follows:
‘Le bananier auquel j’ai attaché une liane est à Kolotai. Des femmes sont venues me
dire qu’il portait un régime. Je suis allé vivre à Sokisokinga, et cinq mois plus tard,
quand on est allé le voir, le bananier avait disparu.’
A somewhat more literal translation could be as follows (verbal markers are given in
parentheses, verbs without overt grammatical marking are italicized):
‘The banana tree which stood previously [NA] at Kolotai, I attached a liana to it.
And women came to inform me that its bunch was still [KOI] at its place. I went
then to stay at Sokisokinga, to count months, they are five [KUA], and we went to
inspect it, and there was no [KUA] banana tree!’
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4 321

What strikes immediately in this text is that, among seven clear instances of past situations,
only one of them is marked with na, while the bulk (four verbs, all belonging to the main
story line) are left unmarked. Other overt markers, such as koi ‘still’ and kua ≈‘already;
unexpectedly’ (usually called “Perfect” in Polynesian grammars; see Hooper 1998 for an
insightful discussion, as well as Moyse-Faurie 2002) convey additional aspectual and/or
modal meanings, so we shall not be concerned with them here.
It can be seen that the only situation marked in (8) with na is characterized by an addi-
tional property (not found with other past situations referred to in the text): it is known as
having obtained at some moment in the past but not obtaining anymore. This is explicitly
stated in the last sentence of the text, but it can also be said to be part of the grammatical
meaning of na, so the hearer is prepared to the dramatic conclusion from the very be-
ginning of (8). We shall revert to this issue in section 3.
The text (8) is typical in many respects – not only for the Polynesian area, but for all
languages with “idle” past tense markers. Let us summarize the relevant properties of such
languages as they appear in grammatical descriptions.
(9a) “Idle” past markers are used only for a minor subset of past situations; they are
practically never used for the situations within the main story line of the narratives.
(9b) The languages which have “idle” past markers more often than not do not dispose
of a coherent system of tense marking – in other words, they tend to lack specific
means for marking present and future. The temporal reference in such languages
is usually established by means of a default interpretation of aspectual and modal
constructions. Thus without contextual indications to the contrary, perfective
markers normally provide the past reading, imperfective markers the present
reading, prospective or intentional markers the future reading, and so on2. An
“idle” past marker thus turns out to be a language’s unique tense morpheme, which
means that a grammatical category of tense (at least, in the standard sense of the
term) can hardly be recognized in such cases, because the idle tense marker is not
opposed to other tense markers.
Thus, the semantics of “idle” past markers is never restricted just to past time reference;
they develop a plethora of other meanings, and even their basically temporal uses are to
be interpreted very carefully (see below). In fact, they are “idle” only with respect to the
standard view of tense; when they are not forced into the accustomed temporal mould,
they appear to have a well-defined semantics of their own.
Now we turn to a more detailed investigation of this semantics. We shall do it in
two steps: first, we consider the basically temporal uses of these markers (referred to as
“discontinuous past”3), and then we shall concentrate on their derived (discourse and
modal) uses.
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
2Futunan is a typical example of such a system; note that the unmarked verbs in Polynesian languages
are most likely to be treated in narrative context as perfective, cf. Hooper 1998.
3As far as we know, the term “discontinuous” has not been used in a typological perspective in this
way, though the notion of “discontinuance” has been applied, for example, to capture the semantics
of Modern English used to constructions (as in Tagliamonte &Lawrence 2000: 331, among others).
The term “discontinuous” in Schwenter 1998, on the other hand, has little to do with what we are
going to describe here, since Schwenter uses it for an extremely rare type of remoteness marker
referring to two distinct – and separated – intervals, namely, remote and hodiernal past.
322

3. Temporal uses:discontinuous past
Generally speaking, discontinuous past uses come in two varieties, depending on the
aspectual properties of the main verb. These varieties can be called “imperfective-based”
and “perfective-based” discontinuous past. With imperfective verbs (states, durative pro-
cesses, habitual situations), the markers of discontinuous past denote situations of limited
duration, which are claimed not to extend up to the moment of speech. This is the type
exemplified by (8). Another Futunan example can be (10):
(10) na koi su’a le li’ua
dp cont flow sp river
‘Il y avait encore de l’eau dans la rivière’ [Moyse-Faurie 1993: 210]
or, more literally
‘at that time, the river still used to flow’,
where the part of the meaning glossed as ‘at that time’ is provided by the dis-
continuous past marker na.
Compare (10) with phrases like koi iai ‘there exists still’ (Moyse-Faurie 1993: 210), where
the marker of the discontinuous past na is absent, and the state is asserted to have present
relevance.
The concept of limited duration in the examples above applies equally to states, pro-
cesses and habits. There is evidence, however, that past habits have a special status. A
language may have no special marker for denying present relevance of past processes and
states, but have one for habitual situations only (the opposite seems not to be attested).
Thus, it seems that discontinuous past marking is found most often for habituals. Note that
as early as in Dahl 1985: 100–102, a special cross-linguistic category of “habitual past” was
isolated as opposed to standard habituals with no specific temporal reference. From the
formal point of view, Dahl’s past habituals are not analyzable as a regular combination of
past and habitual markers and constitute a category of their own. Semantically, Dahl’s
analysis is more cautious: according to him, “the basic semantics of hab<itual> past
appears to be describable as a combination of hab<itual> and past time reference” (Dahl
1985: 101). As we have shown, however, a prototypical past habitual situation obtains only
at some limited period of time in the past – and does not obtain in the period including the
moment of speech.
Habitual pasts are not infrequent in the world’s languages. A nice example of the system
where discontinuous uses are restricted to the habitual contexts is Kisi, a West-Atlantic
language from Guinea (Childs 1995: 231–233). Kisi is claimed to have both Past Progres-
sive and Past Habitual forms; the difference between these is (apart from the aspectual one)
that the Past Habitual “marks an action as continuing for a time in the past but no longer
occurring, or a state as once obtaining but no longer in effect” (Childs 1995: 231), while the
Past Progressive “differs from the Past Habitual in that it says nothing about the present
state of affairs. The Past Habitual conveys that the state no longer obtains or the action is
no longer occurring. The Past Progressive (…) says that an action was once ongoing (and
may still be ongoing)” (Childs 1995: 233). It is clear from this that the Kisi Past Habitual is
a discontinuous past marker (and the only discontinuous marker in the system), while the
Past Progressive conveys a standard past meaning, like that exemplified by (6).
It should be noted that a special marker for the past habitual may coexist with discon-
tinuous past marker(s) having a broader semantics, i.e. applicable both to durative and
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4 323

perfective situations (see below). Such systems are attested al least among the Creole
languages of South America. Thus Guyanese Creole has a habitual past marker yuustu
along with a broader discontinuous past marker bin (Gibson 1992, Sidnell 2002)4.
Another example of a special remote past habitual marker daan (though rarely used in the
modern language) is attested in Wolof, along with a regular discontinuous past marker
[w]oon; see examples (17–19) in section 4 below.
In order to account for the second (“perfective-based”) variety of discontinuous past
uses, it should be borne in mind that the semantic effect of “limited duration” described
above is impossible with perfective verbs, i.e. verbs which refer to dynamic events with
clear temporal boundaries. Thus, a perfective verbal form like they went out asserts that
some momentary change took place in the past. Even perfective verbal forms such as they
wrote or they ran describe situations, which, though not momentary, are still limited in
time. It makes little point to assert their time-limited character with the help of a special
discontinuous past marker, because a similar meaning is already supplied by the perfective
marker itself.
Nevertheless, the combinations of perfective verbs and discontinuous past markers do
exist. The meaning this combination yields can be characterized as the non-existence of a
consequent state at the moment of speech (or its “current irrelevance”).
Obviously, it is not the situation itself which is in the scope of the discontinuous past
marker in these cases (since this situation is completed anyway), but the consequent phase of
it – the only available temporal fragment which can have some duration and be extendable
to the present. Thus, a form like went out alone can be interpreted as ‘went out [and is
probably still absent]’, while went out + discontinuous past marker means something like
‘went out [at some moment], but is no longer absent’. With telic verbs having a clear
resultative component this combination asserts the non-existence of a result at the moment
of speech, hence the labels of “cancelled”, or “reversed” result (Squartini 1999) or “anti-
resultative” (Plungian 2001).
One can point to several contexts which are crucial for distinguishing the value of
cancelled result. The first applies to verbs of (directed) motion such as come or go. The
natural result (in the sense of Nedjalkov &Jaxontov 1988) of these situations is that the
subject is supposed to be located at a place (at or outside the deictic centre) which is
the goal of the previous motion. To deny this result is to say that the subject is no longer
located at that place. Thus, if a language has a special verbal marker which can distinguish
a meaning like (i) ‘you came and are still here’ from a meaning like (ii) ‘you came
[before], but now you are not here any more’, this language can be said to have a gram-
maticalized expression of the “cancelled result”. Examples (11–13) show that Tokelauan,
Seychelles Creole and Korean belong to this type.
(11) Tokelauan [Hovdhaugen et al. 1989]
(11a) Kua pa¯ te vaka ki Niu Hila
pf reach sp:sg boat dir N.Z.
‘The boat has arrived in New Zealand’
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
4The same past habitual marker was borrowed from Guyanese Creole into Dutch-based Berbice
Dutch Creole, yielding a similar system, with an (original) broad marker of the discontinuous past wa
and a more specific marker of the habitual past justu (Kouwenberg 1994: 61–82).
324

(11b) Na kua pa¯ te vaka ki Niu Hila
dp pf reach sp:sg boat dir N.Z.
‘The boat has been to New Zealand’ [i.e., arrived there and came back]
(12) Seychelles Creole [Michaelis 1993: 81–82 et passim]
(12a) mon’ne vine
I pf come
‘I have come’ (sc. ‘I am now here’)
(12b) mon ti vine
Idp come
‘I came/had come’ (sc. ‘I came, and then went back’)
(13) Korean [Chang 1996: 122–123]
(13a) Yong-i w-ass-eyo
Yong-nom come-past-fp
‘Yong came/has come’
(13b) Yong-i w-ass-
´
ss-eyo
Yong-nom come-past-past-fp
‘Yong had been here (but left)’
Note that in Korean the discontinuous past meaning is provided by the use of a double
past marker (-ass- and -
´
ss- in [13] are allomorphs), and an interesting difference between
Tokelauan and Seychelles Creole is that in the former the discontinuous past marker na is
added to the perfect one, while in the latter the discontinuous past marker ti replaces that
for the perfect, [i]ne or fine (though in principle they are not incompatible, cf. Michaelis
1993: 92–95). We shall elaborate on this below in section 6.
A typical context for this use of the discontinuous past is described for sentence Q.61 in
Dahl 1985’s questionnaire:
(14) = Dahl’s Q.61
{It is cold in the room. The window is closed. Question:} You <open> the window
(and closed it again)?
According to Dahl, “quite a few” languages can use pluperfect forms here (with the
notable exception of English), and several languages from his sample have specialized
markers for this meaning (among them, Oneida and Akan, which is not at all surprising,
since both West Africa and North America are areas where the marking of the discon-
tinuous past is widespread, see section 4.1 below).
Hereafter, we shall use the labels “framepast” and “cancelled result” for these aspectu-
ally distinguished varieties of the discontinuous past, imperfective-based and perfective-
based, respectively5. The term “discontinuous past” will remain a cover term for both
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4
5The term “framepast” (or “past temporal frame”) was originally introduced in Dahl 1985: 146–149
to account mainly for what we have called “cancelled result”, but some examples Dahl suggests for
this meaning seem to have more to do with imperfective-based uses. The terminological distinction
we adopt in this paper is closer to that proposed in Squartini 1999, according to whom “framepast”
and “reversed result” are two different – though frequently coexistent – derived values of the plu-
perfect (he is not concerned with their expression outside of the pluperfect domain).
325

values. This type of temporal value got relatively little attention in the typological litera-
ture, but it still did not pass totally unrecognized. As far as we know, Seiler 1971 was the
first to raise this issue, though not directly in the domain of tense, but rather in the domain
of mood. His main focus was on the diachronic connections between optatives and preterits
in a number of old Indo-European languages; Seiler claims that it is the strengthening of
the component of temporal discontinuity (Seiler calls it “dissociative”), which favoured
the diachronic passage from the past through a presumable “dissociative past” to the
optative in the history of Greek and elsewhere. In this way, the existence of a special past
form with a discontinuous meaning (though not attested directly) was first hinted at. In
further work, genuine verbal markers with discontinuous meaning of the type ‘formerly,
but not now’ were in fact isolated. They were reported for a number of native North
American languages – mainly, in Uto-Aztecan (Steele 1975) and Algonquian (James 1991,
who acknowledges early findings of Bloomfield in this domain); interestingly, in both
papers, the main discussion again concerns the diachronic developments of such past
markers towards the markers of irreal modality. (We shall discuss the relation between
temporal and irreal uses of the discontinuous past markers in section 5.2. at length.)
Though the verbal markers of the discontinuous past undoubtedly deserve a wider
cross-linguistic study, the data we dispose of so far allow us to propose some tentative
generalizations about the place of discontinuous past markers in verbal systems, about the
inventory of their other uses and about their morphological properties. Below they will be
considered in turn.
4. The discontinuous past in verbal systems
Generally, languages may have the following type of grammatical expression of the
discontinuous past:
(i) The discontinuous past marker may be the only marker of tense within a basically
non-tensed verbal system.
(ii) The discontinuous past marker may be one among several tense markers in a
system, which thus provides a fine-grained grammatical distinction between the
standard past and the discontinuous past.
(iii) A system may have no specialized (or “dedicated”) discontinuous past markers, but
the meaning of the discontinuous past can be part of the meaning of another verbal
marker. In this case one can also speak about a “discontinuous use” of some marker
(with other basic meaning[s] or use[s]).
4.1. “Atemporal” systems with the discontinuous past
The systems of the first type can be called “atemporal systems”, because they do not
mark standard tense at all. The only temporal value available in such systems has a very
specific tinge, which goes beyond the simple location of some reference interval with
regard to the moment of speech.
Obviously, our systems with the “idle” past markers belong to this type. These systems
are not infrequent; an especially high concentration of them is found in Oceania, as well as
in East and West Africa; they are also found among North American and Amazonian
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
326

languages – all these are areas where the grammatical marking of tense is not common
(rather, we deal here with what can be called “aspect-prominent” or “mood-prominent”
systems, adopting Bhat 1999’s parlance). It should be noted further that “atemporal”
systems with discontinuous past marking are typical for many Creole languages (which
may not be surprising given the African, Oceanic or South American substrate of most
Creoles), though this fact is not recognized in any clear way in most descriptions of
Creoles.
The most widespread terminology in Creole studies goes back to Bickerton’s 1975 and
1981 proposals, which establish a category of “anterior” tense defined as “very roughly,
past-before-past for action verbs and past for stative verbs” (Bickerton 1981: 58). Creole
languages are said to have only this type of tense markers: they usually take the form bin or
ben in English-based Creoles, ti or te in French-based Creoles, and something like ba or va
in Iberian-based Creoles (in Berbice Dutch Creole, it is wa, as already mentioned).
Bickerton is basically correct with the claim that “anteriors” have little to do with the
standard past tense, but his interpretation of their semantics is somewhat problematic. The
literature on creole TAM systems is rather voluminous, so what we can propose below is
only a very succinct summary of recent findings. The main point of recent criticism against
Bickerton’s approach (see, e.g. Givón 1982, Dahl 1993 and especially Winford 1996
and 2000a) is that it imposes a different treatment to stative (= imperfective) and action
(= perfective) situations, interpreting “anteriors” as markers of relative tense in the first
case, and as markers of absolute tense in the second case. In fact, the semantic difference
between the two cases is not that large, and there is no hard and fast boundary between
alleged “absolute” and “relative” uses of anteriors. Moreover, the core meaning of
“anteriors” seems to differ both from pastness in the case of imperfectives and from
temporal anteriority in the case of perfectives. In both cases, the main semantic contribu-
tion of “anterior” markers is an emphasis on the difference between past and present state
of affairs; like in other instances of discontinuous past uses, it can be paraphrased as
‘before – and not now’.
As concerns the temporal dimension of Creole systems, it should be born in mind that
Creole verbs are primarily marked for mood and aspect, and aspectwise usually a marked
imperfective and/or habitual is opposed to an unmarked perfective or “bare verb” – with
stative verbs, however, zero marking may have an imperfective interpretation6. Perfective
verbs have a default past interpretation, while imperfective verbs have a default present
(or concomitant) interpretation, so these systems normally do not need to provide a
temporal reference of the situation – neither absolute nor relative. Against this back-
ground, the notion of discontinuous past can give a better account of the basic uses of
“anteriors”, and detailed and theoretically non-biased analyses of Creole data (as in
Winford 2000a) seem to corroborate this hypothesis. The similarity of Creole “anteriors”
to “framepasts” has been already stated in Dahl 1985 and 1993, though mainly for statives.
There exists, however, an alternative view (mainly, among the Creolists), which attempts to
correct Bickerton’s analysis in another way, emphasizing the component of anteriority. In
that way, anteriors are treated as a (nearly complete) analogue of the “relative pasts” in
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4
6This pattern (identical and mostly zero marking for both perfective actions and imperfective statives)
is largely attested in West African languages; markers of this type have been called “factatives” by
Welmers (1973: 346–348).
327

European languages; thus, all the examples of Haitian Creole anteriors in Lefebvre 1996
(given mostly in isolation, out of context) are translated with the English pluperfect.
Yet temporal precedence is more a pragmatic side-effect of many uses of Creole
anteriors than a basic meaning. As shown in Winford (2000a), there are many cases where
temporal precedence is not marked by anteriors, as well as cases where anteriors are used
in manifest absence of temporal precedence. Consider example (15) (Winford 2000a:
406), which is taken from a dialogue in Sranan about a person who happened to have fallen
out of touch with the dialogue’s participants. Discussing the person’s disappearance, one of
the participants utters (15):
(15) Traesde mi miti en even.
other.yesterday I meet him briefly.
Aben taigi mi a o kon na fesisey baka.
he dp tell me he fut come lfront.side back
‘Day before yesterday, I met him briefly. He (then) told me he would come to the
front again’
Note that miti ‘meet’ here has a zero marking providing the past perfective interpretation.
The next verb, taigi ‘tell’, could have zero marking as well, especially since it introduces an
action which is temporally posterior to the fact of meeting. So, no anteriority is implied.
Winford accounts for this use by saying that “the speaker wishes to background this
information”; perhaps, it would be more precise to say that the speaker wishes to insist
on the fact that intending to come to the front again was not realized in the future (since
the person in question disappeared). This is the proper meaning of the discontinuous past,
since the exact semantic contribution of ben in (15) is to cancel the actual relevance of the
person’s own intentions and promises given in the past. In fact, Winford himself seems to
tend towards the same interpretation, when he eventually states that “ … [T]he reference
point for an event marked by ben is not necessarily another past event in the same sentence
or preceding discourse. Use of ben has more to do with the speaker’s wish to distance that
event from some other situation which s/he wishes to foreground” (Winford 2000a: 406).
If we assume that “distancing the event” is the same as stating its lack of relevance at some
later moment, the difference between our analysis and that of Winford will be termino-
logical only.
Winford’s description is in many ways exceptional, since the author is very careful and
unbiased in applying grammatical terminology to vast and thoroughly prepared material.
Often, however, traditional labels may considerably blur the real picture. Thus Davies
(2000) describes the forms with the suffix -ba- in Palenquero (a Spanish-based Creole of
Columbia) as ultimately “past imperfective”. Judging from the examples provided, it is
more plausible that what we are here dealing with are framepast uses of a discontinuous
past marker, though it requires further study. However, one should agree with the main
point of the author, who argues that -ba- cannot be analyzed as a relative tense marker –
neither as the perfect nor as the pluperfect. For all that, “past imperfect” is perhaps not the
best alternative, since it leads us too far in the domain of the standard tense.
Polynesian languages provide other very interesting examples of basically atemporal
verbal systems with discontinuous past marking. In Polynesian, grammaticalized TAM
values are typically rendered by preverbal particles (often mutually exclusive). The total
number of these particles is usually around ten, and their basic meaning is either purely
modal (as with markers of wishes, requests, apprehensions or intentions) or shows a
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
328

peculiar mixture of modal and aspectual components. Typically, the situations can be
marked for imperfectivity [e7], (unexpected or highlighted) continuation [koi] and (un-
expected or highlighted) beginning or change [kua]. Perfective situations, especially the
ones constituting the main story line of a narrative, tend to be left unmarked. The temporal
reference of the situation is not provided for by any specific means: perfective situations
are interpreted as past out of specific context, while the default interpretation for imper-
fective and continuous markers is present (when, however, an explicit past or future
reference point is given by the context, they are construed as denoting simultaneity rather
than present).
Against this background, mentions of “past” markers in most descriptive grammars
appeal to some caution. Indeed, their behaviour is more typical for what we have described
as “idle” past markers above. According to our data, the most clear cases of what cor-
responds to prototypical instances of discontinuous past marking are found in Futunan and
Tokelauan (the markers are in both cases na); Tahitian and Maori cases are less clear, but
much of what is said about the past marker iin these languages makes us think that we may
deal here with discontinuous past markers too (in Tahitian, the discontinuous past value is
most likely to be expressed by the combination of preverbal iand postverbal na ‘there’,
judging from the examples in Peltzer 1996 and Lazard &Peltzer 2000)8. In what
follows, we shall focus primarily on Tokelauan.
In an insightful description of Tokelauan verbal markers, Robin Hooper (1998; cf. also
an earlier treatment in Hooper 1989) seems to privilege the “relative tense” analysis of
preverbal na. According to Hooper, the main function of na is that of marking anteriority,
and it is “equivalent to an English simple past tense if R[eference time] = S[peech time],
but equivalent to a past perfect if R precedes S” (Hooper 1998: 124–125). As we have seen,
exactly the same treatment has been proposed for Creole “anteriors”, and our objections
will be also the same. Constructions with na are not equivalent to English simple pasts in
their “absolute” uses, witness examples like (11). Rather than marking past time reference,
they introduce a “break” between the situation in the past and the moment of speech. The
same effect dominates the uses which are interpreted as “relative”, so that, exactly like for
Creole discontinuous past markers, there is no need to oppose “absolute” and “relative”
uses as two distinct values (if it is not to reflect the difference in English translations).
Consider (16):
(16) = Hooper 1998’s (16a)
Ko au ka toe talanoa atu, toe fakamatala
la
fc 1sg int again talk dir again explain
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4
7In order to avoid secondary detail, we give only the most widespread form, abstracting away
from phonetic or morphological peculiarities the cognates with the value in question may have in
individual Polynesian languages. Of course, each individual system may have some specific subtle
variations which cannot be taken into account in this very sketchy presentation.
8The situation in Samoan is somewhat different, because, unlike most Polynesian languages, it has
two preverbal particles with the basic meaning of the past tense: na and sa¯. The difference between
them is described in Mosel &Hovdhaugen 1992: 339–344 as mainly aspectual, with additional modal
components, na being the marker of “temporarily limited” or “uncommon” situations, unlike sa¯,
“aspectually neutral” and referring mostly to “expected events”. Some examples of the use of these
two particles remind one of contexts typical for cancelled result (na) or framepast (sa¯) values.
329

atu e au te taimi lava te¯ia na talanoa
dir erg 1sg sp time emph that dp talk
atu ai au muamua i toku
dir an 1sg first lposs: 1sg
hola ma te vaka.
run.away lspboat
‘I am going to talk to you again, I will explain further that time I spoke about before,
about my running away from the ship’.
Hooper (1998: 135) comments on this passage as follows:
The (…) passage comes from the abstract section of the second part of a two-
part narrative told on two separate occasions. The phrase na talanoa refers to the
occasion of the telling of the first part – a time subsequent to the event of the narra-
tive, but anterior to speaker-now.
In our view, what is crucial here for the use of na is not the anteriority as such, but a specific
semantic relation between the situation referred to with na (“the first account of the
events”) and the state of affairs at the moment of speech. In fact, the speaker wants to
revert to the same story and to tell it ‘again’ (toe, repeated twice) – most probably, because
the first version has been considered in some respects incomplete or imperfect. Thus the
fact of having already told the story does not preserve its relevance by the moment of
speech: the story was told once before, but it is as if it was not. This is the exact meaning
conveyed by na: the lack of current relevance, or the notional “distance” between the
event referred to and the moment of speech, as scholars like Winford would have it.
It should be added that the peculiarity of Tokelauan consists in that it opposes the
marker of the discontinuous past na to other preverbal particles not only semantically, but
also to some extent formally: na can be combined with at least two of them, kua (as in
example [11] above), and imperfective e. The last combination is usually written in one
word (nae) and regarded as a separate TAM marker (“past” or “anterior imperfective”),
but nothing seems to prevent a compositional treatment (in fact, Hooper 1989 takes up this
possibility). Against this background, it would be most appropriate to consider plain
combinations na + V as the combinations of na and the zero perfective marker – recall that
the same analysis seems to be preferable for Creoles (cf. Dahl 1993 and Kouwenberg
1994: 63ff., among others). Other Polynesian languages seem to be more reluctant con-
cerning such combinations, though sporadic uses of “past” markers attached to various
aspectual particles can be attested9. That typical discontinuous past markers are more
likely to combine with other TAM markers when applied to one and the same stem is not
an accident; this problem is discussed in section 6 at length.
The third case in point is the West African linguistic area, which includes genetically
different languages that show important typological similarities in the organization of
the verbal systems. At least three large genetic groups can provide relevant examples:
Kwa, Mande and West Atlantic, though the phenomenon of “idle” past markers is by no
means restricted to them. What all these languages have in common is the predominantly
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
9As, for example, in Futunan, where at least the combination na kua is possible (Claire Moyse-
Faurie, p. c.); East Uvean (of the same area) also allows several combinations of this kind (see
Moyse-Faurie 2002).
330

aspectual character of core oppositions grammaticalized in their verbal systems. Usually,
perfective and imperfective verbal forms are opposed, with habitual and prospective
values frequently appended to those. The aspectual forms provide a default temporal
reference according to the pattern described above. The systems found in Kwa languages
are especially close to Polynesian and Creole ones in that they often have zero-marked
perfectives (as it is the case of Fongbe, cf. Lefebvre 1996).
Many languages of this area, however, are claimed to have one additional verbal marker,
somehow related to past time reference. As we have already stated, some grammars
assimilate it to the value “past” of the category of tense as it is represented in European
languages, while other authors try to find a more cautious definition, given the “idle”
status of such markers.
Of course, it is not always easy to make a judgement on the exact nature of a verbal
marker relying on what it is said about it in the description. In some cases, it only can be
seen that the authors wish to emphasize a non-standard character of past marking in the
languages they describe, but the exact nature of the marker remains unclear. For example,
it is hard to decide whether Frajzyngier &Shay 2002’s “referential past tense” in Hdi
(a Chadic language of Nigeria and Cameroon) points to a kind of discontinuous past
marker or to another non-standard past meaning. The marker of “referential past tense”
sí is said to denote “events that happened at some well-determined period of time” (op.
cit.: 335) and to be used in conditional protasis clauses as well; this is the only specialized
tense marker of Hdi.
Let us consider Wolof, an important West Atlantic language, one of the major languages
of Senegal and neighbouring areas (our analysis is based mainly on Stewart &Gage 1970,
Church 1981: 195–215 and Robert 1991, as well as on Plungian’s fieldnotes).
Wolof has a rich and complicated verbal system with two main sets of grammatical
oppositions (marked by pre- and postverbal particles, often in cumulation with subject
person and number markers) – those of aspect and focus. Roughly, perfective, punctual,
progressive and habitual aspects are distinguished, as well as predicate, subject and non-
subject focus. Temporal reference, again, is provided by aspectual markers: for example,
by default interpretation, all punctual and dynamic perfective verbs are past, stative per-
fective verbs are present (as in Kwa languages and in most Creole systems), progressive
verbs are present, and so on.
However complicated and self-sufficient this system might be, there is one additional
marker. This is the so-called “past marker”, which can occur either as a verbal suffix
-[w]oon or as an autonomous word woon, depending on the morphosyntactic properties of
the verbal form it modifies. It has a variety of uses, both temporal and non-temporal. Non-
temporal uses of Wolof woon will be accounted for in section 5. Concerning its temporal
uses, they are all related to the prototypical semantics of the discontinuous past: we find
the framepast meaning (including past habitual) with the imperfective forms, and the
meaning of cancelled result with the perfective forms, as in (17)–(18):
(17a) di naa-y jàng
ipf 1sg-hab study
‘I study’
(17b) d-oon naa-y jàng
ipf-dp 1sg-hab study
‘I used to study (at that time)’
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4 331

(18a) dem na
go pf:3sg
‘s/he has gone’
(18b) dem-oon na
go-dp pf:3sg
‘s/he went/had gone (but is back again)’
Church (1981) insists that woon is used in Wolof not for any arbitrary past situation, but
only for those which relate to a limited time in the past (for obvious pragmatic reasons, it is
often a remote past), which do not obtain anymore at present or whose result is cancelled.
Robert (1991) speaks in this connection about a general meaning of “break” (French
rupture) between the situation and the moment of speech proper to woon.
A characteristic property of Wolof is the existence of another discontinuous past
marker daan with a past habitual meaning, though of limited use in the modern language.
According to some descriptions (e.g., Stewart &Gage 1970), this marker refers to events
which are more remote in the past as compared to those referred to by a regular combina-
tion of the imperfective and discontinuous past d-oon represented in (17 b); cf. (19):
(19) Wolof Remote Past Habitual [Stewart &Gage 1970: 375]
daan naa rawante fas
past:hab 1sg compete horse
‘I used to do horse racing (way back then)’
Other West African languages, where the discontinuous past markers are established with
some certainty, are Diola, Fula (both West Atlantic; cf. Sapir 1965 and Koval’ &Gnali-
bouly 1997 respectively), and Bamana (Mande). Thus, a recent detailed description of the
Bamana verbal system (Idiatov 2000) attests for the particle tun all the relevant properties
of a discontinuous past marker, built into an aspect-prominent system with the principal
opposition of imperfective with predominantly habitual uses (marked by an auxiliary b
E
)
and perfective (marked by an auxiliary ye or verbal suffix -ra – one of the very few inflec-
tional affixes in Bamana)10. See a diagnostic context for the value of “cancelled result”
in (20):
(20a) Ataga-ra dugu kXnX
3sg go-pf village l
‘S/he went / has gone to the village’
(20b) Atun taga-ra dugu kXnX
3sg dp go-pf village l
‘S/he went to the village (and came back)’
The phenomenon of basically “atemporal” systems, in which the only marker of temporal
reference is a discontinuous past marker, is by no means confined to the three major
linguistic areas above, and we can expect that their number will increase as soon as
“discontinuous” properties of the tense markers will be checked in a more systematic way.
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
10 Interestingly, Bamana tun is mentioned in Holm 1988: 151, along with Yoruba ti, as an example of
the “anterior” marker which functions “quite similarly” to those of Creoles. The parallel is quite
legitimate indeed, our only reservation being that the discontinuous past interpretation seems to us
more adequate for both markers.
332

One such case may be Lezgian (Lezgic group of the Nakh-Daghestanian family, cf.
Haspelmath 1993: 140–146 and 1994: 270–274). In Lezgian, verbs are grammatically
primarily marked for aspect and mood (as -na Perfective, -z[a]wa Durative, -da Potential,
etc.). This is not very typical for Daghestanian languages in general, which are more likely
to have full-fledged tense systems. The default temporal interpretation of Lezgian forms
is past for the Perfective, present for the Durative, present (habitual) or future for the
Potential. However, a “past” suffix -jis also available, which can apply to all the verbal
forms above. The morpheme combination -na-j‘Perfective [Past] + Discontinuous Past’ is
of special interest at the moment, since it can yield both framepast and cancelled result
values, cf. examples (21–22), based on Haspelmath 1993: 143.
(21a) cˇar-ar kxe-na
letter-pl write-pf
‘S/he wrote / has written (the) letters’
(21b) cˇar-ar kxe-na-j
letter-pl write-pf-dp
‘S/he wrote letters (but now they don’t come anymore)’
(22a) fe-na
go-pf
‘S/he went / has gone’
(22b) fe-na-j
go-pf-dp
‘S/he went (and has returned)’
Lezgian past forms on -jhave other uses as well, which will be considered later on.
4.2. “Temporal” systems with the discontinuous past
Discontinuous past markers are more likely to occur in atemporal systems of the type
described above in 4.1 (or, perhaps, in such systems the discontinuous past marking is more
prominent and more easy to detect). Nevertheless, we can also point to systems where
specialized markers of the discontinuous past coexist with other tense markers. This is
the case of Algonquian and Washo in North America. Many Algonquian languages are
reported to have a so-called “emphasized preterit”, with markers going back to Proto-
Algonquian *-(e)pan (James 1991); its functions are typical for the discontinuous past. The
same functions are characteristic for Washo “defunctive” on -u
N
il (Mithun 1999b: 156). In
these systems regular past tenses are also well attested – often more than one, as in Washo
with its fine-grained system of four past tenses differing in the degree of remoteness.
The discontinuous past markers of these languages can be considered as part of the
tense system. Nevertheless, they often display features setting them apart from other
verbal markers. In particular, they tend to encroach on the nominal domain and, when
applied to nouns, yield meanings like ‘former’, ‘destroyed’, ‘defunct’ (in the case of human
beings) or ‘formerly possessed’ (in combination with possessive markers). This “nominal
expansion” is a rather frequent phenomenon, attested not only in Algonquian (cf. James
1991), but also in Yupik, Wakashan, Yuman, Washo and other native North American (cf.
Mithun 1999b: 154–157), as well as South American languages (see, e.g., Derbyshire
1999: 40–42 on Carib systems of “present/former possession” marking; more data on
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4 333

“nominal tenses” can be found in Sadler &Nordlinger 2001). If a language has some
past tense marker applied to nouns, the meaning of this marker (of the “former” type) is
more likely to relate to the discontinuous than to the standard past tense, because the
semantics of “former X” presupposes that the property of being X is not applicable to the
object in question at present.
4.3. Discontinuous past as a specific value of a standard past marker
Finally, let us consider cases where a language does not dispose of a specialized discon-
tinuous past marker, but the value of the discontinuous past is part of the semantic inven-
tory of some of its past tense markers. Probably, the most obvious candidates for this type
of polysemy are remote past markers. As is well known, in systems with an opposition of
remoteness (or temporal distance) within the category of tense, it is not only the relation of
the situation to the moment of speech, but also a more precise location of the situation on
the time axis which is expressed (cf. Dahl 1985: 120–128 and Comrie 1985: 83–101 for more
detail; cf. also Bhat 1999: 31–35). For our purposes, two values of temporal distance are of
particular relevance: that of remote past tense (≈‘long ago’) and that of non-hodiernal past
tense (≈‘earlier than today’; in very rare cases, a language may distinguish more than one
type of non-hodiernal tense). These values assert that the situation occupies a limited time
span before the moment of speech, and this semantic component brings them very close to
the domain of the discontinuous past. Of particular interest here are systems which have
been described as having only two markers of temporal distance, one labelled “remote
past” and other “recent past”. This is, for example, the case in Mishmi (Tibeto-Burman)
and Tulu (Dravidian), referred to by Bhat, as well as in some Occitan dialects (referred
to by Dahl) and Quechua (referred to by Comrie). One can be in doubt as to what
such systems really mark: the temporal distance or the discontinuity, since the situation
considered as “remote” is, above all, a situation which does not obtain at that part of the
time axis which includes the moment of speech. Some suggestions in the descriptions may
point in this direction. Thus, it has been stated frequently that the notion of temporal
distance has more of a “subjective” than an “objective” dimension, and especially in binary
systems with very unstable criteria of choice between remote and recent interpretation (cf.
especially Dahl 1985: 123–125). At least in some cases, the reinterpretation of remoteness
opposition as an opposition of neutral/discontinuous past is plausible.
The second powerful source of the discontinuous past values are forms of the pluperfect.
The standard meaning of the pluperfect is precedence in the past: the situation is claimed
to take place in the past and before some other reference point in the past. This meaning as
such has no immediate connection to the idea of discontinuity, but in discourse pluperfect
forms are often interpreted as introducing a kind of temporal break between the situation
and the main story line, relegating the situation to the status of (remote) background.
The additional reference point between the situation and the moment of speech can be
interpreted as a kind of notional barrier, acting in the direction of canceling the subsequent
relevance of the situation. These pragmatic connections can explain the frequently observed
drift of the pluperfect forms from purely relative contexts towards various so-called “abso-
lute” uses, where no additional reference point is implied, but rather what we would
interpret as the semantic effect of discontinuity. In a schematic form, this development can
be represented as ‘before X’ →‘before now’ →‘before now and not now’, where ‘X’ is a
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
334

contextually given reference point, replaced by the absolute reference to the moment of
speech at some stage of the semantic development of a pluperfect construction.
Accordingly, we expect to find among derived pluperfect uses attested in the world’s
languages those which correspond to our framepast and cancelled result values. This is
indeed the case; moreover, these values are among the most frequent derived uses of the
pluperfect forms, which in many languages even exceed its basic uses in frequency and
discourse importance.
The discontinuous uses of the pluperfect forms were demonstrated in a typological
perspective already in Dahl 1985: 144–149. The examples there represent mainly the value
of “cancelled result”, as exemplified by (14) above. More examples of cancelled result and
framepast uses of the pluperfects are provided in Squartini 1999. This polysemy has been
attested in many languages – perhaps, in the majority of languages with the pluperfect
forms. Dahl makes mention of Italian, French, Spanish, Modern Greek, Persian, Hindi,
Finnish and several other languages. At least Latin (especially late Latin after the 2nd
century AD; see Mellet 1994 and especially Sitchinava 2003 for numerous examples and
discussion), German (see Hauser-Suida &Hoppe-Beugel 1972: 160–182 and Eroms 1983,
among others; for Swiss German see Squartini 1999), Serbo-Croatian (see Thomas 2000
for a more detailed account) and Santali (see Sitchinava 2001) can be added to the list,
though this polysemy is by no means universal: thus, in English this type of use of the
pluperfect is at best “not very natural” (Dahl 1985). This is not to say that English has no
derived uses of the pluperfect at all, but we shall not be concerned with them now.
Absolute uses of the pluperfect (unlike relative uses) do not require the past context,
since they presuppose the unique reference point coinciding with the present moment.
Sentence (23) can be a simple example of an absolute use of the pluperfect in Latin; it
exemplifies a framepast value:
(23) Non sum ego qui fu-era-m [Properce]
not am I who be:pf-pqp-1sg:act
‘I am not like what I used to be (formerly)’
As a simple illustration from French, a frequently used expression (24) can be given. It can
be ascribed the value of cancelled result: here, the result of ‘being warned’ is supposed to
be cancelled at the moment of utterance. (Further examples and references can be found
in Engel 1994, Squartini 1998 and 1999, among others.)
(24) On te l’avait dit! ≈‘didn’t I warn you?’
lit. ‘they had said it to you’
The pluperfect is not the only verbal form which may develop the discontinuous meaning.
Thus, in Russian and other East Slavic languages (and, to a lesser extent, in Polish) this
meaning is regularly conveyed by past imperfective verbal forms (in Slavic aspectology,
this use of the imperfective aspect falls under the traditional label of “general factual
meaning”, in the sense that such forms appeal to the simple fact of a past occurrence of a
situation and not to its result). For example, the Russian correlates of (14) would, as noted
in Dahl 1985: 148–149, make use of the imperfective aspect. Unlike the perfective past in
(25b), the imperfective past in (25a) implies that the window is closed at the moment of
speech. (In Dahl &Hedin 2000, it is even discussed whether Russian imperfective forms
can be ascribed a meaning of “current irrelevance” construed as an antonym to the main
meaning of the present perfect.)
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4 335

(25a) Kto otkryval okno?
who open:ipf:past:3sg:m window
‘Who opened the window?’
(25b) Kto otkryl okno?
who open:pf:past window
‘Who has opened the window?’
Interestingly, the closest Russian equivalents to the examples (15) from Sranan and (16)
from Tokelauan would also have imperfective forms as at least the most natural (if not the
only possible) choice, cf. (26a) and (26b):
(26a) Pozavcˇera ja vstretil ego.Togda on govoril,cˇto pridët.
‘Day before yesterday, I metPF him. Then he told IPF that he would come’
Given that the promise was not kept, perfective skazal ‘told’ would make (26a) less
natural and would in no way imply that the person did not come afterwards. A similar
effect is observed in (26b): the imperfective suits better, especially if one wants to empha-
size that the first attempt was in some respect insufficient.
(26b) Ja x ocˇu esˇcˇë raz rasskazat’ tebe tu istoriju, cˇto uzˇe rasskazyval/?rasskazal v prosˇlyj
raz.
I want once more tell you that story, which already told IPF/?told PF in last time
‘I want to tell you again the story I already told you the last time’.
Finally, it should be noted that under some special pragmatic conditions “standard”
absolute past tense forms may take the discontinuous interpretation. Usually such cases
are isolated examples with a high degree of lexicalization, as (9) cited in the beginning
of section 3. Another example can be Latin (27), where the perfective past is used (cf.
Pinkster 1990: 229–232; the pluperfect would be also possible in such a context):
(27) fuit Ilium [Virgil]
be:pf:past:3sg:act Troy
‘Troy does not exist any more’ (lit. ‘Troy was’)
In the remainder of this paper, we shall concentrate on two issues: the derived values of
the discontinuous markers and the formal peculiarities of marking temporal discontinuity.
The focus will be on specialized discontinuous markers; derived uses of otherwise “regu-
lar” tense-aspect markers (such as canonical pluperfects or past imperfectives) will also be
considered, though less systematically.
5. Derived uses
5.1. Temporal and discourse uses
Discontinuous past forms may have other uses, both temporal and non-temporal (or,
perhaps, less temporal). Two of them will be briefly considered in this section: namely,
remoteness and discourse backgrounding.
The relation of discontinuity to remoteness has already been discussed in the previous
section. We have suggested that some markers interpreted as expressing remoteness may
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
336

be closer to discontinuous past markers than to those of temporal distance. It should
be added that among the genuine discontinuous past markers some have the remoteness
meaning as well, as daan in Wolof (19). Remoteness is also a frequent derived meaning of
the pluperfects (though sometimes it is not distinguished well from the discontinuous
past uses in the descriptions); this development seems to be particularly typical for the
Indo-Aryan domain (cf. Dahl 1985: 147). Interestingly, the rise of the remoteness value
(or the complete shift from pluperfect to remoteness) is attested in many substandard
varieties of European languages surrounded by languages with the discontinuous past
marking: it has occurred in Cameroonian Pidgin English (Anderson 1979: 102) or in South
American Spanish (Fleischman 1989); there may be many other examples, though attested
with a lesser certainty.
In discourse, the markers of the discontinuous past are likely to play a backgrounding
role (cf. especially Hooper 1998): since they are rarely used within the main story line of
the narratives, this connection is natural. The discontinuous past markers are called for
when the speaker wishes to revert to some previous fragments of his/her experience and
are overt signals of what can be called “retrospective switch”. This narrative procedure
is favoured by the property of the discontinuous past markers which has been called
“notional barrier” or “notional distance” above: since the discontinuous past markers
introduce a notional break between the situation they refer to and a subsequent moment,
they are readily used when the speaker needs to overtly mark a detachment of some –
more ancient or secondary – piece of information from the main story line. Accordingly,
we typically find the discontinuous past markers in digressions, in relative clauses, as well
as in the initial fragments of the narratives which provide a “setting” for the main story.
The two first types of discourse uses correspond to what Givón (1982) has called “out-of-
sequence”. The third type is different, because it does not signal digression from the
normal course of the narrative: it is used in pre-narrative fragments of texts. So rather
than being out-of-sequence it is to be regarded as “before-sequence” (cf. Sitchinava 2001
and 2003, where this type of contexts was first pointed out). Both functions – “out-
of-sequence” and “before-sequence” – are also characteristic of discourse uses of the
pluperfects in many languages. Thus, a nice illustration for Italian is Maiden &Robustelli
(2000: 293–294); Sitchinava (2001 and 2003) discusses examples from Santali and Latin.
Note that English has out-of-sequence uses, but is more reluctant for using the pluperfect
in before-sequence contexts such as introductory fragments of narratives (cf. Salkie 1989).
This shows that a language may treat these two functions differently. In fact, such
examples, however infrequent, are attested. For instance, the description of Aghem (a
Grassfields Bantoid language of Benue-Congo family) mentions a pair of markers called
“narrative” and “consecutive” (Anderson 1979: 89ff.), which behave in the following way.
The “narrative” marker is used “in the first sentence of stories which refer to something
which happened at some distant time”, thus setting the time frame, while all the sub-
sequent events are recounted with the help of “consecutive” markers. Since no other
functions of the “narrative” marker are reported, it can be characterized as a narrow
“before-sequence” marker. In a sense, this is a discourse antonym of the English plu-
perfect, if we assume that the English pluperfect is a narrow “out-of-sequence” marker (at
least, from the discourse-pragmatic point of view).
It is more common, however, that these two functions are performed by one and
the same element. Thus, in a language with a basically atemporal system and “idle”
discontinuous past markers, a typical narrative recounting past events will contain the
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4 337

markers of the past only in the initial fragments of the text and then – sporadically – when
out-of-sequence fragments occur. Otherwise, there is no tense marking. This distribution
has been attested for the discontinuous past markers in many languages, e.g. in Creoles
(Givón 1982, Michaelis 1993, Kouwenberg 1994 and Winford 2000a, among others), and
in Polynesian (Vonen 1994 and Hooper 1998).
As a simple illustration of the same principles applied to somewhat different data, let us
consider a beginning of a narrative (a fairy tale) in Bamana. The text is taken from the
French bilingual collection Recueil de littérature manding (1978: 22ff.); for the sake of brevity,
we provide only the English (literal) translation. The verbal occurrences accompanied by
the marker of the discontinuous past tun (see above) are put in boldface and the sentences
including them are indented; perfective verbal forms are underscored.
(28) The structure of a Bamana narrative
(28a) There was a little orphan.
(28b) Her mother died and left her to a co-wife.
(28c) This co-wife, she had got a daughter,too.
(28d) Formerly, when girls were going to marry, they used to go and get a calabash.
(28e) Everybody tried to find a calabash in order to add it to the dowry.
(28f) The orphan, when her wedding-year came,had to go with the other girls of
her age to the place where the calabashes used to grow.
(28g) So, the girls came to pick her up.
(28h) They came and found that her step-mother had made her working.
(28i) The step-mother had made the girl pound the fonio,to pound her fonio first.
(28j) She asked the other girls to break off some branches on the trees on their way, so that
she could know where to go.
(28k) And so they left.
(28l) They cut off the branches and left them for her, but the wind took their branches and
put them on a different way.
(28m) After the girl had finished pounding the fonio,she left.
(28n) She took the way she had found the branches on.
(28o) And she arrived to a brook.
(28p) There was an old woman,which was washing.
(28q) The girl greeted the old woman.
(…)
It can be seen from this text that the discontinuous past marker appears in the be-
ginning, when the setting is introduced (28a), and disappears as soon as the main narration
starts (b). In all the sentences constituting the foreground – i.e. (28b), (28g)–(28h), (28j) to
(28o) and (28q) – the marker tun is not used. It appears, however, when out-of-sequence
fragments are marked. Thus, (28c) and (28d) are two different comments about some
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
338

previous events; note that in (28d) tun appears again, because (28d) introduces a different
topic as compared to (28c); it is thus a marker of a new fragment (this delimitative function
was mentioned in Givón 1982 for Creole “anteriors” as well). Note that tun is not used in
(28e), though this sentence does not belong to the main story line: it is because (28e) is a
natural continuation of (28d), and is to be construed within the temporal frame of the
previous sentence. The appearance of tun in (28i) is due to a temporal digression, inter-
rupting the main story line (and serving at the same time as a comment), while in (28p) it is
used rather as a pure marker of background (the second occurrence of tun in the same
sentence relates to the context of a relative clause).
It is important to stress once more that the use of tun (as well as of all the other typical
discontinuous past markers) has little to do with marking temporal anteriority as such, and
the identification of the discontinuous past markers with “anteriors” or with European
relative pasts is largely misleading. When the anteriority is part of the main story line and
does not violate the iconical order, it is not marked by the discontinuous past markers. On
the contrary, more specialized relative past tense markers (as, for example, the English
pluperfect) can and sometimes even must be used in these contexts. Compare in this
connection the sentences (28m) and (28n), where English requires a pluperfect form,
while the use of tun in their Bamana equivalents is not appropriate: it is because both
(28m) and (28n) are the description of a homogenous narrative fragment, without any
notional or temporal “break” between its episodes, in which case the simple perfectives
are used in Bamana. So, it could be said that the English pluperfect has a stronger syntactic
orientation (being sensitive to particular types of subordinate clauses, among other things),
while the use of discontinuous past markers like Bamana tun is organized mostly on
discourse-pragmatic grounds.
5.2. Modal uses
We have adopted the label “modal” for this group of values, because semantically they
are related, primarily, to the speaker’s assessment of some parameters of the situation.
The most widespread modal uses of the discontinuous past markers are of two types,
namely, “attenuative” and “irreal”. By “attenuative” uses we mean contexts in which
the speaker expresses a wish or a request presenting it in a somewhat mitigated, less cate-
gorical (and/or perhaps more polite) form as compared to the neutral situation. Witness
example (29) from Wolof (Stewart &Gage 1970: 375):
(29) dama bëgg-oon nga dimbëli ma
fc:1sg:pf want-dp 2sg help 1sg:o
‘I would like you to help me’, lit. ‘I had wanted that you help me’
In attenuative uses, past markers apply only to a specific set of verbal lexemes, denoting
wishes and requests (as want, desire, ask, and the like). Presenting a wish or a request as
pertaining to the past (and not related to the present) is to weaken its illocutionary force
and to give the addressee more freedom for the refusal. This is why expressions like (29)
are less categorical and more polite. In this specific type of contexts, the link between
modality and temporality is still transparent.
Irreal uses, on the contrary, pose more problems for interpretation. Unlike attenuative
uses, they are syntactically rather than lexically restricted, since they appear mainly in
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4 339

hypothetical or counterfactual conditional constructions – more often in the protasis, but
sometimes in the apodosis or in both parts. The use of the discontinuous past markers
within hypothetical or counterfactual conditionals (the boundary between them is not
always very neat, cf. Comrie 1986 and Athanasiadou &Dirven 1997 for more detail) is
widely attested, including, as far as we can judge, almost all Creole and West African
systems, witness Wolof (30a). Example (30b) illustrates the use of the discontinuous past
marker in Bamana in both parts of the counterfactual construction; cf. also Sranan (32)
below.
(30a) buma wax-oon
if:1sg speak-dp
‘If I had spoken’
(30b) su rX ni ne tun t
E
d
E
,
night at if 1sg:fc dp be:neg emph
faama tun b
E
ikun ci ka i faga d
E
king dp ipf 2sg head cut inf 2sg kill emph
‘Truly, if I had not been there in that night, the king would have cut your head off!’
Both groups of derived modal values are not only found in the domain of the discontinuous
past markers, but also in the domain of other past markers: they are characteristic at least
for pluperfects (almost universally, especially in what concerns irreal contexts), and past
imperfective forms of Romance languages (see Steele 1975, James 1982 and especially
Fleischman 1989 for an overview). The literature about the possible links between the
temporal and modal uses is vast, and it is impossible to do justice to all hypotheses. The
most widespread view (advocated, in particularly, in Fleischman 1989) treats the modal
uses as a metaphorical reinterpretation of the temporal distance, thus relating them
primarily to remote past forms (including pluperfects). This hypothesis has several weak
points (see also Dahl 1997 and Michaelis 1998 for an insightful criticism). One of them is
that there are languages like English, in which the pluperfects (as well as the other past
forms) have a wide variety of modal uses, but do not show any trace of remoteness
meaning; so it is not very clear, for example, why a form like had spoken should be inter-
preted as ‘more distant’ (and ‘less real’) than a form like spoke.
Another approach (advocated in Janssen 1993, 1994 and Thieroff 1994, 1995, among
others) is based on the idea that it is the past tense in general which is related to the notion
of remoteness, because the past, unlike the present, always views the situation as distant
from “the speaker’s now” and somehow detached from the actual word. Thus there is
a strong link between pastness and lack of actuality, which can take various forms in
concrete cases. Giving the emphasis on the non-temporal uses of past tenses, this theory
even treats past as a basically non-temporal category with a more abstract semantic
invariant called “disfocal referential concern” by Janssen and “distance” or “remoteness”
by Thieroff.
In our view, an important objection to the “past-as-distance” theory concerns the
existence of the discontinuous past markers. Against the background of the discontinuous
past, it becomes clear that the standard past cannot be unambiguously treated as “distant”
or “disfocal”: in fact, the standard past forms turn out to be basically neutral as regards
their relations to the present, while the idea of distance (or, more precisely, of a break)
between ‘now’ and ‘then’ is more likely to be ascribed to the discontinuous past. This
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
340

problem was discussed at length in sections 1–3 of the article. Consequently, the existence
of modal uses is more difficult to explain for the standard past forms than for the discon-
tinuous ones11.
Leaving the standard past forms apart, let us consider more closely what happens in the
domain of the discontinuous past forms when they are used in various modal contexts. An
important point is that the modal meaning may result not only from the combination of the
discontinuous past marker with a (past) perfective or (present) imperfective form, as we
have observed in the domain of the temporal meanings, but also from the combination
with various aspectual and modal forms with a (default) future time reference (they can be
prospectives, intentionals or potentials; hereafter, we shall refer to them as “futurate
modals”). We can even claim that the combinations of the discontinuous past markers
with the futurate modals always result in a modal value of the type above – while the
combinations with the perfective and imperfective forms may yield both temporal (most
commonly) and modal values, as in (29)–(30).
Thus, the combinations of the discontinuous past markers with the futurate modals
can be used in the attenuative contexts or (most often) in the apodosis of irreal or counter-
factual conditionals. Both uses are possible, for example, in Sranan; (31) provides an
attenuative context, while (32) illustrates the use of a futurate modal in the apodosis. The
modal marker in question is sa (glossed in the examples as “potential”), which expresses a
wide range of meanings centered on (weak) probability; cf. Winford 2000a, 2000 b for
more detail.
(31) Mi ben sa wani meki yu (…) [Winford 2000 a: 410]
I dp pot want make you
‘I would like to ask you (…)’, lit. ‘I might want…’
(32) Efu unu ben tenapu luku, unu ben sa bori en leki mi [Winford 2000a: 409–410]
if you dp stand.up look, you dp pot cook it like me
‘If you had stood and looked, you’d probably cook it [a soup] like me’
In fact, the same semantic effect can be observed with standard past markers, witness the
English would-constructions (where the past tense applies to the future/intentional marker
will) or the so-called conditional forms in Romance languages, where the marker of the
future is modified by what is morphologically the past imperfective, cf. Spanish 1pl forms
of the verb comer ‘eat’: com-er-emos ‘we shall eat’ [Future] ~ com-ía-mos ‘we ate/were
eating’ [Imperfect] ~ com-er-ía-mos ‘we should eat’ [Conditional].
Counterfactual conditionals postulate the existence in the past of an event which is
known not to have taken place and consider a hypothetical consequence of this event.
The discontinuous past marker in this type of constructions is best seen as an operator
introducing a kind of possible world in the past (if a futurate modal is used in the protasis,
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4
11 It should be noted that some authors have rightly pointed out to the fact that the modal uses of differ-
ent past forms develop very differently depending on the aspectual properties of these forms. So the
possible source for the modal values of the standard past forms can be their aspectual meaning rather
than temporal one. In particular, imperfective verbal forms have a strong advantage, as shown
already in James 1982; cf. also Vetters 2001. It is as if the imperfective aspect (and not the past tense)
would supply the same semantic component as the markers of the discontinuous past do. Note that
semantic connections between the imperfective aspect and the discontinuous past have been already
discussed in 4.2, where imperfective verbal forms in Slavic were argued to have derived discontinuous
past uses.
341

it is to stress the meaning of consequence, which is a frequent strategy, witness Bamana
[30]); cf. Dahl 1997 for a similar analysis and argumentation. While the temporal uses of
the discontinuous past markers introduce a break between ‘now’ and ‘then’ (stating that
the event belongs to the domain of ‘then’ and does not belong to the domain of ‘now’),
their irreal uses introduce a similar break, but this time it is a break between the “real
world” and a “possible world”: the event is claimed to belong to a possible world (in the
past) and to be excluded from the real world.
Apart from the meaning of counterfactual consequence, the combinations of the dis-
continuous past with the futurate modals, when used in the independent sentence, may
have secondary values, related to the idea of a weak probability or of a nearly-unrealizable
wish. Sranan (33) is an example of this. Languages may behave very differently as concerns
the possibility of such combinations. In this connection, one can distinguish between
“narrow” and “broad” discontinuous past markers: only the latter apply to the forms
having neither past nor present temporal reference. The existence of “broad” discon-
tinuous past markers seems to evidence a higher degree of their grammaticalization, be-
cause as markers of the modal values they develop the more abstract meaning of lower
probability.
Most languages considered above have “broad” discontinuous past markers, though
restricted to indicative verbal forms. Thus, the Lezgian discontinuous past suffix -japplies
equally well to forms with past, present and future reference. The “past potential” forms
on -da-jare used mostly in conditional clauses, in counterfactual and hypothetical contexts.
Nevertheless, none of the Lezgian non-indicative moods allows combinations with -j.
A more extreme case is exemplified by Udmurt (a Finno-Ugric language of Middle Volga),
where the discontinuous past marker (in this instance, the particle val, a fossilized past
form of the verb to be) can display special modal meanings when applied to imperative
forms. Cf. the following example (from Alatyrev 1970: 121):
(33) Z
ˇadi. Iz´ val,sobere cˇorïganï mïn val –
tired sleep:IMP:2SG DP,then fishing go:IMP:2SG DP
ug luï. Kïsïk uz´ van´.
neg possible. urgent work exist:pres
‘(I am) tired. I would like to sleep a while, to go fishing, but I cannot. I have an
urgent job’
Here, the 2sg imperatives iz´ ‘sleep’ and mïn ‘go’ are transformed into the expression of
an (unrealizable) wish with the help of the discontinuous past marker val: ‘it would have
been good to sleep/go’. In a sense, this development is logical, because imperatives are
often treated as forms with a realis semantics: unlike wishes or requests, they imply a
strong effort of the speaker towards the realization of the situation under the scope of
imperative (cf. Mithun 1995). Since imperatives have a considerable realis component, it
can be denied: this is what the Udmurt discontinuous past marker val does. The output
of the combination of an imperative and a marker of non-reality is a mere expression
of wish, i.e. what can be left from imperatives when the “performative” component is
omitted.
Udmurt seems to represent a system with a maximally “broad” discontinuous past
marker. On the other hand, some languages display very narrow discontinuous past
markers. Thus, a description of Koromfe (a Gur language spoken in Burkina-Faso;
Rennison 1997: 277–282) reports about a “past” suffix -ewhich can modify only the
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
342

(unmarked) form of the perfective. The Koromfe verbal paradigm consists of two basic
forms – perfective and imperfective ones, and of one progressive form morphologically
built onto the imperfective. The combinations of -eand imperfective forms, however,
are not attested. The semantics of e-forms seems close to the discontinuous prototype
(Rennison defines it as basically past, which corresponds better to the English pluperfect
and is used very rarely).
To conclude, we can say that the modal output of the forms modified with the discon-
tinuous past markers can be due both to the semantics of “non-reality” which the discon-
tinuous past markers acquire in these combinations and to the input semantics of the
futurate modals. Since these forms are modal rather than temporal, the “deactualization”
they undergo results in the weakening of the component of certainty.
6. Morphological account: the discontinuous past markers as “secondary modifiers”
In this section, we would like to argue that a closer look at some morphological pro-
perties of the discontinuous past markers may prove useful. The point is that, in the
overwhelming majority of cases, discontinuous past tense uses seem to require a non-
standard type of formal marking. While the standard tense markers display additive
marking, the discontinuous past tense markers behave rather as what can be called
“secondary modifiers” (Plungian 2001).
Standard tenses are viewed in most theories (both formal and functional) as temporal
operators applied to a non-tensed verbal stem and setting its temporal reference on the
time axis; semantically, they are equivalent to temporal adverbs of the type ‘formerly’ or
‘later’ (cf. Hornstein 1990 or Bhat 1999). Morphologically, this view parallels the analysis
of tense markers as affixes or constructions which apply to some linguistic construct
(usually called “verbal stem”). Generally speaking, the formal input of the standard tense
markers does not exist in a language as an autonomous unit: if tense is an obligatory gram-
matical category, there is no (finite) verb without tense. Thus, the tense marker is added to
the verb, which does not dispose of any tense specification before the operation takes
place.
In the domain of discontinuous past marking, however, we find a different situation.
Recall that the prototypical case of the discontinuous past markers is represented by
basically atemporal systems, where verbal forms already have some default time reference
provided by aspectual and modal markers. It seems that the discontinuous past markers,
when applied to verbs, do not perform the same task as standard tense markers are
supposed to do. In fact, the discontinuous past markers do not supply the temporal
reference; they rather modify the existing time reference of the verbal forms. The discon-
tinuous past markers tend to deal with a verb which already has some established tempo-
ral reference; therefore the units they apply to are not stems, but full-fledged verbal
forms. The input of the discontinuous past markers is not a stem-like construct: it always
exists in the language as a legitimate element of the verbal paradigm. That’s why we
prefer to use the term “secondary modifier” as representing the formal properties of
these markers.
From the morphological point of view, the discontinuous past markers are most often
auxiliaries (verbs or invariable clitics), as in Creole and many West African systems, but
they may also be bound affixes, as it is the case of Wolof -[w]oon, Koromfe -e, Lezgian -j
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4 343

and many other12. As affixes, these markers tend to occupy the most marginal position
within the word-form, and they are placed after all the other verbal markers, even those of
person and number. This behaviour is not at all typical for the standard verbal markers of
tense, which tend to an internal position within the word-form, after derivational and
aspectual markers, but before modal and personal ones; the exceptions are very rare (cf.
Bybee 1985: 196–200 and Mithun 1999a for a detailed discussion13).
As auxiliaries, the discontinuous past markers tend to readily combine with the other
(aspectual and modal) auxiliaries in the verbal system. In this respect, the example of
Tokelauan is significant. From the semantic point of view, the Tokelauan particle na shows
the most prototypical behaviour of the discontinuous past marker; it is not an accident that
its combinability is more free as compared to other Polynesian languages. Nevertheless, it
should be borne in mind that even when we find an “idle” past marker with a bare verb
form, this combination is most likely to be interpreted as a combination of the discon-
tinuous past modifier and the zero perfective marker, and not as an additive combination
of the discontinuous past and the grammatically unspecified verbal stem.
The question arises, whether a morphologically non-standard tense marker is always
semantically non-standard, i.e. expresses a kind of discontinuous past value. Such a strong
correlation would be tempting, though the data we dispose of are insufficient to prove or
disprove it. Nevertheless, this correlation may be heuristically useful. For example, the
available descriptions of the Romani verbal systems (cf., for example, Matras 2001 for an
overview) show that the only specialized past tense marker in Romani (mostly, in the form
of a suffix -as) is added to a full-fledged verbal form and is placed after the subject endings.
Matras treats this suffix as a marker of “remoteness”, but this label is used more in
keeping with Thieroff’s terminology, which applies it as a cover term for all past tense
markers with a broad spectrum of temporal and non-temporal values. Yet the list of values
provided for -as is suggestive: it includes “habitual past” with the imperfective verbs,
“anteriority” with the perfective verbs, as well as attenuative and irreal uses. All this points
towards a discontinuous past prototype, but a more extensive study is required.
Generally speaking, this formal peculiarity of the discontinuous past markers can be
accounted for as an iconical device reflecting their semantics. In fact, what the discon-
tinuous past markers do is to change the default temporal interpretation of the verbal form
they apply to. This change amounts in introducing a temporal (or notional) break between
the point of reference and the situation: metaphorically, one can speak about a kind of
“detachment” or “shift”. Since in the majority of cases the verbal forms become “more
past” than they were before the discontinuous past markers applied, the term “retrospec-
tive shift” may be used to account for this semantic operation (Plungian 2001 and
Sitchinava 2001). The terms “discontinuous past” and “retrospective shift” can be
construed as synonymous, the distinction being that the first focuses on the semantic
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
12 Among the affixes, one case is of special interest, viz. the one in which the secondary modifier is
(synchronically) the same as one of the standard additive past markers; hence morphological curi-
osities like “double past” forms, attested, e.g. in Korean (see example ([13)]; for more detail, cf.
Kim 1974, Chang 1996: 122–123 and especially Sohn 1995) or in Diola (a West-Atlantic language
of Senegal). Consider Diola esukey eniilo-
EE
nvillage small-past ‘the village was small’ vs. esukey
eniilo-
EE
n-
EE
n‘long ago, the village was small [now it isn’t]’ (see Sapir 1965 and S
ˇosˇitajsˇvili 1998).
13 It should be noted that the real number of exceptions may in fact be even lower, because quite a few
cases where alleged tense markers display an anomalously marginal position can qualify as instances
of discontinous past and not of standard tense; cf. the case of Romani discussed below.
344

output of the forms in question, while the second reflects the process that creates the
specialized markers.
Another question which needs further investigation is that of the origin of the discon-
tinuous past markers. The evidence available is not vast, though the most plausible sources
for this type of markers seem to be past forms of auxiliary verbs (mostly, to be) merged
with a full-fledged main verbal form in a kind of serial construction. The development
would in this case start from constructions like “it-was they study” towards constructions
with a more or less morphologized discontinuous past marker and a discontinuous seman-
tics of the type “they used to study”. This pattern is exemplified by Creole systems (where
the discontinuous past markers go back to some past form of the verb to be: most often, it is
English been or French était). Other languages where the origin of the discontinuous past
markers is transparent and points in the same direction are Udmurt and Mari. Classical
Arabic (as well as modern Arabic dialects) may have the same systems, with the past
auxiliary ka¯na added to perfective and imperfective forms, though it is not clear whether
the semantics of the ka¯na periphrasis is close enough to the discontinuous past prototype.
Finally, recall that the origin for the “remote past” of Romani is also, most likely, the past
form of the verb to be (Matras 2001: 172).
Polynesian data (especially, Tahitian, according to Lazard &Peltzer 2000: 130–131)
suggest a possible link between the discontinuous past markers and the deictic markers of
remoteness. This source is not uncommon for the past domain (cf. Comrie 1985) and, what
is especially interesting, it has been discussed in the relation to the Indo-European markers
for tense (see, first of all, Kiparsky 1968). The oldest Indo-European past marker may be
related to the so-called prefixal augment e- (probably, deictic in its origin). The evidence
suggested by Kiparsky points strongly at the hypothesis that it initially functioned more
like a secondary modifier than like an affix. In this connection, its original meaning could
be something like the discontinuous past, which was later weakened towards a standard
past meaning.
7. Conclusion
We have tried to show that in the domain of past tense marking one special type of value
(largely ignored in previous descriptions) is worth considering as a cross-linguistically valid
“category type” (Dahl 1985: 31–35). We have proposed to call this value “discontinuous
past”, since the main difference from standard past markers is that it provides an additional
semantic component specifying the situation as non-existent or no more relevant at the
moment of utterance.
In a typological overview, we have shown that discontinuous past value is represented in
quite a few languages from different areas and genetic families. There are verbal systems
which have a specialized discontinuous past marker. More often than not, this marker
remains a single temporal gram (this is especially characteristic of Oceanic, West African
and Creole systems), though it may also coexist with other standard markers of past tense.
A discontinuous past value can also be manifested as a particular meaning (or use) of some
other verbal grammatical markers; this type of polysemy is recurrent for pluperfects, but is
also attested for past duratives or past habituals.
Generally speaking, the discontinuous past value is very rarely realized by a specialized
monosemic marker. Usually, it shows up within a semantic network. The elements of this
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4 345

network seem to be cross-linguistically stable: they include both temporal values (as
remoteness and discourse backgrounding) and modal values (as counterfactual, irreal and
attenuative ones). One can speak about a kind of semantic invariant dominating all these
uses: it can be construed as a notional “break” between “speaker’s now” and the point in
time referred to. Interestingly, this semantic property often finds a parallel in the morpho-
logical characteristics of discontinuous past markers: they tend to behave like “secondary
modifiers”, applying to a pre-existent full-fledged verbal form rather than to an under-
specified verbal stem, like standard TAM markers do.
Abbreviations
1,2,3 1,2,3 person int intentional
act active (voice) l locative
an (locative) anaphore mmasculine
caus causative neg negation
cont continuative (‘still’) nom nominative
dir directional nsp non-specific
dp discontinuous past oobject
emph emphatic particle pf perfect(ive)
erg ergative pl plural
fc focus poss possessive pronoun
fut future pot potentional
hab habitual pres present
imp imperative pqp pluperfect
inf infinitive sg singular
ipf imperfective sp specific
References
Abraham, Werner (1999): Preterite decay as a European areal phenomenon, in Folia linguistica 33.1,
11–18.
Agence de coopération culturelle et technique (ed.) (1978): Recueil de littérature manding. Paris:
Edition Nubia.
Alatyrev, V.I. (ed.) (1970): Grammatika sovremennogo udmurtskogo jazyka:sintaksis prostogo
predlozˇenija [A grammar of Modern Udmurt: the syntax of simple sentences]. Izˇevsk: Udmurtija.
Anderson, Stephen C. (1979): Verb structure, in: Hyman, Larry (ed.), Aghem grammatical structure.
Los Angeles: USC, 73–136.
Athanasiadou,Angeliki &Dirven, René (1997): Conditionality, hypotheticality, counterfactuality, in:
Athanasiadou,Angeliki &Dirven, René (eds.), On conditionals again. Amsterdam: Benjamins,
61–96.
Bhat,D. N. S. (1999): The prominence of tense, aspect and mood. Amsterdam: Benjamins.
Bickerton, Derek (1975): Dynamics of a Creole system. Cambridge: Cambridge University Press.
Bickerton, Derek (1981): Roots of language. Ann Arbor: Karoma.
Bybee, Joan L. (1985): Morphology: a study of the relation between meaning and form. Amsterdam:
Benjamins.
Bybee, Joan L., Perkins, Revere &Pagliuca, William (1994): The evolution of grammar:Tense, aspect
and modality in the languages of the world. Chicago: Univesity of Chicago.
Chang, Suk-Jin (1996): Korean. Amsterdam: Benjamins.
Childs, G. Tucker (1995): A grammar of Kisi, a Southern Atlantic language. Berlin: Mouton de Gruyter.
Church, Eric (1981): Le système verbal du wolof. Dakar: CLAD.
Comrie, Bernard (1985): Tense. Cambridge: Cambridge University Press.
Comrie, Bernard (1986): Conditionals: a typology, in: Traugott Elizabeth C. et al. (eds.), On condi-
tionals. Cambridge: Cambridge University Press, 77–99.
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
346

Comrie, Bernard (1999): Tense, in: Brown, Keith &Miller, Jim (eds.), Concise encyclopedia of
grammatical categories. Amsterdam: Elsevier, 363–368.
Dahl, Östen (1985): Tense and aspect systems. Oxford: Blackwell.
Dahl, Östen (1993): Review of Singler (ed.) 1990. Studies in language 17.1, 251–258.
Dahl, Östen (1997): The relation between past time reference and counterfactuality: a new look, in:
Athanasiadou,Angeliki &Dirven, René (eds.), On conditionals again. Amsterdam: Benjamins,
97–114.
Dahl,Östen &Hedin, Eva (2000): Current relevance and event reference, in: Dahl, Östen (ed.), Tense
and aspect in the languages of Europe. Berlin: Mouton de Gruyter, 385–402.
Davies,Martha S. (2000): The past imperfect in Palenquero, in: Studies in language 24.3, 565–581.
Declerck, Renaat (1991): Tense in English: its structure and use in discourse. London: Routledge.
Derbyshire, Desmond C. (1999): Carib, in: Dixon, R. M. W. &Aikhenvald, Alexandra Y. (eds.), The
Amazonian languages. Cambridge: Cambridge University Press, 23–64.
Dixon, R. M.W. (1988): A grammar of Boumaa Fijian. Chicago: University of Chicago.
Engel, D. M. (1994): Plus-que-parfait: Past anterior or Past punctual? in: Linguisticae Investigationes 18,
223–242.
Eroms, Hans-Werner (1983): Relativer und absoluter Gebrauch des Plusquamperfekts im Deutschen,
in: Askedal, John Ole et al. (eds.), Festschrift für Laurits Saltveit zum 70. Geburtstag. Oslo:
Universitetsforlaget, 58–71.
Fleischman, Suzanne (1989): Temporal distance: a basic linguistic metaphor, in: Studies in language
13.1, 1–50.
Frajzyngier,Zygmunt &Shay, Erin (2002): A grammar of Hdi. Berlin: Mouton de Gruyter.
Frimigacci, Daniel; Keletaona, Muni; Moyse-Faurie, Claire &Vienne, Bernard (1995): Ko le fonu
tu’a limulimua – La tortue au dos moussu.Textes de tradition orale de Futuna. Paris: Peeters.
Gibson, K. (1992): Tense and aspect in Guyanese Creole, with reference to Jamaican and Carriaconan,
in: International Journal of American Linguistics 58.1, 49–95.
Givón, Talmy (1982): Tense-aspect-modality: the Creole prototype and beyond, in: Hopper, Paul (ed.),
Tense-aspect: between semantics and pragmatics. Amsterdam: Benjamins, 114–163.
Gosselin, Laurent (1996): Sémantique de la temporalité en français:Un modèle calculatoire et cognitif
du temps et de l’aspect. Louvain-la-Neuve: Duculot.
Harris, Martin (1983): The “past simple” and the “present perfect” in Romance, in: Vincent,Nigel &
Harris, Martin (eds.), Studies in the Romance verb: Essays offered to Joe Cremona on the occasion
of his 60th birthday. London: Croom Helm, 42–70.
Hauser-Suida,Ulrike &Hoppe-Beugel, Gabriele (1972): Die Vergangenheitstempora in der deutschen
geschriebenen Sprache der Gegenwart: Untersuchungen an ausgewählten Texten. München: Hueber,
Schwann.
Haspelmath,Martin (1993): A grammar of Lezgian. Berlin: Mouton de Gruyter.
Haspelmath, Martin (1994): The tense system of Lezgian, in: Thieroff, Rolf &Ballweg, Joachim
(eds.), Tense systems in European languages. Tübingen: Niemeyer, 267–277.
Holm, John A. (1988): Pidgins and Creoles. Vol. 1: Theory and structure. Cambridge: Cambridge Univer-
sity Press.
Hooper, Robin (1989): Tense and aspect in Tokelauan, in: Harlow, R. &Hooper, Robin (eds.), Papers
from the Fifth International Conference on Austronesian Linguistics. Pt. 1: Oceanic languages.
Auckland: Linguistic Society of New Zealand, 247–270.
Hooper, Robin (1998): Universals of narrative pragmatics: a Polynesian case study, in: Linguistics 36.1,
119–160.
Hornstein, Norbert (1990): As time goes by:Tense and universal grammar. Cambridge (MA): MIT.
Hovdhaugen,Even et al. (1989): A handbook of the Tokelau language. Oslo: Norwegian University Press.
Idiatov, Dmitri (2000): Le sémantisme des marqueurs aspecto-temporels du bambara: une tentative
d’analyse, in: Mandenkan 36, 1–59.
James, Deborah (1982): Past tense and the hypothetical: a cross-linguistic study, in: Studies in language
6.3, 375–403.
James, Deborah (1991): Preterit forms in Moose Cree as markers of tense, aspect and modality, in: Inter-
national Journal of American Linguistics 57.3, 281–297.
Janssen, T.A. J. M. (1993): Tenses and demonstratives: conspecific categories, in: Geiger, R. A. &
Rudzka-Ostyn, B. (eds.), Conceptualizations and mental processes in language. Berlin: Mouton de
Gruyter, 741–783.
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4 347

Janssen, T. A. J. M. (1994): Preterit and perfect in Dutch, in: Vet, Co &Vetters, Carl (eds.), Tense and
aspect in discourse. Berlin: Mouton de Gruyter, 115–145.
Kim, Nam-Kil (1974): The double past in Korean, in: Foundations of language 12, 529–536.
Kiparsky, Paul (1986): Tense and mood in Indo-European syntax, in: Foundations of language 4, 30–57.
Klein,Wolfgang (1994): Time in language. London: Routledge.
Kouwenberg, Silvia (1994): A grammar of Berbice Dutch creole. Berlin: Mouton de Gruyter.
Koval’, I. A. &Gnalibouly, B. A. (1997): Glagol fula v tipologicˇeskom osvesˇcˇenii [Fula verb in a typo-
logical perspective]. Moscow: Institute of linguistics.
Lazard, Gilbert & Peltzer, Louise (2000): Structure de la langue tahitienne. Paris: Peeters.
Lefebvre, Claire (1996): The tense, mood, and aspect system of Haitian creole and the problem of
transmission of grammar in creole genesis, in: Journal of Pidgin and Creole languages 11.2,
231–311.
Lindstedt, Jouko (2000): The perfect: aspectual, temporal and evidential, in: Dahl, Östen (ed.), Tense
and aspect in the languages of Europe. Berlin: Mouton de Gruyter, 365–383.
Maiden, Martin & Robustelli, Cecilia (2000): A reference grammar of Modern Italian. London:
Arnold.
Maslov, Ju. S. (1988): Resultative, perfect and aspect, in: Nedjalkov, V. P. (ed.), Typology of resultative
constructions. Amsterdam: Benjamins, 63–85.
Matras, Yaron (2001): Tense, aspect and modality categories in Romani, in: Sprachtypologie und
Universalienforschung 54.2, 162–180.
Mellet, Sylvie (1994): Les temps du passé: le plus-que parfait, in: Mellet, Sylvie; Joffre, M.D. &
Serbat, G. (eds.), Grammaire fondamentale du latin:le signifié du verbe. Louvain: Peeters, 105–123.
Michaelis, Laura A. (1998): Aspectual grammar and past-time reference. London: Routledge.
Michaelis, Susanne (1993): Temps et aspect en créole seychellois: valeurs et interférences. Hamburg:
Buske.
Mithun, Marianne (1995): On the relativity of irreality, in: Bybee, Joan & Fleischman, Suzanne (eds.),
Modality in grammar and discourse. Amsterdam: Benjamins, 367–388.
Mithun, Marianne (1999 a): The status of tense within inflection, in: Booij, Geert & van Marle, Jaap
(eds.), Yearbook of morphology 1998. Dordrecht: Kluwer, 23–44.
Mithun, Marianne (1999b): The languages of native North America. Cambridge: Cambridge University
Press.
Mosel, Ulrike & Hovdhaugen, Even (1992): Samoan reference grammar. Oslo: Scandinavian Univer-
sity Press.
Moyse-Faurie, Claire (1993): Dictionnaire futunien-français (avec index français-futunien). Paris: Peeters.
Moyse-Faurie, Claire (1997): Grammaire du futunien. Nouméa: CDP Nouvelle-Calédonie.
Moyse-Faurie, Claire. (2002): Tense-aspect markers in Faka’uvea (East Uvean), in: Rongorongo studies
12.1, 3–21.
Nedjalkov, Vladimir P. & Jaxontov, S. E. (1988) The typology of resultative constructions, in: Ned-
jalkov,Vladimir P. (ed.), Typology of resultative constructions. Amsterdam: Benjamins, 3 –62.
Noonan,Michael (1992) A grammar of Lango. Berlin: Mouton de Gruyter.
Peltzer,Louise (1996) Grammaire descriptive du tahitien. Pape’ete: Polycop.
Pinkster, Harm (1990) Latin syntax and semantics. London: Routledge.
Plungian, Vladimir A. (2001): Antirezul’tativ: do i posle rezul’tata [Anti-resultative: before and after
result], in: Plungian,Vladimir A. (ed.), Issledovanija po teorii grammatiki. I: Glagol’nye kategorii.
Moscow: Russkie slovari, 50–88.
Rennison, John (1997): Koromfe. London: Routledge.
Robert,Stéphane (1991): Approche énonciative du système verbal: le cas du wolof. Paris: CNRS.
Sadler, Louisa & Nordlinger, Rachel (2001): Nominal tense with nominal scope: a preliminary
sketch, in: Butt, M. &Tracy Holloway King (eds.), Proceedings of the LFG01 conference. Hong
Kong: University of Hong Kong [http://www-csli.stanford.edu/publications/].
Salkie, Raphael (1989): Perfect and pluperfect: what is the relationship? in: Journal of Linguistics 25.1,
1–34.
Sapir, David J. (1965): A grammar of Diola-Fogny. Cambridge: Cambridge University Press.
Schwenter, Scott A. (1998):Discontinuous remoteness in Dagaare, in: Maddieson Ian &Hinnebusch,
Thomas J. (eds.), Language history and linguistic description in Africa. Trenton (NJ): Africa World
Press, 87–93.Seiler, Hansjakob (1971): Abstract structures for mood in Greek, in: Language 47.1,
79–89.
V. A.Plungian &J. van der Auwera, Towards a typology of discontinuous past marking
348

Sidnell, Jack (2002): Habitual and imperfective in Guyanese Creole, in: Journal of Pidgin and Creole
languages 17.2, 151–189.
Singler, John Victor (ed.) (1990): Pidgin and Creole tense-mood-aspect systems. Amsterdam: Benja-
mins.
Sitchinava, Dmitry V. (2001): Pljuskvamperfekt i retrospektivnyj sdvig v jazyke santali [Pluperfect and
retrospective shift in Santali], in: Plungian, Vladimir A. (ed.), Issledovanija po teorii grammatiki.
I: Glagol’nye kategorii. Moscow: Russkie slovari, 89–114.
Sitchinava, Dmitry V. (2003): K tipologii glagol’nyx sistem s neskol’kimi formami pljuskvamperfekta:
casus latinus [On typology of verbal systems with several pluperfect forms: casus latinus], in:
Voprosy jazykoznanija 2003–5.
Sohn, Sung-Ock S. (1995): Tense and aspect in Korean. Honolulu: Center for Korean Studies.
S
ˇosˇitajsˇvili, I.A. (1998): Funkcii i status pljuskvamperfekta v glagol’noj sisteme [Functions and status of
the pluperfect in the verbal system]. Unpublished PhD thesis. Moscow, Russian State University
for the Humanities.
Squartini, Mario (1998): Verbal periphrases in Romance: Aspect, actionality and grammaticalization.
Berlin: Mouton de Gruyter.
Squartini, Mario (1999): On the semantics of the Pluperfect: evidence from Germanic and Romance, in:
Linguistic typology 3.1, 51–89.
Steele, Susan (1975): Past and irrealis: just what does it all mean? in: International Journal of American
Linguistics 41.3, 200–217.
Stewart, William A. & Gage, William W. (1970): Notes on Wolof Grammar, in: L.V. Nussbaum, Gage,
W. W.,Varre, D. (eds.), Dakar Wolof:a basic course. Washington: CAL, 355–412.
Tagliamonte, Sali &Lawrence, Helen (2000):“I used to dance, but I don’t dance now”: the habitual
past in English, in: Journal of English Linguistics 28.4, 324–353.
Thieroff, Rolf (1994): Inherent verb categories and categorizations in European languages, in:
Thieroff, Rolf &Ballweg, Joachim (eds.), Tense systems in European languages. Tübingen:
Niemeyer, 3–45.
Thieroff, Rolf (1995): More on inherent verb categories in European languages, in: Thieroff, Rolf
(ed.), Tense systems in European languages II. Tübingen: Niemeyer, 1–36.
Thomas, Paul-Louis (2000): Le plus-que-parfait en serbo-croate (bosniaque, croate, monténégrin,
serbe) dans une approche contrastive avec le français, in: Carlier, Anne et al. (eds.), Passé et
parfait. (Cahiers Chronos 6). Amsterdam: Rodopi, 117–131.
Vetters, Carl (2001): Le conditionnel: ultérieur du non-actuel, in: Dendale, Patrick & Tasmowski,
Liliane (eds.), Le conditionnel en français. Metz: Université de Metz, 169–207.
Vonen, A. M. (1994): The expression of temporal and aspectual relations in Tokelau narratives, in:
Bache, Carl et al. (eds.), Tense, aspect and action: empirical and theoretical contributions to lan-
guage typology. Berlin: Mouton de Gruyter, 371–396.
Welmers, Wm. E. (1973): African language structures. Berkeley: University of California Press.
Winford, Donald (1996): Common ground and the Creole TMA, in: Journal of Pidgin and Creole
languages 11.1, 71–84.
Winford, Donald (2000a): Tense and aspect in Sranan and the Creole prototype. in: McWhorter, John
(ed.), Language change and language contact in Pidgins and Creoles. Amsterdam: Benjamins,
383–441.
Winford, Donald (2000 b): Irrealis in Sranan: Mood and modality in a radical creole, in: Journal of
Pidgin and Creole languages 15.1, 63–125.
Vladimir A. Plungian Johan van der Auwera
Department of Linguistic Typology Center for Grammar, Cognition and Typology
Institute of Linguistics University of Antwerp
B. Kislovsky per. 1/12 Universiteitsplein 1
125009 Moscow B - 2610 Wilrijk
RUSSIA BELGIUM
plungian@gmail.com johan.vanderauwera@ua.ac.be
Sprachtypol. Univ. Forsch. (STUF) 59 (2006) 4 349