Chapter The Past Tense

by Östen Dahl and Viveka Velupillai

Please note: A common introduction to Chapters 65, 66, 67, and 68 on Tense and Aspect is available here.

1. Introduction

In English, like virtually all European languages, there is a systematic grammatical distinction between present tenses and past tenses, as in the following sentence pair:



The temperature is below zero right now.


The temperature was below zero yesterday at noon.

In (1) the form of the finite verb (in this case, the copula is/was) depends on what time we are talking about – what we may call the topic time (following Klein (1994)): the present form is  is used if the topic time coincides with the time of speech, and the past form was  if the topic time precedes the time of speech. This is of course a very rough rule in need of a number of further specifications. As was already noted in the introductory chapter, grammaticalized marking of time in the form of tenses is typically independent of considerations of relevance. The fact that time in (1) is also indicated (and more precisely so) by time adverbials does not make tense marking less necessary. Rather, the presence of a deictic time adverbial such as yesterday  renders the use of anything but the simple past tense unacceptable in (1b).

Since it is generally the past tense rather than the present that is overtly marked, we may speak of languages having or not having past marking rather than having a past/non-past distinction. 

It is only from a Eurocentric point of view that the marking of the distinction between present and past appears to be a necessary part of grammar. Languages may or may not distinguish (1) grammatically, and there is no clear majority for either alternative. The following Indonesian example, which translates both ‘The water is cold’ and ‘The water was cold’, illustrates the lack of a present/past distinction:

(2) Indonesian (own data) 







‘The water is/was cold.’

The primary goal of this chapter is to show the geographical distribution of past tenses. There are several things that make this task less straightforward than it might seem prima facie. 

Perhaps most importantly, we must elucidate the relationship between past tenses and grammatical aspect. Given that perfective forms are by default interpreted as referring to the past (see the general introduction to Chapters 65, 66, 67, and 68), any further marking of past time reference would appear redundant. Indeed, it is probably most common for overt past tense marking not to be compatible with perfective aspect. In such cases, we may get the tripartite system represented e.g. by the present-imperfect-aorist system of many Indo-European languages, illustrated here by Eastern Armenian, where a periphrastic construction (copula + converb of the main verb) is used in the imperfective:

(3) Eastern Armenian (own data) 














'He is writing/writes a letter.'














'He was writing/wrote (habitually) a letter.'












'He wrote a letter (a single event).' 

Similar systems are found in many different families, although they are particularly common in the past-marking part of Eurasia. The question is how they should be analysed. The possibility that first comes to mind is probably to assign the labels "present", "past imperfective" and "past perfective" to the three forms. This would imply that (3b) and (3c) are both members of a general past tense category, and that the aspectual distinction is made only there. However, there are at least two arguments against such an analysis. One is that it does not fit the actual make-up of the forms very well. In the Eastern Armenian system, there is nothing that unites (3b) and (3c) in the way they are expressed. Rather, (3b) is naturally seen as an elaboration on (3a), and (3c) has a wholly unique composition. This state of affairs turns out to be typical of tripartite systems (although some exceptions can be found). Another argument is that calling (3c) a "past perfective" obscures the basic cross-linguistic unity of perfectives, which appear to have more or less the same semantics irrespective of whether the language distinguishes past and present in the imperfective or not. Following Dahl (1985: 81-84), we therefore do not regard categories like the Armenian Aorist as “past perfectives” but rather simply as “perfectives”, and categories like the Armenian Imperfect as a particular variety of past, restricted to imperfective forms. In Dahl (1985: 117-118), such forms were labelled “PASTi” but here we use the more intuitive “past imperfective”. Aspectually unrestricted pasts and past imperfectives are not distinguished on the map, the argument being that the latter are found in the overwhelming majority of languages which have both the perfective/imperfective distinction and past tense marking. The most notable exceptions are languages such as Russian, where these two grammatical phenomena are more independent of each other.

Both Dahl (1985: 117) and Bybee et al. (1994: 82) provide support for the claim that a qualified majority of all pasts are marked morphologically. The tendency may be weaker for past imperfectives, but the material is not large enough to make any certain claims. 

Very often, languages make further grammatical distinctions within the domain of past time reference. Thus, many languages have perfects as separate categories – this will be discussed in Chapter 68. In addition, as many as one fifth of the languages in our sample make remoteness distinctions – that is, tense choice is dependent on the temporal distance between the time of speech and the topic time. Remoteness may be more subjectively or more objectively determined; in the latter case, a combination of a “remote” time adverbial with a “non-remote” tense will result in ungrammaticality. (It should be noted that like tenses in general, tenses which distinguish degrees of remoteness do not substitute for adverbials but are used whether or not there is another temporal indication in the sentence.)

Almost universally, if there is one well-defined cut-off point in the past between different forms, the division lies between ‘today’ and ‘before today’. The ‘before today’ range is often divided further. The term hodiernal is commonly used for ‘today’s past’, and tenses that are restricted to the day before the point of speech may be called hesternal.

The richest system in our sample is that of Yagua, which according to the available description (Payne and Payne 1990: 386-388) has five degrees of remoteness in the past, as shown in Table 1: 

Table 1. Remoteness distinctions in Yagua

Name in grammar




Proximate 1

‘a few hours previous to the time of utterance’





‘I went (this morning).’

Proximate 2

‘one day previous to the time of utterance’





‘I saw him (yesterday).’

Past 1

‘roughly one week ago to one month ago’





‘He has died (between a week and a month ago’).

Past 2

‘roughly one to two months ago up to one or two years ago’





‘He has died (between 1 to 2 months and a year ago’).

Past 3

‘distant or legendary past’





‘I was born (a number of years ago).’

Forms used for recent past sometimes coincide with perfects (or are historically derived from them). As the label ‘distant or legendary past’ in the table suggests, it may sometimes be hard to distinguish distant pasts from forms reserved for use in myths and legends, which may function not only as tenses, but also as stylistic or modal markers.

2. Definition of values

The basic dichotomy here is between languages that mark the past/non-past distinction grammatically (including marking by periphrastic constructions) and those which do not. Within the first group, three subtypes are distinguished, depending on the number of remoteness distinctions made, as shown in the feature-value box. 

Values of Map 66A. The Past Tense
Go to map
Past/non-past distinction marked; no remoteness distinction 94
Past/non-past distinction marked; 2-3 degrees of remoteness distinguished 38
Past/non-past distinction marked; at least 4 degrees of remoteness distinguished 2
No grammatical marking of past/non-past distinction 88
Total: 222

3. Geographical distribution

There are quite strong areal tendencies in the distribution of past tenses—perhaps strong enough to deserve another word than “tendency”. This can be seen on the map as homogeneous one-colored areas. 

The largest homogeneous past-marking area is one that stretches from Iceland in the northwest to the Horn of Africa in the south and Bangladesh in the southeast, including much of central Eurasia (but excluding northeastern Siberia). Since the Indo-European phylum also extends between Iceland and Bangladesh, it is tempting to see this as an Indo-European phenomenon. However, the area also includes phyla such as Uralic, Altaic, and large parts of Afro-Asiatic, and isolates such as Basque and Burushaski. Other homogeneous past-marking areas include Australia, northern South America and central New Guinea.

In Africa, the majority of the sample languages are past-marking, but there is also an (almost) homogeneous non-marking area in central and western Africa. Two things are worth mentioning here. One is that most of the languages involved have an imperfective/perfective distinction. The other is that the languages without past in this area include members of all three major phyla that are represented here—Afro-Asiatic, Niger-Congo, and Nilo-Saharan. None of these phyla is consistently non-marking, however. In other words, we are dealing with a fairly clear example of areal convergence. 

The most salient area of homogeneous non-marking of past is found in Southeast Asia, and also includes languages from a number of different phyla—Sino-Tibetan, Austro-Asiatic and Austronesian, to mention the largest ones. The languages in this area lack not only past tenses but also marking of the imperfective/perfective distinction and inflectional futures. These are of course the well-known isolating languages of Southeast Asia, and it is hardly a coincidence that they lack precisely those tense-aspect gram types that cross-linguistically are most often marked inflectionally. What is more difficult to decide is what is cause and what is effect here.

There may be a similar area in West Africa, but it is at any rate much more restricted – in our sample there are only three West African languages that are marked as white on all three of the Maps 65A, 66A, and 67A.