By perfect we mean a category with approximately the same semantics as the English (Present) Perfect in I have read this book, which is used to express events that took place before the temporal reference point but which have an effect on or are in some way still relevant at that point. This includes at least two related but distinguishable uses:
A perfect may be used resultatively, i.e. of an event, often but not always a recent one, which has results that hold at the time of speech (or any other time serving as reference point): Someone has stolen my purse! (=the purse is gone).
A perfect may be used experientially, i.e. to say that a certain type of event took place one or more times over an interval of time, typically one that extends up to the moment of speech (or whatever time serves as the reference point): I have seen worse things in my life.
Constructions that have only one of these two uses (dedicated resultatives and experientials) have not been treated as perfects here. Admittedly, the information given in grammars is not always specific enough to make reliable decisions on this point.
Perfects may have further uses, such as the universal perfect (or perfect of persistent situation), as in I have lived here for five years. It is also common for perfects to develop evidential uses (see Chapter 77; in English this is typical of the perfect progressive – You have been drinking ), and recent past uses (as in several Romance languages).
Note that the terms “perfect” and “perfective” are not synonymous (the terminology is confusing but has a historical motivation). Bybee et al. (1994) use the term ”anterior” in the same sense as ”perfect” is used here.
The English Perfect is cross-linguistically typical in being expressed periphrastically.
The basic distinction is between languages that have perfects and those that do not. As noted above, only constructions or forms that have both resultative and experiential readings are regarded as perfects here. On the other hand, to count as a perfect, a construction or form must not be equivalent to a general past tense. An operational criterion for judging about this is whether or not the form or construction is regularly used in narratives (Lindstedt 2000: 366).
The most interesting areal patterns with regard to perfects are related to their diachronic sources. Perfects derive diachronically from at least three types of sources, the last two of which are singled out for special marking on the map:
‘The train has arrived.’
(2) Yoruba (own data)
‘He has read this book’
It may be, as argued by Ebert (2001), that perfects of type (iii) should be treated as a gram type of their own (Ebert calls it “NEWSIT”), where the primary function is to introduce a “new situation”. There is also some evidence that a NEWSIT and a perfect, with distinct forms, may coexist in the same language (Burmese and Fijian are possible examples in our sample). We have not tried here to make a systematic distinction between perfects and NEWSITs, since the information at hand does not allow this, but languages with perfects that are known to be diachronically derived from (or have the same shape as) ‘already’ or ‘finish’ are marked as such on the map.
There may be further distinctions between perfects in the narrow sense (also called present perfects) on the one hand, and past perfects (or pluperfects) and future perfects on the other. For reasons of space, these are not represented on our maps. It may be noted that in general, past perfects indeed tend to be realized as combinations of pasts and perfects, and that they are to be expected in those languages in which both these gram types appear. However, sometimes a past perfect may survive the demise of a present perfect, as has happened for instance in Romanian. Past perfects have a relatively strong tendency to develop non-compositional readings, that is, they become semantically independent of pasts and perfects. Future perfects tend to play a more peripheral role in tense-aspect systems.
We distinguish three types on the map: perfects derived from a possessive construction (‘have’-perfects), perfects derived from words meaning ‘finish’ or ‘already’, and all other perfects.
|Perfect of the 'have'-type (derived from a possessive construction)||7|
|Perfect derived from word meaning 'finish' or 'already'||21|
If we look at the general distribution of perfects, we can note the following tendencies. The largest homogeneous areas with perfects are found in western Europe and South and Southeast Asia. A high proportion of languages with perfects is also found throughout Africa and in an area comprising Mesoamerica and the northwestern corner of South America. Large areas with virtually no perfects, on the other hand, are found in the rest of South America and Australia.
However, the differences between the different parts of the world are accentuated when we take into consideration the possible diachronic sources of perfects. Perfects deriving from possessive constructions are found almost exclusively in Europe (Chukchi is a possible exception, but we have not treated the construction in question as a perfect.) Furthermore, they are almost exclusively based on constructions with a transitive possessive like English have, and are restricted to a contiguous area in western Europe, as a result of an apparently rapid spread in the Middle Ages. Map 68.1 shows the maximal extent of this spread. In the centre of the area, there is a division of labour between ‘have’-perfects and ‘be’-perfects, the latter being used primarily for intransitive verbs of motion and change. In the centre, there has also been a new development in that the original perfects have come to be used as general pasts or perfectives. The boundaries of these areas are also shown in Map 68.1.
Perfects derived from ‘already’ and ‘finish’—which were noted above as having a special semantics—also show a very marked geographical distribution, being concentrated in Southeast Asia and West Africa. (It is possible that more languages should be marked as having this type, but information is sometimes lacking.) It is striking that this coincides with the areas where we find little or no morphological marking of tense and aspect in general, which is reflected on our other maps by the preponderance of white circles.
We are grateful to Joan Bybee for putting unpublished data from the GRAMCATS database at our disposal.