vidynath at math.ohio-state.edu
Mon Aug 9 18:51:56 UTC 1999
First, some methodological issues.
As I understand it, a cornerstone of historic linguistics, or any historical
science, is that we should try to explain as much as possible by processes
that we can and have observed. This should go for syntactic change as much
as for phonological or morphological.
Secondly, some changes seem to be much more common than their reversals. In
such cases, requiring the reverse change must be supported by
proportionately stronger evidence than we would want for the more common
In particular, the tendency is for new formations to arise to denote
``core'' notions and expand outward to denote allied notions as well.
Correspondingly, older formations tend to become restricted away from one or
more core notions, depending on the new formations. This means that we need
to be careful in distinguishing specilized and general notions in diachronic
syntax. For example, progressive is a special case of imperfective. But we
do not confuse the two, nor see progressive as a natural outgrowth of
imperfective. We need to be just as careful with less familiar categories
such as completives.
Jens Elmegaard Rasmussen <jer at cphling.dk> wrote:
> On Mon, 26 Jul 1999, Vidhyanath Rao wrote:
>>> If my observation that there is an alliance between the sk^-present
>>> type and the s-aorist is correct ... then the s-aorist was originally
>>> inchoative in function. [...]
>> This makes it harder for me to understand how the aorist became the
>> perfective. `Started driving', in contrast to `drove', suggests
>> incomplete action.
> The aorist reports a turn of event that caused a new situation: is that
> not the perfect thing to express a beginning?
Aorist of `start/begin' would be the thing to express a beginning. But why
is the thing to express the whole event? While there are several examples
where an auxillary meaning `finish' generalizes to become a `perfect', and
in a few cases to a perfective, there seem to be none in which `begin' does
> The PIE function of the different derivative categories must be at least
> compatible with that of their later reflexes, and the simplest solution is
> that wherever we find non-trivial correspondences between the daughter
> languages we have a relatively direct reflexion of the protolanguage.
The perfect was rolled into the (perfective) past in Germanic, Italic,
Tocharian and possibly Celtic. Does that mean that the Perfect was
originally a past, and Indic and Greek innovated that into a
resultative/perfect? Or should be pay attention to the fact that the
development is typically the other way?
> Why would a morpheme with lengthening and -s- turn up as the
> expression of non-durative past in Italo-Celtic, Slavic, Armenian,
> Greek and Tocharian if that were not its function in PIE already?
Now the (sigmatic) aorist is non-durative. But,
>> How do you classify ``I learned that chapter in one month?''
> [Jens:] I believe as a job for the aorist.
the aorist is compatible with duration,
>> Could the imperfect report `action pure et simple', or was that reserved
>> for the aorist? How did PIE speakers report a durative action that was
>> done, like ``I walked home''? How did they say ``I made pots
> I guess they walked home in the aorist, made pots (generically) in the
> ipf., but made a specific pot or set of pots in the aorist.
but not with unbounded but whole events. [I am assuming that `generically'
includes ``I made pots yesterday'', rather than just ``I used to make
This looks remarkably like Modern Russian, but unlike perfectum-imperfectum
distinction of Latin, or the aorist-imperfect distinction of Slavic (even in
Modern Bulgarian, the imperfective aorist is regularly used in this case)
etc. The difference between the two, and the historical reasons for such
differences are amply discussed in the references I gave.
It seems to me that Szeremenyi was correct to charge that the syntax of the
Modern Slavic languages (that have lost the aorist-imperfect distinction)
was unjustifiably transferred to other IE languages and thence to PIE. If
the situation in the oldest known stages of languages with a distinct
imperfect were really the guide, we would not assign ``I made pots
yesterday'' to the domain of the imperfect.
> But who am I to know? Please don't demand that I write a fable.
So we don't really know. We are left with having to extrapolate back. That
means that we need to weigh the liklihood of the different possibilites. In
particular, we need to cross-check the traditional view of PIE syntax with
the evidence coming from the studies of grammatization. Perhaps my
Uniformatarian bias is showing, but I refuse to believe that IE evolution
was subject to processes unkown in other languages.
>> The point is that an alternate explanation is possible: Completives have
>> a ``hot news'' value, which makes it plausible to see them develop into
>> recent past. They also can develop into perfectives, Slavic being a
>> usable example There are gaps in the examples, but I find this more
>> plausible than deriving the Vedic usage out of perfecitve-imperfective
> [Jens:] But Vedic _is_ an IE language.
So IE languages evolve in a manner completely different from other
languages? I suspect that what you want to say is something else, namely
aspect for PIE is so well established that we must believe the less
probable. That brings us to
> [Jens:] With reference to PIE they must
[ie, verbs such as *ghenti, (Vedic) ta:s.t.i etc must be imperfect (NR)]
> in case such was the system - and that it was is very well established.
Could please point to me some sources where Szeremenyi's criticisms are
answered. In particular, they must take in account the differences between
aorist-imperfect distinction and the perfective-imperfective distinction of
Russian etc, and explain why the latter is so well established for PIE when
aorist-imperfect distinction worked the other way even in Slavic. I cannot
take seriously any argument that ignores such a basic point. To be blunt, it
is not even clear that those arguing for aspect in PIE are even aware of the
difference, even though it has been pointed out repeatedly, starting from
Jespersen at least.
>> How do you explain that it is the so-called imperfect that is the tense
>> of narration in Vedic?
> Sorry to take so long in coming to the point, this _is_ a very important
> question which deserves more attention than is perhaps mostly accorded to
> it. Now, under no circumstances can we completely divorce the imperfect
> from the present, for they are formed from the same stem - that must mean
> _something_. And I do not think we can disregard the situation- changing
> effect of the aorist stem which turns up in all corners of IE.
First the point that the imperfect and present are formed from the same
stem: This means nothing in those languages which lack aspect, or what seems
to be the more common situation, languages that have only a
progressive-present/nonpast-past system, especially those in which the
present is denoted by zero. An example in the sample used by Bybee et al is
Bari. In such cases, the present is imperfective if anything, covering
habitual and generic. But the past is formed by adding a prefix ( in some
other languages, a particle). Is this so different from PIE?
I am not sure if `situation changing' is same as `moving the narrative
forward' or not. If they are not the same, then we are in partial agreement.
The so-called imperfect can still narrate while the aorist signals an
high-lighted action or a juncture. This is not too different from
past/perfective vs completive. Completives can be situation changing, and
when they become generalized halfway (ie, with the older past/perfective
still being in majority in narration), they come to be seen as providing
sutiable end points for segments of narration. But the older past/perfective
still drives narratives forward. Panjabi uses the older construction based
on the PPP often enough to be classified with the perfectives in Dahl's
study. Yet it uses the compounded verbs (ie, used with a ``vector verb'') in
the most protptypical perfective situations (especially from the Slavic
viewpoint). Which has situation changing effect, both or just the latter?
> So, if the present stem is situation-preserving, and the aorist stem
> situation-changing, how do we explain the Vedic facts? They do not look so
> odd to me: The present is also a narrative form, namely to report what is
> going on: "The horse is turning at the corner, it's emerging in full
> sunlight, and is now approaching the finish line" - this would all be in
> the present indicative in Vedic I guess. If the corresponding past
> narrative is the imperfect, that could simply be due to the status of this
> category as the past of the present stem.
But languages with a perfective do not do that. As soon as the narrative is
in the past, they use the perfective.
> If instead of perfective past you read concluding past which is just a
> natural further development,
If it is so natural, we should be able to find several such examples, at
least as often as the reverse. What we generally find is that forms start
from a restricted ``core'' category to more generalized situations. When
older forms lose ground, they become restricted to the ``periphery'' which
can cover varied ground. In particular, there are several examples of an
auxillary meaning finish becoming the marker of `perfect' or perfective.
[They are not the same: ``I read War and Peace yesterday'' is not the same
as ``I finished reading War and Peace yesterday''.] How many examples are
there of the development you are proposing, of perfectives becoming
restricted to the function of emphasizing the attainment of result?
> The IE imperfect is not just Greek. The Slavic imperfect, which mostly
> translates the Gk. ipf. in OCS, is an almost direct continuation of the IE
> ipf. (in Baltic it has become a preterite pure and simple due to the loss
> of the aorist). The Armenian ipf. has endings in -i- from *-e:- stemming
> from the verb 'be' (all thematic verbs rhyme with 'be' in Arm.) which
> formed *e:st from *e-H1es-t with the augment. The Toch. ipf. is basically
> the optative, but there are some long-vowel imperfects which in my view
> simply copy the old relation *es-/*e:s- of 'be'. Old Irish no-bered 'was
> carrying' is from the middle-voice ipf. *bhereto, notably always
> compounded (if only by the default preverb no-) and so rather obviously
> continuing an augmented form. No matter what one thinks of Lat. ama:bam it
> does contain the same span as Oscan fufans and so adds the same preterital
> marker to the present stem as the latter had added to the perfect stem;
> and ama:ver-a:-s has preteritalized the perfect stem just as er-a:-s has
> the present stem, so here, too, the ipf. is the preterite of the present
> stem. Even Albanian ish or ishte (the C,amian forms) may artlessly reflect
> *est (in part with productive superimposed ending, probably borrowed from
> the aor. qe), i.e. the present stem with secondary ending.
Armenian contains extra material and it is not clear that this was
functionless; the imperfectivity can come from the stative auxillary (<
*eHes-). Synchronically Latin proves nothing because the perfect is the
past, so the imperfect is not present stem plus the past. Diachronically,
era:- contains an auxillary (a stative one at that) and it seems that -ba:-
does as well. Again it is not obvious that the auxillaries were functionless
from the beginning, that is that they were added only at the stage where
other ways of referring to the past [and there must have been since the
perfect and pluperfect originally denoted states] had been eliminated.
Without that, we don't know that the imperfectivity is due to the present
stem and not due to the use of auxillaries.
That leaves Tocharian and Celtic. These do deserve a closer look. But their
typical imperfectives are formed differently We need to weigh this against
the fact that Vedic has several root presents for which imperfectivity of
the past is problematic, and of which you said
> I find no problem with the existence of individual verbs that act in
> individual ways that have to be entered in the lexicon. That sound fairly
Note also that this is the mirror image of the problem of eber in Armenian,
and some aorists in Slavic that look like they are from PIE ``imperfect''
(according to Szemernyi).
Given that Tocharian and Irish texts are from a much latter time from Vedic
or Hittie, how do we know that the former preserve the archaic state and not
the latter. Please don't say that ``we know that is how it was'' because
that is what requires proof.
There is also the long vowel of the Tocharian imperfect: If it is analogical
as you seem to be suggesting, or was based on some verbal noun (Kortlandt),
we cannot directly compare it to the Greek-Indo-Iranian imperfect.
>> Do we really understand the variety of syntactic structures and their
>> diachrony that well? I have mentioned the Tamil -vidu construction a few
>> times. You will find some linguists call that a perfective and the Tamil
>> simple past an imperfective. This is simply wrong as the simple past is
>> and has been the tense of narration for the 2000+ year recorded
>> history, and this distinction is nothing like the
>> perfective-imperfective distintion in Russian or Arabic [...].
> Maybe we have something as important as the explanation of the Indic
> development here. Perhaps Anatolian and Iranian have been influenced by
> some common source which could not distinguish different types of
> synthetic preterites?
Good old substratum explanation to the rescue. But what substratum? Semitic,
a natural candidate if you really menat Iranian, had aspect by 1700 BCE. If
you meant Indian here, it should be noted that the use of auxillaries in
Tamil is post 2nd c. CE. There are reasons to suspect that at an earlier
stage, which had just two forms, the opposition was aspectual and not based
on tense. Since the only forms common to all Dravidian languages are these
two forms, we are in the same situation.
> As long as our mistakes are comparable to calling the expression of
> imperfective action "imperfect" and the expresiion of perfective action
> "perfect" in a language that really has a perfective-imperfective
> distinction, I see little cause for alarm.
BTW, emperical approaches (based on `prototypical' uses) have little
difficulty distinguishing between perfect and perfective.
This could be considered harmless till someone uses the names alone to
decide what the syntax should be. In particular, we really don't know how
the ``imperfect'' was used in PIE. All we can do is use grammatization
studies to evaluate probabilities of different developments that can give
what we observe in the extant texts. This is of course what we do for
phonology and morphology and all I am saying that the same goes for syntax.
More information about the Indo-european