WEQATAL vs. QATAL statistics

Rolf Furuli furuli at online.no
Mon Aug 14 03:51:09 EDT 2000


>> From: Rolf Furuli <furuli at online.no>
>
>> philosophically I subscribe to Karl Popper's system with the
>> followng two maxims:
>>  1) A good scientific hypothesis is one that can be falsified.
>>  2) By help of induction or deduction we can hope to falsify a
>>     theory but we can never verify (prove) it.
>
>> I have given a clearcut and unambiguous definition of tense and
>> aspect (in the English system).  Then I have applied the statistics
>> of QATAL and WEQATAL against this definition, and I have concluded
>> that QATAL and WEQATAL are neither tenses nor aspects (in the
>> English sense).  Those who formed the four-component model - from
>> the middle ages and up to the present - did not follow Popper's
>> principles.  In fact, they did the very opposite, that is, the gave
>> positive meanings to verbal forms on the basis of statistics! And
>> this statistics was not even a good one, because nobody bothered to
>> study *all* the verbs of the Tanach, so the statistics was only
>> partial.  The resoning in the Middle Ages (on the basis of the
>> tenses of Mishna-Hebrew) was that because WAYYIQTOLs tend to have
>> past reference and WEQATAL to have future reference, they represent
>> past and future tense respectively.  Later the past and future
>> tenses were substituted by the perfective and the imperfective
>> aspect although there is no one-to-one correspondence between tense
>> and aspect.  Popper would have shuddered because of such reasoning.
>> The basic error in many studies of Hebrew verbs, is that it is
>> completely rejected that the same verb form can be used both for
>> past, present, and future.  Therefore, when so many WEQATALs have
>> future reference and so many QATALs have past reference, they MUST
>> belong to two different semantic groups! But this is very bad
>> semantics indeed!
>
>Actually, given the lack of any really adequate semantic theory at
>that time, applying an inductive generalization based on the observed
>predominant meaning of these verb forms _was_ in fact a scientific
>procedure (as you yourself admit, "the natural thought we get, is that
>there most likely is some connection between WAYYIQTOL and QATAL on
>the one hand and YIQTOL and WEQATAL on the other, because the temporal
>references of each of the two are similar").  In situations where
>there is no available theory that can explain a small residue of
>problematic cases, but where a theory is available that does cover the
>great majority of cases, then it's actually quite scientific to
>provisionally adopt such a mostly-adequate theory, and to leave the
>small problematic residue aside to be explained later.  To try to
>force an explanation of the problematic residue where no satisfactory
>explanation is in fact possible, would be to engage in ad-hoc
>scholasticism; while just throwing up one's hands and refusing to
>adopt any theory at all (even provisionally) because of the small
>problematic residue would be obscurantist, and would leave no room for
>future refinement and development by building on what is known
>already.  So for example, if linguists had refused to adopt any
>provisional theory at all about the correspondence between Germanic
>and Indo-European consonants (i.e. just thrown up their hands and said
>everything was random and unexplainable) because Grimm's
>generalization had a residue of problematic exceptions, then it would
>not have been possible to discover Verner's law by examining the
>exceptions to Grimm's law.  It was in fact the scientific thing to do
>in the mid 19th century to provisionally adopt Grimm's law, but to
>admit there were exceptions -- since prior to 1876 there was no
>available theory that could explain the exceptions to Grimm's law
>(that after 1876 were explained by Verner's law), and also no
>alternative theory that could explain what Grimm's law did explain as
>well as Grimm's law did.  Similarly, when 19th-century astronomers
>discovered the precession in Mercury's orbit, they didn't suddenly
>suddenly abandon Newton's laws of gravitation, and throw up their
>hands and say that the motion of the planets was random; instead they
>kept on using Newton's laws as their provisional theory, but carefully
>observed certain phenomena which did not seem to be fully explainable
>by Newton's laws, while waiting for a theory to arise which would
>explain these phenomena by revising and extending Newton's laws (as
>the later theory of Relativity in fact did).

Your examples are well taken. We had a seminar in Oslo a year ago where the
Grimm/Verner stuff were discussed. While I try to follow Popper, I realize
that it is often difficult to meet his strict demands when we work with
language. I also accept that good results can be reached without using
theories that can be falsified. But to follow the standard whenever
possible is a good procedure.
However, from a modern linguistic point of view the work with Hebrew verbs
right into the 20th century, has been rather primitive for two reasons. To
the best of my knowledge, no published study has ever collected and
discussed the complete "small residue of problematic cases", so the number
of this "residue" has not been known. And further, no study has
systematically differentiated between "past reference" and "past tense"
(between pragmatics and semantics), but it has been taken for granted that
past reference of finite verbs means past tense.


>
>Anyway, I think it rather ill becomes you to get on an exalted
>philosophical/methodological high horse, since:
>
>1) For you the C/RT/E theory seems to be an unquestionable starting
>point, such that there do not seem to be any observable Biblical
>Hebrew verbal phenomena that could cause you to doubt the C/RT/E
>theory; so for you, the C/RT/E theory seems to be in fact
>"unfalsifiable".

I appreciate your criticisms because I know your excellent Hebrew
background. Moreover, new insight can only come after a number of failures,
and when being criticized, one is forced to rethink one's foundation. It is
true that I find Broman Olsen's linguistic model of a very high quality.
But I am at the same time aware that others think it more problematic to
draw a clearcut distinction between what is semantic and what is pragmatic.
I am always open for new information and am willing to reject my own
hypotheses if good reasons are given. The strength of the C/RT/ET model is
that it is simple, and that it can be applied to any language, even
languages where tense is lacking. There is no dispute among linguists that
this model is a very fine tool for analysing natural language.


>
>2) You seem to display a strong persisting lack of clarity as to
>phonological, morphological, diachronic, and comparative Semitic
>concepts which apply to these verb forms (i.e. whether or not
>WAYYIQTOL is a morphologically distinct conjugation -- you seem to
>hold both points of view on this simultaneously -- ; whether it is
>phonologically plausible that the stress shifts could somehow
>originate by pure "pragmatics" thrusting its way onto the phonological
>plane; whether the difference between _pathah._ + geminate vs.
>_sh at wa_ + non-geminate seen in WAYYIQTOL vs. W at YIQTOL could be a
>non-phonologically significant difference, etc. etc.): --

This point gives me the opportunity to clarify some viewpoints. I use
"conjugation" in the sense of "a major inflectional category of verbs" (see
Waltke/O´connor p 690). Thus the term in my use includes both morphology
and semantics. I expect that a difference in morphology signals a
difference in meaning. I connect "morphology" with the form of single
words, so a prefixed preposition or conjunction will make the particular
word graphically different from the same word without the prefix but will
not necessarily make it morphologically different, because we are dealing
with two words. As to WAYYIQTOL I note that the form is graphically
different from YIQTOL and WEYIQTOL, but it is an open question whether it
is morphologically different. The short forms are of course morphologically
different from the long ones. My working hypothesis is that Classical
Hebrew has only two conjugations, one prefix-conjugation (YIQTOL, WEYIQTOL
and WAYYIQTOL) and one suffix-conjugation (QATAL  and WEQATAL). I think the
difference in temporal reference and modality can be accounted for by
pragmatics (syntax, focus/topic, and lingusitic convention). Whether I am
able to give such a plausible pragmatic  explanation will be seen when my
thesis is completed.

I do not hold the view that WAYYIQTOL is a morphologically distinct
conjugation,so far I view it as a graphically different group. There are,
however,  five graphically different groups of verbs, I deal with these
groups, make statistics regarding them - without viewing them as five
semantically distinct groups, but I am open for good arguments to that
effect.

It may be that I lack clarity in the areas you mention, but I strive to
learn more, and am happy to be corrected when necessary.


>
>
>> Can the waw in the WEQATALs be explained as a simple conjunction?
>> Can we explain why a conjunction is needed where the WEQATALs occur?
>> And we must of course not forget the penultimate stress.  Although
>> this stress is not consistently used, it is done so often that there
>> can be no doubt that either those who wrote the text or some of
>> those who at some point copied it, wanted to use this stress as a
>> distinction.  But is this distinction just phonetic, or is it
>> pragmatic, or is it semantic?  In addition we should ask whether we
>> can account for the fact that so many WAYYIQTOLs and QATALs have
>> past reference and so many YIQTOLs and WEQATALs have future
>> reference, by help of the context in which they occur without taking
>> them as four different conjugation.  Then we get the following point
>> of departure: 1) QATAL and WEQATAL can be parts of one conjugation
>> or be two defferent conjugations, and 2) each Hebrew conjugation can
>> have past, present, future, and perfect reference and signal
>> modality.
>
>Etc etc.  (By the way, the term "conjugation" is a morphological term,
>and what is or is not a conjugation should be determined according to
>morphological criteria.)
>
>
>> And instead of getting the four conjugations as an end product after
>> a longer study of the verbs, the four conjugations are the starting
>> point, it is simply assumed that there *are* four conjugations with
>> different meanings.
>
>Actually, morphological conjugations must be the starting point for
>semantic investigations, since without establishing morphological
>conjugations, you don't even know what it is that you're investigating
>the semantics of.  You're implicitly following this principle when you
>collect data separately on WAYYIQTOL vs. YIQTOL, etc.  Otherwise,
>you're not using the word "conjugation" in the sense in which it is
>usually defined.


As shown above, I use "conjugation" to include both morphology and semantic
meaning. My starting-point is that in unpointed texts there are only two
clearly different groups of finite verbs, the prefix-forms and the
suffix-forms. A great number of both groups have a prefixed waw, and a
number of the verbs of the prefix-group are shorter than others. So I need
not start with more than two different groups. When I go to pointed texts,
it is possible to classify the finite verbs in five graphically different
groups. Because most students of Hebrew believe that four of the five
groups are independent conjugations with different semantic meanings, it is
natural to study each of these groups and their statistics.

>
>
>> If you don't see that the premise for Henry's feeding my data of
>> QATAL and WEQATAL into his SAS statistical program, was that QATAL
>> and WEQATAL are two distinct groups which cannot have both past,
>> present, and future reference, then I cannot see how we can continue
>> an intelligent discussion.
>
>Actually, the formal statistical procedure of calculating the
>chi-squared for the "null hypothesis" that there is no-interaction
>between the QATAL vs. WEQATAL variable and the semantic categories
>variable makes no such assumption as to whether a form "cannot have
>both past, present, and future reference" (and similarly with the
>Cramer's V measure of "effect size" or association, which is a
>proportionally scaled version of chi-squared).

My post of yesterday comments on this. I agree that your analysis per se do
not build on the assumptions I have stated, and neither do you need any
such assumptions when you point out that the temporal reference and
modality is significantly different in the groups with and without prefixed
waw. This is just observation and calculation. But when you apply your
analytic results semantically to show that QATAL and WEQATAL are two groups
with different semantic meaning, then you have to build build on the
mentioned assumptions. This is what I take issue with.



>
>For my first postings, I collapsed the semantic categories into a
>dichotomy of "PAST & PERFECT" vs. "All Other" because I had to do some
>collapsing to get your data in the form of a proper two-dimensional
>table; because I wanted to show that your numbers didn't refute a
>"traditional" view quite as soundly as you thought; and because of
>certain convenient statistical properties of 2x2 tables.  However,
>nothing in the statistical demonstration that the hypothesis of
>non-interaction between morphology and semantics must be rejected
>actually depends on this grouping of semantic categories, as seen in
>my second SAS results posting, where minimal collapsing of semantic
>categories results in a stronger semantic distinction between QATAL
>and WEQATAL (0.7794, where 1 is the maximum) than was observed with a
>collapsing into two semantic groupings (only 0.6030).
>
>
>> Because we take a neutral standpoint and keep open the possibility
>> that each conjugation can have any reference, the percentages of
>> past, present, future, and perfect reference tell us absolutely
>> nothing as to the *meaning* of QATAL and WEQATAL or whether the two
>> constitute one conjugation or two.
>
>Then why the heck did you gather statistics on the distribution of
>20009 such forms?  For my dissertation, I gathered data by examining
>9309 Hebrew words with disjunctive accents, taken from the books with
>"non-poetic" accentuation, and classifying them according to whether
>they showed pausal stress-shift blocking or non-pausal stress shift;
>however, you can be sure that I would not have gone to such effort if
>I had thought that these statistics had no relevance to the
>phonological questions I was looking at...
>
>--

I do not think that all researchers before me were idiots, and that I
follow the only correct approach. I have learned much from all these
previous efforts. True, there have been some who have tried to make the
terrain fit the map instead of the opposite. Birkeland and Blake, for
instance, both had decided that WAYYIQTOL was past tense. They discussed
the same 130 passages which seemed to have non-past meaning. The
conclusions of both were that these 130 *were* preterites as the others.
Both drew this conclusion on the basis of the context
However, according to Birkeland the context showed that they were true
preterites but according to Blake the context showed that they were true
non-preterites, but they had been wrongly pointed by the Masoretes.


When we do research, we have to do so inside a restricted area and from a
particular point of view. We also must make several assumptions. But if
most researchers allways work in about the same area and allways make the
same assumptions, there is the danger that important points are never
studied. While gaining much fine input from previous studies, I discovered
two areas which really needed close scrutiny: 1) Nobody had ever made a
study of, and a comparison of *all* the verbs in Classical Hebrew, and this
is possible today with electronic texts both of the Bible and of the DSS.
2) That Hebrew have four different semantically distinct conjugations have
been assumed by almost all researchers, and nobody had ever tried to
isolate the smallest units of the verbal system and studied which meanings
are semantic (uncancellable) and which are pragmatic (cancellable).

My first task, therefore, has been to make an analysis of all the finite
and infinite forms of Classical Hebrew, regarding temporal reference,
modality and several other syntactic factors. After getting the whole
pattern, my second task is to differentiate between pragmatics and
semantics. In the study of a dead language there are particularly two
dangerous areas: 1) Which assumptions do we make (do they really fit the
makeup of the dead language)? and 2) Are we forcing modern grammatical
models upon a language where they do not fit?

A basic part of my present project is to test the principal assumptions
behind the earlier studies of Hebrew verbs, assumptions which have been
repeated over and over again through the centuries, but never have been
tested. I also try to be careful not to sin against point 2). For instance,
in all modern studies of verbs in the Semitic languages of which I am
aware, Comrie's definition of aspect is applied in a deductive-nomologic
way (as THE definition of aspect). But nobody has designed tests to show if
this definition in its entirety can be applied to Hebrew.

The advantage of Broman Olsen's model in relation to points 1) AND 2), is
that the only assumption necessary to make, in order to use it in the study
of Hebrew verbs, is that Hebrew can be studied the same way as other
natural languages. What I have done, so far, is to compare Hebrew verbs
with the very exact definition of tense and aspect in English. The
conclusion drawn (with good marigins) is that tense and aspect (in the
English meaning) are not grammaticalized in Hebrew. After having shown what
the Hebrew verbs *are not* (something which is quite easy), I will try to
show what the verbs really *are*, and that is much more difficult.





Regards

Rolf



Rolf Furuli
University of Oslo

















More information about the b-hebrew mailing list