WEQATAL vs. QATAL statistics, cleaning up assumptions, 90% regularities

Henry Churchyard churchh at usa.net
Tue Aug 29 11:19:23 EDT 2000

> Subject: Re: WEQATAL vs. QATAL statistics, cleaning up assumptions
> From: Rolf Furuli <furuli at online.no>
> Date: Sun, 20 Aug 2000 20:06:32 +0200

>> From: Randall Buth

>> qaTALti and veqatalTI are two different forms and do not need an
>> assumption in order for them to exist.  They exist.

> The forms exist graphically, that is correct, and the accentuation
> is different in a few forms (enough forms to show that the Masoretes
> distinguished between them).  Your error, in my view, is that you
> *presume* that the stress is phonemic, but this is completely
> unfounded.

Actually, the term "phonemic" has a highly specific technical meaning
within phonological theory: it refers to something which is not
predictable from the surrounding purely phonological/phonetic context.
(In classic 1940's/1950's American structuralist linguistics, things
were considered non-phonemic only if they could be redundantly
predicted from strictly surface phones and junctures in a "bi-unique"
way; but with the rise of generative phonology, an extended definition
of "non-phonemic" has come into use, in which things are also
considered non-phonemic if they can be redundantly predicted from more
abstract phonological representations, such as synchronic underlying

In analyzing Tiberian Biblical Hebrew main-stress positioning, I have
basically followed the descriptive generalization which was apparently
first proposed by Jean Cantineau in his 1931 paper "De la Place de
l'Accent de Mot en He'breu et en Arame'en Biblique" (_Bulletin d'E'tudes
Orientales de l'Institut Franc,ais de Damas_ vol. 1, pp. 81-98).  As I
have formulated it in modern phonological terms for chapter 1 of my
dissertation, this generalization states that forms which end in a
consonant in synchronic underlying phonological representations are
stressed on the underlyingly word-final syllable, while forms which end
in a long vowel in underlying phonological representations are stressed
on the underlyingly penultimate syllable.  ("Underlying" representations
are linguistic constructs posited to explain phonological alternations
between various surface forms derived from the same root or stem, as
explained in chapter 4 of my dissertation.  Here underlying phonological
representations can be defined as representations in which a number of
phonological processes which create alternations between related
morphological forms, and interfere with the clarity of the underlying
stress pattern in Hebrew -- by changing the syllable structure of a
word, the vowel- or consonant-final status of a word, or the position of
stress within a word -- have not yet applied.  Such phonological
processes include segholate epenthesis, change of feminine -at to -aa
word-finally, and wayyiqtol stress-shift.)

So what all this comes down to, is that I consider any surface
main-stress which is on the underlyingly-penultimate syllable of an
underlyingly vowel-final form, or on the underlyingly word-final
syllable of an underlyingly consonant-final form, to be
"non-phonemic"; while any main-stress placement elsewhere in a word is
by definition "phonemic" in terms of my analysis.

Therefore the word-final stress-placement in w at qaatalTII must be
considered phonemic.  However, since this stress doesn't really reveal
the morphological distinctness of all WEQATAL with respect to all
plain QATAL, but rather the morphological distinctness of _some_
WEQATAL (with much theoretical homophony which is not actually very
important in practical terms due to the infrequent occurrence of
"compositional reference" WEQATAL), it is only somewhat indirectly
relevant to the statistical analysis that was being discussed (in
which all WEQATAL, regardless of stress, was lumped together for
research convenience, as discussed in my last posting).

> Henry has correctly argued that the stress distinction between QATAL
> and WEQATAL cannot be diachronically traced back to some older
> forms.

Certainly, but this has almost no relevance to its synchronic status,
since in my analysis (and that of many others cited in the dissertation)
a contrast was created by analogy which had not previously existed in
earlier diachronic stages of Hebrew.  (I wonder why you brought it up
here, considering that I also argue that wayyiqtol phonemic stress
originated in the late 2nd. millennium B.C.E, but this reconstruction of
a long pedigree has not seemed to make you any more willing to accept
such wayyiqtol stress as morphologically relevant.)

> My suggestion is that in the chanting in the Synagogue the 'default'
> understanding ( to use your words) was that QATALs with prefixed WAW
> for the most part was used with future reference, and therefore
> these forms were marked with ultimate stress, in contrast with
> WAYYIQTOLs with penultimate stress.

In other words, a linguistic analogy took place.  (I don't know why you
drag "chanting in the synagogue" into the discussion at all, unless
you're trying to imply that the analogy was in some way a linguistically
artificial or "unnatural" one, which is not obviously the case.)  The
working-out of the analogy actually turned out to be rather complex,
being constrained by Hebrew-specific phonological considerations in
quite a number of ways (as you can read in sections 4.4.2 and 4.5 of my

> This need not be more than a *functional* view, an observational
> view, not a semantic one.

Whether or not the distinction between QATAL and WEQATAL is
semantic according to your own personal individual definition of
the word "semantic" is really up to you.  But the distinction is
certainly a morphological one -- and when your own data shows that
a proportionally-adjusted 90.7% of "suffix-conjugation" forms with
future reference are WEQATAL, 88.8% of all "suffix-conjugation"
forms with OTHER/modal reference are WEQATAL, 90.1% of all
"suffix-conjugation" forms with past reference are plain QATAL,
95.4% of all "suffix-conjugation" forms with perfect reference are
plain QATAL, and 85.1% of all "suffix-conjugation" forms with
present reference are plain QATAL, then I think that there are
rather systematic correlations here, which your theory should
really try to account for, instead of merely sweeping it under the
rug of "pragmatic function" (especially considering that a 90%
result is in fact actually quite impressive, given all the actual
and potential sources of confounding experimental "noise" involved
in the data-gathering).

I'm sure you originally regarded strict and rigid criteria of
"grammaticalization" (based on the temporal categories of the Broman
Olsen model) as liberating in their innovative simplicity, and in
cutting through a lot of the complexities and confusion of previous
theories of the Hebrew verbal system, and as definitively overthrowing
the traditional idea of "waw conversive" (which had long been
recognized as an oversimplification).  However, it now appears that
your own data shows that there are some highly significant
regularities and strong patternings in the temporal/modal reference of
Hebrew verb forms which these strict and rigid grammaticalization
criteria can't account for.  This is where what I said about
"falsifiablity" earlier comes into play -- if you never let the facts
of Hebrew call into question any aspect of your theoretical model,
then you aren't properly testing your theoretical model.  Therefore
when you ignore extremely strong correlations (such as 95.4% of all
"suffix-conjugation" forms with perfect reference being plain QATAL)
because the strict and rigid grammaticalization criteria of your
theoretical model won't let you try to explain them, it seems to me
that your theory is less than helpful in the Biblical Hebrew context
(as also when you let abstract semantics redefine what is and what is
not a distinct morphological conjugation, instead of letting
morphology determine morphology).

| Date: Mon, 28 Aug 2000 12:58:49 +0100
| From: "Peter Kirk" <Peter_Kirk at sil.org>
| Subject: Re: WEQATAL vs. QATAL statistics: a question and a doubt

| I have reached my own conclusions, that there is a clear distinction
| in temporal usage between QATAL and WEQATAL.  I think you now agree.
| If list members disagree about whether this distinction is one of
| meaning or semantics, they are now differing more over the meaning
| or semantics of the terms "meaning" and "semantics" within their
| presupposed theoretical models.


    (1-(965/(965+(13922/6087)*4100))) = 90.6695%
    (1-(397/(397+(13922/6087)*1383))) = 88.8488%
       (7450/(7450+(13922/6087)*357)) = 90.1226%
       (2605/(2605+(13922/6087)*55))  = 95.3935%
       (2505/(2505+(13922/6087)*192)) = 85.0844%

Henry Churchyard   churchh at usa.net   http://www.crossmyt.com/hc/

More information about the b-hebrew mailing list