[b-hebrew] The CV theory and the Biased Sample Logical Fallacy

Yitzhak Sapir yitzhaksapir at gmail.com
Wed Nov 29 17:46:51 EST 2006

On 11/27/06, K Randolph wrote:

> > For example, as regarding the CV syllables,...
> >
> Two problems with your argument:
[... Skipping #1, which was name calling ... ]
> 2) Your counter example is a biased sample logical fallacy.

Now, this is a misapplication of "biased sample logical fallacy."  In fact, the
you are the one using "biased sample logical fallacy".  Simply stated, you
have a theory.  I provided a counterexample.  It doesn't matter how biased,
how rare, my counterexample is.  If my counterexample deals with one
member of the data that your theory is supposed to address, and it shows
that the results of your theory do not hold, then this counterexample stands.
Your theory is refuted.  This is how falsification works.  You can never
prove a theory correct because, as a theory, it is general, and being general,
it addresses too many cases.  However, as a theory, it claims to address
all cases.  So if one case comes up that your theory doesn't hold, it is
refuted.  This is falsification.  This is why if you can't falsify a
theory, you've
got no theory -- then you can't prove it either right or wrong, so what good
does it do?  It is unlikely that theories will always be 100% to start
off.  It is
likely that after they are proposed, other scholars will look at the theory and
test it in various cases.  At some point, they may very well come up with
a problematic case that the theory does not address properly.  Then, the
theory is refuted, but it can possibly be modified to address that contingency.
The resultant theory is more robust as a result.

You apparently started off with a theory/assumption: bgdkpt consonants were
all pronounced hard.  I am not sure what evidence there is to that, but then,
you found yourself happening upon a problem, so you added another theory
assumption, in the way that theories address contingencies as they occur:
because bgdkpt consonants were pronounced "hard" and there is an
apparent problem in pronouncing some bgdkpt consonants together (for
example, [dt] is kind of problematic), you suggested that there was always
a vowel after every consonant.  Now, this is the biased sample logical fallacy.
Just because [dt] is problematic, doesn't mean that [kp] is. You are making
a generalization based on a sample that doesn't apply to all cases.

The reason I chose [dt] is because dt does appear to have been problematic.
For the number one ")xd", we have the feminine ")xt", not ")xdt", because the
dt were apparently assimilated together.  You find a similar phenomenon in
the "hitpael" verb form, when the first verb root is d.  (The
"hitpael" was likely
originally an "hiptael", as in the Mesha inscription).  So there is
evidence that
dt was problematic, and as a result, these consonants were assimilated.  But
just because dt is, doesn't mean all other combinations of bgdkpt are

Now, I provided an example with -tb-, from the root ktb.  What it shows is that
-tb- is not problematic when preceded and followed by a vowel.  What does
this mean?  It means that -tb- may be problematic at the start of a word or at
the end of a word.  But it can't be problematic in the middle of a word, because
Massoretic Hebrew nor any reconstruction of more ancient variants of Hebrew,
are not presumed to have had three consonant clusters in the middle of a word.
This is why a schewa under a dagesh forte is mobile.  It's why a second
schewa is mobile in a sequence of two schewas one after the other except at
the end of a word.  Because -tb- is ok when followed by a vowel and preceded
by a vowel, the only other possibilites left are when it's at the beginning of a
word or end of a word.  But at the beginning of a word, the Massoretic reading
would read a mobile schewa below the -t-, so that we have tVb.  We are left
with the choice of tb at the end of a word.  Now I'm not sure if this is really
that problematic, but first, does such a case even appear in Massoretic
Hebrew?  Because if it does not, then the entire "problem" in having tb together
is a non-issue.

So we have seen the -tb- case, and the -kp- case.  These suggested cases
appear to be ok.  You have taken the bgdkpt, and a generalization based on
some combination of letters from bgdkpt (such as [dt]), and generalized to
the whole bgdkpt scheme without grounds for doing so.  That is the biased
sample logical fallacy.

You gave two examples to support your theory: Capernaum and Rebecca.
You argued that these show that Hebrew had to have had a vowel after
every consonant.  Apparently, the Capernaum example was meant to show
that the /k/ had a vowel after it, unlike the Massoretic rendering.  But this is
a misinterpretation of the Massoretic vocalization.  As for Rebecca, this was
meant to show that b had a following vowel unlike the Hebrew [rivqɔ].  First
of all, let's point off the start that both these examples apparently
function as
one word ("Capernaum" is not "Caper Naum"), and in both these cases we
have a two consonant cluster with no intervening vowel ("rn" and "cc").  So
these two examples seem to disprove your own theory.  Furthermore, if you
bring in Rebecca, why not Abram with "br"?  More likely, Rebecca is an
example of a sound change whereby an unstressed short vowel dropped
when preceded by a short vowel.  One effect of this in Massoretic Hebrew
does not have two mobile schewas one after the other.  This sound change
is also apparently responsible for the loss of the first a in *uxafar ("and the
village") which became [uxfar].  It might also be responsible for other things
like the change of the predicate *qotilat to *qotilt and from this to *qotɛlɛt,
which was discussed a little while ago.  When the vowel dropped in
Rebecca so did the syllable-closing /c/, because the three consonant
cluster could no longer be maintained.  Thus, we got "Rebca".  The first
epsilon probably was used in Greek in this case for a short [i] in the Hebrew,
and the b in place of the Hebrew v, so we really had Rivca.  The final a
underwent an a: > ɔ change, which gives us [rivqɔ], the Massoretic

In short, both your examples, Rebecca and Capernaum can be explained and
found consistent with the Massoretic.  Both of these are internally inconsistent
with the CV theory you proposed (the "cc" in Rebecca, and the "rn" in
Capernaum).  These are examples that are relatively biased, as you did not
include examples such as Abram.  The consideration that led you to choose
this CV theory (the bgdkpt problematic pronunciation problem), was a "biased
sample logical fallacy" since most "problematic" cases are not really that
problematic, and the really problematic ones (such as [dt]) do have an effect
on Biblical Hebrew.  In fact, the effects that we see from such combinations as
[dt] suggests that there was no intervening vowel!

It was sufficient for me to show the problems in the [kɔ:tbu:] example (that it
is not problematic to pronounce), in order to require you to rethink
and "modify"
your theory as theories normally do when a problematic case pops up.  I have
now, however, shown how all these recent examples that you suggested are
not consistent with your theory and are perfectly consistent with the
Massoretic and a non-CV structure of Hebrew.  I have also given other
examples (the [dt] issue and the Abram spelling) that your theory needs to
address.  If you want to maintain your CV theory, you really need to address
these issues.  This is how falsification works.  It is not "biased
sample logical
fallacy" and claiming it is means you misunderstand  what "biased sample
logical fallacy" is all about.

Yitzhak Sapir

More information about the b-hebrew mailing list