phonetic resemblances

[Larry Trask]

----------------------------Original message----------------------------
On Sun, 24 Jan 1999, H. Mark Hubey wrote:

[on phonetic resemblances]

> I always wonder what people by "phonetic resemblance" because sometimes
> things like this don't make sense to me. How is phonetic resemblance
> measured?

In most cases, it is not.  As I have pointed out on various lists,
phonetic resemblances play no part in standard comparative linguistics,
which is based entirely upon patterns.  Therefore it is pointless to try
to define the notion `phonetic resemblance'.  Why should we want to
define a concept we don't use and can't see any value in?

The exception lies in certain statistical approaches.  Some (but not
all) statistical approaches *do* make use of phonetic similarities.
Therefore, the proponents of such approaches are obliged to provide
rigorous definitions of what will be counted as a phonetic resemblance
-- and also, of course, as a semantic match.  Responsible workers do
this, so that readers may have a chance to evaluate their work.

The difficulty with much of the work called `multilateral comparison' is
that it depends crucially upon the identification of both phonetic and
semantic resemblances, and yet no criteria are ever advanced for
identifying these.  Or, if criteria *are* advanced, these are
subsequently ignored.

> Resemblance/Similarity is related to distance/difference. We
> can always measure or mentally use a scale from 0 to 1 in this case.
> For example, if the two words are identical the distance between them
> (i.e. difference) is zero. That means resemblance is 1 meaning maximum
> resemblance.

Oh, no.  Of course, the case of identity is trivial.  But, as soon as we
start looking at non-identical forms, the problem rapidly becomes far
more difficult.  There is a Website which lists the number-names from
one to twn in over 2400 languages around the globe.  It's here:

http://www.tezcat.com/~markrose/numbers.shtml

Take a look at the names for, say, `five' in a collection of languages.
The semantics is rigidly controlled, so we can forget that.  Now try to
decide which words for `five' are phonetically similar and to what
degree.  I promise you, you will not find it easy to come up with a
system that satisfies anybody but you.

> I fail to see how regular sound change can fail to create a phonetic
> resemblance because they are functions of each other.

Not really.

First, even regular changes can accumulate in layers of such depth that
their combined effect produces an output in which the original
regularity is difficult, or even impossible, to discern.

Second, not all sound changes are regular.

Third, sound change is not the only kind of change operating: lexical
replacement, semantic shift, grammaticalization, and other changes are
cutting across the language all the time.  So, even if an ancestral word
had undergone a series of perfectly regular changes, there is no
guarantee that the word will still be a word, that it will still have
its original meaning (or *any* recognizable meaning), or that it will
still be in the language at all.

> It is distressing to find someone who has studied chemistry or chemical
> engineering to fail to make clear what exactly he is opposing.

Would an ostensive definition do?  ;-)


Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk