? "Vocabulary Density"

[Dr. John E. McLaughlin and Michelle R. Sutton]

[ moderator re-formatted ]

Concerning semantic space in reconstructed languages, there is a fundamental
problem.  Just as we cannot reconstruct the exact phonetic nature of any
segment in a proto-language, we cannot possibly reconstruct the exact semantic
range of any given form in a proto-language.  Given the modern phonetic
correspondence set of [e], [i], [e], [e], [i], [I], and [ae], we might conclude
that the proto-segment was probably a mid-front vowel, but be couldn't get any
more precise than that.  Now let's look at a typical Indo-European type of
semantic set (I'm making this set up, but it's not beyond the realm of
possibility based on what I've seen in Proto-Indo-European, Proto-Uto-Aztecan,
Proto-Utian, or any one of a dozen other reconstructions I've looked at).
"run, escape"; "go, run away"; "go away"; "run (of a horse)"; "river flows";
etc.  Now, given prior knowledge that the sound correspondences match what we
already know about the proto-language, what is the exact semantic range of the
proto-form?  It could be a very broad form entailing all these meanings from
which the daughters narrowed the meaning after adding other forms from
elsewhere for the broad form (e.g., 'pig' in OE meant both the live and the
dead animal, but now just means the dead one after the introduction of 'pork').
It could also be that the proto-form was restricted in meaning to, say, "river
flows" and that the other daughters have broadened the meaning to include other
notions of rapid movement away from the speaker (e.g., "ship" expanding from
just a noun for a vessel on the sea to now meaning either a vessel on the sea
or in space as well as the act of moving goods either by sea, truck or train).

I agree with Lloyd's assertion that a strict one-to-one semantic match doesn't
necessarily reflect reality, but the problem in establishing a genetic
relationship between two languages is that once a little semantic leeway is
allowed, it becomes far too easy for that semantic leeway to allow us to match
anything to anything.  Last year sometime I spent an hour in the k section of
my Shoshoni dictionaries and matched about a dozen words between Shoshoni /k/
and English /k/ with plausible semantic leeway.  Two or three of those words
even had a two-consonant match (like kaan 'rat' and con 'deceptive criminal').
With the possibility of chance resemblances high enough as it is (especially
when matching t, p, k, s, m, and n), we can't let our imaginations get carried
away by allowing semantic leeway as well.  Once firm phonological
correspondences have been established with regular shift identified, then we
have a scientific check on the plausibility of any given semantic match.
Without the prior establishment of a genetic relationship with firm sound
correspondences, however, semantic leeway must be avoided.
--
John E. McLaughlin
Assistant Professor

Program Director
Utah State University On-Line Linguistics
http://english.usu.edu/lingnet

English Department
3200 Old Main Hill
Utah State University
Logan, UT  84322-3200

(435) 797-2738 (voice)
(435) 797-3797 (fax)