the Trask-Hubey debate

[Robert R. Ratcliffe]

----------------------------Original message----------------------------
At the risk of getting caught in the crossfire, I'd like to interject
myself in the debate between Larry Trask and Mark Hubey. I'm in
agreement with about 90% of what LT has to say, and have to
say (with no malice intended) that MH's postings reveal a great
ignorance of the field of linguistics. Nonetheless  I do think that MH
is asking some legitimate questions and deserves a better response from
the profession than the old refrain that the esoteric knowledge is only
revealed to those who have joined the secret brotherhood.
 
re the following sequence in particular:
 
H. M. Hubey wrote:
 
> So far I have found no book on historical linguistics
> in which there is a clear algorithm for reconstructing protoforms. And
>
> i have asked many people about this. They don't know either. Sometimes
>
> they can justify it easily sometimes it is a guess and sometimes it is
> no more than a concensus.
 
MH again:
 
> > On the other list and other lists when I ask "experts" to explain
> > what rules are used to construct protoforms and why they can't be
> > found in textbooks, I notice that there is a lot of hemming and
> > hawing.
 
LT's response:
 
> That's because there are no "rules", in the mechanical or algorithmic
> sense of the term.  Performing good reconstructions requires both
> knowledge of the languages and experience of the craft.  Asking a
> linguist how to do reconstruction is not like asking a mathematician
> how
> to solve a differential equation.  It's more like asking a
> professional
> cyclist how to ride a bike.  (Not a good analogy, but the best I can
> do
> off the top of my head.)  Your experts are hemming and hawing because
> they can't find any simple way of explaining the procedure to a
> novice,
> not because they don't know how to do it.
 
My view:
 
That "reconstruction" is (or even could be) a matter of algorithm, or
rule is a widely held misunderstanding outside the field, no doubt
supported by a misinterpretation of the technical term "reconstruction"
in its ordinary sense. We cannot literally "reconstruct" anything. A
"reconstruction" is a hypothesis, neither more nor less. Like any
hypothesis in any field of science we arrive at it by guesswork, by
intuition, by imagination, by accident.  There is no path of deductive
reasoning, no "discovery procedure", no algorithm which leads from the
data to the hypothesis.  The rigor in historical linguistics as in the
natural sciences is not in the way in which hypotheses are reached but
in the way in which the hypotheses are submitted to the test of the
data. A good reconstruction should be testable in principle-- it should
be a specific hypothetical prediction about the way a particular
language was spoken by a particular group of people at a particular
point in time. A reconstruction is virtually never (directly) testable
in practice-- it is so only on those extremely rare occasions when new
texts of ancient languages are unearthed. How do we test it then?-- By
implication. Each reconstruction (of a proto-phoneme for example) has
implications for the whole system of the proto language (the whole
phonological system, eg), for the development path leading from the
proto-language to the attested languages (the sequence of sound changes,
eg), and for the forms of the reflexes in the descendant languages.
Only the last is directly observable, of course, and only this real data
can be used to rule out a proposed reconstruction absolutely.  The
proposed development path implied by the reconstruction cannot be tested
as right or wrong but only as plausible or implausible based on the
statistical frequency of observed patterns of change. Simlarly the
implications which a reconstruction has for the whole system can only be
tested under the assumption of the "uniformitarian principle" - the
assumption that prehistoric languages were not fundamentally different
in kind from attested languages and hence should not show structural
anomalies of a type not found in attested languages (languages shouldn't
be reconstructed with  no vowels, or all verbs, for example).
Probability and statistics is involved in all of this, and much could be
gained from making it more explicit.  Much could also gained if
historical linguists made an effort to base assumptions about plausible
directions of changes on an explicit, accesible, statistically
analyzable body of evidence of changes, rather than on implicit personal
knowledge.
 
 
+++++++++++++++++++++++++++++++++++++++++++++++++
Robert R. Ratcliffe
Senior Lecturer, Arabic and Linguistics,
Dept. of Linguistics and Information Science
Tokyo University of Foreign Studies
Nishigahara 4-51-21, Kita-ku
Tokyo 114 Japan