On 2008-07-24 23:03, gprosti wrote:
> What I'm wondering is: what empirical evidence is this gradient of
> plausibility based on? For example, the empirical evidence for the
> probability of semantic continuity ("wolf" > "wolf") is (I would
> suspect) that linguists constantly see semantic continuity over the
> history of the languages they research, suggesting a high probability
> of occurrence. I have never seen this type of justification offered for
> the probability of a given semantic change. To clarify what I mean:
> one could offer a single example of the change "wolf" > "jackal", but
> that would be one example out of hundreds or thousands of potential
> cases. To establish the likelihood of the change, it seems one would
> have to use a larger sample size than one.
To assess its _probability_ (empirically, and ex post), yes, one would.
But this branch of linguistics is a historical discipline; it often has
to deal with phenomena that don't happen often enough in replicable
conditions to be approached statistically. Semantic change, in
particular, is influenced by various extralinguistic factors dependent
on the local circumstances (whether natural of cultural). How could one
ever quantify such things? On could easily invent a formal semantic
model with a metric according to which the distance between, say, 'wolf'
and 'jackal' (both, say, [+animal, +terrestrial, +carnivore, +canine,
+bad, ...]) would be reasonably close. But what about 'wolf' and
'jarring note'? Here the connection, while real, is extremely indirect.
In fact, to understand it, one has to know some completely extraneous
facts, like the old belief (popularised by philosophers from Albertus
Magnus to Descartes) in the natural antipathy between wolves and sheep,
extending posthumously to their guts used for musical strings. And there
are many similarly accidental connections, e.g. the link between Juno
Moneta in Rome and 'money'/'mint'. The best we can offer here is an
entirely _convincing_ story, which is, however, unique, and so its
probability cannot even be defined.
Other historical sciences have similar limitations. Take evolutionary
biology. What is the probability that a given species of animal will
evolve the ability to fly within the next twenty million years? Well, it
should have some features "preadapting" it to such a career -- whales,
hippos and giraffes, for example, can be ruled out as candidates. But
even such "obvious" constraints are subjective and can be wrong. All
that we can base our reasoning on is the known precedents (just as in
historical linguistics). The required features may, however, evolve in
the future, quite unexpectedly. Looking at the ancestors of whales --
small terrestrial artiodactyls -- who would have thought?
Piotr