miscellaneous (phonemes, flaw in aktionsart, etc.)

From: Mari Broman Olsen (molsen@umiacs.umd.edu)
Date: Thu Mar 19 1998 - 11:12:52 EST

>From the linguistic perspective: phonemes may be used to change
meaning, in that words (or morphemes) are made up of phonemes, and
when you change the phonemes you change the word or morpheme. But it
is not quite right to say phonemes carry meaning. Although there are
interesting associations in onomatopoetic words, in general the
association between sound and meaning is arbitrary. In fact, in
generative linguistics since at least the mid 60s (Chomsky's Aspects
of the Theory of Syntax), it has been observed that syntax, semantics,
and phonology may be described independently, making use of
independent processes and primitives.

is HARD. I and two students undertook to
classify some 600 verbs used by young children (1-5) and adults
speaking to them. We all agreed on just under 50%. Two of us agreed
on another 40%. The process was complicated by the fact that we
needed to consider the meaning in the grossest sense before we could
classify it, with the role of object and subject in the sentences
controlled for. Here are a few examples:

appear: It appears that (state). The sun appeared suddenly
(punctiliar event)

separate: I separated the twig from the branch (arguably punctiliar)
           I separated the fighting children (arguably durative)

The issue was resolved in large part by allowing the sorts of
variation predicted by my privative model of aspect: variation is
limited to the following:
        verbs basically punctiliar may become durative, and not vice
        verbs basically stative may become events, and not vice versa
        verbs basically unbounded may become bounded, and not vice

The classification task is therefore constrained to the minimal
specification of features required to account for the variance. And
here's where I differ from Jonathan (and Edgar? Dale? with whom he
agreed), in allowing SOME aspects of Aktionsart to be affected by
context: marking of unmarked features.

WHO REALLY CARES: It still is a hard task to classify the verbs. One
might ask, in fact, why to bother. One reason is that there is a lot
of evidence that these categories constrain grammatical forms in
interesting ways across languages, and that they are, in some sense,
universal. We can demonstrate this in study of child language
acquisition. In a study of data from 8 children learning English, we
find kids undergeneralize the use of the -ing form to EVENT verbs and
the use of -ed to bounded verbs. I have a long story why that relates
to the association between boundedness and perfectivity on the one
hand and events and imperfectivity on the other (and derivatively
between present-imperfective and past-perfective (and if you can get
postscript files you can get it off my web page:
http://umiacs.umd.edu/~molsen , or I would be happy to send it)).

Similar results are shown in the literature ('properly' reanalyzed) on
acquisition of Italian, French, Greek (modern), Chinese, Polish and
others. This is intended to be a shameless plug for looking at
linguistics when considering the relevant categories for analysis of a
specific language, Greek in this case, even if there is not a single
Greek word in this post ;-).

Mari Broman Olsen
Research Associate

University of Maryland Institute for Advanced Computer Studies
3141 A.V. Williams Building
University of Maryland
College Park, MD 20742

PHONE: (301) 405-6754 FAX: (301) 314-9658
WEB: http://www.umiacs.umd.edu/~molsen

This archive was generated by hypermail 2.1.4 : Sat Apr 20 2002 - 15:39:14 EDT