Íslenskt mál og almenn málfræði - 01.01.1980, Page 23

A Note on Local Ordering
21
out certain grammars that arise as a consequence of that principle on
the grounds that these grammars are complex or unlearnable.
Therefore, a fundamental difference between the two theories is that
one (linear ordering) severely penalizes grammars that require repeated
upplications of formally identical rules, while the other places a
premium upon such grammars. Analogous remarks can be made with
regard to the other types of „ordering paradox“ that local ordering is
designed to handle. Thus we develop a set of criteria for deciding
between the two theories. The linear ordering theory is supported to
the extent that irreducible cases of ordering paradoxes are rare or un-
attested. The local ordering theory is supported to the extent that
ordering paradoxes are commonplace in the languages of the world.
On the basis of available evidence it is clearly the linear ordering theory
that makes the correct predictions in this respect.2
3.2
A second arena within which the two theories can be compared is
that of historical change. A theory incorporating linear ordering ex-
plains the occurrence of rule duplication by the assumption that the
repeated rules entered the grammar twice, at different points in the
historical development of the language. It predicts, moreover, that a
given rule will not spontaneously proliferate into two or more formally
identical rules applying recurrently in derivations (or in different orders
ln different derivations), since it has no means of explaining such a
development. A theory of local ordering gives rise to rather different
expectations. For example, Anderson states: „just as change can result
m the deletion of some feature specification from the statement of a
rule, so it can also result in the deletion of an ordering statement from
the grammar. When that happens, a previously unpredictable ordering
is resolved in favor of a natural ordering“ (1974:147). Such a situation
can be exemplified in the hypothetical language discussed under (3) and
(4) above. Suppose we found that in an earlier stage of the language,
rule A was explicitly ordered before rule B, giving rise to altemations
such as kanamana + t — kanama, as shown in (5):
The most complex phonological system known, that of Classical Sanskrit,
presents no evidence for „ordering paradoxes" in spite of the fact that such para-
doxes were apparently countenanced by Pánini’s views on rule interaction. I am
indebted to Paul Kiparsky for this observation.