Íslenskt mál og almenn málfræði - 01.01.1980, Page 18

16 George N. Clements
hence strong evidence for alternative theories which provide for such
types of rule interaction.
In an intuitive sense, local ordering appears to involve a relaxation
of some of the constraints placed upon possible types of rule interaction
under the linear ordering theory. For this reason it might be concluded
that the theory of local ordering is a weaker (hence, less desirable)
theory than that of linear ordering, in the sense that the set of logically
possible derivational sequences compatible with the latter theory is a
proper subset of the logically possible derivational sequences compat-
ible with the former. A first aim of this note will be to show that such
a conclusion would be strictly erroneous. I will point out that the
linear ordering theory has a means for characterizing cases of „ordering
paradoxes“ of the sort described above, while on the other hand there
is a type of derivational sequence that can be characterized under the
linear ordering theory but not under the local ordering theory. Hence
it is the latter that is the more restrictive, in the formal sense.
The discussion that follows is based upon the perhaps counterfactual
assumption that there do exist language data suggesting that the linear
ordering theory is too restrictive in its usual interpretation. If this pre-
mise is incorrect, then of course any further discussion of the matter is
purely academic: the theory of local ordering is simply wrong. Some
such putative cases are discussed in Anderson 1974. While many of
these cases have been subjected to plausible reanalyses compatible with
linear ordering, a few others have not, at least in print. In particular,
Anderson cites Icelandic M-umlaut as a case which has stood up well
under further examination (Anderson 1979).
2.
The theory of local ordering is founded upon the following, crucial
assumption:
(1) For all phonological rules in a given language, if any two are
identical in form they are identical rules.
In other words, there is no rule duplication in the grammar. If the
theory recognized the possibility of rule duplication, then the type of
rule interactions that potentially support it would always be susceptible
to an altemative interpretation in which distinct, but formally identical
rules are involved. In such cases, it would not be necessary to abandon