Íslenskt mál og almenn málfræði - 01.01.1980, Blaðsíða 22

20
George N. Clements
(4) underlying representation: kanamana
by A': kanaman
by B': kanama
by A": kanam
by B": kana
by A'": kan
by B'": ka
with an (in principle) unlimited number of rules A', A", A'", etc. In fact
it seems safe to say that a single clearly attested case of this sort would
be sufficient to force the abandonment of the linear order theory. On
the other hand, the proponent of local ordering would presumably be
delighted to find such forms, since they are just what his theory leads
him to expect. The simplest and most natural type of ordering relation-
ship between any pair of rules is determined by universal principles of
markedness, which state, for instance, that feeding order is simpler
than counterfeeding order. Such ordering relations are not specially
mentioned in the grammar; rules A and B of (3), for instance, being
mutually feeding, would in the simplest case be subject to no explicit
ordering restrictions and should therefore apply as in (4).
Now in fact few phonologists are likely to display much sympathy
for a theory of rule interaction which leads to such a prediction. In
order to explain the apparent absence of languages exhibiting deriva-
tions such as that given in (4), the proponent of local ordering would
have to invoke auxiliary principles explaining why such types of rule
interaction do not in fact occur. It might be suggested, for instance, that
if rules A and B of (3) were allowed to interact as they do in (4), they
would be extremely difficult to leam, and the underlying forms of
morphemes would be all but unrecoverable. However, the insight pre-
sumed to underly the local ordering theory is the view that some
ordering relations are less marked (more natural) than others precisely
because of their superior functional properties; for instance, it is com-
monly assumed that feeding order is superior to counterfeeding order
because surface exceptions to rules are thereby eliminated (note that
this condition is satisfied by the derivation given in (4)). However, if
one is led to postulate a certain principle on grounds of simplicity or
learnability, one should not have to adopt auxiliary principles to rule