Íslenskt mál og almenn málfræði - 01.01.1980, Side 20
18
George N. Clements
formally distinct, obligatory rules: A, B, and C. Examining a sample
range of derivations, we find that whenever the application of B creates
a representation upon which A is defined, A fails to apply. Moreover,
this failure cannot be subsumed under any independently-motivated
principles of applicational precedence. Thus, under either theory, A
must be explicitly ordered before B. Now let us consider a derivation
of the following form, in which A applies both before and after two
other rules, B and C. In the discussion, „0“ will represent an arbitrary
phonological string, „A(0)“ will represent the string resulting from the
application of A to 0, „B(A(0))“ will represent the string resulting
from the application of B to A(0), and so forth.
(2) i. input string:
0
ii. only A is defined; A applies:
A(0)
iii. only B is defined; B applies:
B(A(0))
iv. both A and C are defined; C applies:
C(B(A(0)))
i v. only A is defined; A applies:
A(C(B(A(0))))
Under the theory of linear ordering, A must be split into two rules, A'
and A". The phonology therefore contains the ordered set of rules A',
B, C, A" (in that order). Derivation (2) must therefore be recharacter-
ized as A"(C(B(A'(0)))). How will the theory of local ordering deal
with such a situation? By assumption (1), it cannot treat A as two
distinct rules. Since B creates representations upon which A is defined,
and A fails to apply to them, the ordering statement „A precedes B“
must be given. But in that case, derivation (2) is ill-formed, since A
has reapplied after B. Hence, derivations of the form of (2) are pro-
scribed under the theory of local ordering.1
1 It should be noted in passing that it is possible to interpret the statement „A
precedes B“ in two ways: (a) A may never apply in a derivation once B has applied;
(b) A may not apply to a string B(0), i.e., to the immediate output of B. Under
the second interpretation, local ordering allows derivation (2). The first interpreta-
tion is the one intended by Anderson, as shown by the discussion of Faroese ob-
struent deletion, glide formation, and dissimilation (Anderson 1974:173-4).