|Title:||Output-Output Correspondence in Optimality Theory|
|Authors:||Mark Hale, Madelyn Kissock, Charles Reiss|
|Comment:||14pp. To appear in Proceedings of WCCFL XVI|
|Abstract:|| Output-output correspondence (OOC) constraints,
constraints which demand correspondence between
independently occurring surface forms, have recently
been added to the set of constraint types invoked in
much OT work. In this paper we examine some of the
better-known arguments originally adduced in support
of OOC constraints, and argue that adoption of such a
powerful mechanism is not justified, at least in the
This paper deals specifically with the following:
the incomplete/complete phase distinction of Rotuman,
as analyzed by McCarthy (1995); English truncated
hypocoristics, as discussed by Benua (1995); and the
treatment of Base Identity and Uniform Exponence in
Three criticisms are leveled at the OOC-based
analyses cited above. First, we find cases of
'opportunism.' For example, there is an unprincipled
culling of the data and an unprincipled choice of bases
in correspondence relations. Second, there is
misanalysis, in that clearly significant generalizations
are overlooked, technical aspects of the theory are
improperly treated and implausible generalizations are
accepted. Third, we believe that the analyses based on
OOC lead to problematic predictions, some of which are
strongly contraindicated by existing data, and others of
which we consider highly suspect. We offer simple,
principled solutions which we hope will contribute to a
more constrained theory of phonology--one that perhaps
has no place for OOC.
The main example we discuss is McCarthy's (1995)
study of Rotuman metathesis--a cover term for morphemic
alternations of the type in (1) termed by Churchward (1940)
McCarthy followed, in its basic outlines, the original
analysis of Churchward (1940) in attributing the synchronic
conditions on metathesis to "syntactico-semantic principles".
However, a detailed study of the conditions on incomplete
phase formation in Rotuman reveals that "syntactico-semantic
principles" are not involved in these alternations. Instead,
we show that the phases are PHONOLOGICALLY conditioned.
Producing the phases via output-output correspondence fails
to capture the correct, purely phonological conditioning of
the phases. Such an analysis also ignores an additional
alternation in Rotuman roots known as 'broad' and 'narrow'
versions. We show that there is no principled way of selecting
a base for output-output correspondence in Rotuman if one wants
to account for both the incomplete/complete and broad/narrow
In the conclusion to a later paper, Kenstowicz (1995:433)
raises some fundamental questions regarding the use of OOC.
"Can [OOC] be restricted to situations in which one structure
is a substring of the other? Or should we allow identity
constraints to hold among a family of related words, e.g. to
get the effects of paradigm leveling?" Kenstowicz goes on to
note the vagueness of terms like ' family of related words'
and 'isolation form'. Clearly, these terms need to be defined
in order to select a base against which identity can be evaluated.
As far as we can tell, these fundamental questions have yet to
receive a satisfactory solution in the literature. The failings
of the specific cases of OOC which we discuss, those of McCarthy,
Benua and Kenstowicz, are related to the absence of clear guiding
principles concerning these fundamental questions. If phonological
theory wishes to be constrained by standards of explicitness and
rigor, OOC should be eschewed until these fundamental questions
receive a more satisfactory treatment.