Deconstructing Affectedness:  A Hierarchical Approach

In this paper I discuss the notion of affectedness, intuitively
defined as when a participant in an event comes to change state in
some way (through a change in some property, its position, its
existence, etc.).  Affectedness has played a role in a range of
sometimes disparate strands of literature, including argument
realization (Fillmore 1968, S. Anderson 1971, 1977, Jackendoff 1990),
transitivity (Hopper and Thompson 1980, Tsunoda 1981, 1985, Testelec
1998, Naess 2003), aspect (Tenny 1992, 1994), and NP-preposing
(M. Anderson 1979, Jaeggli 1986), among others. Yet these various
strands of work often posit sometimes conflicting definitions for what
affectedness is and its role in the given phenomenon.  I begin by
reviewing some of this previous work and showing that these
conflicting results demonstrate that (a) affectedness is not a single
category but instead should be broken down into various types (or
``degrees'') and (b) it is not reducible to aspect, as is often
suggested (cf. Tenny 1992, Cornips and Hulk 1999), despite being
intimately tied to it.

I then briefly sketch an analysis of affectedness developed in Beavers
(2006, 2007) as a ternary relationship between a theme, an event, and
a scale of change that defines the change undergone by the participant
(following Krifka 1998, Hay et al. 1999), in which the scale mediates
between the theme and the event in a way that correctly predicts a
disjoint correlation between affectedness and aspect.  I define
several different degrees of affectedness in terms of how specific a
predicate can be regarding the progress of the theme along the scale
of change (following Beavers 2006).  The degrees of affectedness are
related to one another implicationally in a way that form an
Affectedness Hierarchy.  I show that many of the constraints relevant
for various linguistic phenomena discussed above (and other phenomena
as well) can be stated quite simply as constraints along the
Affectedness Hierarchy, thus tying a range of phenomena together under
a single theoretical construct.  I conclude by suggesting that many of
the phenomena that have been tied to affectedness actually demonstrate
gradient effects and implicational relationships between predicate
classes, something that only becomes obvious on the hierarchy-based
analysis suggested here.