A Less Informative Approach to Argument/Oblique Alternations

Typical analyses of Argument/oblique alternations in English (Levin,
2003) have tended to focus on differences of specific thematic
properties, such as affectedness for the "spray/load" and conative
alternations in (1) and (2) (Dowty, 1991) and intention of possession
for dative shift in (3) (Levin and Rappoport Hovav, 2002).

1. a. John sprayed the paint on the wall. (paint totally affected;
      wall underspecified)
   b. John sprayed the wall with the paint. (wall totally affected;
      paint underspecified)
2. a. John slashed the canvas. (canvas affected)
   b. John slashed at the canvas. (canvas underspecified for affectedness)
3. a. John mailed Mary a letter. (Intention of possession)
   b. John mailed a letter to Mary. (Intention of possession underspecified)

Furthermore, some alternations (e.g. dative shift) have been shown to
be sensitive to animacy, definiteness, heaviness, and information
structural properties (see e.g. Wasow, 2002), which often don't have
repercussions in the thematic role of the participants.  However,
there has been no unified proposal of the nature of the various
thematic differences observed in many alternations, and certainly no
proposal that is compatible with non-thematic factors like heaviness.

I propose that one unifying thematic relation exists between argument
and oblique alternants: oblique markers encode a subset of the
thematic role entailments that direct arguments encode.  For instance,
in both alternants in (1) the direct object is necessarily
holistically (totally) affected, i.e. the paint is all used up in
(1a), and the wall totally covered in (1b).  Crucially, the oblique
participants are underspecified for holistic affectedness, as in (4).

4. a. John sprayed the paint on the wall, covering the entire wall
      with a fresh new coat/but there wasn't enough to finish the job.
   b. John sprayed the wall with the paint, using every last
      ounce/saving enough for the ceiling.

Adopting Dowty's (1991) theory of proto-roles, I take a proto-role
entailment to an n-ary relation entailed by a lexical predicate of one
of its arguments, where the n-ary relation holds between the argument,
the (neo-Davidsonian) event argument, and possibly other arguments.
Proto-roles are sets of proto-role entailments.  Following Primus's
(1999) extension to Dowty's approach, I assume that verbs encode
proto-Agent, proto-Patient, and proto-Recipient roles respectively
into the subject, direct object, and indirect object positions, each
role being a specific set of proto-role entailments.  I further assume
that the obliques in the alternations in (1)-(3) are encoding subsets
of these three roles, i.e., obliques are monotonically less
informative thematic role markers.  For instance the direct objects in
(1) are known to be canonical proto-Patients and thus are holistically
affected by the agent, but the obliques "with/on" encode all the same
entailments as canonical proto-Patient except for the holistic
affectedness, thus encoding only a subset of the proto-Patient
entailments.  Similar subset relations hold for the remaining
alternations.  Functionally, the uses of the alternations in (1)-(3)
can be viewed on this approach as a way of signaling that more or less
entailments are known about each participant.

This approach has two advantages: (1) it generalizes over previous
work on argument/oblique alternations by giving a uniform, universal
structure to the relationship of direct arguments and obliques and (2)
since obliques encode proto-roles that are subsets of the proto-roles
encoded by direct arguments, obliques are compatible with the stronger
sets of entailments encoded by direct argument positions, so that
alternations may also be triggered by non-thematic operations like
heavy-NP shift without requiring a difference in thematic roles.

References

Dowty, David.  1991.  "Thematic proto-roles and argument selection".
Language 67(3):547-619.

Levin, Beth. 1993. English Verb Classes and Alternations:  A
Preliminary Investigation.  University of Chicago Press, Chicago, IL.

Levin, Beth and Malka Rappoport Hovav. 2002.  "What alternates in the
dative alternations?"  Presented at the 2002 Conference on Role and
Reference Grammar:  New Topics in Functional Linguistics:  The
Cognitive and Discoursive Dimensions of Morphology, Syntax, and
Semantics, Universidad de la Rioja, Longrono, Spain, July 27-28.

Primus, Beatrice. 1991.  Case and Thematic Roles:  Ergative,
Accusative, and Active.  Tuebingen:  Max Niewmeyer Verlag.

Wasow, Thomas. 2002.  Postverbal Behavior.  CSLI Publications,
Stanford, CA.