Gesture Resolution and Definiteness

According to a proposal by Lascarides & Stone (2006, 2009) co-speech gestures, gestures
co-occurring with simultaneous speech, can be analyzed by a semantic formalism in the
framework of SDRT (Segmented Discourse Representation Theory; Asher & Lascarides,
2003). Gestures are analyzed by formulas which relate the depicted content to the spoken
one, but several gestures may also be connected with each other. One example they
provide for gestures building on other gestures is the discourse in (1):

(1) a. and [took his] [HATCHET and
First: Speaker’s right hand grasps left hand, with wrists bent.
Second: Speaker lifts poised hands above right shoulder.

b. with] [a mighty SWEEP]
First: Hands, held above right shoulder, move back then forward slightly.

c. [(pause 0.4 sec)] [SLICED the wolf ’s stomach open]
First: Speaker turns head
Second: Arms swing in time with sliced; then are held horizontally at the left

(cited as in Lascarides & Stone, 2009, p. 400; originally from Kendon, 2004,
p. 136)

As Lascarides & Stone (2009, pp. 425-427) analyze the discourse, the gestures co-
occurring with the bracketed speech segments in (1) described by First: and Second: are
connected by rhetorical relations. The crucial observation is that certain characteristics
of one gesture are picked up by the following gesture: The imaginary object the hands
are taking in the first gesture in (1a) is the handle of the hatchet, and it is still the same
handle of the same hatchet in all subsequent gestures which is depicted to be held by the
two hands of the speaker.

One of the concerns of Lascarides & Stone (2009) is therefore the proper way to
formalize when properties of similar looking gestures (e.g. holding an imaginary object
with both hands) can be interpreted in a coherent or even uniform way. One other is how
to formalize constraints on the coherence of gestures with spoken content. Both points
are addressed by the usage of discourse referents, semantic representatives which license
the usage of pronouns. The main point of the analysis is therefore formalizing gestures
in a way that allows for anaphoric dependencies, the relationship found with pronouns
and many instances of definite descriptions. Definite descriptions are taken here to be
phrases of the form the NP or NPs preceded by a possessive determiner. The properties
of pronouns and definite descriptions are collapsed by the term definiteness in some
accounts (e.g. Kadmon, 1990). The term definiteness as used in the title of this thesis is
intended in this sense. In a less formal way they can be said to have in common that they
refer to given entities salient from context or previous discourse.

With the framework of Lascarides & Stone (2006, 2009) as a starting point this
thesis investigates two possibilities of capturing instances of gestures involving
definiteness: An approach where gestures behave like pronouns, and one where
they behave like definite descriptions. In order to provide the necessary background
information, Section 2 discusses the nature of discourse referents, how the discourse
theory of DRT (Kamp & Reyle, 1993a, 1993b) incorporates them, how the framework
SDRT (Asher & Lascarides, 2003) is built on top of DRT and how Lascarides & Stone
(2006, 2009) use this background to account for gestures. This background is then put
to use in Section 3, which examines the possibility of analyzing the discourse given
parts of a gesture’s semantics as pronouns. By employing large monologues defying the
pronoun resolution constraints of SDRT and by minimal pairs with speech pronouns and
definite descriptions it is established that an analysis as pronouns does not capture the
data in question. Likewise rhetorical relations are shown not to capture all anaphoric data.

In Section 4 an alternative account based on definite descriptions is developed.
After some background on previous literature (Section 4.1) and how to distinguish
gestures as definite descriptions from other analyses (Section 4.2) empirical data is
used to establish which conception of definite descriptions is plausible (Section 4.3).
First it is established that in case of a plural antecedent the gestural data indicates
that often only some entities of the antecedent are involved with an event and that
every account involving definite descriptions must do so by a partitive analysis. The
next part presents evidence that some gestures can only be used felicitously at all
if the context provides entities suitable as an antecedent, while others only have to
obey a bridging constraint of Lascarides & Stone (2006, 2009). Since partitives do
not allow for testing the definiteness of gestures by simply testing if they contribute
the unique maximal antecedent as such, a maximal antecedent will be justified by
examples where a notion of relevance leaves open important implicit questions with
speech partitives and gestural partitives alike. Having established that some notion of
a maximal antecedent is involved with the gestures of interest, subsequent discussion
allows to decide that the denotation of a gesture should be a contextually unresolved
predicate rather than an interpreted version. In order to complete the fundamental work
for a definite description approach first steps towards a more formal account of the
notion of relevance involved with testing maximal antecedents are undertaken. During
the course of Section 4 an algorithm for (mainly definite) gestural resolution is developed.

Section 4.4 applies the findings to Lascarides & Stone’s (2006, 2009) version of
SDRT, while Section 4.5 makes concluding remarks about the two kinds of definite
description uses, attributive and referential (Donnellan, 1966), in gestures.

Towards the end Section 5 discusses problems encountered with the account of
Section 4, and while possible solutions to some aspects are pointed out, still some points
will have to remain unexplained or at least slightly ad hoc. The phenomena involve
non-exhaustive inferences with definite gestures and antecedents the algorithm of Section
4 predicts to be available but which are not. Section 6 concludes the thesis with an overall
evaluation.