The Role of Telicity in
Sentence Comprehension
Steve Seegmiller,
Karen Ingraffea, and
David Townsend
Montclair State University
Presented at It's About Time,
LSA Summer Conference,
July 2003
Research Questions
"What role does event structure (specifically
telicity) play in sentence processing?
"When does telicity come into play
during processing?
"What is the linguistic nature of
telicity?
Data
"
Results
of processing experiment
"
Results of experiment eliciting acceptability
judgments of sentences
"
Results
of examination of a large corpus of
English
Telicity: An Overview
Our Assumptions about
Telicity:
–Telicity is a semantic property that
reflects the boundedness of events: verbs that denote
bounded events are telic; those that denote unbounded events are atelic
–Telicity is partly an inherent
lexical property of verbs and partly a compositional property of predicates
–Telic predicates need a delimiter,
such as an object NP, to provide boundedness
–There are two useful diagnostics for
telicity: the in/for test and the homogeneity test
Diagnostics for Telicity
"The in/for test:
–Telic predicates take in +
temporal NP:
She won the race in/*for
an hour
–Atelic predicates take for +
temporal NP:
She ran for/*in an hour
2. The homogeneity
test:
–Atelic events are homogeneous
–Telic events are not homogeneous
Verbs
For the processing
experiment we used a set of 64 verbs 16 of each of the following:
Telic, transitive only (Tto):
e.g. identify
Telic, potentially intransitive (Tpi): e.g. heal
Atelic, transitive only (Ato):
e.g. ridicule
Atelic, potentially transitive (Api): e.g. assist
Self-Paced Reading Experiment
The self-paced reading
experiment used sentences like
The woman (that was) identified at the clinic rejoined
the demonstration
The woman (that was) healed at the clinic
...
The woman (that was) ridiculed at the clinic...
The woman (that was) assisted at the clinic...
Reading Time Measurements
Reading times were measured at the
first verb, the preposition, the article, the noun, and the second verb:
–
The woman healed at the clinic rejoined the
demonstration
V1 P Art
N V2
Predicted Results
Predicted results:
Reading times will be
longer for reduced than for unreduced relative clauses (the garden path
effect)
The garden path effect will be smaller for telic verbs
than for atelic verbs
Explaining the Results
Possible explanations for
the failure to find the predicted effect:
–Telicity plays a role in sentence processing,
and it comes into play late rather than early
–Our materials were faulty perhaps
we misclassified the verbs in our sentences
–Our assumption(s) about telicity are
mistaken maybe telic verbs are not particularly inclined to take objects
Forced-Choice Test
As a check on our verb
classifications, we asked subjects to consider sentence pairs like
The mother burped the baby in three minutes
The
mother burped the baby for three minutes
They were asked whether sentence
(a) or (b) is better, whether both are equally good, or whether neither is good
Summary of Results on
Forced Choice Test
The in/for test is
sometimes effective in distinguishing telic and atelic verbs and sometimes not.
Conclusion: the in/for
tests must be used with caution when creating materials.
Corpus Data
SOURCE: 56-million word
Collins COBUILD Corpus
(http://www.cobuild.collins.co.uk)
QUESTIONS:
Are telic verbs usually transitive?
Have we classified our verbs correctly as
telic or atelic?
Are verbs consistently either telic or atelic
in their behavior?
Conclusions from the
Corpus Data
"Verbs can vary greatly in their meanings and their
uses: a given verb may be telic in some uses and atelic in others,
or unaccusative in some and unergative
in others.
"Telic verbs are not, in fact, more likely to have
objects than atelic verbs
"While there are no significant differences between
telic and atelic verbs in their likelihood to switch telicity, between 12 and
21 percent of the time, an atelic verb is used with a telic meaning and a telic
verb is used with an atelic meaning.
Summary of Results
On the self-paced reading
task:
"
Telicity has a significant effect on reading
times on V2, the verb of the main clause: telic verbs are read more quickly
than atelic verbs.
"
Telicity has no significant effect on the garden
path effect.
"
Transitivity has a significant early effect on
the garden path effect (on the preposition).
Summary of Results 2
Based on the forced-choice test:
The in/for test is not
always reliable in distinguishing between telic and atelic verbs.
Based on the corpus data:
" The classification of verbs as telic or atelic, or as unaccusative or unergative, does
not seem to be categorical, i.e. a given verb may switch from one use to
another.
" There is only a weak association between telicity and
transitivity.
Conclusions
"
The binary-feature model for classifying verbs
may not be the right one with respect to features like telicity.
"
Telicity
is, at least in part, contextually determined.
"
Both telicity and transitivity have effects in
sentence processing, but they have different effects: transitivity comes into
play early, while telicity comes into play late.
Acknowledgments
We would like to thank Tom Bever and Errin OBryan for enlightening discussion on the topics of
telicity and processing.
Lisa Indovino, Keri Woelpper,
Jacky Pita, Katherine Thomas, and Marcela Pacesova
contributed greatly to this project in the preparation of materials and in the
gathering, coding, and analysis of the data.
