Académique Documents
Professionnel Documents
Culture Documents
Speakers must choose whether to communicate explicitly, and say exactly what
they mean, or implicitly, and let the listener do the work. We investigate whether
this choice can be primed (e.g., Bock, 1987).
Participants engaged in a communication game in which they took turns with a
confederate describing and identifying a card. Speakers could be implicit,
requiring the addressee to use a conversational implicature (Grice, 1975), or
explicit. The DV was whether the participant produced an implicit or an explicit
construction. Experiment 1 used a between-subject manipulation, in which a
confederate described target cards using either explicit or implicit expressions.
Participants in the explicit condition used more explicit expressions than those
in the implicit condition. Experiment 2 manipulated the priming within-subject
and found similar results.
Our findings demonstrate that the decision to communicate implicitly depends
on the style embraced by the conversational partner. The only previous
suggestion about why people communicate implicitly is that they do so for
reasons of politeness (Brown & Levinson, 1978) and our results cannot be
explained in this way. We suggest our effects arose because we primed
abstract, meaning-based structures associated with conversational implicatures,
similar to syntactic frames (Branigan & Pickering, 1998).
Posters
A pragmatic account of implicit causality
van den Hoven, E. & Ferstl, E.
University of Freiburg
In sentences like the following, the preferred interpretation of the pronoun 'she'
depends on the matrix verb: 'Mary frightened/criticized Sue because she was
unpredictable.' This phenomenon is known as implicit causality (IC). Whereas
recent studies on IC attribute causality biases to lexical semantics, we introduce
a pragmatic account of IC. On our account, causality biases are due to
inferences to the stereotype. That is, verbs do not entail anything about likely
explanations. Rather, likely explanations are inferred. In a stereotypical
situation, A may criticize B because she thinks B did something wrong, but this
is a defeasible inference, unlike the requirement that A communicates
something. Hence there is no inconsistency in the sentence 'Mary criticized
John, although she knew he didn't do anything wrong.', like there is in the
sentence '??Mary criticized John, although she didn't communicate anything.'
The pragmatic account predicts a strong influence of discourse context. Data
from a story completion study support this account, showing that IC biases can
be altered with a discourse context that runs counter to the stereotype (e.g. A
has reason to be jealous prior to criticizing B). Data for a follow-up visual world
study investigating the online comprehension process are pending.
[PS-1.6] All the time you need to calculate implicatures
Dotlacil, J.
University of Groningen
It is well-known that when interpreting, we consider what has been said, as well
as what has not been said. Thus, if someone mentions that some boys are
tired, it is common to assume that the fact that the speaker did not use "all"
instead of "some" suggests that the universal statement is not true (implicated
meaning; implicature). Recently, several frameworks have been developed that
model the way humans arrive at implicatures (e.g., Bayesian models, RSA
theory). However, such statistical models are atemporal -- they only predict
whether an implicature is drawn, not what amount of time it would take.
I present an ACT-R model that aims to capture the data from Bott and Noveck
(2004) (B&N), who studied implicatures in sentences like "Some elephants are
mammals". The model starts by learning implicatures related to "some" (by
learning what quantifiers are relevant alternatives). This process arguably
happens as part of acquisition. The resulting model applied to B&N and
supplemented with simple reasoning module for category relations (elephants
are mammals) correctly predicts mean RTs for target sentences with and
without implicatures, as well as several mean RTs for baseline conditions which
use "all" instead of "some".
[PS-1.12] Contextual effects on online pragmatic inferences
King, J. , Loy, J. & Corley, M.
University of Edinburgh
when instructed to 'Click on the tall glass' when no short glass was present. But
when a short glass was introduced, low-load participants showed a strong
preference to look at the tall glass after hearing 'tall', and high-load participants'
displayed a similar preference but one significantly weaker 200-300ms after the
adjective. Listeners are thus able to use context to derive contrastive meaning
even while under cognitive load--suggesting contrast is lexically encoded--but
this ability can be impaired under high load---suggesting some aspect of CI
computation is pragmatic.
PS-2.6] Allocation and use of processing capacity in language production are
reflected in P300 amplitude and theta power
Shitova, N. 1, 2 , Roelofs, A. 1 , Coughler, C. 1 & Schriefers , H. 1
1 Donders Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen
2 International Max Planck Research School for Language Sciences, Max Planck Institute for
Psycholinguistics Nijmegen
There is some evidence that the allocation and use of processing capacity in
spoken language production are reflected in the amplitude of the P300
component. However, little is known about their relation to reaction time (RT)
and theta power, which reflects capacity demand outside the language domain.
To examine this, we had participants switch every second trial between
describing pictures using noun phrases with one adjective (size only; simple
condition) or two adjectives (size and colour; complex condition) in the format
determiner + adjective(s) + noun. RTs were longer for complex than for simple
trials. Moreover, complexity and sequence interacted: RTs were longer on
switch than on repeat trials for simple phrases (switch cost) but shorter on
switch than on repeat trials for complex phrases (switch benefit), while there
was no main effect of sequence. P300 amplitude and theta power increased
with complexity. There was no main effect of sequence in electrophysiological
measures, but both P300 amplitude and theta power decreased on switch trials
as compared to repeat trials for simple phrases. These results provide evidence
that the allocation and use of processing capacity in language production are
reflected in P300 amplitude and theta power.
configurations; in Match, the correct target for retrieval fully matches the
retrieval cues, in Mismatch there is only a partial match. The studies also vary in
whether interference is proactive (PI) or retroactive (RI), and in type of
dependency: antecedent-reflexive/reciprocal, subject-verb, and agreement
attraction configurations. We present the first-ever quantitative synthesis of
these studies using a Bayesian random-effects meta-analysis. The metaanalysis shows that in Match, overall a slowdown is observed (posterior mean
2.5 ms, 95% credible intervals (CrI) -2.75,8.49). In Mismatch, too, a slowdown is
seen (mean 8.43, CrI -0.8,17.74). A meta-regression further showed that (a)
number-cue based agreement attraction is different from other dependencies as
it leads to a speedup compared to other retrieval cues (Match: mean -5.28, CrI
-10.27,-0.35; Mismatch: -8.35, CrI -16.64,-0.35), (b) in Match, PI leads to a
slowdown compared to RI (mean 5.19, CrI 0.2,10.38), (c) in Mismatch, PI leads
to a speedup (mean -5.48, CrI -14.05,4.09), (d) in Mismatch, if the distractor is
more prominent, a greater slowdown is observed (mean 13.8, CrI -5.74,33.57).
Non-actuality implicatures (NAIs; Grant, Clifton & Frazier 2012) highlight the
contrast between a described state of affairs and the actual world, introducing
implicit Questions under Discussion (QUD; Roberts 1996) as to whether the
described state holds. NAIs are triggered by predicates like want or modals like
should; e.g. ''Alice should've eaten breakfast'' implicates Alice did not eat
breakfast. This study asks whether the question structure of the discourse
containing a NAI trigger influences how likely the inference is to survive, or
alternatively, be cancelled. Experiment1 varied whether NAI sentences
appeared as an answer to a QUD followed by another QUD (creating the
impression of a complete answer to the first question), or as one of multiple
continuations all partially answering the same QUD. Even though NAIs were
never explicitly cancelled, discourses where NAI sentences were construed as
partial answers gave rise to more ''cancellation'' interpretations (''Alice should've
eaten breakfast'' meaning that she did eat breakfast). Experiment2 asks
whether, even in structurally congruent discourses, the type of questions
following a NAI can shift interpretations toward implicature or cancellation
readings. NAIs followed by ''Why not?'' questions reinforced the implicature,
whereas ''Why?'' (''Why is it important to eat breakfast?'') questions increased
cancellation interpretations.