Title : Vagueness in
Implicature: The Case of Modified Adjectives
Researchers : Timothy Leffel
Alexandre Cremers
Nicole Gotzner
Jacopo Romoli
Year
of Publication : 2019
Source : Journal of Semantics,
00, Pages 1 – 32
DOI:10.1093/jos/ffy020
Objectives : To investigate
tall and late, which for current purposes we took to be
representative of the classes of relative standard gradable adjectives and
minimum standard absolute gradable adjectives, respectively.
Method : The method to
investigate tall and late is asking the
participants to indicate whether they agreed or disagreed with Mary by adjusting
a slider whose position was encoded as a value ranging from 0% to 100%, where 100%
is interpreted as complete agreement/acceptance, 0% as complete disagreement.
The method to analyze the responses for not very Adj and not
Adj statistically, the researcher used a nonparametric cluster permutation
test to identify significant differences in agreement-% along the scales.
Direct comparison at each scale point would run into serious multiple comparisons
problems, while applying a Bonferroni correction would be too conservative since
responses to adjacent scale points tend to be highly correlated. The
researcher, therefore, adopted a more sophisticated method traditionally used
in eye-tracking and EEG studies, treating the curves along our scales as a “signal.”
Results : The results from the curves associated with tall and late can
be seen as visualizations of scale structure: acceptance of the predicate tall
follows an approximately sigmoidal shape, increasing as height increases
for nearly the entire interval of heights. This reflects uncertainty about the
exact value of the tallness threshold. For late, acceptance remains
relatively constant and close to zero prior to 9am, and then shows a rapid
increase as time moves beyond 9am, eventually leveling off near 100% acceptance.
This reflects the virtual certainty that the threshold for lateness is located at
9am.
The
results from the analysis conducted yielded two
significant clusters for “late”: from 21min early to exactly 9:00am, not
very late was significantly degraded compared to not late (mean
agree 59% (sd = 32) versus 89% (16), respectively); and from 2min late
to 36min late, we observed the opposite (46% (34) versus 20% (23)). Both
clusters were larger than anything obtained from the 10,000 random
permutations; hence p < .0001. For “tall,” acceptability of the
constructions not very tall and not tall coincided in most intervals.
But the analysis revealed a small significant cluster ranging from 6ft 2in
(188cm) to 6ft 5in (196cm), during which not very tall was more
acceptable than not tall (16% (20) versus 10% (13)), with p = .0003.
Comments : This research consists of experiment
one and experiment two. The one I mentioned is the experiment one because the
following experiment studied a similarly case of experiment one but it uses
post-hoc analysis. I think this research has incredible report of those two
experiments have done before. However, for me, the researcher can provide the
whole conclusion from two experiments to help readers getting the point of the
whole part of this journal article.
Komentar
Posting Komentar