Abstracts 2005 - The Psychonomic Society
Abstracts 2005 - The Psychonomic Society
Abstracts 2005 - The Psychonomic Society
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Saturday Noon Posters 4107–4112<br />
University of Western Ontario—<strong>The</strong> triad classification task presents<br />
participants with a target concept and two possible associates. <strong>The</strong>se<br />
associates are related to the target differently, allowing the researcher to<br />
contrast two classification alternatives. Participants choose which alternative<br />
to pair with the target concept. This task has been used to demonstrate<br />
many conceptual and perceptual preferences, including whether<br />
adults might prefer taxonomic groupings, such as bed and crib (Smiley<br />
& Brown, 1979) or thematic groupings, such as baby and crib (Lin &<br />
Murphy, 2001). We explore this issue further by examining the mode of<br />
stimulus presentation (words or pictures) and the sex of participants.<br />
Several experiments showed an overall preference for thematic classifications,<br />
but this was attenuated when the stimuli were presented as pictures.<br />
In particular, male participants tended to make taxonomic classifications<br />
when presented with pictures and thematic classifications when<br />
presented with words. We argue that classification preferences are dynamic<br />
and are influenced by various stimulus properties.<br />
(4107)<br />
Modeling Categorization Response Times Using Decision Trees: A<br />
Goodness-of-Fit Comparison of Rule-Based and Exemplar Models.<br />
DANIEL LAFOND & YVES LACOUTURE, Université Laval (sponsored<br />
by A. A. J. Marley)—Lamberts (2000) has modeled individual<br />
categorization response data from 10 participants, using the extended<br />
generalized context model and the exemplar-based random walk<br />
model. In the present study, we used three rule-based classification<br />
models proposed by Trabasso, Rollins, and Shaughnessy (1971) to fit<br />
the same data. A decision tree consistent with the observed response<br />
choices was first derived for each participant. <strong>The</strong> sequence of decisions<br />
within each tree provides the basis for response time modeling.<br />
Minimizing SSE with a quasi-Newton search algorithm allowed us to<br />
estimate individual parameters. Two goodness-of-fit indicators were<br />
used, MS e and R 2 . Results show that one of the rule-based models,<br />
which has fewer free parameters than both exemplar models, provides<br />
overall best fit. Some implications for modeling categorization<br />
processes are discussed.<br />
(4108)<br />
Exemplar Models Seen as a Mixture of Serial Decision Rules. FABIEN<br />
MATHY, Rutgers University, Piscataway (sponsored by Jacob Feldman)—This<br />
poster reports a study of a decision tree model in Boolean<br />
concept learning. <strong>The</strong> model aims to determine the minimal decision<br />
tree corresponding to a concept. <strong>The</strong> time needed to classify a stimulus<br />
(i.e., response time) is linked to the number of decisions that must<br />
be made in order to know its category. When mixing all equivalent decision<br />
trees for a given concept, the mean predicted response times are<br />
perfectly correlated with those given by exemplar models. We present<br />
an experiment including 84 subjects who learned and applied<br />
Shepard’s (1961) Type IV concept. <strong>The</strong> mean response times fit both<br />
decision tree and exemplar models. However, when the individual patterns<br />
are looked at, the decision tree model explains the results better<br />
than does the exemplar model in 69 out of 84 subjects.<br />
(4109)<br />
Knowledge Partitioning in Categorization Without Context. LEE-<br />
XIENG YANG, National Chung Cheng University, & STEPHAN<br />
LEWANDOWSKY, University of Western Australia (sponsored by<br />
Stephan Lewandowsky)—Knowledge partitioning refers to the notion<br />
that knowledge can be held in independent and nonoverlapping<br />
parcels. People use a context cue to choose a knowledge parcel for responding,<br />
while ignoring information held in other parcels. A number<br />
of studies have reported evidence for such context-gated knowledge<br />
partitioning in function learning (Lewandowsky, Kalish, &<br />
Ngang, 2002) and categorization (Yang & Lewandowsky, 2003,<br />
2004). In addition, knowledge partitioning has been observed in function<br />
learning in the absence of context cues (Kalish, Lewandowsky, &<br />
Kruschke, 2004). We report a study that examined whether knowledge<br />
partitioning can occur without context cues in categorization. Partic-<br />
121<br />
ipants were trained with separate clusters of stimuli located in different<br />
regions of the category space and were found to learn two partial<br />
rules for categorization, as is predicted by the knowledge-partitioning<br />
framework. We show by simulation that a pure exemplar-based model,<br />
such as the GCM, has difficulty accommodating these results.<br />
(4110)<br />
QMQ Shows Male Adolescent Sex Offenders Can Respond to <strong>The</strong>rapy.<br />
R. CHRIS MARTIN, University of Missouri, Kansas City, DEBBIE<br />
K. MARTIN, Quality Motivation Corporation, BONNIE E. POOLE<br />
& LEAH K. GENSHEIMER, University of Missouri, Kansas City, &<br />
SCOTT U. LEGRAND, Devereaux Florida Network—Adolescent male<br />
sex offenders are recalcitrant therapy subjects for many reasons. A<br />
midwestern Boy’s Town facility provided a structured environment<br />
with many group sessions and several individual sessions weekly, with<br />
an emphasis upon behavioral change. <strong>The</strong> Quality of Motivation<br />
Questionnaire (c) is a true/false, computer-scored instrument that provides<br />
38 scales, including amounts of positive and negative motivation,<br />
energy level, emotional stability, depression, life styles, and<br />
skills. Upon presentation, they were deceitful, depressed, and high in<br />
maladaptive skills, such as self-sabotage, self-punishment, and a lack<br />
of problem-solving abilities. Success of treatment was shown in statistically<br />
significant changes in five areas: emotional stability, depression,<br />
decision making, responsibility, and problem solving. Recommendations<br />
are made regarding how to further improve adaptive<br />
skills and reduction in the maladaptive ones. Depression scores, although<br />
reduced, were still high, and specific changes are needed. <strong>The</strong><br />
attitude that sex offenders are immutable might need to change.<br />
(4111)<br />
Accessibility of Taxonomic and Script Knowledge in the Domain of<br />
Food. ANNA Z. VITKIN, JOHN D. COLEY, & KONSTANTIN FEIGIN,<br />
Northeastern University—Ross and Murphy (1999) showed that priming<br />
improved the accessibility of script knowledge about food (apple<br />
as snack), but not taxonomic knowledge (apple as fruit), relative to<br />
neutral priming, suggesting that taxonomic knowledge is more accessible<br />
to reasoning about food. We examined the effects of cross<br />
priming, asking whether incongruent priming would differentially interfere<br />
with accessibility of script versus taxonomic knowledge. Participants<br />
read a sentence about a food that primed either taxonomic or<br />
script knowledge. <strong>The</strong>y then verified the food’s taxonomic or script<br />
category membership (congruent or incongruent with priming). Overall<br />
taxonomic decisions were faster and more accurate than script decisions.<br />
Incongruent priming had no effect on taxonomic decisions,<br />
whereas script decisions were significantly slower and less accurate<br />
for incongruent primes. Taken together with Ross and Murphy (1999),<br />
this suggests that not only is script knowledge less accessible overall,<br />
but also it is actively inhibited by the activation of taxonomic knowledge.<br />
(4112)<br />
Who Do You Look Like? Further Exploration Into the Relationship<br />
Between Names and Faces. MELISSA A. LEA, Union College, ROBIN<br />
D. THOMAS, AARON T. BELL, NATHAN A. LAMKIN, & JEFF H.<br />
SHOLL, Miami University—Many cognitive researchers support the<br />
notion that names and facial features have no relationship with each<br />
other (Bruce & Young, 1986; Burton & Bruce, 1992). However, Lea<br />
(2002) showed that when participants are given a name and are asked<br />
to provide a description of that person by rating second-order relational<br />
features, they give similar descriptions of how that face should<br />
look. For example, the name “Bob” provoked a description of a widefaced,<br />
round-featured man, and “Tim” a thin-faced, thin-featured<br />
man. Conversely, the present study was conducted to determine<br />
whether participants would consistently label a set of prototype faces<br />
with those names. That is, would participants give the wide-faced,<br />
round-featured face the name “Bob”? <strong>The</strong> results demonstrate that<br />
participants are reliable at giving names to prototype faces and support<br />
the idea that faces and names can have a direct relationship.