Abstracts 2005 - The Psychonomic Society

Posters 2112–2118 Friday Noon
(2112)
Motivational Influences on Rule-Based and Information Integration
Category Learning. GRANT C. BALDWIN, W. TODD MADDOX,
& ARTHUR B. MARKMAN, University of Texas, Austin—Markman,
Baldwin, and Maddox (in press; Psychological Science) have shown
more nearly optimal decision criterion learning when the motivational
focus “fit” the environmental conditions. Promotion focus people are
more sensitive to gains in the environment and performed best when
gains were emphasized. Prevention focus people are more sensitive to
losses in the environment and performed better when avoiding losses
was emphasized. We report results that extend this work to classification
learning—in particular, to rule-based and information integration
category learning. One set of studies examined deterministic
rule-based and information integration category learning, focusing
primarily on trials to criterion. Prevention subjects reached criterion
sooner than promotion subjects in an information integration, but not
in a rule-based, task. A second set examined probabilistic rule-based
and information integration category learning, focusing on percent
correct. Promotion subjects performed better in rule-based, but not information
integration, category learning. Implications for the neurobiology
of motivation and category learning are discussed.
• PROBLEM SOLVING •
(2113)
Individual Differences in Cognitive Abilities: A Component
Processes Account. MEREDYTH DANEMAN, University of Toronto,
Mississauga, & BRENDA A. HANNON, University of Texas, San Antonio—In
this study, we examined the extent to which the ability to integrate
prior knowledge with new information can account for the variance
that is typically shared among tests of cognitive abilities. Participants
completed Hannon and Daneman’s (2001) component processes task
that includes a measure of knowledge integration ability, as well as a
battery of cognitive tests that assessed general fluid intelligence (e.g.,
Raven’s matrices) and specific abilities (e.g., verbal and spatial abilities).
Not only was knowledge integration an excellent predictor of
performance on the general and specific abilities tests, but it also
tended to have better predictive power than two measures of working
memory capacity (reading span, operation span). We argue that integrating
prior knowledge with new information is a fundamental
process that underlies skill in a range of complex cognitive tasks.
(2114)
The Adaptiveness of Strategy Choices and the Effect of Strategy
Use Experience. AYAKA WATANABE & YUJI ITOH, Keio University
(sponsored by Kathy Pezdek)—This study investigated the effect
of strategy use experience on strategy choice. Participants were required
to determine the number of target cells in a 7 � 7 matrix of
black and gray cells. Two strategies could be applied: the counting
strategy, in which the number of the target cells are counted, and the
subtracting strategy, in which the number of the nontargets are subtracted
from the total number of cells. When participants could
choose the strategies freely, they adaptively chose the strategy that
made the RT shorter. However, letting the participants use one of the
strategies tended to increase the choice of strategy inadaptively in the
subsequent phase, in which they could choose strategies freely. These
tendencies were observed only for the condition in which there were
no preferences between two strategies in the free-choice condition,
and were greater when the participants had never used the other strategy
in the previous phase.
(2115)
Verbal Intelligence or Verbal Expertise? Explaining Individual
Differences in Scrabble Problem-Solving Performance. MICHAEL
TUFFIASH, ROY W. RORING, & K. ANDERS ERICSSON, Florida
State University—Although numerous laboratory studies have found
strong relations between general cognitive abilities and verbal problemsolving
performance, few investigators have studied experts in socially
86
recognized skill domains. We examined the relationship between general
cognitive abilities and verbal problem-solving skill in the game
of Scrabble. Elite- and average-level tournament-rated Scrabble players
were asked to solve a series of realistic Scrabble game problems,
as well as a series of anagrams of varying length. Each player also
completed a battery of standardized verbal ability measures and a selfreport
survey of their past and present Scrabble-related practice activities.
Preliminary analyses (23 elite and 14 average players) indicated
strong positive relations between measures of Scrabble tournament
performance and Scrabble-related problem-solving skills, but much
weaker relations between measures of Scrabble problem-solving skills
and aspects of general verbal intelligence. Additional analyses of biographical
data suggested that Scrabble experts’ exceptional domainspecific
problem-solving skills were largely a product of accumulated
deliberate practice.
(2116)
Predicting Expert and Novice Anagram Solution. LAURA R.
NOVICK, Vanderbilt University, & STEVEN J. SHERMAN, Indiana
University—Several experiments examined the predictors of two measures
of anagram solution: solution time and the proportion of subjects
who generated pop-out solutions (defined as solutions that occurred
within 2 sec of stimulus presentation). The subjects were
college students who were selected on the basis of their performance
on a scrambled words pretest: They were either quite good or not good
at solving anagrams. Of particular interest is whether different aspects
of the solution words (e.g., fit to the constraints of English spelling,
word frequency, number of syllables) and the anagrams (e.g., similarity
to the solution word) predict measures of performance for experts
versus novices.
(2117)
Traveling Salesman Problem: The Role of Clustering Operations.
ZYGMUNT PIZLO, JOHN SAALWEACHTER, & EMIL STEFANOV,
Purdue University—We will present a pyramid model, which provides
near-optimal solutions to the traveling salesman problem (TSP) in a
time that is a low-degree polynomial function of the problem size. In
this model, hierarchical clustering of cities is first performed by using
an adaptive method. The solution tour is then produced by a sequence
of coarse-to-fine approximations, during which the model has highresolution
access to the problem representation only within a small
area, simulating the nonuniform distribution of visual acuity across
the visual field. The tour is produced by simulated movements of the
eyes. We will discuss the role of two types of clustering operations:
detecting blobs and detecting smooth contours. The model was applied
to TSPs of size 6, 10, 20, and 50, and its performance was fitted
to that of 5 subjects by using one free parameter representing the
extent of local search.
(2118)
Models of Human Performance on the Traveling Salesperson
Problem: The Shortest Route to Falsification. SUSANNE TAK,
MARCO PLAISIER, & IRIS VAN ROOIJ, Technische Universiteit
Eindhoven—The task of finding the shortest tour visiting a set of
points, known as the traveling salesperson problem (TSP), is notorious
for its computational complexity. Nonetheless, people show remarkably
good performance on the task. MacGregor, Ormerod, and
Chronicle (2000) have proposed a computational model to describe
the human strategy for solving TSPs. Studying only random point
sets, these authors claimed empirical support for their model. We
argue that random point sets do not instantiate critical tests of model
performance, since both the model and people are known to perform
well on random TSPs. We designed six (nonrandom) point sets as a
critical test bed for the MacGregor et al. model. We observed a systematic
misfit between human and model performance for five of the
six points sets. Our findings falsify the MacGregor et al. model. We
discuss some methodological lessons for testing computational models
of human problem solving in general.