Abstracts 2005 - The Psychonomic Society

Friday Morning Papers 54–60
model. Thus, participants appear to be able to use various heuristic
and normative decision procedures, depending on task conditions.
10:40–10:55 (54)
Prediction and Change Detection. MARK STEYVERS & SCOTT D.
BROWN, University of California, Irvine—We measure the ability of
human observers to predict the next datum in a sequence that was generated
by a simple statistical process undergoing change at random
points in time. Accurate performance in this task requires identification
of changepoints and prediction of future observations on the basis of
the observations following the last changepoint. We assess individual
differences between observers both empirically and using two kinds
of models: a Bayesian approach for change detection and a family of
cognitively plausible fast and frugal models. Some individuals detect
too many changes and, hence, perform suboptimally, due to excess variability.
Other individuals do not detect enough changes and perform
suboptimally because they fail to notice short-term temporal trends.
11:00–11:15 (55)
Modeling Binary Choice Patterns of Individuals. DOUGLAS H.
WEDELL, University of South Carolina—Several extant models of
binary choice differ in their conceptualization of (1) the comparison
process, (2) contextual sensitivity of valuation and weighting, (3) the
importance of the covariance structure of alternatives, (4) the role of
bias, and (5) the ability to predict choice latencies and effects of time
pressure. Two experiments reported here were designed to provide insights
into the applicability of these models. Both studies were presented
via the Internet, with repeated choices enabling the modeling
of choice proportions of individuals. Experiment 1 replicated and extended
the gambles experiment of Tversky (1969), designed to produce
violations of weak stochastic transitivity (WST). Robust violations
of WST were found and were well explained by models that
assumed bias toward choosing the higher outcome gamble. Experiment
2 was a simulated shopping study based on a design that sampled
a choice space across which models showed markedly different
patterns of choice behavior.
11:20–11:35 (56)
Signal Detection Analyses of Deductive and Inductive Reasoning.
EVAN HEIT, University of California, Merced, & CAREN M. RO-
TELLO, University of Massachusetts, Amherst—Reasoning can be
conceived of as a signal detection task in which the goal is to discriminate
strong from weak arguments. Hence, analyses more commonly
used in other research domains, including estimation of sensitivity
and bias and the evaluation of receiver operator characteristic
(ROC) curves, can also be applied to reasoning. Two experiments
were conducted, following Rips (2001), in which subjects judged a set
of 16 arguments in terms of either deductive validity or inductive
plausibility. It was found that deductive and inductive judgments differed
in many ways. In general, deduction did not simply reflect a
stricter criterion for saying “yes,” as compared with induction. There
were also differences between deduction and induction in terms of
sensitivity and preferred ranking of arguments, as well as the form of
the ROC curves. Overall, the results point to a two-process account of
reasoning, rather than to a single process underlying both deduction
and induction.
11:40–11:55 (57)
An EZ-Diffusion Model for Response Time and Accuracy. ERIC-JAN
WAGENMAKERS & HAN VAN DER MAAS, University of Amsterdam—
The EZ-diffusion model for two-choice response time takes response
times and accuracy as input. The model then transforms these data via
three simple equations to unique values for the quality of information
(i.e., drift rate), response conservativeness (i.e., boundary separation),
and nondecision time. This redescription of the observed data in terms
of latent variables addresses the speed–accuracy tradeoff and thereby
affords an unambiguous quantification of performance differences in
9
two-choice response time tasks. Performance of the EZ model was
studied with Monte Carlo simulations. The advantage of the EZ model
over the full diffusion model (e.g., Ratcliff, 1978) is that the EZ model
can be applied to data-sparse situations, facilitating the use of individualsubject
analysis.
Visual Perception
Civic Ballroom, Friday Morning, 10:00–12:00
Chaired by J. Farley Norman, Western Kentucky University
10:00–10:15 (58)
The Perception of Distances and Spatial Relationships in Natural
Outdoor Environments. J. FARLEY NORMAN, CHARLES E. CRAB-
TREE, ANNA M. CLAYTON, & HIDEKO F. NORMAN, Western
Kentucky University—The ability of observers to perceive distances and
spatial relationships in outdoor environments was investigated. In Experiment
1, the observers adjusted triangular configurations to appear
equilateral, whereas in Experiment 2, they adjusted the depth of triangles
to match their base width. The results of both experiments revealed
that there are large individual differences in how observers perceive
distances in outdoor settings. The observers’ judgments were
greatly affected by the particular task they were asked to perform. The
observers who had shown no evidence of perceptual distortions in Experiment
1 (with binocular vision) demonstrated large perceptual distortions
in Experiment 2 when the task was changed to match distances
in depth to frontal distances perpendicular to the observers’
line of sight. Considered as a whole, the results indicate that there is
no single relationship between physical and perceived space that is
consistent with observers’ judgments of distances in ordinary outdoor
contexts.
10:20–10:35 (59)
Generalization of Perceptual and Cognitive Prism Adaptation.
GORDON M. REDDING, Illinois State University, & BENJAMIN
WALLACE, Cleveland State University—Prism exposure produces
two kinds of adaptive responses. Recalibration is ordinary strategic
remapping of spatially coded movement commands to rapidly reduce
performance error produced by prismatic displacement, a kind of cognitive
learning. Realignment is the extraordinary process of transforming
spatial maps to bring the origins of coordinate systems into
correspondence, a kind of perceptual learning. Realignment occurs
when spatial discordance signals noncorrespondence between spatial
maps. The two kinds of aftereffects were measured for three test positions,
one of which was the exposure training position. Recalibration
aftereffects generalized nonlinearly, whereas realignment aftereffects
generalized linearly. Recalibration and realignment are distinct
kinds of adaptive processes, both evoked by prism exposure, and require
methods for distinguishing their relative contributions.
10:40–10:55 (60)
Perception of Slopes of Hills Is More Accurate in Near Space.
BRUCE BRIDGEMAN & MERRIT HOOVER, University of California,
Santa Cruz—We have found that slopes of hills are perceptually
overestimated in far space (15 m), but not in near space (1 m). A motor
measure, however, shows similar and more accurate estimates at both
ranges. Slopes were measured with observers standing at the bases of
real hills; the perceptual measure required estimating the slope verbally
in degrees, whereas for the motor measure, observers matched
the slope with the forearm. Perceptual overestimates might be due to
near space being handled by different brain structures than far space,
or the brain might calculate a combination of the real slope and the
work required to reach the far point. This was tested by measuring perceived
slope at 1, 2, 4, 8, and 16 m. Perceived slope increased linearly,
indicating that work required is the best description of the result. Perception
gives us the environment’s layout combined with predictions
of the work required to interact with it.