Teton Village, Jackson Hole, Wyoming
                             February 5 - February 10, 2006
        Organizer:  George Sperling, University of California, Irvine



Greg Appelbaum
Smith-Kettlewell Eye Research Institute

Figure and Background Cortical Networks
Authors:  L. G. Appelbaum, V. Vildavski, M. Pettet, A. Wade,  & A. M. Norcia

  The rapid and accurate segmentation of visual scenes into objects and their
surrounds is a fundamental task performed by the visual system.  Figures can be
extracted from their backgrounds based on the detection of spatial gradients in
a number of local cues such as luminance, texture, color, or temporal structure.
To assess the role of local boundary cues in figure-ground segmentation, we
constructed displays in which figure and background regions were separately 
"tagged" with periodic modulations of their local texture elements.  Textures 
were defined on orientation, phase, and temporal cues and were tested over a 
range of retinotopic presentations.  Using these synthetic texture-defined 
objects and an electrophysiological paradigm that allows us to monitor figure-
and background-region activity separately, we have found distinct neuronal 
networks for figure and background processing.  The first is a recurrent 
network of ventral stream visual areas including the lateral occipital cortex 
that is selectively active in response to the figure region of our displays.  
The activated sites and temporal sequence of these activations are invariant 
with respect to the cues used to define the figure, and do not depend on its 
spatial location or extent.  A separate network, extending from primary visual 
areas through the dorsal visual pathway is observed in response to the 
background region of our displays.  


Ben Backus
University of Pennsylvania

Further Studies of Cue Recruitment in Visual Appearance
Authors:  Ben Backus & Qi Haijiang

  After Pavlov (1927), it seemed clear that the visual system should be 
trainable by means of paired association (Fieandt, 1936; Hebb, 1949; Brunswik, 
1953, 1956; Smedslund, 1955).  But this prediction was not confirmed and today 
perceptual learning is often defined as an improvement in the ability to 
discriminate that comes with practice.  Yet instances of associative learning 
have been documented for perceptual appearance and the modern view of perception
as near-optimal inference requires this form of learning.  Last year we reported
a positive result in a direct test of this proposition, using "cue recruitment"
experiments adapted from Pavlov's classical conditioning paradigm.  Trainees 
viewed movies of perceptually bistable Necker cube stimuli.  On training trials,
perceived rotation direction was forced by the addition of trusted cues (stereo
and occlusion).  Critically, arbitrary signals were also added, contingent on 
the trusted cues.  On test trials stimuli contained the new signals but not the
trusted cues.  The new signals often became cues: they acquired the ability to 
disambiguate the percept on their own.  We now report that the learning is 
incremental (cf. Gallistel, 2004), that it becomes slower after initial training
("learned irrelevance"), and that in at least some conditions there is no 
"learning to learn":  after 36 blocks of 80 trials each (spread across 6 days),
with contingency reversed every block, the visual system's utilization of the 
new cue stopped tracking the reversals.


Geoff Boynton
Salk Institute

Attention Within and Across Modalities:  The Effects on Human Visual Cortex

  Attention to a visual or auditory stimulus involves filtering out irrelevant 
information, both within the attended sense, and in other sensory modalities.  
Attending to a cell phone conversation reduces the ability to react to visual 
events, and those engrossed in a novel can fail to hear their name being called.
We used functional magnetic resonance imaging to examine brain responses to
ignored visual and auditory stimuli while subjects performed either a visual or
an auditory task.  Early visual areas are traditionally considered uni-modal, 
yet the response to an ignored visual stimulus was reduced when subjects were 
distracted by attending to auditory stimuli.  This was not a general effect due
to attending elsewhere; responses to ignored stimuli in V1 were weaker when
subjects attended to a different visual stimulus than when they attended to an
auditory stimulus, and the effect of an auditory stimulus was differentially 
gated depending on whether or not it was on the same side of space as the
ignored visual stimulus.  More central visual areas tended to show increasingly
large effects for auditory as compared to visual distraction.  Similarly, 
responses to ignored auditory stimuli within auditory cortex were smaller when 
subjects performed a separate auditory task than when they performed a visual 
task.  Unattended sensory information is not passively propagated across early 
sensory areas; the extent to which unattended information is reduced by 
attention to another stimulus depends on whether or not subjects are attending 
to the same modality, or to the same region of space.


Thomas Busey
Indiana University

On the Nature of Privileged Visual Stimuli:  Immunity From Within-Class
Authors:  Bethany Schneider, Jordan DeLong, and Thomas Busey

  In four experiments we examine the neural correlates of the interactions 
between upright faces, inverted faces and visual noise.  Faces were chosen as 
stimuli since they reliably show strong neural response in a component termed 
the N170, and upright faces may be processed at least in part using configural 
mechanisms.  In Experiment 1, we examined upright and inverted faces across two
contrast levels and one signal-to-noise ratio, yielding a crossover interaction
of the N170 amplitude: when presented in noise, the amplitude of the inverted 
face was smaller than the upright face, while the reverse is true without noise.
In Experiment 2, we showed that the amplitude reversal was robust for full but 
not partial faces across all noise levels.  In Experiment 3, we varied contrast
to see if reversal was a result of degrading a face.  We observed no reversal 
effects.  Thus, across conditions, adding noise to full faces was a necessary 
and sufficient condition for the N170 reversal.  This is consistent with a model
in which inhibition occurs between neurons processing noise and inverted faces,
but not upright faces. In Experiment 4, we delayed onset of the upright/inverted
face presented in noise.  We replicated the smaller N170 for inverted faces at 
no delay, but observed partial recovery of the N170 for inverted faces at longer
delays.  These data are consistent with a model in which neurons responding to 
noise inhibit those responding to inverted faces, with inhibition waning to 
produce selective recovery of the inverted face response at longer SOAs.  Thus,
inverted faces and noise may be processed by the same population that is subject
to within-class inhibition; upright faces may be processed in part by a separate
population of neurons that are immune to this within-class inhibition.


Clara Casco
University of Padova

Region- and Edge-Based Configural Effects in Texture-Segmentation
Authors:  Clara Casco & Enrico Giora

  Visual analysis of textures results in an instantaneous representation of
texture surfaces and, where the output of local filters changes, in the
perception of a texture edge. We investigated the role of region- and edge-based
mechanisms of texture segmentation using Gabor patch matrices of 8x8 stripes, 
with a texture edge at the two outer stripes. In one of the two stripes, textels
were always either parallel (collinear) or orthogonal (non-collinear) to the 
edge. Three results were not consistent with edge-based discrimination models.
First, textel overall orientation was perceived when the edge was not.  Second,
keeping configural effects at the edge fixed, saliency varied, depending on 
whether collinear textels were external or internal: when external, performance
improved with duration (from 62 to 90%); when internal (i.e. collinear textels 
belonged to the field) edge orientation could barely be discriminated").  Third,
configural effects in the field may cancel those at the edge: if duration was 
long enough (80 and 100 ms), edges defined by non-collinear textels on both 
sides were better discriminated than edges with internal textels collinear, with
external textels both oblique and orthogonal.  These results suggest that
region-based analysis by lateral interactions results, before segmentation, in
perceptual groups always oriented as the edge in the external stripe and 
oriented as the textels in the texture field.  Detection of orientation contrast
at the edge is difficult when these groups feed 2nd order detectors all with 
the same orientation (when internal textels are collinear) and easy when 2nd 
order detectors' orientations on the two sides of the edge are different .  


Amy H. Criss
Carnegie Mellon University

An Empirical Test of Differentiation in Episodic Memory

  When items on one list are studied longer than items on another list, the 
improvement in performance typically manifests as an increase in the hit rate
and a decrease in the false alarm rate.  This finding is referred to as the 
strength-based mirror effect and has been accounted for by assuming that 
participants adopt a more strict criterion following a list containing items 
studied several times.  An alternative account is found in differentiation 
models where longer study leads to a more accurate memory representation for the
studied item.  The more accurate the stored representation, the less it is 
confusable with a randomly chosen foil, resulting in a decrease in the false 
alarm rate.  Differentiation models make additional a priori predictions that 
the level of differentiation of a study list interacts with the similarity 
between the studied items and the foils.  These predictions are empirically 
tested and confirmed.


Greg DeAngelis
Washington University School of Medicine

Linking Neural Representation to Function: Roles of Area MT in Depth Perception

  A fundamental task of the visual system is to reconstruct the 3D layout of the
environment from the images formed on the two retinas.  Binocular disparity 
provides a powerful cue for depth perception and neurons tuned to disparity are
found in several areas of visual cortex.  The roles that these different areas 
play in 3D vision are unclear, however.  We have been studying the contributions
of area MT to stereo vision using a combination of single-unit recordings, 
electrical microstimulation, and reversible chemical inactivation.  I will 
describe data which show that MT contributes to coarse discrimination of 
absolute disparities but is not involved in fine discrimination of relative 
disparities.  Moreover, training animals make fine disparity discriminations 
dramatically alters the contribution of MT to coarse depth perception.  I will 
argue that these findings are explained by the fact that MT does not contain an
explicit representation of relative disparities that is needed for fine depth 


Jian Ding
University of California, Irvine

SSVEP Studies of Spatial Attention
Authors:  Jian Ding, Ramesh Srinivasan, & George Sperling

  Frequency-tagging is an experimental design for EEG or MEG in which two or more
stimuli are presented simultaneously but flickered at different frequencies.
Steady-state responses (SSVEP) elicited by the flicker are detected at each 
stimulus frequency by Fourier analysis of EEG responses.  In SSVEP studies of 
spatial attention, two competing stimuli are presented simultaneously and 
"tagged" with different frequencies.  An observer is instructed to attend to one
stimulus and perform a target detection task while ignoring the other one.  
Attentional modulation is studied by the comparison of SSVEP responses when a 
stimulus is attended versus when it is unattended (its competing stimulus is 
attended).  We studied attentional modulation of SSVEP responses using 15 "tag"
frequencies ranging from 2.5-20 Hz.  The stimuli were confined to two color-coded
concentric annuli.  On each annulus, a sequence of search arrays (concentric 
disks with, possibly, a target triangle) was superimposed.  The two sequences 
were updated independently, one updated at a fixed frequency (flicker) and the 
other updated at random intervals to generate a white noise temporal-frequency 
distribution.  On each trial, observers were instructed to attend one of the two
annuli and to detect target triangles that appeared occasionally.  
  Results:  The choice of whether the annulus with flicker is attended or not, 
whether the unattended annulus (the competing annulus) was inner or outer, and 
flicker frequency all influence SSVEP amplitude and cortical distribution.  There
are two cortical response configurations (networks) with different sensitivities
to attention.  At low flicker frequencies in the delta band (2-4 Hz), and in the
upper-alpha band (10-11 Hz), an occipital-frontal network appears to phase-lock 
to the flicker when attending to the flicker, increasing the steady-state 
response.  When attention is directed away from the flicker in the lower-alpha 
(8-10 Hz) band and towards a competing stimulus in the fovea, a global resonance,
including parietal cortex and posterior-frontal cortex, responds preferentially 
to the unattended flicker.


Barbara Dosher
UC Irvine

The Dynamics and Specificity of Perceptual Learning

  Perceptual learning reflects the improvements of perceptual task performance 
with training or practice.  Recent proposals have suggested that perceptual 
learning may often reflect learned reweighting of associations to decision 
structures (Dosher & Lu, 1998, 1999), and that an augmented Hebbian learning mode
may account for perceptual learning (Petrov, Dosher, & Lu, 2005), even without 
feedback.  This approach may be compatible with specific forms of the learning 
function, and these functions suggest single-process learning for basic visual 
tasks, rather than cascades of different learning processes at different levels 
of the visual system, and specificity of learning depends upon the precise nature
of the training and transfer tasks.


James Elder
York University

Estimating Nonlinear Mechanisms Underlying Detection of Narrowband Stimuli
  Using Classification Image Analysis
Authors:  James Elder & Yaniv Morgenstern  

Detection of low-contrast luminance-defined stimuli can involve spatial 
summation over a large portion of the visual field.  For example, contrast 
thresholds for grating detection decrease as a function of the width of the 
grating, up to a width of about 8-10 cycles.  However, prior psychophysical 
results suggest that the summation region may shrink substantially in the 
presence of high-contrast masking gratings or noise (Legge & Foley, 1980; 
Kersten, 1984).  
   We used a classification image technique (Ahumada & Lovell, 1971; Ahumada, 
Marken, & Sandusky, 1975) to study the mechanisms underlying grating detection 
and to directly test for variations in the extent of spatial summation as
a function of noise contrast.  We conducted a yes/no grating detection 
experiment in which stimuli were large (24.2 deg) vertical gratings in noise.  
Classification images derived from signal-present trials were found to be
closely matched to the signal, with no systematic variation in the extent of 
summation as a function of noise contrast.  In all conditions, the estimated 
summation fields extended over many cycles of the stimulus.  However,
classification images derived from signal-absent trials were featureless, 
bearing no discernible relationship to the signal, indicating that the linear 
detection model on which standard classification image analysis is based does 
not apply.  We have tested a number of alternative models, and find a model 
based on spatial pooling of squared responses from local linear mechanisms to be
most consistent with the human data.  We show that this model is linear in the 
power spectrum domain, and introduce a novel classification image analysis 
technique that allows direct estimation of the transfer function of the 
underlying local mechanisms.  These local filters are found to be relatively 
small and broadband, as predicted by Kersten (1984), and quantitatively similar
in their spatial and spatial-frequency properties to neurons in early visual 
cortex of primate.


Ione Fine
University of Southern California

Vision in the Blind
Authors:  Ione Fine, Alan Horsager, & Scott Greenwald

  Photoreceptor loss is one of the major causes of blindness in the Western 
world, and the prevalence is likely to increase dramatically as the population 
ages.  Currently there are several groups trying to develop retinal prostheses, 
analogous to cochlear implants, in which photoreceptor input is replaced by 
direct electrical stimulation.  Recently six patients have been implanted 
chronically with simple 4x4 retinal prostheses lying over the inner retinal 
layer.  I will talk about behavioral data examining the perceptual consequences
of electrical stimulation in these patients.


Wilson S. Geisler
University of Texas at Austin

Contour Statistics in Natural Scenes:  Contour Grouping and Border Ownership
Authors:  W. S. Geisler, J. S. Perry, D. Ing, & A. J. Wilson

  A potentially powerful approach in measuring natural scene statistics is to 
analyze natural images that have been hand segmented by human observers (e.g.,
Brunswik & Kamiya 1954; Balboa & Grzywacz 2000; Geisler, Perry & Super 2001; 
Elder & Goldberg 2002; Martin, Fowlkes, & Malik 2004; Konishi, Yuille, Coughlan
& Zhu 2003).  The central assumption is that humans can, under some 
circumstances, segment images veridically and hence provide an approximate 
"ground-truth."  These ground-truth segmentations can then be used to determine
the conditional and prior probability distributions of image properties needed 
for optimal (Bayesian) perceptual inference.  We have carried out two hand 
segmentation studies.  In one, observers assigned edge elements (which were 
extracted with an automatic algorithm) to unique physical contours, for a 
diverse collection of natural images (Geisler et al. 2001).  In the other, 
observers hand segmented a large number of calibrated, close-up images of 
foliage into unique objects (leaves, branches, flowers, etc.).  Analysis of 
these hand-segmented images reveals a number of strong statistical regularities
that could be used to support contour grouping and perceptual decisions of 
border ownership.  Using these statistical regularities, we derived a Bayesian 
ideal observer for a particular "naturalistic" contour-grouping experiment.  Our
preliminary results suggest that humans perform similar to ideal on a 
trial-by-trial basis; i.e., the trial-by-trial agreement exceeds 90%.


Norma Graham
Columbia University

A New (At Least to Us) Kind of Contrast Adaptation
Authors:  Norma Graham & S. Sabina Wolfson

  We started out to investigate the dynamics of a contrast-gain control process
of the normalization type.  This process had proved necessary to explain our 
previous results with static textured patterns.  We decided to sinusoidally 
modulate the contrast of a background adapting pattern and to measure the 
observer's ability to identify short-duration test patterns superimposed at 
various phases.  We had good reason to hope that the results would reveal the 
dynamics of the normalization gain-control process, and we computed predictions
from several candidate models for normalization dynamics.  We collected the 
experimental results.  They bore no relationship to our expectations or to the 
model predictions.  We could make no sense of them.  For weeks we looked for 
computer errors.  Finally, with the help of an overdose of Buffy the 
Vampire-Slayer (brought home by visiting offspring) our false expectations were
overcome.  And we realized that the results were explainable by a form of 
adaptation we had not foreseen:  the adaptation of a "contrast comparison 
level."  Apparently, at each spatial position, the current contrast is processed
relative to a comparison level.  And this comparison level adapts to equal the 
recent time-averaged contrast at that position.  One can model this adaptation 
within the context of a channel composed of three linear filters with two 
rectification-type nonlinearities interposed:  The zero-point of the second 
rectification adapts to equal the recent time-averaged input.  For lack of a 
more respectable name for these channels that is also short and informative, we
are currently calling them Buffy channels.  (Suggestions for alternate names are
welcome.)  Adaptation within Buffy channels generally makes a test pattern
formed by increases in contrast from the adaptation level (or decreases from the
contrast level) easy to perceive.  But a test pattern formed by both increases 
and decreases is quite hard to perceive.  Is this characteristic adaptive for 
the observer or is it a side effect of something that is adaptive?


Erin Harley

Functional MRI Can Measure Timing of Neural Response With High Precision
Authors:  Engel, S. A., & Harley, E.

  Functional MRI (fMRI) has traditionally been thought of as a tool for 
measuring human cortical function with relatively good spatial precision, but 
relatively poor temporal precision.  Nevertheless, a small number of prior 
studies have used fMRI to measure the onset of neural activity from a resting 
baseline with a temporal precision on the order of 100 msec.  Many important 
neural processes, however, reveal themselves as increases in activity that lie 
on top of another neural response.  We attempted to determine whether fMRI could
accurately measure the timing of an increase that occurred 200 msec after 
initial response.  Five subjects viewed stimuli from three conditions.  In our 
baseline condition, a low contrast flickering checkerboard was presented for 600
msec.  We added a 400 msec contrast increment to this stimulus either at a 200 
msec delay or immediately upon stimulus presentation.  We acquired BOLD fMR 
images every 100 msec (TR) while these conditions were presented in a rapid 
event-related design.  We averaged data within primary visual cortex and 
computed conditional hemodynamic timecourses using standard linear methods.
Subtracting the baseline response from the responses to the delay condition and
the immediate condition generated difference timecourses that isolated the 
transient increase in response.  The two difference timecourses were similar in
shape, and in all five subjects the delay difference timecourse peaked later 
than the immediate difference timecourse.  To quantify the delay, we temporally
cross-correlated the two time courses from each subject, and took the size of 
the temporal shift that yielded the highest correlation as an estimate of delay.
The average delay was 166 msecs with a standard error of 33 msecs.  These 
results demonstrate that fMRI can detect relatively small differences in the 
timing of increases in neural activity, even when the increases lie on top of a
substantial response.


David J. Heeger
New York University

Coding the Arrow of Time
Authors:  David Heeger, Uri Hasson, Eunice Yung, Ignacio Vallines, & Nava Rubin

  We presented human subjects with two repeated presentations of original 
("Forward", F) and time-reversed ("Backward", B) movie clips while acquiring 
measurements of brain activity using functional magnetic resonance imaging 
(fMRI).  We computed the correlation between the response time-courses at each 
location in the brain across the repeated presentations.  Specifically, we 
computed the correlations between:  1) the responses to repeated presentations 
of the forward movies (CF1,F2), 2) the responses to repeated presentations of 
the backward movies (CB1,B2), 3) the responses to the forward versus backward 
movies (CF,B), and 4) the responses to the forward clips versus the time-reverse
responses to the backward clips (CrB,F), corrected for hemodynamic delay.  We
reasoned that some brain regions would respond similarly to the moment-to-moment
content within a short time interval (~1-3 sec), for both the forward and 
backward movie clips.  Specifically, the response time-courses acquired from the
backward movies would be similar, but reversed in time, to those acquired from 
the forward movies (strong CrB,F).  For these brain regions, time-reversing the
response time-courses would be analogous to undoing the time reversal editing of
the movie, by editing the brains' responses to the backward movie clips.  We 
expected to find other brain regions, by contrast, that would respond 
differently to the forward and the backward movies.  Specifically, regions that
are optimized for analyzing temporal sequences, should exhibit consistent 
responses to repeated presentations of the forward clips (strong CF1,F2) but not
between the reversed-backward and forward responses (weak CrB,F).  Moreover, 
brain regions that are optimized with respect to the statistical regularities of
natural temporal sequences might not respond consistently to repeated 
presentations of the backward clips (weak CB1,B2).  Our findings suggest that 
most of the posterior cortex responded primarily to the moment-to-moment content
within short time intervals, regardless of the direction of time.  In contrast,
the responses in the left temporal-parietal junction (TPJ) and right 
intraparietal sulcus (IPS) were more consistent during the forward presentations
relative to the backward presentations, indicating that these regions were 
sensitive to the temporal order in which the movie was presented.  Moreover, our
use of two different kinds of movies, one composed of unrelated events and the 
other composed from plot-related events, revealed that these two regions were 
sensitive to the different aspects of time reversal.  The TPJ results, in 
particular, are consistent with previous reports that this brain region plays a
critical role in interpreting the intentions of others (theory of mind).


John Jeka
University of Maryland

Properties of Multisensory Integration and Human Spatial Orientation
Authors:  John Jeka and Tim Kiemel

  While generally accepted as an ongoing adaptive process for the control of 
movement, the properties underlying multisensory fusion are poorly understood.
We approach sensory fusion through two basic processes:  estimation and control.
Dynamic characteristics of body movements identified with time series models 
have led to measures that distinguish estimation from control.  Using these 
measures to then develop models that hypothesize underlying mechanisms has led 
to two important findings.  First, most of the variability observed in movements
during quiet standing can be linked to the process of estimating the center of 
mass from multisensory information.  Standard control theory algorithms cannot 
account for how multisensory information is fused for center of mass estimation 
without a process we refer to as "noisy computation."  Computation noise may 
represent a major distinction between the control of engineered systems and the
neural processes involved in human spatial orientation.  Second, while much  
is known about the information provided by individual sensory modalities,
little is known about the information provided by fused sensory signals.  New 
results provide evidence that velocity information (rather than position or 
acceleration) is the most accurate form of information for upright stance 


Holly Jimison
Oregon Health and Science University

Computer Games as a Tool for Cognitive Monitoring
Authors:  Holly Jimison and Misha Pavel

  Maintaining cognitive performance is a key factor influencing elders' ability
to live independently with a high quality of life.  We have been developing 
computer games with inherent metrics of various dimensions of cognitive 
performance that are also fun for elders to play on a daily basis.  There are
several distinct advantages to using performance information from the computer 
games, as compared to traditional office-based cognitive assessments that are 
typically performed yearly at best.  The frequently sampled indices of cognitive
performance obtained from computer game play on our research system allow us to
estimate variability and track within-subject trends over time.  This offers the
possibility of detecting cognitive changes at an earlier point in time.  
Additionally, within-subject trends are less susceptible to biases due to 
educational level, language proficiency, and cultural background.  In our early
pilot studies, we have also found that measures of performance variability (only
available with frequent samples) are important in differentiating elders who are
cognitively healthy from those with mild cognitive impairment.


Shu-Chen Li
Max Planck Institute for Human Development

Neurocomputational Models of Cognitive Aging:  Relating Neuromodulation to Noise
  Tuning, Neuronal Dynamics and Representation
Authors:  Shu-Chen Li, Timo von Oertzen, & Ulman Lindenberger

  In most developed countries, average life expectancy at birth has increased 
from about 45 years in 1900 to about 75 years in 2000.  Alas, this impressive 
30-year gain in longevity so far has not been matched by equally enhanced 
maintenance of psychological or bodily functions in old and very old age.  In
particular, brain aging remains the most serious threat to successful aging.
Hence, interrelations among sensory, cognitive, and brain aging have become a 
focal point for interdisciplinary research in the 21st century.  This
presentation focuses on neurocomputational models of brain aging at the 
neurochemical level.  Based on empirical findings demonstrating a massive 
reduction of dopaminergic neuromodulation in both striatal and extra-striatal 
regions of the brain throughout adulthood and old age, we model aging-related 
decline in neuromodulation as suboptimal stochastic gain tuning in connectionist
neural networks.  Our computational and mathematical analyses show that 
deficient gain tuning leads to noisier neural information processing, which, in
turn, results in less distinctive cortical stimulus representation and less 
efficient neuronal stochastic resonance (Li et al., Neurocomputing, in press).
Informed by the recently proposed equivalence relation between gain tuning in 
connectionist and gain tuning in firing-rate models (Brown et al., 2005), we 
explore age-related changes in stochastic resonance in terms of functional 
correspondences between connectionist networks and firing-rate models with 
time-varying drift and diffusion coefficients, and tentatively suggest possible
avenues for remediation.


Ken Malmberg
University of South Florida

On the Cost & Benefit of Taking it out of Context: A Model of Directed Forgetting

  Forgetting can occur as the result of unconscious or automatic memory processes
or as the result of their conscious control.  The latter form of forgetting is 
often referred to as suppression, repression, or inhibition, and it is 
investigated in the laboratory using the directed forgetting procedure.  The
authors describe and empirically test the first formal model of directed 
forgetting, implemented within the framework of the Search of Association Memory
Theory (SAM).  The critical assumption is that episodic memory can be suppressed
by a conscious attempt to alter the mental context in which new memories are 
encoded.  The present model accounts for both veridical and erroneous free recall


Tim McNamara
Vanderbilt University

Roles of Egocentric and Allocentric Spatial Representations in Navigation and

  Human navigation must depend on both egocentric and allocentric representations
of the environment.  Skiing through closely-spaced trees, for example, requires 
the computation of precise self-to-object spatial relations to guide locomotion.
But planning a route to a distant goal, and maintaining a sense of orientation in
large-scale environments, would seem to require enduring representations of the 
locations of objects relative to other objects.  Contemporary models of human 
spatial memory and navigation specify roles for both egocentric and allocentric 
representations of space.  The models differ in the nature of those 
representations and in how they are used during navigation.  I will discuss the 
results of four experiments that investigated the nature of the spatial 
representations used in navigation and reorientation by examining the effects of
disorientation on subjects' abilities to point to objects in a recently learned 
environment.  Results suggest that human navigation in familiar environments 
depends heavily on allocentric representations.  Egocentric representations may 
be used for obstacle avoidance, and their role in navigation may be greater when
allocentric representations are not of high fidelity.


Jeff Mulligan
NASA Ames Research Center

Temporal Summation in Trajectory Perception

  Temporal summation in perception can arise from a variety of mechanisms; at
the lowest level, sensory mechanisms integrate stimuli linearly, with time 
constants measured in tens or hundreds of milliseconds.  For longer 
presentations, additional (sublinear) integration may occur, sometimes 
attributed to probability summation, still within preattentive mechanisms.  For
the longest delays, explanations may be couched in terms of memory and cognitive
decision processes, and temporal nonuniformities can be observed such as 
"primacy" and "recency" effects.  Here we examine subjects' ability to 
discriminate clockwise and counterclockwise circular trajectories when 
superimposed on a "pedestal" random-walk trajectory.  An ideal observer exhibits
the familiar square root law, with sensitivity increasing by a factor of 2 when
the stimulus duration is increased by a factor of 4.  Subjects, while less 
sensitive than the ideal observer, show the same dependence on stimulus 
duration, over a range from 250 milliseconds to 4 seconds.  Various models to 
account for subjects' performance are considered.


Tatiana Pasternak
University of Rochester

Remembered Direction Modulates Responses to Visual Motion in MT and Prefrontal
Authors:  Tatiana Pasternak, Daniel Zaksas, and Nick LaMendola.

  There is evidence during a task in which monkeys compare directions of two 
consecutive 500ms stimuli, sample and test, separated by a memory delay, that
neurons in area MT contribute not only to the perception of visual motion but 
also to its storage.  We examined whether these neurons also participate in the 
comparison and decision phases of this task by analyzing neuronal activity in 
response to the test.  We found that responses to the test were significantly 
attenuated when its direction matched that of the preceding sample.  This
attenuation, seen only relatively early in the response (100-200ms after test
onset), decreased with coherence of the remembered sample and disappeared when 
the sample had no net direction (random motion) even though the monkey made a 
decision about its direction.  The absence of modulation by sample direction in
the later part of the test response and following random motion suggests that 
this modulation is unlikely to reflect the decision about the directions of 
sample and test made by the monkey.  It is more likely that this attenuation is
a reflection of the initial comparison of sensory input with a stored template 
of the preceding sample.
   To examine whether decision-related signals are carried by neurons at later
stages of processing, we analyzed motion responses in prefrontal cortex (PFC) 
recorded during the same task.  Many PFC neurons were strongly direction 
selective, though this selectivity arose, on average, ~40ms later than it did in
MT.  PFC responses to the test, like those in MT, were attenuated when the 
preceding sample direction was a match.  However, in PFC this attenuation began
later and persisted through the end of the stimulus, even when the preceding 
sample was not coherent, suggesting that this attenuation may be associated with
the decision or upcoming motor response.  Similarities in the response 
modulation in MT and PFC suggest a shared functional pathway.  Conversely, 
differences in the effect's timing and in the dependence on coherence imply 
unique roles for direction selective neurons in both cortical areas during task


Misha Pavel
Oregon Health and Science University

Home-Based Psychophysics:  Unobtrusive Assessment of Mobility and Cognitive
Authors: M. Pavel, A. Adami, M. Morris, J. Lundell, T. Hayes, H. Jiimison, and
  J. Kaye

  Mobility and speed of processing are important components underlying the 
mental and physical functionality of elders.  Currently, these capabilities of 
elders are generally not assessed until the elder or the family notices 
significant symptoms.  The variability of the subsequent infrequent (once a 
year) measurements interfere with the detection of these changes.  We propose a
set of approaches to measuring mobility, e.g. speed of walking and possibly 
speed of processing, by unobtrusively monitoring elders' activities and by 
measuring response times to specific events.  We illustrate this approach using
subjects' response times to telephone rings while in their normal environment. 
A key idea put forth in this talk is that if the elders' location distribution 
is stable over time, response times can be used to assess the "instantaneous"
speed of walking.  The feasibility of this approach is illustrated using data 
collected in a study in homes of several elderly subjects.


Zygmunt Pizlo
Purdue University

What's New in 3D Shape Perception?

  The last 25 years of research on 3D shape perception have been dominated by
"Marr's Paradigm."  This paradigm refers to several assumptions about the 
underlying perceptual mechanisms:

1.  Perceived 3D shape is a result of information processing.

2.  3D shape is derived from depth relations represented in the viewer-centered
coordinate system, and called 2.5D sketch.

3.  Shapes on the retina provide no useful information and, thus, figure-ground
organization is not involved in 3D shape perception.

It will be shown that assumptions 2 and 3 have to be rejected and the paradigm
has to be changed.  The new paradigm involves the following assumptions:

(a) Perceived 3D shape is derived from 2D retinal shape.

(b) Figure-ground organization is critical because it establishes shapes on the

(c) Three-dimensional metric properties of the perceived shape, which are
missing from the 2D shape on the retina, are produced by the application of a
simplicity principle, not by the reconstruction of depth relations.

The nature of the simplicity principle involved in 3D shape perception will be
discussed and illustrated by examples.


Roger Ratcliff
Ohio State University

Diffusion Model Account of Individual Differences and Training Effects in Choice
  RT Tasks
Authors:  Roger Ratcliff, Anjali Thapar, & Gail McKoon

  We examined fits of the diffusion model to data for three groups of 10 subjects
(college students, 65-75 year olds, and 75-90 year olds) tested for four sessions
on each of four different tasks:  signal detection, letter discrimination, 
brightness discrimination, and recognition memory.  The model was fit to each 
subject individually.  Subjects' decision criteria and boundary separations were
consistent across tasks and, surprisingly, so were drift rates.  In other 
experiments, we examined the effects of training on the two perceptual tasks and
a signal detection task.  There were only small changes in nondecision components
of processing training, generally moderate changes in decision criteria (with
larger effects for older subjects in the perceptual tasks), and generally larger
changes in drift rates for older subjects than young subjects.


Josef P. Rauschecker
Georgetown University Medical Center

Parallel Processing Streams in the Auditory Cortex

  Higher visual areas of humans and nonhuman primates are often thought of as 
being organized into two distinct processing streams for the identification of 
objects and their localization in space.  A similar argument has been made more
recently for higher processing stages in the auditory cortex.  While the 
evidence from nonhuman primates is still being sorted out, an explosion of data
from functional magnetic resonance imaging (fMRI) in humans has corroborated the
existence of dual pathways of higher auditory processing.  Parallel evidence 
from several labs, including our own, for segregated processing of "what" and
"where" information will be presented and compared, and an integration of these
results with studies of language will be attempted.  A particular challenge (and
ambition) in fMRI is to reach the necessary spatial resolution that permits the
visualization of clusters of neurons in a quasi-columnar organization.  Visual 
cortex studies have now reached the resolution of ocular dominance columns.  Our
own studies are aiming to demonstrate a similar clustering and mapping of vowel
and consonant features in human speech.


Paul Sajda
Columbia University

Single-Trial Neuroimaging for Identifying Neural Correlates of Trial-to-Trial
  Behavioral Variability

  Single-trial analysis of electroencephalography (EEG) and functional magnetic
resonance imaging (fMRI) data potentially enables identification and 
characterization of neural activity responsible for trial-to-trial behavioral 
variability -- e.g. variability in response accuracy and/or response time.  In
this talk I will describe current research in single-trial analysis of
neuroimaging data, with emphasis on our work using rapid serial visual 
presentation (RSVP).  I will discuss the signal processing methodologies as well
as our specific findings regarding the timing and localization of neural 
activity associated with perceptual processing and decision making.  Finally I
will discuss implications of such single-trial methods for clinical assessment,
including discrimination of normal and abnormal cognitive aging.


John Serences
Salk Institute

Representing and Reconfiguring Attentional Priority in Human Visual Cortex
Authors:  John Serences and Steven Yantis

  The human visual system seamlessly integrates new sensory information with the
current contents of awareness to provide a continuous and coherent experience of
the visual environment.  However, this is a complex computational task; limits 
on information processing require that some stimuli win cortical representation
at the expense of others.  Local computations in the retina immediately begin to
transform sensory input so that the most salient (i.e. high-contrast) stimulus 
in the scene evokes the strongest neural response.  In addition, current goals
and expectations influence neural activity at almost every level of the visual
system to either enhance or attenuate stimulus-driven modulations based on the 
behavioral relevance of each stimulus.  Thus, the convolution of stimulus-driven
and top-down factors determines the competitive advantage, or the attentional 
priority, of each stimulus in the environment.  Here, we use functional magnetic
resonance imaging to show that regions of human occipital, parietal, and frontal
cortex constitute a series of attentional priority maps in which the magnitude 
of neural activity reflects both the sensory strength and the behavioral 
relevance of objects presented in the contralateral visual field.  Moreover, 
distinct regions of parietal cortex become transiently active whenever a new 
location is selected.  This transient 'shift' signal may play a role in 
initiating new acts of selection by gating input from the sensory environment or
from working memory, where current task goals are maintained.


Shihab Shamma
University of Maryland

The Role of Salient Behavioral Cues in Receptive Field Plasticity

  How does the auditory cortex adapt to different salient cues in multiple 
auditory tasks?  We studied task-related dynamic cortical change while the 
animal was engaged in: (1) spectral tasks such as tone detection or tone 
discrimination or in (2) temporal tasks such as gap detection or tone duration 
discrimination and also in (3) spectrotemporal tasks such as FM detection and FM
directional discrimination.  By measuring the spectrotemporal receptive field 
(STRF) of single neurons in A1, we could quantitatively describe the STRF 
changes in shape which resulted as the animal went from a passive condition to 
different active behavioral conditions.  We initially trained five ferrets, 
using aversive conditioning, to detect variable tonal targets against a 
background of rippled noise stimuli.  They quickly learned a variety of other 
tasks, all of which were variations on a basic task paradigm, in which the 
animal learned to discriminate between a set of reference stimuli and distinct 
target stimuli.  We studied adaptive responses by comparing STRFs in the awake,
but non-behaving ferret (passive condition) vs. STRFs measured while the ferret
performed various auditory tasks.  The STRFs were derived using standard reverse
correlation techniques.  Neuronal responses to the same ripple stimuli were then
measured in the context of an active detection or discrimination task.  We
recorded single units in A1 of trained ferrets and found that >60% of neuronal 
STRFs changed during behavior in a consistent pattern.  Our working hypothesis 
is that these adaptive changes may serve to enhance task performance.  For
example we would predict that in (1) spectral detection task, excitatory 
responses in the STRF would be enhanced and inhibitory responses reduced at the
tonal target frequency; (2) spectral discrimination task, responses would change
in an opposite direction: at the reference frequency compared to the target 
frequency (differential effect); (3) temporal detection task: STRF excitatory 
regions would be sharpened, and latency would decrease; (4) spectrotemporal 
tasks: changes would occur both along the spectral and temporal dimensions.  We
shall illustrate these effects at the single-unit and population levels and 
critically examine these predictions.


Steve Shevell
University of Chicago

Binding Color to Form: New Insights from Binocular Rivalry
Authors:  S. K. Shevell and S. W. Hong

  An object has a particular size, shape, color, location and depth but how are
these features integrated to give a unified percept of the object?  This is the
question of binding, which was investigated using binocular color rivalry.  
Previous studies show that color rivalry can be resolved by dominance 
(perception of one or the other color of a rivalrous pair) or fusion (binocular
mixture of the two colors).  Here, we report that both colors of a rivalrous 
pair can be seen simultaneously in separate spatial locations.  That is, the hue
of the "suppressed" stimulus is perceived in a visual area that does not 
correspond to its location in the visual field.  In a prototypical experiment, a
vertical equiluminant red/gray grating was presented to one eye and a horizontal
equiluminant green/gray grating to the other eye.  During 60 seconds of 
rivalrous viewing, an observer often perceived a red/green grating, oriented 
either vertically or horizontally.  Control experiments rule out vergence errors
or simultaneous contrast.  The results show that color and form rivalry are 
processed separately, and that the neural representation for color from the 
form-suppressed eye can be expressed with full vividness and saturation but in a
location distinct from its retinotopic position.  This is a fundamental error in
binding color to form, and demonstrates the neural representation of color prior
to binding is not strictly localized to the stimulated area of the visual field.


Richard Shiffrin
Indiana University



Philip L. Smith
The University of Melbourne

An Integrated Theory of Attention and Decision Making in Visual Signal Detection

  Response times to detect luminance stimuli are typically faster if the stimuli
appear at locations to which attention has been summoned by means of spatial 
cues.  However, with well-localized stimuli, attention typically increases 
detection accuracy only if the stimuli are followed by backward masks, but not 
otherwise.  These effects are predicted by a theory that attributes them to an 
interaction among visual encoding, masking, short-term memory, attention, and 
decision mechanisms.  The theory ascribes a central role to attention in the 
process of visual short-term memory trace formation and incorporates an 
explicit, dynamic model of how trace formation occurs.  The theory correctly 
predicts the effects of attentional cues and backward masks on accuracy and on 
the distributions of response times for correct responses and errors.  The
empirical data can be predicted by either of two attention models, an orienting
model and a gain model, and by either of two decision models, a single and a 
dual diffusion model.

George Sperling
University of California, Irvine

A Neurally-Based Theory of Binocular Combination
Authors:  Jian Ding and George Sperling 

  In binocular combination, light images on the two retinas are combined to form
a single "cyclopean" perceptual image, in contrast to binocular rivalry which 
occurs when the two eyes have incompatible ("rivalrous") inputs and only one
eye's stimulus is perceived.  We propose a computational theory for binocular 
combination with two basic principles of interaction:  In every spatial 
neighborhood (1) each eye exerts gain control on the other eye's signal in 
proportion to the contrast energy of its own input and (2) each eye additionally
exerts gain control on the other eye's gain control.  For stimuli of ordinary 
contrast, when either eye is stimulated alone, the predicted cyclopean image is 
the same as when both eyes are stimulated equally, coinciding with an easily 
observed property of natural vision.  The gain-control theory is contrast 
dependent:  Very low-contrast stimuli to the left- and right-eye add linearly to
form the predicted cyclopean image.  The intrinsic nonlinearity manifests itself
only as contrast increases.  To test the theory more precisely, a horizontal 
sinewave grating of 0.68 cycles/degree is presented to each eye.  The gratings 
differ in contrast and phase.  The predicted (and perceived) cyclopean grating 
also is a sine wave; its apparent phase indicates the relative contribution of 
the two eyes to the cyclopean image.  For 48 measured combinations of phase and 
contrast, the theory with only one estimated parameter accounts for 95% of the 
variance of the data.  Conclusion:  A simple, robust, physiologically plausible 
gain-control theory accurately describes an early stage of binocular combination.


Courtney Stein
Dartmouth College

The Role of Metaphor in Causal Reasoning
Authors:  Courtney Stein and George Wolford

  There has been much debate over how abstract concepts such as love, justice, 
and time are represented and understood.  The most recent evidence in support of
a metaphoric structuring view suggests that abstract concepts are initially 
structured by more concrete concepts but that these metaphorical mappings come 
to be stored in the abstract domains after frequent use.  The current series of
studies has attempted to determine (1) if metaphorical concepts are semantically
related, (2) what factors predict, when faced with concepts represented by 
multiple metaphors, which metaphor will be used, and (3) whether or not these
metaphoric structures influence other areas of cognition such as causality.


Shun-nan Yang
Smith-Kettlewell Eye Research Institute

Neurons That Call Balls And Strikes: A Neural Network of Visuomotor Control
  Based On Neural Interaction

  To survive in the real world, one's sensorimotor responses have to adhere to 
rules and strategies.  Little has been revealed about the underlying neural 
processes supporting such capacity.  Here we reported some modeling work based 
on empirical studies in which nonhuman primates performed a novel oculomotor 
baseball while the neural activity was recorded from their supplementary eye 
fields (SEFs), a neural region located at the dorsomedial wall of the frontal 
cortex that was traditionally defined as part of the supplementary motor area. 
In the task, a circular target moved from the edge of the screen to the center 
with certain degrees of horizontal deviation.  The subjects had to hold their 
gazes at a central fixation cross that was surrounded by a visible square plate
while the target moved across the screen.  The primates had to determine whether
the trajectory of the moving target was going to intercept the plate: when it 
was to intercept the plate, referred to as a strike trial, the gaze had to be 
shifted to track the target once it crossed the near edge of the plate; when it
was not, referred to as a ball trial, the gaze had to be held until the target
disappeared.  Targets with trajectories closer to the edge of the plate made the
task more difficult.  Our results show that different types of neurons (strike-
and ball-related) in the SEF responded differently for ball and strike trials 
before plate crossing.  Task difficulty affected whether and when the separation
in neural activity for ball and strike trials would occur.  The strike- and
ball-related activity did not respond differentially for eye movements of 
different directions.  These results indicate that a single act of sensorimotor
control may require the involvement of both movement- and inhibition-related 
neurons.  Based on these results, we propose a neural model of visuomotor 
decision based on the interaction of different neurons.  In this model, the
decision of initiating a movement is made by removing or lowering a general 
inhibitory signal that temporally blocks a nonspecific movement-related signal 
associated with the intended saccade.  The implication of this type of neural
network on rule-based behavioral control will be discussed.


Li Zhaoping
University College London

Bottom-Up Saliency by a Single Stage V1 Process Rather Than a Combination of 
  Feature Maps
Authors:  Li Zhaoping & Keith A. May

  Attentional selection of visual inputs integrates the top-down and the 
bottom-up mechanisms.  Saliency, defined as the extent to which a stimulus 
attracts attention, provides a useful platform to study the attentional 
mechanisms.  Highly salient visual locations, e.g., a red item among green ones,
or a horizontal bar among vertical ones, attract attention through bottom-up or
stimulus-driven mechanisms.  The standard view assumes that visual inputs are
processed by separate feature maps such as red and green maps, each for a 
feature value in a few basic dimensions like color and orientation, which are 
then summed to a spatial master map of bottom-up saliencies.  Any assumptions 
about this bottom-up saliency map can greatly influence assumptions about the 
top-down mechanisms, and should therefore be confirmed experimentally.  We show,
using psychophysical experiments on visual search and segmentation tasks, that 
summations or other simple combinations of the feature maps cannot explain the 
bottom-up saliency.  Instead, a single stage computation by the primary visual 
cortex, using intracortical interactions (Li TICS 2002), is adequate to explain
the data, including the aspects of the data often associated with visual 
grouping.  While V1 mechanisms suffice to account for our data, our framework 
does not exclude other cortical areas from contributing additionally to 
computing the bottom-up saliency, and we will discuss when and how it could 
happen.  We will discuss how our work relates to other works, and its 
implications on the top-down attentional mechanisms.