Annual Interdisciplinary Conference, Jackson, WY, February 3-8, 2002. ABSTRACTS ================================================================================ Greg Appelbaum University of California, Irvine TBA ================================================================================ Harry P. Bahrick Ohio Wesleyan University A Metacognitive Theory of the Spacing Effect Current theories of the spacing effect are based upon encoding variability or diminished processing. These theories explain why massed presentation of memory content is less effective than presentations separated by short time intervals. They do not explain why presentations spaced at 60 days are more effective than presentations spaced at 30 days. We present data that support an explanation based upon conscious monitoring and control of encoding strategies. ================================================================================ Tom Busey Indiana University Set-Size Effects in Identification and Localization: Theory and Data Authors: Tom Busey and John Palmer The effect of divided attention is different for identification and localization. We ask whether this difference is due to perceptual processing capacity or to the decision process. Using visual search, we measured set-size effects for finding a target grating (left-leaning) among distractor gratings (right-leaning). The identification task was yes-no detection and the localization task was to specify the target location. The observed set-size effects were larger for localization than for identification. This difference was shown for several spatial and temporal frequencies and controls ruled out explanations based on task difficulty, sensory factors, and response measures. The different decision requirements for the two tasks was modeled using signal detection theory and by assuming unlimited capacity for both tasks. This model predicted much of the observed difference between tasks. Thus, the observed difference may be due to the differences in the decision process. ================================================================================ Lawrence K. Cormack University of Texas at Austin Relating Image Properties and Eye Fixation Author(s): L. K. Cormack, U. Rajashekar, A. C. Bovik and W. S. Geisler Efficient selection of fixation points must ultimately be based on image data, yet the image properties that attract gaze are largely unknown. It is thus difficult to, e.g., implement good fixation strategies in foveated artificial vision systems. We therefore sought to elucidate the image properties that attract gaze by combining accurate eye tracking with modern image analysis techniques. In one paradigm, subjects searched for targets embedded in 1/f noise. The noise in a region of interest (ROI) around each fixation was averaged over many trials yielding gaze attraction images analogous to the discrimination images of [1]. In a second paradigm, subjects studied several hundred natural images, and ROIs around each fixation were accumulated. These ROIs were then subject to Principle Components Analysis (PCA) to reveal commonalities (other techniques including Independent Component Analysis are currently being explored). Gaze attraction images indicate that subjects fixate likely targets (as opposed to randomly sampling the image), and they often search for a characteristic feature instead of the entire target. Results from the second paradigm indicate that statistics of ROIs in natural scenes are often different from those of randomly selected regions from the same images. The pixel-averaging technique of [1] can be successfully combined with accurate eye tracking to reveal image structure that attracts gaze. This technique can potentially reveal image structure that draws fixation in a wide variety of search tasks. Accurate eye-tracking can also be combined with image analysis techniques such as PCA to reveal statistics of natural images at the point of fixation. This promises to complement recent work on natural image statistics and their relationship to the neurophysiological properties of the visual system. [1] Beard and Ahumada (1998) SPIE Proc. Human Vis. and Elec. Im. III, v3299. ================================================================================ Denis Cousineau University of Montréal Blocking the Search and Other Illusory Conjunctions I will present two sets of intriguing results. The experiment involved a visual search for one of four possible artificial targets among from 1 to 4 other artificial distractors. The subjects were well trained (75 hours). Unbeknownst to them, we manipulated the moment at which diagnostic information was presented using asynchronous presentation with very rapid rates. Apparently, in one condition, the participants "froze," not initiating the search as soon as the first object is presented, questioning the pop-out assumption. In addition, in the final session, we modified the presentation so that after half of the information was presented, it started disappearing while the second half was appearing. Whereas the errors were almost non-existent before (near 2.5%), it now jumped to 25%, suggesting massive illusory conjunctions. This questions our above questions on pop-out and calls for a better definition on what is the relevant level of information in the stimuli. ================================================================================ Simon Dennis University of Queensland Category Effects in Episodic Recognition: Are Item Noise Accounts Sufficient? Authors: Angela Maguire, Michael Humphreys and Simon Dennis Item noise accounts assert that interference in episodic recognition is generated by overlap between the representation of a test item and the items that appeared in the study list. If a study list contains items from categories, such accounts predict that discriminability between targets and distractors from a presented category should degrade as the number of items in the category increases. In addition, item noise models have no mechanism for distinguishing between taxonomic categories, such as Brisbane, Sydney and Melbourne, and associate categories such as bed, snooze and night. Furthermore, item noise models suggest that within category discrimination should be easier than between category discrimination where both categories were presented at study. In a series of eight experiments, the nature of the category (taxonomic versus associative), the number of items from the category (one versus five), presentation type (blocked versus distributed) and presentation location within the category (first item versus last item) were manipulated. Recognition memory was assessed using both yes/no and two-alternative forced choice procedures, where the distractor items were either the most representative exemplar from the taxonomic categories, or the prototype from the associative categories. For taxonomic categories, while there were strong bias effects as a function of number of items in the category in yes/no recognition, the only effect on discriminability was a lower A' for items in the distributed, first presentation, five item category condition. Note this is the one condition for which participants may be unaware that the study list contains categories during encoding. Furthermore, in the forced choice paradigm, there was no effect of the between/within category manipulation. For associative categories, again there were strong bias effects as a function of number of items in yes/no recognition. However, there was also a decrease in both yes/no and forced choice discriminability as the number of items increased. In addition, there was a decrease in yes/no discriminability for the last item in the five item categories for blocked presentation. Such a pattern is inconsistent with item noise accounts, but would occur if participants were activating the prototypical item during study (i.e. an implicit associative response), storing an item to context association, and then using the item to retrieve contexts at test (i.e. a context noise model). =============================================================================== Nick Donnelly University of Southampton Searching for Targets in Configurations: Reaffirmation of the Role of Collinearity in Determining Search Efficiency Donnelly, Humphreys and Riddoch (1991; see also Donnelly, Weekes, Humphreys and Albon, 1998, and Humphreys and Donnelly, 2000) reported that search for vertex targets was more efficient when the configuration of distractor vertices formed a shape on absent trials when they grouped but did not form a shape. The authors attributed the difference between conditions to the rapid computation of collinearity between distractors, via edge-interpolation, although other interpretations were possible. In the present study, an alternative account based on shape-template matching is examined in a series of experiments using Kanisza-type inducers rather than the line vertices used in earlier studies. The results show that Kanisza-type inducers do not support the efficient detection of targets. The data are consistent with edge-interpolation, but not with shape-template matching, mechanisms enabling efficient search. Donnelly, N., Humphreys, G. W. and Riddoch, M. J. (1991). Parallel computation of primitive shape descriptions. Journal of Experimental Psychology: Human Perception and Performance, 17, 561-570. Donnelly, N. Weekes, B., Humphreys, G. W. and Albon, A. (1998). Processes involved in the computation of a shape description. Journal of Experimental Psychology: Human Perception and Performance, 24, 1119-1130. Humphreys, G. W. and Donnelly, N. (2000). 3D constraints on spatially parallel shape perception. Perception and Psychophysics, 2000, 62, 1060-1085. ================================================================================ Barbara Dosher University of California, Irvine TBA ================================================================================ Mario Fific Indiana University Parallel vs Serial Processing and Individual Differences in Visual and Memory Search Task Revealed by the Systems Factorial Technology Author(s): Mario Fific and James Townsend Methodology involving factorial variation in order to determine mental architecture and to assess processing capacity has been greatly expanded over the past several decades. However, it has never been adapted to study the realm where much of the interest began in the 1960s: short-term memory search. We present a new method of manipulating probe-to-memory item processing speed and our initial results for loads n=2 for both memory and visual search tasks. Three variables were manipulated in this experiment: number of processing elements (2), phonemic (memory search) and graphemic (visual search) dissimilarity of a target to the particular memorized element (high, low) and temporal characteristics of the tasks (short and long ISI in memory search and duration of target exposure in visual search). We employ the recent results involving the distribution functions rather than means alone. Our results in memory search suggest that some observers really are serial whereas others are strongly parallel. Thus, these fine grained analyses portend quite striking individual differences in this basic cognitive task. However results from visual search suggested architecture that is not compatible to independent channels parallel or serial processing. Further, less individual differences between observers have been observed. ================================================================================ Greg Francis Purdue University Quantitative Models of Visual Backward Masking: They are all Correct...No wait...They are all Incorrect Visual backward masking refers to a class of phenomena where the percept of a briefly presented target stimulus is greatly reduced by the subsequent presentation of a mask stimulus. Backward masking has been studied for over a century, so it was significant when Di Lollo et al. (2000) reported experimental findings on backwards masking that they concluded were inconsistent with all current theories of masking. I show that their conclusion is not valid. Analysis and simulations of four current quantitative models of backward masking show that three of the four models can account for the experimental data that Di Lollo et al. used to reject existing models. The fourth model could be easily altered to account for the data. Thus all of the models are correct. When the stimulus onset asynchrony (SOA) between the target and mask stimuli is experimentally varied, one can measure reports of the target percept as a function of SOA to produce what is called a masking function. Depending on the properties of the target and the mask, the masking function may be monotonic (with the maximum masking occurring for SOA equal zero) or u-shaped (with the mask having its strongest effect for a positive SOA). Analysis of the models demonstrates that the models predict that monotonic masks should occur when the mask is strong and u-shaped masks should occur when the mask is weak. Thus, for a fixed target and task the masking functions should never intersect: monotonic masking functions must always lie below u-shaped masking functions, regardless of the properties of the mask. A new experiment rejects this prediction, thereby indicating that all current quantitative models of backward masking are incorrect. =============================================================================== Bill Geisler University of Texas Bayesian Natural Selection and the Evolution of Perceptual Systems In recent years, there has been much interest in characterizing the statistical properties of natural stimuli in order to better understand the design of perceptual systems. A fruitful approach has been to compare the processing of natural stimuli in real perceptual systems with that of ideal observers derived within the framework of Bayesian statistical decision theory. While this form of optimization theory has provided a deeper understanding of the information contained in natural stimuli as well as of the computational principles employed in perceptual systems, it does not directly consider the process of natural selection, which is ultimately responsible for design. Here we propose a formal framework for analyzing how the statistics of natural stimuli and the process of natural selection interact to determine the design of perceptual systems. The framework consists of two complementary components. The first is a maximum-fitness ideal observer, a standard Bayesian ideal observer with a utility function appropriate for natural selection. The second component is a formal version of natural selection based upon Bayesian statistical decision theory. Maximum-fitness ideal observers and Bayesian natural selection are demonstrated in several examples. We suggest that the Bayesian approach is appropriate not only for the study of perceptual systems but for the study of many other systems in biology. =============================================================================== Joetta Gobell University of Califoria, Irvine TBA ================================================================================ Jason M. Gold Indiana University Characterizing Visual Memory Decay with External Noise Authors: J. M. Gold, R. Sekuler, R. F. Murray, A. B. Sekuler and P. J. Bennett The ability to recall visual patterns from short-term memory often declines with the passage of time. There are two possible sources for this decline. One possibility is that internal variability or "noise" introduced during the process of storage and retrieval grows over time. A second possibility is that internal noise remains constant, but the non-stochastic parts of the internal operations performed during storage and retrieval become less optimal over time. One way to distinguish between these possibilities is to measure changes in performance as varying amounts of externally added noise are introduced into a task. We applied this "external noise masking" technique to a visual pattern discrimination task that involved the use of short-term visual memory. The task required observers to perform same/different discriminations with pairs of randomly generated noisy textures, separated by one of three time delays (100, 500, or 2000 ms). Increasing the delay between stimuli had little or no effect on internal noise, but reduced the efficiency of the non-stochastic parts of the internal operations by about 200%. In a subsequent experiment, externally added noise was used to determine which spatial frequencies observers relied upon to perform the pattern matching task at the long and short delays. The results showed that at least part of the reduction in efficiency with longer delays was due to observers' greater reliance on uninformative high spatial frequencies outside of the stimulus band. Possible explanations for the shift to higher frequencies will be discussed. ================================================================================ Todd Handy Dartmouth College Visual Sensory Gain in Action-related Processing Authors: Todd C. Handy, Sarah Ketay, Scott T. Grafton, & Michael S. Gazzaniga Research on sensory gain in early visual cortex is typically predicated on the assumption that sensory gain serves to enhance the perception of stimuli in attended visual field locations. In a recent series of event-related potential (ERP) experiments we have been investigating whether sensory gain may be involved as well in visuomotor processing. In particular, evidence suggests that there is a tendency to orient one's visual spatial attention towards objects that afford manual interactions (i.e., "graspability"), relative to objects that don't. The effect -- measured at the level of the lateral occipital P1 ERP component -- is non-volitional and is stronger in the right visual field. The results have direct implications for attention-related models of sensory gain, object competition, and visuomotor processing. =============================================================================== Erin Harley University of Washington Is Hindsight 20/20?: Evidence for Hindsight Bias in Visual Perception Tasks Authors: Erin Harley and Geoffrey Loftus This research addresses whether or not a hindsight bias exists for visual perception tasks, and if so, under what conditions. In two experiments participants were asked to search for digits hidden in visual noise. Experiment-1 participants were given information to aid them in the search process (digit, location, or both), and then estimated what their performance would have been without the aid. Estimates were compared to performance on trials in which no aid was given. Experiment-2 participants were shown the hidden digits (i.e., were provided outcome information) and predicted the performance of other observers who received no outcome information. A traditional hindsight bias was found in Experiment 1 and in Experiment 2 when outcome information was presented last, but a reverse bias was found in Experiment 2 when outcome information was presented first. =============================================================================== David Heeger Stanford University Neural Basis of the Motion Aftereffect Authors: David Heeger and Alex Huk Several recent fMRI studies have reported response increases in human MT+ correlated with perception of the motion aftereffect (MAE). However, MT+ responses can be strongly affected by attention, and subjects may naturally attend more strongly during the MAE than during controls without MAE. We found that requiring subjects to attend to the motion of the stimulus on both MAE and control trials produced equal levels of MT+ response, suggesting that attention may be a major confound in the interpretation of previous fMRI MAE experiments; in our data, attention appears to account for the entire effect. After eliminating this confound, we sought to measure direction-selective motion adaptation in human visual cortex. We observed that adaptation produced a direction-selective imbalance in MT+ responses (as well as earlier visual areas including V1), and yielded a corresponding psychophysical asymmetry in speed discrimination thresholds. These findings provide physiological evidence of a population-level response imbalance related to the MAE, and quantify the relative proportions of direction-selective neurons in human cortical visual areas. ================================================================================ David E. Huber University of Colorado, Boulder How is the Brain Able to Identify Items with Minimal Interference from Prior Presentations? Author(s): David E. Huber and Randall C. O'Reilly In perceptual areas of the brain, rapid neural adaptation is ubiquitously observed. We propose that this rapid adaptation serves a crucial function, reducing interference between items presented in succession. The biological mechanism of synaptic depression, which causes a short-term, transient decrease in synaptic efficacy as a function of neural activity, is likely a major contributor to this adaptation. Thus, we specifically propose that synaptic depression reduces item-specific interference by diminishing the activation of already-identified items. We derive a rate-coded version of synaptic depression and implement this activation function within a neural network processing hierarchy. In the hierarchy, early levels integrate and identify more rapidly than later levels. Mutual inhibition at each level of the hierarchy explains a variety of u-shaped general interference results in which, following the presentation of a first item, identification for a dissimilar second item is spared for short and long delays but not for intermediate delays; because synaptic depression results in an n-shaped identification response, this produces u-shaped interference. We account for general and item-specific interference effects in the domain of brief, near-threshold word identification and discuss broad applications of the theory to other perceptual interference phenomena. =============================================================================== Petr Janata Dartmouth College Cognitive Binding of Auditory and Visual Objects into Semantic Concepts Authors: Petr Janata and Reginald B. Adams This paper presents a brief history and introduction to the idea of environmental sounds representing "auditory objects" in semantic memory, along with new behavioral and functional neuroimaging data from a study in which subjects performed a common name verification task. This task requires that subjects verify the match between a presented written word and either a corresponding sound or picture. It therefore requires integration of semantic concepts that are accessed via the lexicon and semantic concepts accessed via sensory-specific object representations. While the mechanisms underlying this integrative process have received significant attention in the visual domain, the auditory parallel, i.e. the manner in which environmental sounds are bound with associated semantic concepts, has not. Consistent with a theory of the inferior frontal gyrus that emphasizes semantic, phonological, and working memory functions, our results support a complementary hypothesis that the IFG region of the prefrontal cortex in both hemispheres is also involved in binding sensory-specific object representations into polymodal aggregates, i.e. more general concepts in semantic memory. =============================================================================== Geoffrey R. Loftus University of Washington Perceptual Interference in Face Recognition Bruner and Potter (Science, 1964) measured observers¹ ability to recognize pictures of objects, initially seen blurred, then gradually focused. Pictures initially seen very blurred were harder to eventually recognize than pictures initially seen moderately blurred. I report research that replicates this ³perceptual interference effect² using recognition of celebrity faces. This work is an offshoot of the idea that a distant face can be represented by blurring: Combining face size, face distance, and the human modulation transfer function allows construction of a theoretically equivalent filtered (blurred) face. Therefore one can visually represent a face seen at any given distance either by shrinking it to simulate the visual angle or blurring it to simulate the spatial-frequency composition corresponding to the distance. Representing distance by either blurring or shrinking produced the perceptual-interference effect: Faces beginning at 500 feet away needed to move closer for eventual recognition than faces beginning at 250 feet away. =============================================================================== Zhong-Lin Lu University of Southern Califonria TBA =============================================================================== Kenneth J. Malmberg Indiana University Effects of Similarity, Repetitions, and Normative Frequency on Memory: Registration With Learning? Author(s): Kenneth J. Malmberg, Jocelyn E. Holden, Richard M. Shiffrin Explicit memory improves as the number of times an item is studied increases. Given this, the situation that Hintzman, Curran, and Oppy (1992) described -- in in which repetitions of an item apparently do not much improve recognition performance -- is noteworthy. In this paradigm, subjects study items up to 25 times at spaced intervals. At test, old items (targets; e.g., "TOAD"), new items (foils; e.g., "ROPE"), and items that are perceptually very similar to studied items (similar foils; e.g., "TOADS") are presented, and the task is to determine how many times the test item was studied (i.e., a judgment of frequency or JOF). JOFs for targets and similar foils increase steadily as the number of repetitions increases, but the ability to discriminate between targets and similar foils does not improve after the first 2 or 3 presentations. Hintzman et al. (1992) termed this finding "registration without learning" because the increase in JOFs indicate that each presentation is "registered" in memory, but the features critical for discriminating targets and similar foils are not "learned" after the first 2 or 3 presentations. In this study, we test two hypotheses: 1.) The registration without learning observed by Hintzman et al. was due to ceiling effects, and 2.) The word-frequency effect for recognition is due to differences in the way high- and low-frequency words are represented in memory. A "dual-process" REM model (cf. Malmberg, Steyvers, Stephen, & Shiffrin, in press; Shiffrin & Steyvers, 1997, 1998) is described that accounts for our findings. ================================================================================ Richard Murray University of Toronto Snapshots of Perceptual Organization Authors: Richard F. Murray, Jason M. Gold, Patrick J. Bennett, and Allison B. Sekuler Visual patterns can differ greatly in their perceived simplicity and their perceptual organization, even if they are very similar in their local properties (e.g., a Kanizsa square vs. a random placement of Kanizsa inducers). These differences in perceptual organization have an enormous influence on how well observers perform visual tasks. Why is this? How do differences in perceptual organization affect performance of visual tasks, and what does this tell us about visual processing? To answer this question, we carried out psychophysical experiments using the response classification method, which reveals how an observer uses various parts of a visual stimulus to perform a task. We found that observers used only one or two simple features of a stimulus to perform visual tasks, even if the stimulus had many features that gave information relevant to the task. In stimuli with simple perceptual organizations, observers used large parts of the stimuli that were perceptually grouped into a single edge or contour. In stimuli that appeared as a collection of unorganized fragments, observers used only smaller, local edges or contours. These results indicate that the features observers use to perform visual tasks are strongly influenced by perceptual organization: observers use only one or two stimulus features, and the features that are available depend on the perceptual organization of the stimulus. I will also discuss some surprising inefficiencies that these response classification experiments revealed about observers' judgements of even simple stimulus properties, such as the orientation of edges. =============================================================================== Thomas Nelson University of Maryland Analysis in Metacognition and Memory: Data Collection vs. Data Analysis =============================================================================== Miguel Nussbaum Catholic University of Chile Developing Technical Support for Establishing Face-to-Face Relations Between People Who Do Not Know Each Other Author(s): Miguel Nussbaum and Roberto Aldunate Our question is how to support social encounters between people who don't know each other, are near each other, and share similar mental models and matching needs. For example, suppose there are several people in a bus, each with a PDA (personal digital assistant, e.g., a Palm Pilot, Ipaq, etc.), and that each PDA is equipped with a short-range data communicator (e.g., Blue Tooth, WI-FI). Through the wireless network, the PDAs could communicate (i.e., send msgs to each other) and form an Ad Hoc network within the bus. That is, each PDA runs a piece of software named Agent; each of these Agents communicates and interchanges data. In this way you would have a distributed net of agents that is running inside the bus. Such a scheme should allow people mobility, using a distributed net of Agents, that constantly enables people in their vicinity to monitor and to communicate with each other. Mental models are built using principles of social psychology. Agents constantly monitor other agents' mental models and, in the models' intersection, heuristically verify if their current necessities are shared. The network is built using portable devices (Compaq Ipaq) with a wireless network communication card, Wi Fi (IEEE 802.11 B). Since Wi Fi is a short range network (around 300 feet), it restricts interactions to Agents that are physically close. In this way technology becomes a facilitator of casual encounters and not only as a communication medium. The current status of this project is to identify the constituents of the mental models on one side, and the construction of the distributed Agents in a wireless network on the other. ================================================================================ Tatiana Pasternak University of Rochester Activity of MT Neurons is Affected by Remote Visual Stimuli Used in a Working Memory Task Area MT neurons have relatively small receptive fields representing the contralateral hemifield and have been shown to play an important role in processing of visual motion. Recent lesion, microstimulation and psychophysical studies have implicated MT in temporary storage of visual motion (Bisley and Pasternak, 2000; Bisley et al., 2001; Zaksas et al., 2001). We recorded the activity of MT neurons during the performance of a working memory task in which the remembered (sample) and the comparison (test) stimuli were placed in opposite hemifields. This manipulation allowed us to determine whether the activity of MT neurons during the performance of the memory task is strictly retinotopic and generated locally or reflects the top-down influences of cortical areas that have access to the information from the entire visual field. During the performance of the task, many MT neurons altered their activity when stimuli were presented in the portion of the visual field contralateral to their receptive fields. After the sample was presented in the receptive field, MT neurons showed transient elevated activity during the early portion of the delay, followed by decreased activity in the middle portion of the delay. Furthermore, many neurons showed activation to the test stimuli presented in the remote location. In contrast, many MT neurons showed inhibition during the presentation of the sample in a remote location. Under these conditions, activity early in the delay was reduced or absent but increased during the middle portion of the delay. Since area MT is strongly retinotopic, changes in the neural activity produced by stimuli presented in the contralateral hemifield suggest that this activation is unlikely to be generated locally through connections within MT. Rather, this activity may be indicative of the top-down influences of cortical areas with access to information from the entire visual field. These areas, along with area MT, may form the circuitry underlying the ability to remember visual motion. Bisley JW, Pasternak T (2000) The multiple roles of visual cortical areas MT/MST in remembering the direction of visual motion. Cerebral Cortex 10:1053-1065. Bisley JW, Zaksas D, Pasternak T (2001) Microstimulation of Cortical Area MT Affects Performance on a Visual Working Memory Task. J Neurophysiol 85:187-196. Zaksas D, Bisley JW, Pasternak T (2001) Motion information is spatially localized in a visual working-memory task. Journal of Neurophysiology 86:912- 921. =============================================================================== Misha Pavel Oregon Health & Science University Augmented Cognition: New Design Paradigm We describe a novel approach to the design of future systems and devices that is based on the idea of augmentation of human cognitive abilities. These resulting systems are expected to significantly improve human performance on a number of demanding, high-level cognitive tasks. In this presentation we will first compare our approach to the traditional design of human-computer interfaces that has been focused on specific tasks such as editing, searching, etc. In contrast, our starting point is based on the characterization of human cognitive limitations, and on the quantitative modeling of cognitive bottlenecks and limitations. These quantitative models are then used to design interfaces that could alleviate these limitations. We will argue that the notion of an ideal observer, e.g. as used in signal detection theory, that has been used to characterize human limitations, can be extended to support the design of cognitive amplifications devices. We will discuss several examples that are related to the human limitations of resources including attention, fusion of abstract information, and incorporation of uncertainty in complex decisions. An additional key idea of this approach is that the human operator is typically not aware of his sub-optimal strategies until his performance deteriorates and he makes errors. To augment an operator's cognition, we continuously monitor his behavior, help him allocate his resources, and make abstract or missing information available to him in a convenient perceptual form prior to his own awareness of the problem. ================================================================================ Adina Roskies MIT Are Ethical Judgments Intrinsically Motivational? Lessons from "Acquired Sociopathy" Metaethical questions are typically held to be a priori, and therefore impervious to empirical evidence. Here I examine the metaethical claim of belief-internalism, the position that moral beliefs are intrinsically motivating. I argue that belief-internalists are faced with a dilemma. Either their formulation of internalism is so weak that it fails to be philosophically interesting, or it is a substantive claim, but can be shown to be empirically false. I then provide evidence for the falsity of substantive belief- internalism. I describe a group of brain-damaged patients who sustain an impairment in their moral sensibility: although they have normal moral beliefs and make moral judgments, they are not inclined to act in accordance with those beliefs and judgments. Thus, they are walking counterexamples to the substantive internalist claim. In addition to constraining our conception of moral reasoning, this argument stands as an example of how empirical evidence can be relevantly brought to bear on a philosophical question typically viewed to be a priori. ================================================================================ Richard M. Shiffrin Indiana University Relating Recognition Memory to Categorization with Bayesian Modeling The REM model for recognition memory was developed on the assumption that memory storage is 'noisy' (incomplete and error prone), but that retrieval is 'optimal' in the sense that the best decision is made in the face of poor data in memory. The theory is instantiated by assuming memory traces are separate vectors of feature values, and that retrieval operates by comparing the test vector to the traces, feature value by feature value. The result of such matching is the assignment to each trace in memory of a likelihood ratio (lambda), giving the probability that the trace had been stored as a copy of the test item, divided by the probability that the trace had been stored due to something else. The optimal Bayesian decision is the average of the lambdas, called the odds (with the default criterion being an odds of 1.0). A similar approach can be used for categorization studies, with the same lambdas playing a key role. The relations between such models for the two paradigms, and some relevant data from a study of categorization based on animacy, are presented. ================================================================================ George Sperling University of California, Irvine TBA ================================================================================ Mark Steyvers Stanford University Bayesian Networks and Causal Reasoning Authors: Mark Steyvers, Josh Tenenbaum and Eric-Jan Wagenmakers The ability to infer causal relationships is crucial for scientific reasoning and, more generally, for our understanding of the world. Recently, Bayesian networks (e.g., Pearl, 2000; Glymour & Cooper, 1999) have been proposed as a theoretical framework for how causal knowledge can be represented and learned on the basis of observational and/or experimental data. In this research, we provide a detailed computational account of human learning of causal networks. We will show that people can learn complex causal networks on the basis of both observational trials and intervention (experimental) trials where subjects are allowed to manipulate a variable. =============================================================================== Bosco Tjan University of Southern California Limitation of Ideal-Observer Analysis in Understanding Perceptual Learning Authors: Bosco Tjan, S. T. L. Chung and D. M. Levi Performance for a variety of visual tasks improves with practice. Various attempts have been made to determine the functional nature of perceptual learning. Previously, using an ideal-observer analysis (similar to Gold et al., 1999, Nature), we reported that the improvements in identifying letters in peripheral vision following 6 days of training was due to an increase in sampling efficiency, but not a reduction in intrinsic noise (Chung et al., 2001, pre-OSA). Here we reexamine the functional nature of perceptual learning using the perceptual template model (PTM, Lu & Dosher, 1999, JOSA-A). We used the data set collected by Chung et al (2001) in which the change in contrast thresholds for identifying single letters embedded in external noise were tracked over a period of 6 days. Six levels of external noise were tested each day. Data were collected using the Method of Constant Stimuli. Thresholds at d'=0.8, 1.7, 2.7 were determined per noise level by fitting a psychometric function to the raw data. This resulted in three threshold-vs-noise-contrast (TvC) functions for each day and each observer. Fitting PTM to the TvC functions revealed that the performance improvements observed in 4 of 5 observers were due to a reduction in internal additive noise. Three of these 4 observers also showed a template retuning. No significant change in the internal multiplicative noise was observed. (The fifth observer did not show any significant change in performance.) Despite individual differences, both internal noise reduction and template retuning appear to be the common mechanisms of perceptual learning in peripheral vision. This interpretation of the results contrasts sharply with the one obtained with an ideal-observer analysis. The discrepancy is due to the omissions of non-linearity and a stimulus-driven stochastic component (multiplicative noise or uncertainty) in the linear amplifier model use to interpret the data human. Philsophical implications of the current result on the use of ideal-observer analysis will be explored. =============================================================================== Chia-huei Tseng University of California, Irvine TBA =============================================================================== Eric-Jan Wagenmakers Northwestern University Estimation and Interpretation of l/f Noise in Human Cognition =============================================================================== Heather A. Wild Indiana University Poster: Density in the Face Space: What Do We Really Mean? Authors: Heather A. Wild and Thomas A. Busey Valentine (1991) proposed the face-space framework to account for differences in other- versus same-race face perception. He claims that these deficits result from differential representation of such faces relative to faces of one's own race, such that other-race faces are more densely clustered, i.e., closer together, in the space. The face-space framework, however, was proposed as a metaphorical model and was not formalized. Thus the differential density hypothesis also was not formalized and, while widely cited in the face perception literature, has never been directly tested. In the present study, I collected pairwise similarity ratings of Asian and Caucasian faces from Caucasian and Asian observers. Analyses of these data yielded face-space representations. I developed six different mathematical formalizations of density measures, partially based on previous work (e.g., Krumhansl, 1979; Zaki & Nosofsky, 2001). I directly examined whether there is actually increased density for other-race faces in the face-space representations. The results show that whether other-race faces are actually more densely clustered depends crucially on the definition of density that is employed, and moreover that these differences are not consistently present for all groups of subjects. This casts doubt on Valentine's (1991) initial hypothesis, and challenges researchers to precisely quantify hypotheses regarding differences in the representations of same- and other-race faces. ================================================================================