CSJ Archive > Contents > Conference Proceedings: 2008 (30th Annual Conference; Washington, DC)

CogSci 2008

30th Annual Meeting of the Cognitive Science Society

July 23-26, 2008    Washington, DC, USA

Poster Session II -- Attention & Memory

(Friday, July 25, 2008, 5:00-7:00 PM, Exhibit Hall)


Verbal Overshadowing as Perceptual Interference

Ava Santos   (santos_a@fortlewis.edu)
Fort Lewis College, Emory University

Lawrence Barsalou   (barsalou@emory.edu)
Emory University

Verbal overshadowing is a phenomenon where verbalizing a nonverbal experience produces an inaccurate memory of the experience. Drawing from perceptual symbol systems, we postulate that verbal overshadowing belongs to the larger phenomenon of perceptual interference. Verbalizing or mentally imaging nonverbal experiences produces inaccurate memories of those experiences. This is because people initially form prototypical representations of nonverbal experiences. These prototypical representations are not always identical to later exemplars of the experience. On perceiving an exemplar and later having to remember it, people may retrieve inaccurate prototypical representations instead. To test this account, we developed a visual interference paradigm that uses fictional spy devices. Two experiments demonstrated that we can train people to have specific prototypical representations of spy devices. When later presented with an exemplar of the spy devices and asked to remember it, participants remember the prototypical spy device instead. These experiments let us better understand verbal overshadowing and perceptual interference.


(Return to Top)

Evaluating Mechanisms of Fatigue Using a Digit Symbol Substitution Task

Larry Moore   (larry.moore@mesa.afmc.af.mil)
Lockheed Martin @ Air Force Research Laboratory

Glenn Gunzelmann   (glenn.gunzelmann@mesa.afmc.af.mil)
Air Force Research Laboratory

Kevin Gluck   (Kevin.Gluck@mesa.afmc.af.mil)
Air Force Research Laboratory

The effects of sleep deprivation on human performance can be as impactful as a military operation and as broad reaching as the daily commute. With the ACT-R architecture, we have already demonstrated how simple fatigue mechanisms can influence cognition to produce performance comparable to human subjects. This research has led to the new ?micro lapse? theory of fatigue, and paves the way for a priori predictions of human performance under fatigued conditions. This poster describes recent updates in the mechanisms that implement the theory and latest progress in evaluating the generalizability of the theory, this time in the context of the Digit Symbol Substitution Task. In this scenario, subjects must correlate a random symbol with a paired digit presented in a list. Fatigue mechanisms for procedural memory (for scanning and responding) and declarative memory (for remembering digit-symbol pairings) will be tested in the same model.


(Return to Top)

How Memory Guides Strategy Selection

Julian N. Marewski   (marewski@mpib-berlin.mpg.de)
Max-Planck-Institute for Human Development

Lael J. Schooler   (schooler@mpib-berlin.mpg.de)
Max-Planck-Institute for Human Development

We contribute to answering the question of how people select among different strategies to make decisions. Our proposal is shaped by three theories, the fast and frugal heuristics framework, the ACT-R cognitive architecture, and J. J. Gibson s theory of perception. From the heuristics framework we adopt the thesis that people make decisions by selecting from a repertoire of heuristics that exploit regularities in our environment and basic cognitive capacities such as human memory. Gibson leads us to ask the question about how our environment provides opportunities for selecting different heuristics. The ACT-R architecture provides a quantitative theory about how memory works. In a series of experiments and computer simulations with ACT-R, we show how memory determines which opportunities the environment provides for selecting different heuristics, and how in doing so, memory guides strategy selection. In addition, we provide a theoretically grounded method for predicting response time distributions from Internet statistics.


(Return to Top)

Memory for Musical Tone Intervals and Tonality

Charles Barousse   (charlesbarousse@gmail.com)
University of Louisiana At Lafayette

Memory for tone intervals is an important capacity in both speech and music. This memory is stronger for tonal melodies in comparison to atonal melodies. The majority of past research investigating tonality and memory, however, has failed to make an important distinction- the distinction between the musicality (the quality of being musical) of a melody and the scale that has been used as the basis of the melody. Both musicality and underlying scale contribute to the perception of tonality in a melody. Four experiments examine the contributions of scale and musicality to a tone interval memory task. Memory is stronger for melodies based on a familiar scale than for melodies based an unfamiliar scale. Memory is also stronger for musical melodies when compared to unmusical melodies. The contributions of musicality have a stronger effect on memory than the contributions of scale.


(Return to Top)

Autobiographical Memory and Motor Action

Katinka Dijkstra   (k.dijkstra@fsw.eur.nl)
Erasmus University

Daniel Casasanto   (djc@psych.stanford.edu)
Stanford University

Retrieval of autobiographical memories is facilitated by activation of perceptuo-motor aspects of the experience, for example a congruent body position at the time of the experiencing and the time of retelling (Dijkstra, Kaschak, & Zwaan, 2007). The present study examined whether similar retrieval facilitation occurs when the direction of motor action is congruent with the valence of emotional memories. Consistent with evidence that people mentally represent emotions spatially (Casasanto, in press), participants moved marbles between vertically stacked boxes at a higher rate when the direction of movement was congruent with the valence of the memory they retrieved (e.g., upward for positive memories, downward for negative memories) than when direction and valence were incongruent (t(22)=4.24, p<.001). In addition, valence-congruent movements facilitated access to these memories, resulting in shorter retrieval times (t(22)=2.43, p<.05). Results demonstrate bidirectional influences between the emotional content of autobiographical memories and irrelevant motor actions.


(Return to Top)

A Memory Model for Cognitive Agents

Guilherme Bittencourt   (gb@das.ufsc.br)
Universidade Federal De Santa Catarina

In this paper we present a memory model that, syntactically, consists of logical propositions but whose semantic description includes, besides the usual truth value assignments, what we call emotional flavors, that correspond to the state of the agent's body translated into cognitive terms. The combination between logical propositions and emotional flavors allows the agent to memorize relevant propositions that can be used for reasoning. These propositions are represented in a specific format -- prime implicants/implicates -- which is enriched with annotations that explicitly store the internal relations among their literals. Based on this representation, the details of the proposed memory mechanism is described and some arguments to support its psychological plausibility are presented.


(Return to Top)


Evaluation of the Efficacy of the Delacato?S Neuropsychological Method in the Treatment of 7-12 Years Old Boys With Adhd

Farzad Momeni   (momenifarzad@yahoo.com)
Vrije University of Brussels- Phd Student in Children and Adolescents Clinical Psychology

Farzaneh Mehrabi Mansour   (farzaneh124000@yahoo.com)
Ma Student in Psychology

The study was designed to evaluate the effectiveness of the Delacato method in the treatment of hyperactive 7-12 year old boys. 60 out of 120 boys visiting psychiatric offices and children's psychiatric hospitals were randomly selected. These subjects had a diagnosis of attention deficit hyperactivity disorder on the basis of DSM-IV criteria and Conner's Parental and Teachers Scale. 30 of the subjects were randomly allocated to experimental group. These children were treated with the Delacato's Neuropsychogical method (1992). The remaining 30 subjects served as control group, receiving no treatment. Dalacatos Neuropsychological method appears as an effective method for the treatment of children with ADHD.


(Return to Top)

Trait or Situation? Cultural Differences in Judgment of Emotion

Megumi Kuwabara   (mekuwaba@indiana.edu)
Indiana University

Ji Y. Son   (jys@ucla.edu)
University of California Los Angeles

Linda B. Smith   (smith4@indiana.edu)
Indiana University

Traditional research in cognition assumes that fundamental processes such as memory and attention are universal. However, a growing number of studies suggest cultural differences. One proposal is that Westerners focus on a central single object whereas Easterners integrate their judgment of the focal object with surrounding contextual cues. For emotions, several studies suggest that young children in western cultures judge people s emotions based primarily on expressive facial expressions, and not context. Two experiments with preschool children in the US and Japan show that this inattention to context may be specific to Western children. Japanese children judged emotions based more on contextual information than facial expressions whereas the opposite was true for American children. The addition of linguistic cues to the task (labeling the emotion) increased these cross-cultural differences.


(Return to Top)

Context and Induction: The Impact of Background Context on Children'S Category Learning

Haley A. Vlach   (haleyvlach@ucla.edu)
University of California, Los Angeles

Catherine M. Sandhofer   (sandhof@psych.ucla.edu)
University of California, Los Angeles

Categorization research paradigms, models, and theory of category acquisition and induction have focused on features of the stimulus itself (i.e., shape, function, etc.). To explore variables beyond the stimulus, we investigated the role of background context in children s category induction. In each condition, children learned about exemplars in three phases: a learning phase, distractor phase, and test phase. During the learning phase, children were presented with exemplars in specific context(s). At the test, children were asked to generalize to a novel category exemplar in either the same or a different context. Parallel to other literatures, results revealed that the changes in context had a significant impact on children s ability to generalize to a novel exemplar. Because subtle changes in context had a strong effect on performance, future research should examine how contextual effects operate with properties of the stimulus in order to inform theories of category induction.


(Return to Top)

Do Preschoolers Track a Character's Mental Perspective While Listening to a Story?

Agnieszka M. Fecica   (ampolano@uwaterloo.ca)
University of Waterloo

Daniela K. O'Neill   (doneill@uwaterloo.ca)
University of Waterloo

Previous research suggests that while reading, adult readers track, and often adopt, a narrative character's mental perspective (e.g., Zwaan, 1999). We explored preschoolers' ability to track a narrative character's mental perspective by measuring children's reaction times in a computerized story-listening task. Children listened to a story one sentence at a time and were asked to press a button to hear the next sentence in the story. Throughout a series of sentences describing a character getting ready to go to an event, children pressed the button faster when the character was described as thinking the event was "really great" versus "really horrible." These findings suggest that while listening to a story, 5-year-olds track a character's mental perspective. Children's performance on the story listening task was not related to their general linguistic ability or general emotion knowledge but was related to their ability to repeat sentences presented in a narrative context. This study contributes to our understanding of the development of narrative cognition.


(Return to Top)

Development of Synonym-Based Induction

Bryan Matlen   (bmatlen@andrew.cmu.edu)
Carnegie Mellon University

Anna Fisher   (fisher49@andrew.cmu.edu)
Carnegie Mellon University

As early as the second year of life, children can reference objects by multiple labels, such as taxonomic labels (e.g., animal-rabbit) and synonyms (e.g., bunny-rabbit). Children's ability to generalize properties of objects relying on taxonomic labels has received much attention in the literature, however it has only been sparsely investigated within the domain of synonyms. The present study was motivated by this gap in the literature. Participants performed an induction task either with lexically associated synonyms (e.g., bunny-rabbit) or with non-associated synonyms (e.g., rock-stone). Results suggested that: (1) less than 20% of 4-year-olds could use non-associated synonyms to generalize object properties, whereas 57% of 5 year-olds and the vast majority of 7-year-olds were able to do so; (2) proportion of 4-year-olds able to rely on synonyms during induction more than doubled in the associated synonyms condition. These findings are discussed with regards to two competing theories of early induction.


(Return to Top)

Causal Supports for Early Word Learning

Amy E. Booth   (a-booth@northwestern.edu)
Northwestern University

What factors determine the likelihood with which new words will be acquired by young word learners? Although there are surely numerous contributors, the current investigation highlights the role of causal information. Three-year-old children were taught 6 new words for unfamiliar objects and animals. Items were described in terms of their non-obvious causal or non-causal properties. Performance in these experimental conditions was also compared to a baseline control condition in which no specific information about the items was provided. When tested only minutes after training, no significant differences between the conditions were evident. However, when tested several days after training, children who were exposed to causal information revealed superior performance. These results suggest that more than simple associations are required to account for early word learning. Instead, the well-documented effect of causal information on learning and categorization appears to extend to word learning in young children.


(Return to Top)

Poster Session II -- Learning

(Friday, July 25, 2008, 5:00-7:00 PM Exhibit Hall)


Temporal Continuity in Cross-Situational Statistical Learning

George Kachergis   (gkacherg@indiana.edu)
Indiana University

Chen Yu   (chenyu@indiana.edu)
Indiana University

Previous studies show that subjects typically learn a surprising number of word-object pairings solely from their co-occurrences in small sets of individually ambiguous trials. An unrealistic assumption in those studies forbids the same pairing from appearing in two consecutive trials, as this trivializes the role of statistics in learning this pairing. However, it may be that attention to the repeated pairings is necessary to take advantage of them. On average, we find that temporal continuities do not significantly improve average performance, but the highest overall performance was achieved in these conditions. Items that appeared twice successively were learned twice as often as unrepeated items in that block, whereas thrice-repeated items had no advantage over unrepeated items, suggesting that longer repetitions simplify the learning of other pairings by reducing the complexity of the situation. A model fed with eye movement data is constructed to predict individual differences in learning.


(Return to Top)

An Embodied Approach to Achieving Mastery and Learning While You Work

Brian Krisler   (bkrisler@cs.brandeis.edu)
Brandeis University

Richard Alterman   (alterman@cs.brandeis.edu)
Brandeis University

When a professional achieves mastery of technology, she develops a comprehensive understanding of that technology, allowing her to more productively and efficiently accomplish the tasks required to achieve her goals. The development of mastery is a problem of skill acquisition that requires both situated awareness and proper concept grounding. Our research introduces a training method that helps increase technology specific skills by automatically generating task specific learning events that are introduced into the flow of the activity. Thus as the professional engages in her computer-mediated work activities she gradually gains mastery of the articulation work required to perform her job, reducing her overall cognitive load. An initial study provides evidence for the utility of this approach to training and skill acquisition. The theoretical model that underlies this research depends on a characterization of cognition as both situated and embodied.


(Return to Top)

The Importance of Ordinary Experience: Providing Girls with Time for Regular Practice of Mathematical Cognition

Robin C. Flanagan   (FlanaganR@wcsu.edu)
Western Connecticut State University

Theresa Canada   (CanadaT@wcsu.edu)
Western Connecticut State University

Advances in educational theory and practice are often aimed at finding cognitive shortcuts. We wondered, though, whether girls actually get enough ordinary experience with mathematical thinking to support their mathematical education. We hypothesized that girls (grades 3 through 8) who were given an extra hour of ordinary mathematical experience in such work as scale modeling, origami, Chinese abacus and computer programming, would develop a more solid foundation on which to build their education. Fifteen undergraduate women worked with 42 school-aged girls for one hour a week during an eight-week after school program. Girls had significantly higher post-program ratings of confidence and enjoyment in math and using computers than their pre-program ratings, although pre- and post-program scores on the WRAT-4 math subtest were not significantly different.


(Return to Top)

The Stability and Strength of Knowledge Representation Acquired During Artificial Grammar Learning

Michal Wierzchon   (wierch@apple.phils.uj.edu.pl)
Jagiellonian University, Institute of Psychology

Dariusz Asanowicz   (d.asanowicz@gmail.com)
Jagiellonian University, Institute of Psychology

The aim of the presented studies was to investigate the characteristics of implicit knowledge using artificial grammar learning task. Two experiments were design to investigate the stability and the strength of the representation of acquired knowledge. We present participants with additional ungrammatical strings during learning phase to test the knowledge stability (manipulation was based on procedure of adding the noise on input of the connectionist network). The representation strength was probed by speeded presentation of the learning material. The classification of the group presented with additional strings was impaired. However, the speeded presentation has no influence on the classification accuracy. The results are discussed with regard to Cleeremans & Jimenez (2002) assumptions about the graded representation of knowledge, confirming that the implicit representation could be weak and is relatively unstable. Such conclusions deny the common in the filed assumption that implicit processes are robust and stabile in time.


(Return to Top)

Guided Learning by Reading (Lbr) as a Cognitive Growth Model

Alexei V. Samsonovich   (asamsono@gmu.edu)
George Mason University

New capabilities emerge from integration. Example: the present state of the art offers strong capabilities in syntactic and semantic parsing of natural language and in formal reasoning based on well-defined representations of knowledge. Bridging the two by developing the capability to capture natural text as formal knowledge enables autonomous cognitive growth. This also requires acquisition of natural language that cannot be achieved via passive reading. Solution includes conversational guidance and the ability for the system to try out learned concepts in a virtual environment, leading to a concept of LBR system consisting of a cognitive architecture embedded in a microworld and equipped with linguistic interface (parsers). The output is written to a formal knowledge base. Training is done by a human based on a curriculum and a selected limited domain. This work addresses detection of a critical mass of intelligence that enables cognitive growth and design of efficient meta-learning techniques.


(Return to Top)

Sub-Functions of Human Learning Process During a Sequential Task

Sergey Tarasenko   (sergey@cog.ist.i.kyoto-u.ac.jp)
Jst Erato Asada Project

Toshio Inui   (inui@i.kyoto-u.ac.jp)
Kyoto University, Jst Erato Asada Project

Abdikeev Niyaz   (nabd@rea.ru)
Plekhanov Russian Academy

This study investigates how amount of information (complexity) about a stimulus influences human performance of sequential learning. We used modified Sequence Learning task, which allows to test participants providing them no prior information about a stimulus structure. Usually, a single logistic curve (LC) provides good approximation of the learning dynamics. It was found that complexity increase causes to use more than one LC for a perfect approximation. Moreover, when complexity is high, LC no longer describes an actual learning dynamics, but only illustrates a general trend. The results imply appearance of more components for approximation. We consider this reflects changes in an underlying learning process. Therefore, we hypothesize existence of sub-functions in human learning, which develop themselves on various levels of complexity.


(Return to Top)

Coding by Demand": Identifying the Dimensions of Student Dialogue That Underlie Theories of Learning"

Gwendolyn Campbell   (gwendolyn.campbell@navy.mil)
Navair Orlando Tsd

Natalie Steinhauser   (natalie.steinhauser@navy.mil)
Navair Orlando Tsd

Myroslava Dzikovska   (M.Dzikovska@ed.ac.uk)
Hcrc, University of Edinburgh

Johanna Moore   (J.Moore@ed.ac.uk)
Hcrc, University of Edinburgh

Charles Callaway   (Charles.Callaway@ed.ac.uk)
Hcrc, University of Edinburgh

While there are a variety of well-documented and empirically-supported theoretical descriptions of the events and experiences that lead to learning, little work has been done to integrate and/or compare the relative impact of the causal factors described by those perspectives. In this paper we propose that a set of five dimensions collectively forms a foundation underlying a number of the most prevalent theoretical perspectives. We show how these dimensions can be translated into a coding scheme and reliably assessed on student contributions to an instructional dialogue. Finally, we provide preliminary validation evidence for our coding scheme and discuss the potential value of such an approach to analyzing student behavior under a variety of learning environments.


(Return to Top)

Poster Session II -- Expertise and Explanation

(Friday, July 25, 2008, 5:00-7:00 PM Exhibit Hall)


Effects of Scaffolding Problem Formulation Phase During Multifaceted Physics Problem-Solving

Serkan Toy   (serkan@iastate.edu)
Iowa State University

Dale Niederhauser   (dsn@iastate.edu)
Iowa State University

John Jackman   (jkj@iastate.edu)
Iowa State University

Craig Ogilvie   (cogilvie@iastate.edu)
Iowa State University

Sarah Ryan   (smryan@iastate.edu)
Iowa State University

Aliye Karabulut   (aliye@iastate.edu)
Iowa State University

In this study we investigated effects of scaffolding the problem formulation phase on physics students problem-solving behavior. Students worked in small groups to solve five multifaceted problems in a Web-based problem-solving environment during regularly scheduled lab times. Students were assigned to either a scaffolded or unscaffolded condition. Both conditions were presented with the same problems to solve and tasks to complete. Students in the scaffolded condition received information and leading questions that modeled a successful problem formulation process. Scaffolding was faded requiring students to take increasing responsibility for formulating the problem. Student web-log files and videotape also served as data sources. Overall problem-solving performance, and problem formulation task performance served as dependent variables. Preliminary findings suggest that, scaffolding can promote more effective problem formulation strategies that persist over time, and that problem formulation tends to have a positive effect on general problem-solving performance.


(Return to Top)

The Content of Self-Explanations While Studying Incomplete Worked-Out Examples

Robert G.M. Hausmann   (bobhaus@pitt.edu)
University of Pittsburgh

Brett Van De Sande   (bvds@pitt.edu)
University of Pittsburgh

Kurt Vanlehn   (vanlehn@cs.pitt.edu)
University of Pittsburgh

The present experiment conducted a detailed analysis of the content of self-explanations generated while studying worked-out examples in the context of solving problems with an intelligent tutoring system (ITS) for physics called Andes. Two variables were crossed: Activity (paraphrase vs. self- explanation) and Example (complete vs. incomplete). The number of paraphrases and self-explanations were coded for each condition. The Paraphrase condition generated more paraphrases than the Self-explanation condition; however, the reverse was not true. Therefore, a finer-grained coding scheme counted the number of meta-cognitive statements about the Andes user interface. The Self-explanation condition demonstrating more meta-cognitive statements about the user-interface than the students in the Paraphrase condition (d = .45). Moreover, the main effect was qualified by a significant interaction, with Incomplete/Self-explanation producing more meta-cognitive statements than the Complete/Self-explanation condition. These results suggest the content of self-explanations may be contingent on the completeness of the examples that are being studied.


(Return to Top)

Comparing Similar or Dissimilar Examples for Analogical Transfer

Young Hoan Cho   (yhc2k2@mizzou.edu)
University of Missouri

The purpose of this study is to investigate how comparing similar or dissimilar examples affects schema construction and analogical transfer. The similar examples had the same structure and similar surface features, whereas the dissimilar examples had the same structure but dissimilar surface features. Undergraduate students compared two similar negotiation examples (n = 5) or two dissimilar negotiation examples (n = 5), and then they solved an isomorphic problem. Think-aloud protocols were recorded while they solved the problem and participants were interviewed. Results showed that schema quality was not different between two groups, but participants who compared dissimilar examples focused more on structural features than participants who compared similar examples. Furthermore, participants in the dissimilar example condition were more successful in analogical transfer and generated less justification statements in the problem solving process. The effect of comparing examples on analogical transfer seems to be moderated by the surface similarity between examples.


(Return to Top)

Examining First Grade Students' Reading Skill Growth Through a Culturally-Responsive Vocabulary Intervention

Phyllis Swann Underwood   (punderwood@fcrr.org)
Florida State University

Carol Mcdonald Connor   (cconnor@fcrr.org)
Florida State University

The purpose was to investigate whether students whose teachers use culturally-responsive practices with greater frequency are more likely to demonstrate stronger reading skill growth than students whose teachers use culturally-responsive practices less frequently. To begin to understand the underlying causal mechanism associated with culturally-responsive practices, first grade students were randomly assigned within classrooms at ethnically and socioeconomically diverse schools to treatment or alternate treatment conditions (n=2 schools, 7 teachers, and 140 students). Treatment condition students received small group vocabulary instruction using culturally-responsive teaching practices with an episodic memory event component. Alternate treatment condition students received small group vocabulary instruction using business as usual, not culturally-responsive, teaching practices without the episodic memory event component.  Preliminary results suggest that treatment condition student groups demonstrated stronger reading skill growth compared to alternate treatment condition student groups, suggesting that explicit vocabulary instruction incorporating culturally-responsive teaching practices may be associated with stronger student outcomes.


(Return to Top)

The Effects of Skill Diversity in Peer Feedback: It's What You Don't Know

Melissa M. Nelson   (mmnelson@pitt.edu)
University of Pittsburgh

Brandi N. Melot   (bnm6@pitt.edu)
University of Pittsburgh

Christopher A. Stevens   (cas107@pitt.edu)
University of Pittsburgh

Christian D. Schunn   (schunn@pitt.edu)
University of Pittsburgh

In peer learning, research has been found to support two arguments: matched pairs (Nathan and Koedinger, 2000) and mismatched pairs (Vygotsky, 1978). This study attempts to determine whether one is truly better. Using the SWoRD interface, comments given to writers from their peers were coded for several relevant categories: Type of Feedback, Type of Criticism, Level of Problem, Level of Solution, and Implementation. Using a two-tailed, independent t-test, we discovered that mismatched pairs did in fact give more feedback in the specific areas of criticism and solution, and that mismatched pairs were more likely to implement their criticisms. Though more research is needed, this could lead to the shift of deliberately placing peer learners into mismatched pairs.


(Return to Top)

Toward a Process Model of Explanation

John E. Hummel   (jehummel@uiuc.edu)
University of Illinois

David H. Landy   (dlandy@uiuc.edu)
University of Illinois

Derek Devnich   (devnich2@uiuc.edu)
University of Illinois

The ability to generate explanations plays a central role in human cognition. Generating explanations requires a deep conceptual understanding of the domain in question and tremendous flexibility in the way concepts are accessed and used. Together, these requirements constitute challenging design requirements for a model of explanation. We describe our progress toward providing a such a model. The model is based on the LISA model of analogical inference (Hummel & Holyoak, 1997, 2003). We augment LISA with a novel representation of causal relations more explicit than simple associative links but less explicit than propositions and with the capacity to dynamically construct novel schemas by assembling elements from multiple sources of knowledge in LTM. We demonstrate how the resulting model serves as an explicit process model of explanation.


(Return to Top)

Transitions, Analogical Processes, and Expertise in Contemporary Art: a Detailed Case Study

Jude Leclerc   (jude.leclerc@umontreal.ca)
University of Montreal

Takeshi Okada   (okadatak@p.u-tokyo.ac.jp)
University of Tokyo

Sawako Yokochi   (t74sawa@interlink.or.jp)
University of Tokyo

Frederic Gosselin   (frederic.gosselin@umontreal.ca)
University of Montreal

The current study aims to apply the framework developed by Okada, Yokochi, Ishibashi, Namba, and Ueda (2007) and Yokochi and Okada (2006, 2007) to better understand, and generate hypotheses about, the nature of transitions observed in artists' long-term production of artworks; this framework consists of a multi-level model of artistic expertise acquisition and a set of cognitive operations (i.e., analogical modifications) used by artists as they produce new artworks. We describe a detailed case study of a Canadian-based video artist. Our results highlight the impact of changes in artistic medium on the art-making process and the role played by art-making strategies as mediating factors of expert artistic performance.


(Return to Top)

Poster Session II -- Judgement and Reasoning

(Friday, July 25, 2008, 5:00-7:00 PM, Exhibit Hall)


Approaches on Neurocomputational Self-Organizing Behavioral Modeling

Spyridon Revithis   (revithiss@cse.unsw.edu.au)
University of New South Wales

It has been proposed in previous case studies, by the author and by other researchers, the claim that neurocomputational self-organizing behavioral modeling, based on the class of SOM neural networks, can be realized in two approaches: behavioral anticipation and exegetic prediction. In the former approach, the model is utilized as a behavioral space compression tool, which depicts statistically dominant behavioral patterns and enables the accurate temporal projection of domain specific behavior. The latter approach follows cognitive modeling methodologies, in which behavioral patterns represented within the model are linked to interpretation-rich neuropsychological empirical evidence; such empirical data point to verbal theories of cognitive-behavioral phenomena, such as autism, and are matched to quantitative aspects of the model s mechanism and structure. The various studies that have been proposed, if viewed under these two approaches, support the analysis exemplified in this work and suggest a novel and productive categorization dichotomy for guiding future research.


(Return to Top)

Logical Thinking, Deontic Reasoning, and the Fairness Principle: Exploring the Relationship Between Selection Tasks and the Ultimatum Game

Kuninori Nakamura   (knaka@ky.hum.titech.ac.jp)
Tokyo Institute of Technology

Previous studies reveal that people follow the fairness principle even when it deviates from subjective expected utility maximization. Thus, the following question arises: How does a rational person who can correctly solve logical tasks behave when his/her maximization of the expected utility is not compatible with social rationality? We examined this question by using the ultimatum game (Güth et al., 1982) and various versions of Wason's selection task (Wason, 1966). In this study, participants first attempted to solve the selection tasks and then performed the ultimatum game. The results indicated that regardless of their performances on the selection tasks, the participants opted for fair allocation.


(Return to Top)

Truth-Based Or Possibility-Based Compatibility Judgments and Handley Et Al.'s (2006) Litmus Test of the Suppositional Conditional.

Walter Schroyens 
University of Gent

In their critiques of mental-models theory, Handley et al. (2006) adopt a possibility-based definition of compatibility: Propositions are compatible when they share at least one possibility. Handley et al. explicitly claim their suppositional-conditional theory characterizes false-antecedent, i.e, cases as possible vis-ŕ-vis conditionals. It follows by definition that suppositional-conditional theory should characterize and as compatible propositions. This is contradicted by their findings and stands in direct contradiction with their own claim that suppositional-conditional theory characterizes the conditionals as incompatible. We suggest modal reasoners adopt a truth-based notion of compatibility: two propositions are compatible when they can be true at the same time. We show that mental-models theory is consistent with the notion of truth-based compatibility and present two studies that demonstrate people do not reason according to a possibility-based notion of compatibility.


(Return to Top)

Memory Judgments of Relative Order in Short Lists: Multiple Strategies Are Available, Depending on Wording of Instructions

Michelle Chan   (mc3@ualberta.ca)
University of Alberta

Jeremy B. Caplan   (jcaplan@ualberta.ca)
University of Alberta

Short-term judgments of relative order are important for human cognition as many activities involve knowing precise temporal sequences of events. Studies have found order search to be self-terminating; however Sternberg (1969) reported forward memory search whereas Muter (1979), Hacker (1980), and McElree and Dosher (1993) reported backward memory search. Here we ask what determines these strategies. Participants judged which of two probe items was presented earlier later, or whether they were in the same or reversed order as on studied lists of 3-6 consonants. With earlier" and "later" instructions, forward and backward self-terminating scanning were observed, respectively. Results for "intact" instructions were comparably more variable, but scanning rates were comparable across instructions. Our findings suggest that participants can perform short-term judgments of relative order with multiple comparably effective strategies. Implications for constraining both serial and parallel search models are discussed."


(Return to Top)

Poster Session II -- Language and Concepts

(Friday, July 25, 2008, 5:00-7:00 PM, Exhibit Hall)


Similarity Between Vowels Influences Response Execution in Word Identification

Jason D. Zevin   (jdz2001@med.cornell.edu)
Sackler Institute for Developmental Psychobiology

Thomas A. Farmer   (taf22@cornell.edu)
Sackler Institute for Developmental Psychobiology, Dept. Psychol., Cornell Univerisy

Bruce D. Mccandliss   (bdm2001@med.cornell.edu)
Sackler Institute for Developmental Psychobiology

We have adapted arm-movement measures previously developed for lexical and grammatical ambiguity resolution to study inter-category similarity of speech sounds. Participants listened to multiple recordings of naturally produced words ("pin," "pen," and "pan") and used a computer mouse to select the matching stimulus from a two-picture array. The same participants performed a dissimilarity judgment task. Both tasks revealed evidence for graded effects of inter-category similarity, albeit in the context of strongly categorical classification. Accuracy in the word identification task was nearly perfect, reflecting categorical processing of stimulus type. However,  arm-movement dynamics revealed attraction toward the alternative response. The magnitude of attraction was related to perceptual similarity between categories measured in the offline metalinguistic judgment task. The results demonstrate that even relatively subtle patterns of stimulus similarity can influence motor responses by increasing the amount of competition between motor programs corresponding to two competing choices in a perceptual decision task."


(Return to Top)

The Duck/Rabbit Illusion: Re-examination of Information Encapsulation

Aysu Suben   (aysu.suben@fandm.edu)
Franklin & Marshall College

Michael Anderson   (michael.anderson@fandm.edu)
Franklin & Marshall College and University of Maryland

Tony Chemero   (tony.chemero@fandm.edu)
Franklin & Marshall College

In his article "Is the mind really modular?" Jesse Prinz argues that the persistence of visual illusions in the face of knowledge of their illusory character (normally taken as evidence for the encapsulation of visual processes) can instead be explained by a trumping mechanism that is consistent with the view that perceptual processes are not encapsulated. In this paper, we offer an alternate model of the trumping mechanism that is consistent with information encapsulation. According to our model\\ the trumping mechanism selects which representations cross the threshold into conscious awareness, without needing to interfere with the proprietary representations produced by visual processing. Our model requires that bi-stable or ambiguous stimuli generate multiple sub-conscious representations; each of these representations ought to have detectable down-stream effects. Thus\\ we propose a picture priming experiment to test our model."


(Return to Top)

Examining the Hidden Factors That Underpin Semantic Representation: What Functional Brain Imaging Reveals about the Neuroarchitecture of Object Knowledge

Kai-Min Kevin Chang   (kaimin.chang@gmail.com)
Carnegie Mellon University

Tom Mitchell   (Tom.Mitchell@cmu.edu)
Carnegie Mellon University

Marcel Adam Just   (just@cmu.edu)
Carnegie Mellon University

Recent multivariate analysis of fMRI activities have shown that discriminative classifiers, such as Support Vector Machines (SVM), are capable of decoding fMRI-sensed neural states associated with the visual presentation of categories of various objects. However, the lack of a generative model of neural activities makes these discriminative classifiers less appealing when studying the underlying neural mechanism. In this study, we propose a generative classifier which models the hidden factors that underpin semantic representation of object knowledge with a multivariate multiple linear regression model. Our results indicate that object features obtained in an independent feature norming study can capture some of these hidden factors and explain a significant portion of the systematic variance in the neural activity observed in an object-contemplation task. Furthermore, the resulting regression model is useful for classifying a previously unseen neural activation vector, indicating that the distributed pattern of neural activities encodes sufficient signal to discriminate differences among stimuli. More importantly, the generative classifier outperforms SVM which does not utilize such intermediate representations in a between-participants analysis, indicating that this set of intermediate semantic features generalizes across people. Finally, this intermediate representation allows us to extrapolate the neural activity for previously unseen words and classify on a leave-two-words-out classification task, which simply cannot be done with a discriminative classifier.


(Return to Top)

Different Mechanisms Control the Allocation of Perceptual Processing Resources and Decisional Resources in Perceptual Categorization

Duncan Guest   (D.Guest@warwick.ac.uk)
University of Warwick

In a perceptual categorization task, a cue was used to direct visual attention toward stimulus features that differed in category diagnosticity. Cueing influenced perceptual processes but not decisional processes. Comparing feature processing rates in the categorization task with those in perceptual matching with the same stimuli also revealed that highly diagnostic features in the categorization task received enhanced processing. Thus, directing perceptual processing resources to the cued location did not prevent the strategic allocation of some of these processing resources elsewhere. Interestingly, the features that received enhanced processing were not those that were weighted heavily in category decision. This indicates that separate mechanisms control the allocation of decisional resources and perceptual processing resources in categorization. This is consistent with the idea that these different systems are used to optimise categorization performance at different time points, depending on the level of perceptual information available.


(Return to Top)

The Ideal Representation of Role-Governed Categories

Micah Goldwater   (micahbg@gmail.com)
University of Texas At Austin

Hunt Stilwell   (mixingmemory@gmail.com)
University of Texas At Austin

Arthur Markman   (markman@psy.utexas.edu)
University of Texas At Austin

Role-governed categories (Markman & Stilwell 2001) are defined by the roles their members play in events, e.g. guests and hosts play different roles in visiting events. These are ubiquitous natural categories, but little is known about them. The current study demonstrates that conceptual ideals are more prominent in the representations of role-governed categories than in the representations of feature-based categories, e.g. truck , which in turn have more prominent prototypes. The study had two phases. During the first phase, participants listed characteristics for typical and ideal instances of the categories. We constructed lists based on the top five most frequent unshared features of each typical and ideal instances. A second set of participants, naďve to the origins of the two characteristics lists, chose which of the two were more illustrative of the category. Participants chose more lists constructed from the ideal characteristics for the role-governed categories than the feature-based.


(Return to Top)

Is Prototypical Typical?

Wolf Vanpaemel   (wolf.vanpaemel@psy.kuleuven.be)
University of Leuven

Eef Ameel   (eef.ameel@psy.kuleuven.be)
University of Leuven

Gert Storms   (gert.storms@psy.kuleuven.be)
University of Leuven

According to the prototype view on category learning, people abstract a prototype when presented with category examples and use this prototype to decide on the category membership of new items. Often, it is assumed that a prototype is the category's central tendency. However, for superordinate categories, such as 'fruits' or 'herbivores,' averaging over all exemplars seems rather implausible. Instead, a more reasonable conception of a superordinate prototype is the most typical exemplar, such as 'apple' or 'cow.' To evaluate this hypothesis, we fitted different prototype models to empirical categorization data involving foods (fruits vs vegetables) and animals (carnivores vs herbivores). Each prototype model assumed a different exemplar as its prototype, which allowed us to select, for each category, the exemplar constituting the best prototype. Comparing the results of the model fits with empirically collected typicality ratings indicated that the best prototype does not correspond to the most typical exemplar.


(Return to Top)

Poster Session II -- Words and Word Learning

(Friday, July 25, 2008, 5:00-7:00 PM, Exhibit Hall)


Identifying Cognitive and Linguistic Strategies in Successful Nonfiction Writing

Gregory Aist   (Gregory.Aist@asu.edu)
Arizona State University

This poster presents initial steps towards an intelligent environment for constructive writing support. The eventual goal is to build software to support authors in research-based writing, a broad class consisting of a narrative or argument structure combined with rich supporting detail. The initial genre is creative nonfiction, specifically travel or memoir writing. Thirteen strategies have been identified based on their use in nonfiction of recognized appeal, including: pinpointing times, dates, and places; using specific-level words (thunderhead) rather than general-level words (cloud); specifying size, material, color, or other properties of objects (little green cucumber pickles); introducing similes for unfamiliar events or concepts; inserting definitions for unfamiliar words; including names and naming-stories for vehicles and animals; contrasting an item with a previously mentioned item of the same type; and incorporating manner or instrument prepositional phrases. Related cognitive issues include how and why such strategies support constructing, remembering, communicating, and learning from narratives.


(Return to Top)

Words Or Word Boundaries? Examining Performance on Statistical Word Segmentation Tasks

Jeremy J. Glick   (jjglick@stanford.edu)
Stanford University

James L. Mcclelland   (mcclelland@stanford.edu)
Stanford University

Early word learners are hypothesized to use statistical cues to segment the speech stream into words, and it is known that both adults and infants use statistical information to learn linguistic, musical, and visual sequences. Nearly all previous work in this area has used languages with a few words of a single fixed length, allowing the learner to restrict the range of possible words under consideration. In order to address this lack, we investigated segmentation in a language containing two-, three-, and four-syllable words, using a word-or-nonword classification task rather than the traditional forced-choice task. Our results suggest that participants successfully segment words of all lengths, with improved performance on two-syllable words. However, for all word lengths, short sub-words of longer words are thought of as more wordlike than other types of nonword items.


(Return to Top)

Structuring the Vowel Space: An Investigation of Turkish and Inuktitut

Brian Dillon   (brianwdillon@gmail.com)
University of Maryland Department of Linguistics

William Idsardi   (idsardi@umd.edu)
University of Maryland Department of Linguistics, Neuro and Cognitive Sciences Program

Colin Phillips   (colin@umd.edu)
University of Maryland Department of Linguistics, Neuro and Cognitive Sciences Program

Recent work on phonological acquisition has generated optimism that sound categories can be statistically induced from acoustic data. Vallabha et al [2007 doi:10.1073/pnas.0705369104] present an expectation-maximization algorithm for learning a subset of Japanese and English vowel categories using Gaussian mixture models. We test this approach with over 50,000 vowel tokens of spoken Turkish, as well as field recordings of Inuktitut. These datasets exemplify the challenges facing the language leaner: noisy data, speaker variability, uncertain category frequencies, and allophonic variation. We analyze the categorizations induced for several representational dimensions of the data (duration, spectral change, various audio transforms, etc.). We find that allophonic variation in the acoustics leads to spurious category formation. Moreover, the variable frequencies of the underlying categories consistently impede acquisition. The results clarify how phonetic data might be mapped to phonological representations during acquisition, and reiterate the importance of informative priors for (machine-)learning processes.


(Return to Top)

Cross-Situational Statistical Learning from Noisy Input

Brian Riordan   (briordan@indiana.edu)
Indiana University

Chen Yu   (chenyu@indiana.edu)
Indiana University

Previous studies have shown that learners can solve the referential ambiguity problem in ambiguous environments by tracking word-referent mappings across trials, even in the absence of social and other cues (Yu and Smith, 2007). In these studies, on each learning trial, the number of words and referents always matched. In natural language learning environments, many learning situations may be characterized by so-called sparse coding" of the scene, where the number of named entities is fewer than the number of entities present. We report three experiments that investigate language learners' abilities to track word-referent co-occurrence statistics under such conditions. Despite the greater ambiguity involved, learners demonstrated robust word-learning performance. For example, given 3 referents, a mix of one, two, and three-word trials actually improved learning over the original three-word/three-referent condition (82.5% vs. 76.1%). Results indicated that, in the space of possible ambiguous environments, there may be unexpected attractors that facilitate word-learning."


(Return to Top)

Sound Versus Meaning: What Matters Most in Early Word Learning?

Sarah Devi Sahni   (sdsahni@gmail.com)
University of Wisconsin - Madison

Timothy T. Rogers   (ttrogers@wisc.edu)
University of Wisconsin - Madison

Previous work suggests that phonological neighborhood density is a key factor in shaping early lexical acquisition. Such studies have, however, focused largely on production rather than comprehension, and have not considered how semantic neighborhoods may influence word-learning. We studied how phonological and semantic neighborhood densities affect both comprehension and production of nouns from the Macarthur-Bates Communicative Development Inventory (MCDI). New measures of semantic and phonological densities, along with child-directed word frequency counts from the CHILDES database, were used to predict the percentage of children who know each word at different ages (8 - 30 months) as indicated in MCDI lexical norms. Production was predicted by frequency and phonological density at all time points, replicating previous research. Semantic density predicted production only at 30 months. However, comprehension norms were predicted by frequency and semantic density, and never by phonological density. Implications for theories of word learning are discussed.


(Return to Top)

The Automaticity of Statistical Word Learning

George Kachergis   (gkacherg@indiana.edu)
Indiana University

Chen Yu   (chenyu@indiana.edu)
Indiana University

Richard M. Shiffrin   (shiffrin@indiana.edu)
Indiana University

The remarkable ability to learn many word-object pairings in ambiguous situations is exhibited by both infants and adults. We investigate the automaticity of cross-situational statistical learning: Are explicit attempts to learn the pairings required, or might learning occur implicitly; requiring only attention to the stimuli? Adults were shown trials with four spoken words and four objects, and instructed to indicate the number of words that were louder than usual and the number of objects that appeared grainier than usual. Subjects were then given a surprise test to evaluate whether they acquired correct word-referent pairings from the set of individually ambiguous trials. In this implicit learning condition, subjects were at chance (18AFC: ~6%), compared to 40% when instructed to do both the signal detection and word learning tasks. A second experiment disguised as a memory task showed no correlation between actual co-occurrences of words and pictures and subjects ratings.


(Return to Top)

Gradations in Phonological Learning

Stephanie Packard   (stephanie-packard@uiowa.edu)
University of Iowa

Prahlad Gupta   (prahlad-gupta@uiowa.edu)
University of Iowa

Language usage depends on the ability to learn new words, which entails mastering not only its meaning but also its phonological form. However, little is known about the processes underlying phonological word-form learning, or about the nature of the phonological word-form representation that is formed as a result. We hypothesize that such phonological learning is a graded process ? that is, not all phonological learning is equal. From this, we make the following two predictions: 1) the robustness of a word-form representation depends on the learning situation; and 2) simple exposure may not support all operations/tasks. Stem-completion ability was used as a measure of robustness in three experiments using a nonword repeated stem-completion task. Results indicate that such phonological learning is item-specific and can vary in robustness as a function of the learning situation.


(Return to Top)

If You Haven'T Got a Head, Get a Label!

Vanja Kovic   (vanja.vucetic@psy.ox.ac.uk)
Department of Experimental Psychology, University of Oxford, UK

Kim Plunkett   (kim.plunkett@psy.ox.ac.uk)
Department of Experimental Psychology, University of Oxford, UK

Gert Westermann   (gwestermann@brookes.ac.uk)
Department of Psychology, Oxford Brookes University, UK

The present study, an extension of Rehder and Hoffman's (2005) eye-tracking study testing Medin and Schaffer's (1978) 5-4 categorisation task, aimed at answering the following question: Can labels facilitate category learning? The study involved two groups of participants, one of which was trained to learn two categories without any labels whilst in the other group, two novel labels were correlated with category membership. The results demonstrated that labelling led to an improvement of performance when assessed through error rates and the number of training blocks needed for solving the task. Furthermore, labels also influenced learning strategies assessed through participants' eye-movements. These results support the idea than labels can facilitate category learning and that even adults can more easily acquire named categories because they have a name.


(Return to Top)

Word Sense and Sensibility: Mental Representations of Polysemy

Susan Brown   (susan.brown@colorado.edu)
University of Colorado-Boulder

The semantic ambiguity of lexical forms is pervasive: Many words have multiple meanings. For example, one can draw a gun, draw water from a well, or draw a diagram. Despite the frequency of this phenomenon, how human beings store and access these meanings is an open question. Do we have a separate representation in our mental lexicon for each ?sense,? or do we store only one very generalized or core meaning for each word? If the latter, do we generate the nuances of each separate sense by rule or by accessing subrepresentations? To even speak of senses in this way implies that we can clearly identify the separate senses of a word. In this study, we use priming in a semantic decision task to investigate the effect of different levels of meaning relatedness on language processing. Both response time and accuracy followed a linear progression through four categories of meaning relatedness. These results suggest that the distinction between a single phonological form with unrelated meanings (homonyms) and a single form with related meanings (polysemes) may be more one of degree than of kind. They also imply that related word 'senses' may be part of a continuum or cluster of meaning rather than discrete entities. In addition, results from specific comparisons between groups do not support the theory that each sense of a word has an entirely separate mental representation.


(Return to Top)

Statistical Co-Learning of Visual and Linguistic Regularities to Improve Word-Learning

Brian Riordan   (briordan@indiana.edu)
Indiana University

Chen Yu   (chenyu@indiana.edu)
Indiana University

Prior studies have found that language learners can solve the referential ambiguity problem in ambiguous environments by tracking word-referent mappings across trials, even in the absence of social, linguistic, and other cues (Yu and Smith, 2007). Here we present evidence that such statistical word learning can co-occur in parallel with two different kinds of statistical learning ? one to extract statistical regularities in visual scenes (Fiser & Aslin, 2001), and one to extract statistical regularities in language (Newport & Aslin, 2004). Across a variety of learning conditions, subjects successfully mapped words to referents at the same time as they discovered and represented multiple aspects of the statistical structure from both the linguistic input and nonlinguistic context. Further, our results indicate that the additional statistical regularities in visual scenes and in language can not only be acquired in the same learning session but also spontaneously facilitate and bootstrap concurrent word-referent statistical learning.


(Return to Top)

Mutual Exclusivity in Adjective Learning: The Case of Bilingual Children and Monolingual Children

Hanako Yoshida   (yoshida@uh.edu)
University of Houston

Megumi Kuwabara   (mekuwaba@indiana.edu)
Indiana University

Maria Guerrero   (mafe042000@yahoo.com)
University of Houston

To what extent are early word-learning constraints universal across different learning environments? Using a novel adjective learning task, the present study investigates whether bilingual children show a lesser degree of mutual exclusivity (ME). Sixty-one 3-year-old English monolingual and bilingual children of various language combinations (e.g., English-Spanish, English-Japanese, English-Romanian) participated in the study. They were randomly assigned into three conditions in which they either mapped (1) a novel label to an unfamiliar property, (2) a known label to an unfamiliar property, or (3) a known label to a corresponding familiar property. The critical condition (2) tested the use of ME in their adjective mapping by requiring the acceptance of multiple labels for a single referent. That successful mapping in this critical condition was achieved only by monolingual children supports the different degree of use of ME across monolingual and bilingual children. The possible mechanisms behind this difference are discussed.


(Return to Top)

Poster Session II -- Modeling and Experimental Approaches of Cognitive Processing

(Friday, July 25, 2008, 5:00-7:00 PM, Exhibit Hall)


The Phylogenetic Roots of Cognitive Dissonance

Jennifer Vonk   (jenvonk@gmail.com)
University of Southern Mississippi Gulf Coast

Samantha West   (sweet_samantha_ms@yahoo.com)
University of Southern Mississippi Gulf Coast

Stephanie E. Jett   (FEMMEFROG@mchsi.com)
University of South Alabama

We presented old-world monkeys, chimpanzees and American black bears with a cognitive dissonance paradigm modeled after Egan et al. (2007). In experimental trials subjects were given choices between two equally preferred food items and then presented with the unchosen option and a novel, equally preferred, food item. In control trials subjects were presented with one accessible and one inaccessible option from another triad of equally preferred food items. They were then presented with the previously inaccessible item and a novel member of that triad. As we predicted, subjects, as a group, preferred the novel item in experimental but not control trials, indicating that they perceived and resolved cognitive dissonance by de-valuing the unchosen option only when an option was derogated by their own free choice. Only monkeys, but not apes or bears showed this effect suggesting that this phenomenon might exist within but not outside of the primate order.


(Return to Top)

Predicting Cognitive Driver Distraction with Threaded Cognition Theory

Dario D. Salvucci   (salvucci@cs.drexel.edu)
Drexel University

Joanna Beltowska   (joannabeltowska@gmail.com)
Drexel University

Threaded cognition theory describes how people integrate and perform multiple concurrent tasks, providing a computational model of human multitasking situated in the ACT-R cognitive architecture. We present further evidence for this theory by predicting performance in a driver distraction task and validating these predictions with data from human drivers. Specifically, we show how threaded cognition accounts for the subtle but significant effects of a simple list-memory task on driver behavior. These results extend previous results for typical perceptual-motor tasks (e.g., dialing a phone) in demonstrating that even purely cognitive secondary tasks can adversely affect driver performance.


(Return to Top)

Storage and Recall in Simple Recurrent Neural Networks

Christo N. Kirov   (kirov@cogsci.jhu.edu)
Johns Hopkins University

On average, simple recurrent networks (SRNs) trained to recall strings in serial order make increasingly fewer errors toward the end of each recalled string. Networks trained to recall the mirror image of strings make increasingly more errors toward the end of each recalled string. Presumably, this pattern is caused by differences between the queue-like and stack-like memories required to perform these two tasks. Past analytical and empirical examination of SRNs has suggested that these differences are localized in the way the networks encode the input string. The decoding required to recall the input is assumed to proceed identically in both cases. We train networks to perform both serial recall and mirrored recall tasks concurrently - reading symbols from either the beginning or end of an input string given a corresponding queue. Examination of these networks suggests that they employ both a stack-like and queue-like memory, implemented independently across different dimensions in hidden unit space.


(Return to Top)

How Perception and Mapping Interact During the Analogy-Making Process and the Process of Reinterpretation

Boicho Kokinov   (bkokinov@nbu.bg)
New Bulgarian University

Svetlin Kosev   (skosev@cogs.nbu.bg)
New Bulgarian University

This paper describes an experiment that tests the theoretical assumption that perception and mapping interact during the process of analogy-making. This is an issue that has been a focus of heated debates in the past, but has been never empirically tested. Kokinov, Bliznashki, Kosev, Hristova (2007) presented evidence that analogy-making can cause the stimuli to be reinterpreted and therefore the mapping process can influence the representation building process. However, one can still argue that the two processes are working consequently, simply after the analogical mapping a new perceptual process is started (like in a loop) that caused the reinterpretation. This paper presents new data coming from a eye-tracking study that can potentially track the dynamics of the processes. The data show that finding some specific local mappings triggers the reinterpretation to take place, and than the mapping continues further, i.e. the parallel processing assumption is backed up.


(Return to Top)

Timecourse of Recovery From Interruptions: Searching for Common Trends Across Multiple Environments

David Cades   (dcades@gmu.edu)
George Mason University

Raj Ratwani   (rratwani@gmu.edu)
George Mason University

J. Gregory Trafton   (greg.trafton@nrl.navy.mil)
United States Naval Research Laboratory

Deborah Boehm-Davis   (dbdavis@gmu.edu)
George Mason University

A solid body of research exists examining the disruptive effects of interruptions. One relatively robust finding is that people take longer to resume a task following longer interruptions. It is not clear if the pattern of recovery is different across tasks or settings. The current meta-analytic approach attempts to integrate these interruption duration findings and bring to light any trends that exist across tasks, methodologies, and settings. Data from multiple laboratories, experiments, environments, and a computational cognitive model were compiled and analyzed to evaluate the effect of interruption length on primary task resumption time. The data suggest a similar pattern of recovery following interruptions of varying lengths across the studies analyzed. Furthermore, interruption length led to similar amounts of disruptions across tasks. These results suggest that across settings or tasks, the interruption duration can be used to accurately predict the disruptiveness of any interruption.


(Return to Top)

An Integrated Model of Action Video Game Play

Marc Destefano   (destem@rpi.edu)
Rensselaer Polytechnic Institute

Wayne D. Gray   (grayw@rpi.edu)
Rensselaer Polytechnic Institute

We report the latest progress on a cognitive model capable of playing the real-time action video game Space Fortress. The model runs on an customized extension of ACT-R, and communicates with an external simulation of the task environment through rapid inter-process communication. The model integrates all modules of ACT-R, in addition to manual control of a second-order system, time interval estimation, and precise visual tracking.


(Return to Top)

Spatial Modeling Using a Bimodal Cognitive Architecture

Unmesh Kurup   (kurupu@rpi.edu)
Rensselaer Polytechnic Institute

B Chandrasekaran   (chandra@cse.ohio-state.edu)
The Ohio State University

In earlier work, Kurup and Chandrasekaran have reported on biSoar, a general bimodal cognitive architecture. The focus of this paper is on certain methodological advantages of using biSoar for cognitive modeling. We build a biSoar model for a concrete spatial reasoning task intended to study the influence of the spatial goal on memory. While the task itself is of intrinsic interest, our goal in modeling this task is to demonstrate two advantages of the architectural approach - that architectural models can exhibit emergent phenomena and provide novel predictions, and that it is easy to distinguish architectural causes of observed behavior from causes related to content. Since biSoar has no commitments to perceptual and motor mechanisms, we also introduce an interface that allows it to interact with an external representation.


(Return to Top)

A Computational Model of Repetition Blindness Using a Liquid State Machine

Patrick Michael Hynes   (phynes@cs.nuim.ie)
National University of Ireland Maynooth

Ronan Reilly   (ronan@cs.nuim.ie)
National University of Ireland Maynooth

Repetition Blindness occurs when two identical visual stimuli, in this case words, appear in quick succession during Rapid Serial Visual Presentation (RSVP). Depending on how far apart the stimuli are separated, the second stimuli will not have been recognised by the subject. We describe a novel computational framework known as a Liquid State Machine to model this psychological phenomenon. The Liquid State Machine is a computational framework that exhibits the properties and behaviour's of liquid. It is perturbed by inputs similar to a pond of water being perturbed by stones. This paper outlines how the Liquid State Machine is utilised to simulate repetition blindness. The parameters of these experiments and their results are described and discussed


(Return to Top)

Computational Perception of Sizes

Julia Taylor   (tayloj8@email.uc.edu)
University of Cincinnati

Lawrence Mazlack   (mazlack@uc.edu)
University of Cincinnati

Human computer interaction and communication is becoming more and more frequent. This means that computers must understand the meanings of words used to describe objects and events, as well as comprehend the meaning of modifiers used to describe them. This is often a difficult task: the word large means two different things when describing a chair and a building. This paper investigates the use of size-adjectives for objects of different physical proportions and proposes a method for deriving the meaning of such adjectives suitable for computational purposes.


(Return to Top)

The Speed/Accuracy Tradeoff in Estimating Means: The Role of Data Characteristics

Bradley Morris   (morrisb@gvsu.edu)
Grand Valley State University

Amy Masnick   (Amy.M.Masnick@hofstra.edu)
Hofstra University

Christa Natschke   (natschkc@student.gvsu.edu)
Grand Valley State University

Adrianne Spenner   (adriane12886@hotmail.com)
Grand Valley State University

Stephanie Hammond   (hammonst@student.gvsu.edu)
Grand Valley State University

Deardra Kearney   (speedy1412@chartermi.net)
Grand Valley State University

Previous research has indicated that adults automatically represent means and variance given number sets. However, it is unclear how this representational format influences approximate mathematical operations such as mean estimation. In this study we asked twenty subjects to estimate the mean of 80 sets of data in which the number of observations (2-6) and coefficient of variation (10% or 20% of the mean) were varied; estimated means and reaction times were measured. Each subject was assigned to one of two conditions: a fast condition in which subjects were asked to calculate means as rapidly as possible, ignoring accuracy and an accurate condition in which subjects were asked to calculate means as accurately as possible, ignoring time. Results indicate that although reaction times were significantly longer in the accurate condition, accuracy did not differ between conditions, and reaction times increased linearly with the number of observations, and with larger relative variance.


(Return to Top)

Discrete Measurement of Sensory Information Using Bayesian Networks

Chris Thornton   (c.thornton@sussex.ac.uk)
University of Sussex

In principle, information theory can be used to measure the amount of information generated by sensory apparatus. This can be the basis for evaluating the viability of a cognitive model. In practice, however, such checks are rarely made due to the complexity of agent-level, informational analysis. Where it is the agent itself which is the `receiver', measurement of sensory information involves determining the way interpretive processes affect stimulus probabilities. No practical method for performing this type of analysis has been developed. The paper shows, however, that Bayesian networks can be adapted for this usage. Illustrative examples are given in three domains but the method is completely general and can be applied to any model which has a sensory component.


(Return to Top)

Poster Session II -- Spatial Cognition

(Friday, July 25, 2008, 5:00-7:00 PM, Exhibit Hall)


Effects of Social Information on Distance Estimation

Justin L. Matthews   (jmatthews@ucmerced.edu)
University of California, Merced

Teenie Matlock   (tmatlock@ucmerced.edu)
University of California, Merced

How do people think about space, both actual and social? What's the connection between these constructs, if any? As limited research has explored this issue, our goal is to explore this area further. In this study, 236 UC Merced students viewed spatial scenes and read stories about a person walking to a stranger or a friend, and then estimated walking distance. On average men's distance estimates (feet) were twice as long as women's estimates (M=1200.31 vs. M=623.29, p<.01). Moreover, differences pertaining to familiarity (whether protagonist walked to friend or stranger) varied according to type of spatial domain. The results suggest that spatial distance and social distance are conceptually linked. As suggested by Matthews and Matlock (2007), the link may be more complex than a simple positive relationship, where increased social distance is analogous to greater physical distance.


(Return to Top)

Encoding Spatial Layout in the Dark: Robustness of Visual Spatial Learning

Naohide Yamamoto   (nyama@gwu.edu)
George Washington University

John W. Philbeck   (philbeck@gwu.edu)
George Washington University

Previous studies have suggested that the advantage of vision over other modalities in encoding spatial layout stems from vision's field of view: It uniquely allows simultaneous access to a wide area of an environment with high fidelity, enhancing spatial learning by effective environmental frames of reference. The present study examined the effects of these environmental reference frames on visual spatial learning by having participants learn spatial layouts both in a lit room and in a darkroom (with phosphorescent objects). After learning each layout, they performed judgments of relative direction among objects. Results showed that memory performance was not modulated by the lighting conditions, suggesting that the advantage of vision does not arise from the presence of environmental reference frames during learning. Rather, the present findings demonstrated the robustness of visual encoding of spatial layout, which calls for further investigation on the true origin of the visual advantage in spatial learning.


(Return to Top)

The Impact of Attentional Shifts on Spatial Memory in Early Childhood

Anne R. Schutte   (aschutte2@unl.edu)
University of Nebraska-Lincoln

Brian Keiser   (bkeiser2@unl.edu)
University of Nebraska-Lincoln

Chelsie Kobza-Guerrero   (cnkobza@hotmail.com)
University of Nebraska-Lincoln

Margaret Ortmann   (mortman1@bigred.unl.edu)
University of Nebraska-Lincoln

Work with adults has shown that spatial attention influences the maintenance of a location in spatial working memory (SWM). Awh and Jonides (1998) proposed that adults use selective spatial attention as a rehearsal mechanism" for SWM, because when attention is manipulated during the delay of a SWM task, adults make more memory errors. The relationship between attention and SWM has not been systematically studied in young children even though several developmental disorders show deficits in both (e.g., ADHD). This study examined whether spatial attention influences SWM in 3- to 6-year-olds. Children completed a SWM task where they had to remember target location on large monitor with distracter dot appearing during the delay on half the trials. Over development there was an increase in children's ability to inhibit the distracter, and there appeared to be a narrowing of their "spotlight" of attention. Results are discussed in terms of Dynamic Field Theory."


(Return to Top)

Segmention of Inside-Outside Relations and Complex Contours in the Parietal Lobes

Nabeela Akhtar   (Nxa153@bham.ac.uk)
University of Birmingham

M Jane J. Riddoch   (M.J.Riddoch@bham.ac.uk)
University of Birmingham

Glyn W W. Humphreys   (G.W.Humphreys@bham.ac.uk)
University of Birmingham

The performance of two groups of neurological patients: (bilateral parietal and right parietals) on contour segmentation and inside outside relations was compared. The patients completed a series of tasks requiring them to visually segment contours and make decisions with regard to inside-outside relations. In Experiment 1 patients were required to make inside-outside boundary distinctions with complex open contours. The results show the bilateral parietal group to be significantly impaired on this task, relative to the right parietal group. In Experiment 2, on a task of closed contours the bilateral parietal patients showed a similar level of performance to the right parietals. We conclude that patients with bilateral posterior lesions have significant impairments in segmenting contours and making inside-outside discriminations. It is suggested that Bilateral patients may have a high level deficit in processing and integrating local and global aspects of shape.


(Return to Top)

The Role of Animacy in Imagined Spatial Transformations

Alfred B. Yu   (alfredyu@wustl.edu)
Washington University in St. Louis

Jeffrey M. Zacks   (jzacks@artsci.wustl.edu)
Washington University in St. Louis

Perspective transformations (PTs) are imagined spatial transformations that involve updating one's egocentric reference frame relative to fixed environment-centered and object-centered reference frames. We examined whether stimulus animacy affects the use of PTs. Subjects performed spatial tasks in which the stimuli depicted either animate things like humans and non-human animals, or inanimate things such as pieces of furniture. The use of PTs was detected by measuring the correlation between reaction time and stimulus orientation. The data suggest that animacy had an effect on the use of PTs, such that images of human beings elicited PTs more frequently than did images of non-human animals. This effect was partially explained by differences in the spatial characteristics of the images, such as skinniness.


(Return to Top)

Spatial Location Uncertainty As Modifier of Attentional Asymmetries

Dariusz Asanowicz   (asanowicz@apple.phils.uj.edu.pl)
Jagiellonian University

Piotr Wolski   (pw@apple.phils.uj.edu.pl)
Jagiellonian University

Spatial neglect is deeper and more persistent after the right hemisphere lesions. However, whereas the clinical and imaging data unequivocally support the notion of right hemisphere predominance in attention, behavioral data from research with normal subjects does not demonstrate such a clear asymmetry. The Lateralized Attention Network Test (LANT) was designed to evaluate the alerting, orienting and executive networks separately in each himsphere. The introductory study of Greene et. al. (2007) has confirmed the expected independence of the three networks, but did not show as clear hemisphere asymmetry as expected. We replicated the LANT task employing the similar procedure with additional the spatial uncertainty factor. Preliminary data show that detection of less frequent stimuli was slower when they were presented in the right visual field. These data suggesting that the right hemisphere is equally efficient with high and low probability stimuli, whereas the left has trouble processing unexpected stimuli.


(Return to Top)

Pointing Out the Role of Gesture in Spatial Development

Megan Sauter   (m-sauter@u.northwestern.edu)
Northwestern University

David Uttal   (duttal@northwestern.edu)
Northwestern University

Susan Goldin-Meadow   (sgm@uchicago.edu)
University of Chicago

Susan Levine   (s-levine@uchicago.edu)
University of Chicago

Maps can influence the way people think about space, allowing map users to integrate multiple pieces of spatial information. Maps are useful in spatial development, as young children have trouble integrating relations. Gestures, like maps, convey spatial information in a symbolic way. The purpose of our work is to determine whether gesture can influence spatial thinking and how this might proceed developmentally. In our study, six- and eight-year-old children learned the locations of animals in a six-room playhouse, communicated this information to their parent, and then answered location and inference questions. Our results show that gesture helps children think about and communicate spatial information. Eight-year-olds were better than six-year-olds at integrating spatial information, but the six-year-olds? performance could be improved if encouraged to gesture while communicating the space. Children encouraged to gesture communicated more spatial relations than those who were not, showing that gesture can also improve spatial communication.


(Return to Top)

Spatial Skills As Predictors of Geometry Achievement

Yvonne S. Kao   (ykao@andrew.cmu.edu)
Carnegie Mellon University

John R. Anderson   (ja@cmu.edu)
Carnegie Mellon University

Many researchers have found moderate correlations between various spatial skills and mathematics achievement. However, the mere existence of this correlation does not provide useful information to researchers and educators who seek to improve students? mathematics via spatial training. Both spatial skills and mathematics are large categories comprised of many diverse components. More research needs to be done to examine how specific spatial skills contribute to learning specific concepts in mathematics, above and beyond other cognitive measures such as verbal ability and general reasoning ability. In this study, we are tracking high-school students' performance in their geometry class, unit-by-unit. We have taken baseline measures of spatial reasoning, verbal ability, and general reasoning, as well as past standardized test scores and prior math grades. We will use these measures as predictors of performance in the different geometry units.


(Return to Top)

You Drive All the Way to ...?!! Effects of Previous Environment and Travel Patterns on Spatial Scaling

Penney Nichols-Whitehead
Grand Valley State University

Stephanie Smith 
Grand Valley State University

Paige Werner 
Grand Valley State University

Tara Amarose 
Grand Valley State University

Hilary Swaney 
Grand Valley State University

Tiffany Rowe 
Grand Valley State University

Most of us have encountered situations in which our personal perceptions of space are at odds with the perceptions of others, i.e., a location we deem nearby and easily accessible, another might perceive as quite distant or inaccessible. A common folk theory is that these differences are due to growing up or living in environments of differing spatial scales. This study empirically tested whether or not the scale of, or habitual movement patterns within, the environment of origin might influence later perceptions of spatial scale. 118 incoming freshman, tested within the first month of their fall semester, completed surveys regarding characteristics of past and present environments, movement patterns within those environments, and/or made judgments categorizing pairs of locations as near or far apart. Results indicate that past movement patterns, rather than the scale of the environment of origin, impacted spatial scaling and movement patterns within the new/university environment.


(Return to Top)

The Relationship Between the Perception of Symmetry and Spatial Memory

Margaret R. Ortmann   (mortman1@bigred.unl.edu)
University of Nebraska - Lincoln

Anne R. Schutte   (aschutte2@unlnotes.unl.edu)
University of Nebraska - Lincoln

In early childhood a transition occurs in the direction of spatial working memory (SWM) biases around symmetry axes. Specifically, young children are biased toward the midline symmetry axis of a space while older children are biased away. One factor that might influence this transition is a developmental change in perception of the midline symmetry axis. To test this possibility 3-year-olds participated in a SWM task and a discrimination task over several weeks where judgments were made as to an object?s location in relation to midline. Performance in this discrimination task was correlated initially with performance in a SWM task. Children also showed significant improvement in the discrimination task over the course of the sessions. However, improvement in the discrimination task did not correlate with changes in performance in the SWM task. Thus, improvement in the perception of symmetry axes may not be the only change underlying the transition in SWM.


(Return to Top)

Mental Rotations and Spatial Cognition: Comparisons Between Vision and Touch

André F. Caissie   (andre.caissie@gmail.com)
Université De Poitiers

Lucette Toussaint   (lucette.toussaint@univ-poitiers.fr)
Université De Poitiers

Yannick Blandin   (yannick.blandin@univ-poitiers.fr)
Université De Poitiers

The present study examines the differences between vision and touch when subjects are submitted to a classical visuo-spatial task of mental rotations (MRT; Vandenberg & Kuse, 1978). For this purpose, we used visual stimuli, and its physical replica, using physical three-dimensional geometrical figures as tactile stimuli. Comparisons between active visual perception and active touch perception gathered information pertaining to commonalities, and differences, when these two modalities function in a complex spatial task. Physical auscultations of the item figures in video format and ocular movement recordings rendered several comparable latency measures, accuracy measures, and qualitative strategy markers between the visual and tactile task performances. Findings suggest a functional approximation between vision and touch. Several issues are discussed as to the limits of this approach and attention is allotted to the prospective question regarding the possibility of measuring similar/dissimilar cognitive constructs with visual and tactile tasks.


(Return to Top)

Spatial Reasoning in Cognitive Architectures

Michael Matessa   (mmatessa@alionscience.com)
Alion

This poster will describe dynamic spatial reasoning enhancements to the Graph-Based Interface Language tool (GRBIL). GRBIL can be used to evaluate interfaces by creating operator models by demonstrating operator actions. The models are created in the ACT-R cognitive architecture, and an integration of ACT-R with a diagrammatic reasoning theory (DRS) allows ACT-R to perform spatial reasoning. This spatial reasoning ability can range from the simple detection of an intersection of projected lines to complex pathfinding. Previous work that integrated DRS with the Soar cognitive architecture found that embedding spatial reasoning in a cognitive architecture produces psychologically plausible side-effects such as the automatic simplification of path representation. This simplification of representation and also transfer of past solutions to novel situations will be demonstrated in a robotic control task.


(Return to Top)

Spatial Cognition in Different Spaces

Harry Haroutioun Haladjian   (haladjian@ruccs.rutgers.edu)
Rutgers University

Carlos Montemayor   (montemayor@philosophy.rutgers.edu)
Rutgers University

In this paper, we explore the cognitive representation of space and how this representation corresponds to physical space. The problem of coordinating mental structures with physical ones continues to be an active topic of study in cognitive science, especially in the field of spatial cognition, spatial navigation, and in attempts to solve the binding problem in mental representations. Cognitive scientists generally refer to two distinct types of spatial representations: metric and conceptual. By merely posing different representations of space, we cannot fully address the problem of how a spatial representation is coordinated with physical space. We propose that a referential system, such as the indexing mechanism in visual processing, is critical for coordinating mental space and physical space. Theoretical and empirical support for this referential mechanism will be discussed, as well as how this mechanism might be the crucial link for connecting mental space to physical space.


To Poster Session III Member Abstracts, Saturday, July 26, 2008, 5:00-7:00 PM, Exhibit Hall

To Poster Session I Member Abstracts, Thursday, July 24, 2008, 5:30-7:00 PM, Exhibit Hall

 

Cognitive Science Journal Archive

The Cognitive Science Journal Archive currently contains electronic versions of 459 articles (of 98 issues and 24 years) of the Cognitive Science Journal and collects materials published in the Proceedings of the Annual Cognitive Science Conference. It is maintained by the CogWorks Laboratories of RPI's Cognitive Science Department.

CogWorks Labs, RPI