Abstracts 2005 - The Psychonomic Society

Abstracts 2005 - The Psychonomic Society Abstracts 2005 - The Psychonomic Society

psychonomic.org
from psychonomic.org More from this publisher
29.01.2013 Views

Papers 265–271 Sunday Morning Event Cognition Grand Ballroom Centre, Sunday Morning, 8:00–10:00 Chaired by Kathy Pezdek, Claremont Graduate University 8:00–8:15 (265) Eyewitness Guessing: Does Postevent Guessing During Police Interrogation Suggestively Influence Eyewitness Memory? KATHY PEZDEK, KATHRYN SPERRY, SHIRLEY LAM, & SHANA OWENS, Claremont Graduate University—Two experiments confirm that encouraging eyewitnesses to guess when they are unsure of the answer to a question suggestively influences their memory. In Experiment 1, subjects viewed a 5-min video of a crime, followed by 22 questions: (1) 16 answerable questions and (2) 6 unanswerable questions about information not observed in the video. Subjects in the “forced guess” experimental condition were instructed to answer all questions; they were not given a “don’t know” response option. Subjects in the “no guess” control condition answered the same 22 questions, but with a “don’t know” response option. One week later, subjects answered the same 22 questions with the “don’t know” response option provided. Thirty-eight percent of the subjects who were forced to answer unanswerable questions at Time 1, gave the same answer at Time 2, even when the “don’t know” response option was available. Experiment 2 replicated Experiment 1 and further demonstrated that the subjects who guessed answers to unanswerable questions three times at Time 1 were even more likely to provide that same guessed answer at Time 2. Encouraging eyewitnesses to guess answers to questions about which they report having no memory is a forensically inappropriate procedure. 8:20–8:35 (266) Effects of Age and Response Deadlines on Unconscious Transference in Eyewitness Memory. ALAN W. KERSTEN, JULIE L. EARLES, & ARIAN R. LOMER, Florida Atlantic University—This research examined the ability of young and older adults to associate actors with their actions. Participants viewed a series of video clips involving 60 different actors performing 60 different actions. One week later, participants were tested for recognition of static frames from the video clips. Participants were given either a short or a longer period of time to respond. The critical test type involved an actor seen at encoding performing an action that had been performed by a different actor at encoding. Young adults given a short deadline resembled both groups of older adults in their false recognition of these conjunction stimuli, whereas young adults given a longer deadline showed less false recognition of the conjunction stimuli. These results are consistent with the theory that unconscious transference results from the familiarity of an actor and an action in the absence of recollection of the contexts in which they were encountered. 8:40–8:55 (267) Effects of Aging and Dementia on Event Perception and Event Memory. JEFFREY M. ZACKS & NICOLE K. SPEER, Washington University, JEAN M. VETTEL, Brown University, & LARRY L. JA- COBY & MARTHA STORANDT, Washington University—To stay oriented, to remember the past, and to plan for the future, people segment ongoing activity into discrete events. Two experiments studied how event segmentation and event memory change with healthy aging and very mild dementia of the Alzheimer type. After watching movies of everyday activities and segmenting them into meaningful events, participants were given tests of memory for the events in the movies. Segmentation was impaired in older adults in comparison with younger adults and impaired further in people with dementia. Memory for temporal order and for visual details showed a similar pattern. Within the older participants, poor segmentation was associated with poor memory. Poor knowledge for the typical structure of everyday events also was associated with poor later memory. These measures of event perception and memory were partially dissociable from general cognitive fitness and thus may indicate a unique source of cognitive changes in aging and dementia. 42 9:00–9:15 (268) Schema Consistency and the Misinformation Effect. ROBERT F. BELLI & ROBERT J. NEMETH, University of Nebraska—Our study explored the combined effects of schematic knowledge, contradictory/ additive misinformation, and retention interval on the production of false memories. Shortly following exposure to scenes, subjects were misled with items that were schema consistent (a toaster in a kitchen) or inconsistent (a stapler in a kitchen) and which either did or did not contradict an original item (a tape dispenser in a kitchen). A cued recall test with remember and know judgments followed either 10 min or 1 week later. A 2 (consistent, inconsistent) � 2 (contradictory, additive) � 2 (remember, know) � 2 (10 min, 1 week) ANOVA revealed only two main effects: Schema-inconsistent misinformation led to significantly more false memories, as did the shorter retention interval. Results are inconsistent with those of Roediger et al. (2001), who found that items with higher schema consistency produce more false memories, and they do not replicate those of Frost (2000), who found that remember responses increased with a longer retention interval for additive misinformation. 9:20–9:35 (269) Color Change as a Cause of Object Movement: Revisited. MICHAEL E. YOUNG & OLGA NIKONOVA, Southern Illinois University, Carbondale—Michotte (1946/1963) concluded that color “has no bearing whatever on the question of qualitative causality” (p. 235). Surprisingly, this claim has received little empirical investigation in the 60 years since its publication. In a series of experiments involving the launching effect, we investigated people’s judgments of delayed causation when the delay was filled with a sequence of discrete or continuous color changes to the launching object. Judgments of causation were significantly stronger when the delay was filled by color changes rather than unfilled, but the relationship between the type of color change and judgments was context dependent. 9:40–9:55 (270) Asymmetries in Motion Event Representations. LAURA M. LAKU- STA, ALLISON WESSEL, & BARBARA LANDAU, Johns Hopkins University (read by Barbara Landau)—In language there is a bias to represent goals (end points) over sources (starting points). The present studies explored the possibility that this bias originates in the nonlinguistic representation of events using a nonlinguistic change detection method. Four-year-olds and adults were shown pairs of motion events in which the second event changed in terms of source, goal, figure, or motion or did not change at all. After viewing the second event, participants judged whether the events were the same or different. In Experiment 1, goal changes were correctly detected more often than source changes, suggesting a nonlinguistic goal bias. However, this bias became weaker when the actor gazed at the source rather than at the goal while motion occurred (Experiment 2) and when the events contained only inanimate objects (Experiment 3), suggesting that intentionality plays a key role in the representation of events and that of goals are important. Errors in Judgment and Decision Making Grand Ballroom East, Sunday Morning, 8:00–10:00 Chaired by George L. Wolford, Dartmouth College 8:00–8:15 (271) The Effect of Context on Evaluating Mammograms. GEORGE L. WOLFORD & ANNE P. ROWLAND, Dartmouth College—Misread mammograms underlie one of the most common types of malpractice suits. The evaluation of earlier mammograms in such suits takes place in the face of subsequent knowledge about the existence and often the location of a tumor. In a controlled experiment, we examined the effect of such knowledge on the identification of tumors. We found that knowledge about presence and location led to significant increases in identifications of tumors in earlier stages. Our work suggests that

Sunday Morning Papers 272–278 many malpractice suits in mammography are influenced, perhaps wrongly, by this hindsight bias. 8:20–8:35 (272) “As Soon As the Bat Met the Ball, I Knew It Was Gone”: Accuracy and Hindsight Bias in Predicting Action Outcomes. ROB GRAY, Arizona State University, SIAN L. BEILOCK, University of Chicago,& THOMAS H. CARR, Michigan State University—Hindsight bias (the “knew-it-all-along effect”) provides a measure of strength and resistance to distortion of memory for predicted outcomes. A virtual-reality batting task was used to compare novice and expert baseball players’ ability to predict their swing outcomes, as well as players’ susceptibility to hindsight bias. During each swing, the simulation stopped when bat met ball. Batters marked where on the field they thought the ball would land. Correct feedback was then displayed, after which batters attempted to re-mark the location they had indicated prior to feedback. Expert batters made more accurate initial predictions and showed less hindsight bias in their postfeedback marking. Furthermore, prediction accuracy and hindsight bias were inversely related to number of hits in the previous block of trials. These results suggest that experts pay attention to different real-time performance information than do novices and that experts’ attention changes dynamically as a function of performance success or failure. 8:40–8:55 (273) Paradoxical Optimism About What Might Have Been. AIDAN FEENEY, IAN M. DAVISON, & VICTORIA M. TILLETT, Durham University—When things go wrong, people often consider how events might have turned out better. One view is that beliefs that things could easily have been better amplify negative affect. In the first study described here, we found that people expressed more confidence that a preferred alternative outcome could have happened than did a group of observers judging the same outcome. In a second study, we showed that for recent and distant regrets, high self-esteem (HSE) participants were more confident that things could have been as they would have preferred than low self-esteem (LSE) participants. However, LSE participants reported more intense regrets than did HSE participants. Degree of confidence did not predict intensity of regret. Taken together, these results suggest that people hold optimistic beliefs that things could have been better and that rather than amplifying negative affect, such beliefs make people feel better about themselves. 9:00–9:15 (274) A Statistical–Ecological Account of the Effects of Sample Size on Correlational and Causal Inference. RICHARD B. ANDERSON, MICHAEL E. DOHERTY, & JEFF C. FRIEDRICH, Bowling Green State University—Research on the distributional characteristics of correlation coefficients, and of other measures of statistical association, suggests that when the task is to detect the presence of a population correlation, the environment can favor organisms that are limited in their capacity to gather and process information. When the task is to estimate the strength of a population correlation, however, such organisms may be disadvantaged. In a behavioral study of correlation detection, if the decision criterion was extremely liberal, detection was more accurate when based on a small rather than a large sample; if the criterion was moderate or conservative, accuracy was greater for large than for small samples. A second study examined people’s ability to estimate (rather than simply decide on the presence or absence of) population correlations, and yielded preliminary evidence suggesting that small samples may lead to exaggerated estimations of relationship strength. 9:20–9:35 (275) Comparison-Induced Anchoring Effects. JESSICA M. CHOPLIN & MARK W. TAWNEY, DePaul University—We propose a comparisoninduced distortion theory (Choplin & Hummel, 2002) account of anchoring effects, wherein biases created by verbal comparisons mediate the effects of anchors on estimation. This model, like previous 43 models of anchoring effects, usually predicts biases toward anchor values, but unlike previous models, it sometimes predicts biases away from anchor values. Furthermore, unlike previous models, this model predicts that the words used to express comparisons will influence estimation. The predictions of this model were tested in two experiments. In Experiment 1, participants were asked to compare the tobe-estimated value to the anchor, and the words used to express the comparison were manipulated before the participants estimated the value. Experiment 2 featured the same task, but the range of acceptable estimates was constrained in order to test the prediction that anchors will sometimes bias estimation away from anchor values. The results of these two experiments suggest that some anchoring effects are comparison induced. 9:40–9:55 (276) Understanding and Modeling Human Sequential Decision Making Under Uncertainty Using Bayesian Statistics. BRIAN J. STANKIE- WICZ, CHRIS GOODSON, & ANTHONY R. CASSANDRA, University of Texas, Austin—The present study investigates human performance in a sequential decision making with uncertainty task in which the observer’s task is to localize a target using reconnaissance and “destroy” the enemy using artillery. Each of these actions comes with a cost, and the outcomes of these actions are probabilistic. Using a partially observable Markov decision process (Bayesian model), we calculated the optimal performance and compared this performance to the humans’ performance. We found that subjects performed at about 60% efficiency in this task. Further studies revealed that one of the primary limitations preventing human subjects from acting optimally was an inability to accurately update the likelihood of the true state of the system given the previous actions and observations. By providing subjects with an external representation that updated the likelihoods, subjects’ efficiency values increased to 80%–90%. Associative Representations Grand Ballroom West, Sunday Morning, 8:00–9:40 Chaired by Curt Burgess, University of California, Riverside 8:00–8:15 (277) A Turing Test of a Generative Word Association Model. JON WILL- ITS & CURT BURGESS, University of California, Riverside (read by Curt Burgess)—The (free) association strength between words is used as a basis for many theories of semantic memory and as an explanatory construct for countless memory retrieval effects. Word associations represent a set of relationships between words but provide little theoretical insight into exactly how and why certain words are associated and others are not. A model of word association productions is presented that is based on the probability of two words co-occurring within language and the contextual substitutability (or global co-occurrence similarity) of the two words. The model was used to generate word associations for 30 stimuli, and these associations were compared against those produced by human subjects. Experts (cognitive psychologists) and novices (undergraduates) were tested on their ability to distinguish the human and model-generated associations, constituting a Turing test of the model. This understanding of the word generation process provides a theoretical understanding of this long-used technique. 8:20–8:35 (278) Using Global Co-Occurrence to Predict Word Associations, Features, and Semantic Priming. JON WILLITS & CURT BURGESS, University of California, Riverside, & CATHERINE DECKER, Chaffey College (read by Catherine Decker)—Featural similarity and word association strength (both via production norms) can be used to model most semantic memory effects. However, they provide little theoretical insight into how representations based on these relationships come about. We suggest that both are emergent by-products of statistical relationships in the learning environment. Two simulations were performed, modeling human-generated feature productions and word as-

Sunday Morning Papers 272–278<br />

many malpractice suits in mammography are influenced, perhaps<br />

wrongly, by this hindsight bias.<br />

8:20–8:35 (272)<br />

“As Soon As the Bat Met the Ball, I Knew It Was Gone”: Accuracy<br />

and Hindsight Bias in Predicting Action Outcomes. ROB GRAY,<br />

Arizona State University, SIAN L. BEILOCK, University of Chicago,&<br />

THOMAS H. CARR, Michigan State University—Hindsight bias (the<br />

“knew-it-all-along effect”) provides a measure of strength and resistance<br />

to distortion of memory for predicted outcomes. A virtual-reality batting<br />

task was used to compare novice and expert baseball players’<br />

ability to predict their swing outcomes, as well as players’ susceptibility<br />

to hindsight bias. During each swing, the simulation stopped<br />

when bat met ball. Batters marked where on the field they thought the<br />

ball would land. Correct feedback was then displayed, after which batters<br />

attempted to re-mark the location they had indicated prior to feedback.<br />

Expert batters made more accurate initial predictions and<br />

showed less hindsight bias in their postfeedback marking. Furthermore,<br />

prediction accuracy and hindsight bias were inversely related to<br />

number of hits in the previous block of trials. <strong>The</strong>se results suggest<br />

that experts pay attention to different real-time performance information<br />

than do novices and that experts’ attention changes dynamically<br />

as a function of performance success or failure.<br />

8:40–8:55 (273)<br />

Paradoxical Optimism About What Might Have Been. AIDAN<br />

FEENEY, IAN M. DAVISON, & VICTORIA M. TILLETT, Durham<br />

University—When things go wrong, people often consider how events<br />

might have turned out better. One view is that beliefs that things could<br />

easily have been better amplify negative affect. In the first study described<br />

here, we found that people expressed more confidence that a<br />

preferred alternative outcome could have happened than did a group<br />

of observers judging the same outcome. In a second study, we showed<br />

that for recent and distant regrets, high self-esteem (HSE) participants<br />

were more confident that things could have been as they would have<br />

preferred than low self-esteem (LSE) participants. However, LSE participants<br />

reported more intense regrets than did HSE participants. Degree<br />

of confidence did not predict intensity of regret. Taken together,<br />

these results suggest that people hold optimistic beliefs that things<br />

could have been better and that rather than amplifying negative affect,<br />

such beliefs make people feel better about themselves.<br />

9:00–9:15 (274)<br />

A Statistical–Ecological Account of the Effects of Sample Size on<br />

Correlational and Causal Inference. RICHARD B. ANDERSON,<br />

MICHAEL E. DOHERTY, & JEFF C. FRIEDRICH, Bowling Green<br />

State University—Research on the distributional characteristics of<br />

correlation coefficients, and of other measures of statistical association,<br />

suggests that when the task is to detect the presence of a population<br />

correlation, the environment can favor organisms that are limited<br />

in their capacity to gather and process information. When the task<br />

is to estimate the strength of a population correlation, however, such<br />

organisms may be disadvantaged. In a behavioral study of correlation<br />

detection, if the decision criterion was extremely liberal, detection<br />

was more accurate when based on a small rather than a large sample;<br />

if the criterion was moderate or conservative, accuracy was greater for<br />

large than for small samples. A second study examined people’s ability<br />

to estimate (rather than simply decide on the presence or absence<br />

of) population correlations, and yielded preliminary evidence suggesting<br />

that small samples may lead to exaggerated estimations of relationship<br />

strength.<br />

9:20–9:35 (275)<br />

Comparison-Induced Anchoring Effects. JESSICA M. CHOPLIN &<br />

MARK W. TAWNEY, DePaul University—We propose a comparisoninduced<br />

distortion theory (Choplin & Hummel, 2002) account of anchoring<br />

effects, wherein biases created by verbal comparisons mediate<br />

the effects of anchors on estimation. This model, like previous<br />

43<br />

models of anchoring effects, usually predicts biases toward anchor<br />

values, but unlike previous models, it sometimes predicts biases away<br />

from anchor values. Furthermore, unlike previous models, this model<br />

predicts that the words used to express comparisons will influence estimation.<br />

<strong>The</strong> predictions of this model were tested in two experiments.<br />

In Experiment 1, participants were asked to compare the tobe-estimated<br />

value to the anchor, and the words used to express the<br />

comparison were manipulated before the participants estimated the<br />

value. Experiment 2 featured the same task, but the range of acceptable<br />

estimates was constrained in order to test the prediction that anchors<br />

will sometimes bias estimation away from anchor values. <strong>The</strong><br />

results of these two experiments suggest that some anchoring effects<br />

are comparison induced.<br />

9:40–9:55 (276)<br />

Understanding and Modeling Human Sequential Decision Making<br />

Under Uncertainty Using Bayesian Statistics. BRIAN J. STANKIE-<br />

WICZ, CHRIS GOODSON, & ANTHONY R. CASSANDRA, University<br />

of Texas, Austin—<strong>The</strong> present study investigates human performance<br />

in a sequential decision making with uncertainty task in which<br />

the observer’s task is to localize a target using reconnaissance and “destroy”<br />

the enemy using artillery. Each of these actions comes with a<br />

cost, and the outcomes of these actions are probabilistic. Using a partially<br />

observable Markov decision process (Bayesian model), we calculated<br />

the optimal performance and compared this performance to<br />

the humans’ performance. We found that subjects performed at about<br />

60% efficiency in this task. Further studies revealed that one of the<br />

primary limitations preventing human subjects from acting optimally<br />

was an inability to accurately update the likelihood of the true state<br />

of the system given the previous actions and observations. By providing<br />

subjects with an external representation that updated the likelihoods,<br />

subjects’ efficiency values increased to 80%–90%.<br />

Associative Representations<br />

Grand Ballroom West, Sunday Morning, 8:00–9:40<br />

Chaired by Curt Burgess, University of California, Riverside<br />

8:00–8:15 (277)<br />

A Turing Test of a Generative Word Association Model. JON WILL-<br />

ITS & CURT BURGESS, University of California, Riverside (read by<br />

Curt Burgess)—<strong>The</strong> (free) association strength between words is used<br />

as a basis for many theories of semantic memory and as an explanatory<br />

construct for countless memory retrieval effects. Word associations represent<br />

a set of relationships between words but provide little theoretical<br />

insight into exactly how and why certain words are associated and others<br />

are not. A model of word association productions is presented that<br />

is based on the probability of two words co-occurring within language<br />

and the contextual substitutability (or global co-occurrence similarity)<br />

of the two words. <strong>The</strong> model was used to generate word associations<br />

for 30 stimuli, and these associations were compared against those<br />

produced by human subjects. Experts (cognitive psychologists) and<br />

novices (undergraduates) were tested on their ability to distinguish the<br />

human and model-generated associations, constituting a Turing test of<br />

the model. This understanding of the word generation process provides<br />

a theoretical understanding of this long-used technique.<br />

8:20–8:35 (278)<br />

Using Global Co-Occurrence to Predict Word Associations, Features,<br />

and Semantic Priming. JON WILLITS & CURT BURGESS, University<br />

of California, Riverside, & CATHERINE DECKER, Chaffey<br />

College (read by Catherine Decker)—Featural similarity and word association<br />

strength (both via production norms) can be used to model<br />

most semantic memory effects. However, they provide little theoretical<br />

insight into how representations based on these relationships come<br />

about. We suggest that both are emergent by-products of statistical relationships<br />

in the learning environment. Two simulations were performed,<br />

modeling human-generated feature productions and word as-

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!