|
The Cognitive Lunch talks will be on Wednesdays from 12:10 pm - 1:25 pm in the Psychology conference room (PY 128) located behind the main office.
- 09/08 Tom Busey and Chen Yu - Abstract
- 09/15 Wenyi Zhou and Jonathon D. Crystal - Abstract
- 09/22 Stephen E. Denton and Richard M. Shiffrin - Abstract
- 09/29 Dan Little and Rich Shiffrin - Abstract
- 10/06 Cameron Buckner - Abstract
- 10/13 Drew Hendrickson, George Kachergis, Todd Gureckis, and Rob Goldstone - Abstract
- 10/20 Chris Donkin, Denis Cousineau and Rich Shiffrin - Abstract
- 10/27 Jim Walker - Abstract
- 11/03 A. George Wilson and Jonathon D. Crystal - Abstract
- 11/10 Asuka Terai - Abstract
- 11/17 Bill Timberlake, Indiana University - Abstract
- 11/24 Holiday - Abstract
- 12/01 Mark Mon-Williams - Abstract
- 12/08 Linda Smith and Chen Yu, Indiana University - Abstract
Abstract 9/8: Tom Busey and Chen Yu
Characterizing Perceptual Expertise In Fingerprint Examiners.
Fingerprints recovered from crime scenes must be matched to prints stored in databases. The latent prints from the scene are often corrupted by visual noise and are distorted, partial copies of stored images. This requires matching by human experts since computers have difficulty dealing with the visual noise. However, the human matching procedure is considered by critics to be a subjective procedure that is lacking scientific rigor. In this talk I will describe how we use eyetracking methodology to determine which features draw the gaze of experts and novices, and apply techniques developed for machine translation to determine correspondences across print pairs. Finally, I will describe our initial attempts to exact a feature set that might distinguish visual features attended by experts from those attended by novices. This involves extracting pixel patches centered on each fixation and then using data reduction and classification procedures from machine learning. Our current procedures have met with limited success, and it is my hope that the audience will have additional suggestions for techniques to apply.9/15: Wenyi Zhou and Jonathon D. Crystal
Validating A Rodent Model Of Episodic Memory
Episodic memory involves encoding and retrieving the contents of a personal past episode, including what happened, where it took place and when it occurred (what-where-when memory). A rodent model of episodic memory is important because it may provide insight and tools to understand human memory disorders. In our approach, rats were tested in an eight-arm radial maze, either in the morning or in the afternoon. Each session had a study phase and a test phase, separated by a constant retention interval. A distinctive flavor (chocolate) was given at a daily-unique location to each rat at only one of these times and chow pellets were baited in other arms. In the test phase, the trial continued with chow available at arms that were inaccessible during the study phase. Chocolate replenished at its study-phase location at one time of day but not at the other time of day (counterbalance across rats). Optimal performance in the test phase is to revisit the chocolate location at the time when it was about to replenish and reduce the tendency to revisit when no chocolate was available. Meanwhile rats needed to avoid visiting previously visited chow locations in the same day because these locations did not provide additional chow. Solving this task required knowledge of what and where an event occurred in addition to the time of day the episode took place. Modeling episodic memory in animals requires careful examination of alternative explanations of performance. Here we present three projects, which attempted to rule out three non-episodic alternatives in rats. In the first project, we documented that at the time of memory assessment, rats could remember when an earlier event occurred rather than judging relative familiarity of recent events. In the second project, we examined an encoding-failure hypothesis and concluded that failing to encode the content of an episode is not used as an alternative strategy to solve what-where-when tasks. In the third project, we documented that rats are able to retrieve a memory of an earlier episode when they are unexpectedly asked to do so, ruling out the use of expectations derived from well-learned semantic rules established by extensive training. The three projects provide converging lines of evidence that rats remember an earlier episode.9/22: Stephen E. Denton and Richard M. Shiffrin
Short-Term Visual Priming Across Eye Movements
The apparent clarity of the visual world is a result of inference processes. Visual priming provides a simple and direct task useful in exploring the way the visual system performs inference in the face of successive perceptual events. In short-term visual priming, a (non-diagnostic) prime word is followed by a brief (threshold) target, a mask, and two choices (the target and a foil, both varying in similarity to the prime). Several variants of the ROUSE model explain a wide range of findings. The basic model assumes that prime features join the perceived target features, without source knowledge. The evidence assigned to a perceived feature is discounted if the feature could have arisen from the prime. Positive and negative priming are respectively the result of too little discounting for short primes and too much discounting for long primes. The present research explores the possibility that features will move from prime to target across eye movements, and if so, whether discounting will operate. The critical conditions presented a prime at fixation, of short or long duration. A signal indicated the location to which the eyes should move, and a brief target was presented there, masked, and followed by two choices. The target choice, the foil choice, or neither choice matched the prime. The results suggest that features migrate from prime to target percept across eye movements (losing source identification) and that prime evidence is then discounted. Several (expanded) parameterizations of the ROUSE model provide an excellent account of our results, with the best providing an illuminating explanation for the observed differences between eye movement conditions.9/29: Dan Little and Rich Shiffrin
Scientific Inference: Explanations for Noisy Data
An important part of the scientific process are inferences based on viewing of two dimensional data plots. We form mental models of likely functional explanations, and use these mental models to interpret our own data and data presented by others, and to judge the adequacy of models that we or others develop to explain the data. Our project, still in early stages, studies aspects of this process. We present noisy data graphs and ask scientific observers (usually grad students and postdocs) to provide us with their best judgment of the underlying causal model. The data plots vary in number of data points, amount of added noise, type of function, the existence of gaps and extrapolation regions, and the existence of odd points that seem deviant from the pattern that explains the rest of the data points. In one study we also vary the instructions given observers about the likely existence of glitches in the measurement systems that might occasionally produce essentially random or meaningless data.
We develop normative Bayesian models that consider certain representative classes of functions (low order polynomials and functions that smoothly track the data), and use such normative models to generate prior probabilities of model classes that properly balance good fit and complexity. We transform these models into equivalent Gaussian processes for the purpose of analysis, and estimate for each observer the tendency to choose different model classes, and the tendencies to adjust their inferences in accord with their mental models of the noise processes that produce the observed data.10/6: Cameron Buckner
Reforming The Distinction Between Cognition And Mere Association
Comparative psychology has long relied on a distinction between cognition and “mere association”. Recently, however, several high-profile critics have called for the distinction’s ouster. Against these critics, I argue that “cognition” and “mere association” function there as mutually-exclusive kind terms. These kinds can be characterized by reflecting upon the empirical tests that comparative psychologists currently perform to differentiate between them. An analysis of these tests evinces a family resemblance of properties, centered on the notion of behavioral flexibility, which comparative psychologists attribute to cognition and deny to mere association. While there is plenty of room for reform, a comparative investigation into the neural bases of this flexibility may finally place the distinction on firm foundations.10/13: Drew Hendrickson, George Kachergis, Todd Gureckis, and Rob Goldstone
Is Categorical Perception Really Verbally Mediated Perception?
Recent research has argued that categorization is strongly tied to language processing. For example, language (in the form of verbal category labels) has been shown to influence perceptual discrimination of color (Winawer et al., 2007). However, does this imply that categorical perception is essentially verbally mediated perception? The present study extends recent findings in our lab showing that categorical perception can occur even in the absence of overt labels. In particular, we evaluate the degree to which certain interference tasks (verbal and spatial) reduce the effect of learned categorical perception for complex visual stimuli (faces). Contrary to previous findings with color categories, our results show that a verbal interference task does not disrupt learned categorical perception effects for faces. A second experiment examines categorical perception effects when the faces vary along verbalizable dimensions. Our results are interpreted in light of the ongoing debate about the role of language in categorization. In particular, we suggest that at least a sub-set of categorical perception effects may be effectively “language-free”.10/20: Chris Donkin, Denis Cousineau and Rich Shiffrin
A Model Of Visual Search.
How we find objects in a visual scene relies on a number of (often competing) factors: 1) endogenous observer decisions to search in some preferred order, 2) automatic attraction of attention to learnt stimuli, 3) perceptual differences between targets and foils large enough to attract attention without practice, and 4) attraction of attention due to stimulus onsets. We present a model for visual search in which endogenous observer decisions control a serial-terminating process that is in competition with a separate, parallel, process governed by automatic attention attraction. We first show that the model can account for full reaction time distributions in the control conditions of a visual search study. We then present the results of experimental conditions in which display objects were presented sequentially, at speeds fast enough to appear simultaneous, but slow enough to allow us to model the effects of stimulus onsets. Finally, by fitting distributions collected by Jeremy Wolfe, we show that the model can account for the influence of changing stimuli on the strength of automatic attention.10/27: Jim Walker
Trust and Trust Worthiness: Extending the Decision Environment in Trust Games and Public Goods
Numerous experimental studies involving private property endowments have demonstrated that individuals’ decisions, in a variety of situations, reflect complex and diverse motivations beyond simple own-income maximization. Pertinent to the research presented here, extensive research has been generated showing that subjects in “Trust Game” experiments and “Public Goods” experiments achieve higher levels of efficiency than predicted by the stage game equilibrium for self-regarding preferences. The research summarized in this talk is designed to examine the robustness of previously reported studies across two dimensions: a) how endowments are assigned - private versus common property and b) in sequential one-shot public goods experiments in which the second mover has information about first movers decisions and/or an expanded strategy space that allows the second mover to take advantage of cooperative decisions by first movers.11/3: A. George Wilson and Jonathon D. Crystal
Prospection in the rat
Prospective memory in people is defined as holding an intention to act in the future. A hallmark of prospective memory is that a future intention has a deleterious effect on current ongoing activity. We tested the hypothesis that rats would show a similar deleterious effect of prospective memory. Rats (n=20) were trained in a standard temporal bisection task (90 min/day). Immediately after the bisection task, half of the rats received an 8-g meal (meal group) and other rats received no additional food (no-meal group). The meal was earned by breaking a food-trough photobeam. Sensitivity to time in the bisection task was reduced as the 90-min interval elapsed for the meal group but not for the no-meal group. This time-based prospective memory effect was not based on response competition or an attentional limit. Rats form an intention to act in the future, which produces a negative side effect on ongoing activity.11/10: Asuka Terai
A Computational Model of Metaphor Understanding based on Language Statistical Analysis
The purpose of this study is to construct a computational model of metaphor understanding based on statistical corpora analysis and that includes dynamic interaction among features. The constructed model consists of two processes: a categorization process and a dynamic-interaction process. The categorization process model, which is based on the class inclusion theory (Glucksberg 1990), represents how a target is assigned to an ad hoc category of which the vehicle is a prototypical member. The dynamic-interaction process model represents how the target assigned to the ad hoc category is influenced and how emergent features are emphasized by dynamic interactions among features. The dynamic interaction is realized based on a recurrent neural network. The constructed model is able to highlight the emphasized features of a metaphorical expression.
Finally, psychological experiments are conducted in order to verify the psychological validity of the constructed model of metaphor understanding with dynamic interactions. The results from the psychological experiments support the model incorporating dynamic interaction.11/17: Bill Timberlake, Indiana University
How General and Focal Search Modes in Foraging Systems Produce Pigeon Superstitions in the Laboratory and Stereotypies in Polar Bears and Walruses at Zoos.
To explain learned behavior and misbehavior in laboratories, General Learning Models, from cognitive to behavioral to neurophysiological, focus on repetition and computation of relations among stimuli, responses, rewards, and special neurophysiological events. However, evolution appears to select for more specific aspects of foraging behavior, including appetitive search modes (general and focal), and attentional biases, memory forms, and behaviors. I will clarify how evolved general and focal search modes and related perceptual-motor modules in foraging systems interact to account for presumed accidental learning effects, such as superstitious behavior in pigeons, and apparent agitation-driven stereotypes in polar bears and walruses. Whole organism behavior will be shown.11/24: Holiday
TBA
TBA12/1: Mark Mon-Williams
The art of predicting human behaviour and why we get stuck in a rut with decision making
Humans are expert decision makers capable of assimilating information rapidly and tailoring behaviour optimally, according to task constraints and context. The flexibility and rapidity of the decision-making mechanisms has led to humans being described as the ultimate opportunistic learning machines. Nevertheless, humans can sometimes select sub-optimal behaviours, with decision-making appearing to become ‘stuck in a rut’. How can we reconcile these apparently conflicting features of the system? A model of learning will be presented to show that the inertia sometimes evident in decision-making is a naturally emergent feature of a learning system.
Empirical tests of the model will be reviewed together with a series of experiments that allow us to predict various aspects of reaching behaviour.12/8: Linda Smith and Chen Yu, Indiana University
Grounding Toddler Learning in Sensory Motor Dynamics.
Most theories of cognitive development are "cognitive" in the sense of being about internal models, propositions, and inferences. It is not at all clear that these theories can explain real world learning.
Child learn in a physical world - about objects, actions, and other social beings, and language - through their second-by-second, minute-by-minute sensorimotor interactions in that world. They create their own experiences through their own actions. This talk considers how the body - and physical actions - may play a special role in - and indeed simplify - learning object names. The body's momentary actions appear to play a direct role in what might seem to be cognitive operations - attention and binding - bind objects in the physical environment to internal cognitive operations. The domain used to illustrate these points is toddler word learning. Using tiny videocameras placed low on the forehead of the child to capture the dynamic first person view, measure of eye-gaze direction, motion sensors on heads and hands, and success in word learning tasks, the experiments show learning that is inseparable from - and made in - embodied interaction in the world.
Previous Cognitive Lunch
|
