James Townsend - US grants

A fruitful approach to the study of human behavior is the use of the relationships between the environment, the perceptual information flow, and the consequent behavior to infer psychological structure and function, the architecture of the mental system. From this approach, psychological functions such as perception, motivation, cognition, decision making, learning, memory, and problem solving are seen as embedded in abstract models of the internal psychological states. An important approach to the modeling of these psychological functions is to do so with closed-form mathematical models (which may, of course, need to be studied at times using computer simulations). At least as important as the individual models of individual processes is the mathematical exploration of the qualitative and quantitative behavior of mental systems which exhibit particular properties. Townsend's laboratory has been, with the aid of prior NSF support, a main source of such mathematical results. This project is designed to investigate the behavior of mental systems which have the following properties: (1) information flows only forward, with no feedback loops; (2) flow is continuous, with partial information being transmitted throughout the system on a continuous basis, rather than each subprocess being completed before it sends the results on down the line; and (3) the outputs of each process are stochastic, i.e. involve elements of randomness. In addition, results will be extended to systems that meet conditions (1) and (3), but not (2). Experimental work will be directed toward discovery of qualitative properties involving reaction time and accuracy that will make it possible to identify the system architecture. Other experiments will involve exploration of theoretical results obtained recently. Finally, a mathematical model will be constructed for a recent experimental paradigm that acquires continuous measures of response tendencies. This research, though basic, impacts in an indirect way on education and on the design of machines with which people must interact.

DESCRIPTION (provided by applicant): This proposal requests support to continue the training program in Modeling of Cognitive at Indiana University. The program, now finishing its tenth year, has provided support for two postdoctoral fellows, and two predoctoral trainees per year. The program has twelve core faculty, ten of whom are in the Psychology Department, and are internationally recognized for their expertise in the modeling of cognition; they have contributed some of the major models of cognition in the field, also many important methodological innovations and techniques. There are in addition 21 affiliated faculty within the Cognitive Science Program at Indiana who model cognition in their research, and provide auxiliary support for the program. Mathematical and computer simulation models of cognitive processes are increasing in importance year by year as the field begins to tackle the true complexity of cognition, and apply the results to problems of important national priority (in education, health, and problem solving in government and business). The unpredictability of outcomes from highly interactive systems makes it impossible to rely upon intuitive predictions, and makes modeling imperative. Our training program provides a wide range of experience in many techniques, including formal mathematical models, stochastic processes, non-linear dynamical systems, and neural net modeling, also provides examples of applications in many areas of cognition, including memory and learning, psycholinguistics, visual and auditory processing, speech production and perception, skill acquisition, categorization and conceptualization, decision making, problem solving, attention, and automatization. The four pre-doctoral trainees will be selected from applicants for graduate study to the Psychology Department, and take a five year program resulting in a joint Ph.D. in Psychology and Cognitive Science, and a Certificate in Modeling in Cognitive Science. Their program will emphasize laboratory research and modeling of the empirical results, with experience in the labs of at least two of the core faculty. The four post-doctoral trainees will visit for two years each, taking modeling courses and carrying out research involving modeling in the laboratories of one or more of the core faculty. Training of both groups will be given in a wide variety of mathematical and computer simulation modeling techniques, the application of models to data (including requisite techniques of data analysis), the testing and comparison of models, and the exploration of models and their implications. They will participate in and present their research in weekly seminars, attend colloquia, submit their research to professional journals, and present the research at meetings.

DESCRIPTION (Applicant's Abstract): The study of how shapes are perceived and otherwise cognitively employed is among the most fascinating, and challenging, of any in cognitive psychology today. There is a sizable and impressive literature dealing with general questions of configurality, although these researches are fairly disparate and a firm definition of the construct is currently lacking. The proposed research is composed of a set of major, interlocked aims: 1. Apply theoretical, empirical, and methodological tools developed by the applicant and colleagues explicitly to the problem of configurality. 2. Synthesize two lines of the applicant's prior theoretical developments and construct new theory and methodology for application to configurality. 3. Generate tentative definitions based on historical and recent hypotheses concerning configurality within the confines of prior efforts and the synthesis described in (3). 4. Run a set of experiments using these nascent and provisional definitions and taxonomies for configurality. 5. Extend the theoretical work to explicitly consider the underlying dynamics and increasingly more sophisticated pattern spaces.

[unreadable] DESCRIPTION (provided by applicant): This project is a continuation of work, begun in 1998, and renewed in 2003, that investigates the general properties of phenomena that have been labeled as reflecting "configural," "holistic," and "gestalt" representations and processing, with a particular emphasis on issues in facial perception and cognition. The majority of published research on these types of effects has worked from behavioral patterns to theoretical implications. This project represents a significant departure from that approach. Specifically, it involves the establishment of a set of theoretically-grounded definitions for configurality, holism, and the notion of a gestalt, and has worked from those definitions to the behavioral implications. The present proposal builds on earlier work in three important ways: (a) It provides a comprehensive analysis of the behavioral regularities that have been taken as strong indicators of configural/holistic representation and processing of human faces, (b) It extends the effort with faces to consider an analysis of a set of complex "naturalistic" configural phenomena, (c) It will analyze the changes in encoded representations, and the manner in which they are processed, as observers learn to work with complex configural patterns. The project as a whole will pursue refinements to both theory and methodology, in order to go beyond our current definitions of configurality, and to extend our ability to connect theory and data. Work toward these goals will involve theoretical, methodological, and empirical efforts. [unreadable] [unreadable] [unreadable]

Instant by instant, we take in all of the information that we can gather with our senses, and we organize that information into objects and events. This is the foundation for our ability to function in the real world. While we can talk in general terms about how and why we function with such effectiveness and efficiency, a satisfactory scientific explanation requires much more. In particular, it requires a way of relating the things we can measure---specifically, the frequency with which we choose certain to choose certain actions and the time it takes to make those choices---to each other and to our hypotheses for how and why we do these things. Currently, there is no general way of doing this: the work to be accomplished with this support will fundamentally change this state of affairs.

The creation of a unified theoretical approach to characterizing the ways in which we organize our perception of the world will open up numerous avenues of research linking behavior and neurobiology. With the aid of the theoretical language that this project will produce, there will be systematic and well-defined ways of testing competing hypotheses for both what we do with perceptual information and how we do it. Critically, because this language will be extremely general and will not be tied to a particular set of ideas or theory, scientists from competing perspectives will be able to frame their ideas using a common vocabulary, something that is not currently possible. In addition, since this language will be mathematical, it will possess an exceptional level of rigor and internal consistency. Finally, this language will demonstrate its utility and power by being applied to a set of thorny issues in the contemporary study of perceptual organization.