Welcome to the lab
What is Psychophysics?
Psychophysics is concerned with describing how an organism uses its sensory systems to detect events in its environment. This description is functional, because the processes of the sensory systems are of interest, rather than their structure (physiology). One psychophysical theory, the Theory of Signal Detectability (TSD), uses a combination of statistical decision theory and the concept of the ideal observer to model an observer's sensitivity to events in its environment. TSD is stimulus-oriented, because properties of the stimuli are used to determine the theoretically best, or ideal, observer for a given detection task. This observer may then be used to compare the performance of human, and other, observers. For instance, the ability of humans to detect simple acoustic waveforms can be modeled as a linear system consisting of a filter, rectifier, integrator, and sampler.
Our research interests
Our main research interest is in psychophysical methodology, including measures of detectability, multi-event and multi-dimensional tasks, reducing the effects of observer inconsistency, evaluating correct and incorrect decisions (the Type 2 task), and the development of algorithms for these new methods.
Green's Relationship, AKA ASIFC=P(C)2IFC
The "PAPC" project began in 1987 when John W asked Sue to derive the discrete version of Green's proof that the area under the ROC curve in the SIFC task was equal to the proportion correct in the 2IFC task. What started as an afternoon's mathematics, ended up being a research project lasting over a decade. In 1991 John W thought that it would be a good idea to write up Sue's derivation as a letter to the editor. After some discussion we decided that evaluating the relationship experimentally would also be interesting and so we ran experiments where the underlying distributions were discrete. For fun we then ran more experiments where the underlying distributions were continuous. Meanwhile the theoretical development was undergoing major development, including Brian's derivations of new measures of detectability and spin-offs into multi-event multi-dimensional detection tasks. Aspects of the project have appeared in one form or another in most of our recent papers, theses, and presentations. A paper on this project was published in the Journal of Mathematical Psychology in 2002.
Group Operating Characteristic (GOC) Analysis
GOC analysis is a technique used to remove the effects of observer inconsistency from psychophysical data. By using GOC analysis to remove this source of error, the underlying ability of the observer to detect an event is revealed. GOC analysis has shown that in some cases human observers may perform as well as an ideal observer.
A starting point to understanding and using GOC analysis is available in this poster presentation given at the 2001 Acoustical Society conference. The definitive work on GOC analysis to date is Vit's PhD thesis, which is available here for downloading. In his thesis, Vit presents a theory of GOC analysis. He also details the history of GOC analysis, which started with Charles Watson's PhD thesis in 1962.
Type 2 decision making
A Type 1 decision is a decision an observer makes about the occurrence of an event in its environment, for example, whether a signal was presented in a background of noise. A Type 2 decision is a decision an observer makes about the correctness of their Type 1 decision. Sue Galvin has developed a mathematical theory of Type 2 decision-making that was published in 2003.
Multiple-event multi-dimensional psychophysics
Most psychophysical methods are limited to two events (e.g.,
signal-plus-noise and noise-alone) and one dimension (e.g., frequency). Many
detection tasks, however, are multiple-event and/or multi-dimensional. Brian has
developed a framework to analyze these tasks using TSD and information theory.
His measure of detectability, Dn (where n is the number of events) is
applicable to multiple events, multi-dimensions, is nonparametric, independent
of the prior probabilities and payoffs associated with the events, and is
independent of the ordering of the events. It has a true zero (i.e., it is zero
when the events are indistinguishable), and its upper value depends on the
number of events. It is infinite only when there are an infinite number of
perfectly discriminable events. For the two-event unidimensional case, Dn,
is simply a transform of the Area under the ROC curve:
To find out more about multiple-event and/or multi-dimensional psychophysics see Brian's abstracts.
The Psychophysics Lab produced 6 PhD theses and 3 Masters theses.
Psychophysical Poetry and Humor (Really!)
Taylor and the racks (c. late 1970s)
"Bird number 7" in action in the operant chamber
Alan Taylor with "bird number 7"
chamber used for psychoacoustic experiments with pigeons
Last updated 27 Feb 2010 12:43 PM
This psyber lab
webpage is maintained by Judi Lapsley Miller