Tuesday, 6 March 2018

Cognitive seminar by Prof Jim Townsend, Indiana University

Many of us know about Signal Detection Theory (SDT) and perhaps even use it in our research. It was historically important in separating the bias in human response from perceptual factors such as signal to noise ratio. There is much more we can do within this framework nowadays, and Prof Jim Townsend from Indiana University is one of the leading figures worldwide in the forefront of these advances. He will present recent advances in General Recognition Theory, which can be very crudely viewed as an extension of SDT to complex decisions that involve multiple dimensions. .

The Cognitive Research Group is proud to host a talk by Distinguished Professor James Townsend from the Department of Psychological and Brain Sciences, Indiana University:

WHEN: Thursday 8 March, 12-1pm.

WHERE: Keats reading room, AVLG17. VC link to Ourimbah available on request.

TITLE: Response Time General Recognition Theory (RTGRT):  The Parallel Class of Systems.

ABSTRACT: GRT (Ashby & Townsend, Psych.Rev., 1986) is, like classical signal detection theory, static in the sense that there is no stochastic process defined on the perceptual detection process itself.  However, it still comprises the major theory-driven methodology for identification of multi-dimensional perception and classification in the field with hundreds of cited applications.  Nonetheless, its static quality is theoretically limiting because:

1. It cannot encompass exceedingly important observables such as probability correct conditional on response times (RTs).
2. As we have repeatedly argued and proven mathematically [e.g., Eidels, A., Townsend, J. T., Hughes, H. C., & Perry, L. A. (2015). Evaluating Perceptual Integration: Uniting Response Time and Accuracy Based Methodologies. Attention, Perception, & Psychophysics, 77, 659-680.], accuracy rather than RTs is optimal for assessing independence of various types but RTs are optimal for identification of mental architectures such as parallel vs. serial processing.

Thus, by extending GRT to a stochastic environment, we simultaneously

1. Probe varieties of independence in terms of the RT dynamics as well as the overall probabilities of response patterns.
2. Set the stage for a grand unification of GRT and SFT, thereby permitting the assessment of independencies and invariances at the same time as identification of architecture and stopping rule.
3. Unify GRT also with A(t), the generalization of the capacity function, C(t) to data containing errors.

If time permits, the following readings will be beneficial preparation for Jim’s talk: 

Ashby, F. G., & Townsend, J. T. (1986). Varieties of perceptual independence. Psychological Review, 93, 154-179.
Townsend, J. T., Houpt, J., & Silbert, (2012). General recognition theory extended to include response times: Predictions for a class of parallel systems. Journal of Mathematical Psychology, 56, 476-494. 

We look forward to seeing you there!