ABSTRACT
The majority of the chapters in this volume consider large-scale behavioral phenomena (e.g., emotional states) at long time intervals (e.g., weeks and months). In this chapter, we take something of a departure. Two differences are of particular importance. The first is with respect to the time scale of interest, and the second is with respect to what we might think of as the epistemological goals of a dynamic modeling effort. With respect to time, whereas the majority of the chapters in this volume are concerned with time at the level of hours, days, months, and so forth, we concern ourselves with milliseconds. With respect to epistemic goals, whereas the majority of the chapters in this volume have been focused on (for example) the use of dynamic conceptions to reveal underlying structures in data, we use methods of dynamic systems modeling to construct formal representations of hypotheses regarding rather specific mechanisms for the production of observable data. Although it turns out that the first of these differences holds little import (assuming, as we do, that measures of time exist on a ratio scale; see Townsend, 1992), the second of these differences deserves some discussion.As such, we begin by presenting a very brief description of the general approach to modeling that we adopt here, an approach that has as its goals the development of formal representations of the mechanisms and representations that support elementary perceptual 243
and cognitive acts (more extended discussion of these issues can be found, in O’Toole, Wenger, & Townsend, 2001). We then focus on precedents in this area for the use of formalisms that address dynamics. This sets the stage for a description of the specific modeling approach that we use here, and that we have used in other applications (see in particular Townsend $z Wenger, 2004; Wenger & Townsend, 2001). The description of this approach raises two critical questions. The first is methodological: Standard response measures in perceptual and cognitive experiments convey no information about the dynamics of internal processing. Consequently, can we develop a response methodology that possesses dynamics that are both interpretable (with resepct to theory), and consistent with standard measures of performance? The second question is statistical: Given a suitable response method, are there statistical methods that address the theoretical form as well as the empirical predictions of the theoretical models? These are questions that we address in a novel (to our knowledge) way in the present effort.
