The article by Dunwoody, Haarbauer, Mahan, Marion, and Tang (this issue) represents an important test of Hammond's (1996) Cognitive Continuum Theory (CCT). Cognitive Continuum Theory has been around for a while (e.g. see discussions in Cooksey, 1996, andDoherty andKurz, 1996) but has receive little attention in terms of rigorous research evaluation. Hammond et al.'s (1987) ground-breaking study of highway engineers provided the ®rst real test of the theory. Hamm (1988) followed up this study with his investigation, based on the same data as Hammond et al., of moment-by-moment variations in cognitive activity of highway engineers. Since then, tests of the theoretical predictions of CCT have been few and far between as Dunwoody et al. indicate. The work by Dunwoody et al. is important for two reasons: (1) it attempts to concretely test speci®c predictions arising from CCT, and (2) it showcases the methodological issues one confronts when trying to test a dynamic decision theory. The predictions arising from CCT tested by Dunwoody et al. focus on the interconnectedness between decision maker and decision task Ð a focus that Brunswik tried to instill in psychology over 40 years ago (see, for example, Brunswik, 1952Brunswik, , 1956)). Emerging out of this focus were predictions such as: mode of cognition being induced by task characteristics; judgment achievement or accuracy being linked to congruence or match between mode of cognition and task demands; and variations in the distribution of errors made using each mode of cognition (intuition tending to produce normally distributed errors; analysis tending to produce leptokurtic error distributions). Dunwoody et al. have provided some experimental evidence to support, at least partially, each prediction. For example, their study produced results consistent with the expectation of more leptokurtic error distributions being associated with analytic cognition; a ®nding that has been substantiated in the past (e.g. Peters, Hammond and Summers, 1974). Dunwoody et al. also provided partial support for the predicted relationship between task characteristics and mode of cognition, but primarily at the level of surface task characteristics rather than depth characteristics. However, all of Dunwoody et al.'s evidence was obtained under circumstances that, while probably necessary in order to obtain the degree of contextual control needed to produce the data, simultaneously circumscribed those very data. This underscores the diculties associated with making speci®c tests of predictions from a dynamic theory grounded in Brunswikian principles.
CCT embodies two key principles arising from Brunswikian psychology: representative design and the principle of parallel concepts. Representative design focuses on the necessity for understanding human behavior under conditions that are representative of the types of tasks, ecologies, and behaviors that the person would normally encounter and exhibit in daily life. Representative design can be accomplished at two dierent levels Ð formal task characteristics (statistical properties of a task and its features) and substantive task characteristics (contentrelated properties of a task and its features Ð Cooksey, 1996). A complete implementation of the principle of representative design would require simultaneous and appropriate attention to both formal and substantive characteristics. Typically, though, either formal or substantive task characteristics become the prime focus of much judgment research. I should note here that the use of task-surface and task-depth characteristics by Dunwoody et al. (as well as by Hammond et al., 1987) does not directly correspond to the distinction between formal and substantive task characteristics; surface' and depth' each implicitly encompass some of the aspects of CCC 0894±3257/2000/010055±05$17.50