and the subjective description will be addressed as well. In a number of cases, forensic researchers also provide a certainty score to their judgement. Another aim of the project is therefore to examine the interpretation and quantification of uncertainty by forensic researchers.
working closely with scholars from both the Brunswikian and Gibsonian ecological traditions for ten years, and my attempt to view these traditions, among others, as complementary rather than as conflicting. While this work was motivated by a desire to create a coherent ecological psychology, I conclude by discussing the methodological implications of this work for the broader domain of psychological science, broadly conceived."
Toward a Coherent Ecological Psychology
Alex Kirlik, University of Illinois at Urbana-Champaign E-mail: email@example.com
Configural judgement strategies and MCPL
I spent the 2001-2002 academic year visiting Haskins Laboratories, New Haven, CT, the PACE Center (Psychology of Abilities, Competencies and Expertise) at Yale University, and the Department of Psychology at the University of Connecticut (gracefully received by Jim Holzworth and the rest of the I/O Program, and also by the Center for the Ecological Study of Perception and Action). After my initial interactions with scholars such as Holzworth, Turvey, Sternberg, Mace, Michaels, Shaw, etc., (and especially their students), I couldn't help but to believe all these theorists agreed on fundamental issues, but that historical accidents in the development of ecological theory had caused fractures, perhaps self-serving in the short run, but truly destructive in the longer term. I thus assigned myself the task of doing the detective work of trying to understand the reasons underlying the divide between the two most prominent schools of ecological psychology (Brunswikian, Gibsonian), and how I might present my findings in a way that was perceived as constructive by members of both intellectual communities. My year culminated with a presentation at Haskins Laboratories, in a talk with the following abstract (the full Powerpoint slides can
http://homepage.mac.com/alexkirlik/FileSharing1.html). "The selection of basic research questions is driven by judgments of what it is that makes research fundamental. Whatever else (elegance, etc.) might be meant by this idea, generalizability is a key, and many would say, defining, element. Fundamental research findings should generalize beyond the context of discovery, either to a specified domain, or in the ideal, apply universally. I describe how, unlike
different logics for generalizing findings beyond the context of discovery. A failure to recognize the legitimacy of all three engines of generalization, and to distinguish between them when required, lends confusion to both theoretical and methodological discussions. I illustrate this thesis by discussing, and attempting to resolve, the tremendous amount of confusion that has grown up around the concept of "ecological validity." This work is a product of
Pieter Koele Universiteit van Amsterdam Email: ml firstname.lastname@example.org
Mick de Niet (an undergraduate student of mine) has just completed an MCPL experiment in which the judgement task consisted of predicting the severity of Brunswik's Disease of fictitious patients on the basis of two labelled cues, reflecting the concentration of certain substances in the blood of the patients. OF was provided after each trial, and subjects had to judge 90 patients. In condition I (n = 46) the cues (X1 and X2) and the criterion Ye
were related thus:
Ye = 0.125 X1 + 9.2 Ye = X1
if X2 < Mean X2 if X2 > = Mean X2
In condition II (also n = 46) the relations were:
Ye = - X1 + 21 Ye = X1
if X2 < Mean X2 if X2 > = Mean X2
X1 and X2 values were randomly drawn from a uniform [1; 20] distribution. So, in both conditions the criterion was perfectly predictable, but not by consistently using a linear additive combination of the two cues. In condition I the best fitting linear combination of X1 and X2 yields a multiple correlation of 0.82, in condition II it is only 0.12.
Analyses of the experimental results (based on dividing the 90 trials into three blocks of 30 trials each, and calculating the lens model measures in each block) showed that in condition I many subjects managed to find a strategy similar to the best fitting linear additive strategy (mean achievement in block 3 is 0.74, mean Rs = 0.88, mean G = 0.99). Configural cue use did not reach a substantial level (mean C = .24 in block 3). In condition II subject overall achievement was low in block 3 (mean achievement = 0.36, mean Rs = 0.52, mean G = 0.05), but substantially better than that of the best fitting linear model. And indeed, this was the result of configural cue use: mean C = 0.42 in block 3. We think this shows that subjects are able to detect to a certain extent configurality in
Newsletter 2002 page 16 of 28