European Journal of Personality 2012 - in preparation Scale Issues in Causality David M. Condon, Ashley Brown-Riddell, Joshua Wilt, and William Revelle Northwestern University Abstract Elaboration of the manner by which graphical frameworks of causality can benefit personality research is a much-needed contribution. We argue that attempts to apply these frameworks in personality will benefit from consideration of two concepts related to scale. The first is that the appropriate scale on which to evaluate causality depends upon the level of analyses on which the research is conducted. Second, the distal scale between typical expressions of personality and their possible causes limits discussion of causality to probabilistic rather than mechanistic factors. By virtue of even attempting to integrate Judea Pearl’s innovative work on causality (Pearl, 2009) into the personality psychology literature, James Lee is to be commended for initiating a conversation that is - in truth - uncomfortable for a field of “correlators” (Cronbach, 1957). But it is Lee’s concise elaboration of so many nuances of the graphical framework that make the target article an invaluable contribution to the field, particularly for those unfamiliar with Pearl’s work. To the extent that personality psychologists increasingly focus on the development and evaluation of predictive causal models, we consider it likely that the influence of this work by Lee will grow over time. Though many aspects of Lee’s review merit further exploration, our commentary primarily focuses on two aspects of the relationship between the causal framework that Lee describes and considerations of scale. The first point of note pertains to Lee’s comments on the role of psychometric factors in graphical conceptualizations of psychological causality. We agree with Lee that one of the goals of personality re- search should be to recursively “expand a directed edge in one graph into an entirely new subgraph” (p. 47) in order to understand the mechanistic relationship between two nodes, an endeavor that would ideally result in richly detailed causal diagrams similar to those found in biology texts. Achieving this level of mechanis- tic detail may well resolve debates about the causal status for many psychometric factors but it would not mitigate the functional utility that factors provide. By analogy, a graph is something like an online geographic map. Psychometric factors, like many cartographic features, are not physically observable; researchers would no more benefit from “discarding the convenient fictions of folk psychology” (p. 48) than travelers would if the town, state, and nation labels were stripped from maps. The familiar experience of “zooming” an online map to the appropriate viewing level illustrates the contextual utility of complexity, which is itself a function of scale. Practically infinite detail is possible in mapping of both human behavior and geography, but this would rarely be functional. contact: William Revelle@northwestern.edu Draft version of February 17, 2012 Please do not cite without permission