Functional relations are not models: A note on covariation detection.
Journal of Experimental Psychology: Human Perception and Performance, 12, 110-111.
Surber's (1986) arguments about testing covariation detection models and about assessment of variable importance are essentially correct but inappropriately applied to analyses of functional relations. Our main points were as follows: (a) Past theorizing in several domains assumed that all rs that are equivalent in Pearson's terms are perceived as equivalent; (b) our data invalidate this assumption; (c) functionally, error variance may be the most important determinant of covariation judgment; (d) several models may account for these functional relations, but we have not tested these models; (e) researchers comparing human judgments with normative models such as Pearson's need to be aware of the obtained functional relations to avoid misinterpretation of deviations from normativeness.
©1986 by the American Psychological Association.
For a pdf version of the article, click here.