Support for the cue-heuristic model is based on suboptimal observer performance: Response to Gilden and Proffitt (1994)

Gilden and Proffitt (1994) have derived support for their updated cue-heuristic model of mass discrimination in observed collisions from two experiments reported by Runeson and Vedeler (1993). However, these experiments are inadequate for theory testing because of confounding and lack of representative variation in parameters. Instead, extended analyses of the third of Runeson and Vedeler’s experiments are presented. It is shown that observers exclusively use collision-axis velocity components in conditions where the heuristic model asserts that only trajectory speeds are available. Moreover, the patterns of error predicted by the heuristic model do not occur. Overall, performance is closer to the ideal than to the predictions from any conceivable model based on elementaristic cues; hence an explanation in terms of advanced informative invariants is the more viable approach. The final discussion concerns the distortion of theory development that might be engendered by empirical data that represent suboptimal observer performance due to unfavorable testing conditions.