The authors hypothesized that (1) attraction toward a stranger based on attitudinal similarity is automatic, but cognitive evaluation of the stranger's quality before the measurement of attraction can make attraction nonautomatic or controlled; (2) personal evaluations from the stranger activate automatic attraction and cognitive evaluation; (3) controlled attraction from attitudes and automatic attraction and cognitive evaluation from personal evaluations engender reverse-causal effects (i.e. they mediate each other); and (4) attraction and cognitive evaluation are distinct constructs. Attitudinal similarity between the participant and the stranger or personal evaluations of the former by the latter were varied in Experiment 1 (N=96), and were crossed with each other in Experiment 2 (N=240). Orders of response measurement were either cognitive evaluation followed by attraction or attraction followed by cognitive evaluation. Results confirmed the hypotheses. Implications of the findings are discussed.