Participants and Design. Forty-eight students from the University
of Göttingen, Germany, participated in this experiment. Half
of them received diagnostic instructions (common-cause model), and
half received predictive instructions (common-effect model). As a
second factor, the number of predictive cues was manipulated (a
single predictive cue vs. two predictive, mutually redundant cues).
Procedure and Materials. The participants received initial
written instructions (in German) in which they learned that a new
allergic disease, Midosis, had been discovered in animals. In the diagnostic
condition, it was mentioned that this disease might produce new types of substances in the blood, whereas, in the predictive
context, participants were told that new types of substances in the
blood that come from food items may be the cause of this disease.
During the learning phase, individual cases were presented on a
computer screen. Three substances, alpha, beta, or gamma, were
listed one above the other and described as either being present
( yes) or absent (no). In the two-cue overshadowing conditions with
two redundant predictive cues, these two substances were perfectly
correlated with each other and the outcome, whereas the third substance
was uncorrelated. This cue was always absent. There were
two types of cases: If all substances were labeled as being absent,
the disease was also absent; when the two predictive substances
both were simultaneously present and the uncorrelated substance
absent, the presence of Midosis was indicated. In the one-cue control
conditions, only one predictive cue was present when Midosis was present, and was absent when the disease was absent. The other
two cues were always absent. The assignment of the three substances
to the three cue types was counterbalanced in both the one- cue and two-cue conditions.