Quality & Quantity 36: 219–238, 2002.
© 2002 Kluwer Academic Publishers. Printed in the Netherlands.
Take The Best, Dawes’ Rule, and Compensatory
Decision Strategies: A Regression-based
Department of Psychology, University of Bonn, Römerstraße 164, D-53117 Bonn, Germany
Abstract. Strategy descriptions like the “Take The Best”-heuristic (G. Gigerenzer et al., 1991), the
weighted additive rule, and the equal weight decision rule are competing theories about information
integration in multi-attribute decision tasks. Behavioral decision research is confronted with the prob-
lem of drawing conclusions about unobservable decision strategies from behavioral data. Although
there has been considerable progress due to methodical traditions like ‘Structural Modeling’ and
‘Process Tracing’, these paradigms have certain limitations in testing speciﬁc hypotheses about indi-
vidual strategies. Some of these problems are summarized brieﬂy. A deductive method for classifying
individual response patterns is introduced. Predictions about regression coefﬁcients are deduced from
competing substantial hypotheses about strategies for decision making. These can be tested at the
level of individual participants. The validity of this classiﬁcation procedure is demonstrated with a
Monte Carlo simulation. Some useful applications of the method are described, limitations of the
method and potential generalizations are discussed.
Key words: decision making, take the best, probabilistic inference, hypothesis testing, strategy
1. Take The Best, Dawes’ Rule, and Compensatory Decision Strategies:
Problems of Behavioral Decision Research
Like any other ﬁeld in cognitive psychology, the special branch called “Behavioral
Decision Research” (BDR; e.g., Maule & Svenson, 1993; Payne et al., 1992) is
concerned with formulating theories about cognitive processes. In the case of BDR,
process models are invented to describe human thinking in various situations in
which judgments or decisions have to be made. Sometimes, the BDR branch is
characterized as descriptive as opposed to normative models of decision making.
The aim is not merely to describe actual data, but to describe cognitive processes
which are theoretical terms within cognitive theories. Of course, these theoretical
claims have to be confronted with actual data in order to test their validity and
predictive power. Linking unobservable theoretical constructs to potentially ob-
servable data patterns is the central problem of psychology in general, and it can
be termed the ‘measurement problem’.