This paper reports an experiment that extends previous findings of the disj
unction effect, sometimes described as a violation of Savage's sure-thing p
rinciple. Evidence of the disjunction effect is observed using elicited bel
iefs about others' actions (rather than controlled beliefs) in a prisoners'
dilemma studied by Shafir and Tversky (1992) as well as in an asymmetric v
ersion of it and in a nondilemma game with a unique equilibrium. Debiasing
techniques as well as implications for these extensions are discussed. (C)
1999 Academic Press.