The conjunction fallacy in probability judgment is described in the literature by the representativeness heuristic; its stability is emphasized. Two determinants are deduced from a general problem-solving model (existence and availability of the conjunction rule), which should reduce the inadequate use of the heuristic. Additionally, a distinction is made between errors in probability judgment which can be explained by the heuristic and those which can not. 32 subjects (16 female, 16 male), between 22 and 28 years old, participated under four experimental conditions (with eight subjects each). They solved two problems (one with high and one with low representativeness) concerning the conjunction rule. The experimental variables (existence and availability) substantially reduce - as postulated - the erroneous decisions. Only 55% of the remaining errors can be described by the representativeness heuristic. (orig.)