A robot player sent intermittent threat messages of 10 percent, 50 percent, or 90 percent credibility to 180 subjects, 90 of each sex, during the course of a mixed motive conflict game. The magnitude of the threatened punishment was either low or high. The simulated player was either 0 percent, 50 percent, or 100 percent accommodative in his use of coercive power on the threat-related trials. The results indicated a strong effect of source accommodation on a behavioral measure of target compliance. Supplementary analyses of the compliance data according to an expected value model of social power indicated that males complied as a direct function of the difference between the expected values for compliance and noncompliance, while females did not. Furthermore, a threat credibility by sex interaction was found on the proportion of cooperative responses made by targets on nonthreat occasions. The combined compliance and cooperation results indicated that targets' responses to threats are highly dependent upon both the words and the deeds of the threatener.
cc3
LTHOUGH Deutsch and Ilrauss (1962) A have reported evidence supporting the commonsense assumption that the possession and use of threats function to exacerbate interpersonal conflicts, other investigators (Kclley, 1965; Shomer, Davis & Kelley, 1966) have examined the possibility that threats may be employed as signals to coordinate cooperative gestures and ameliorate conflicts. Nardin (1968), pursuing the latter line of reasoning, has suggested that a target's attributions of hostile or accommodative intent to a threatener are highly dependent upon the threat-related behaviors of the source. That is, the context and thc content of threatening communications are presumed to affect the target's impressions of the source and consequent conflict exacerbation or amelioration. Using an extended form of the prisoner's dilemma game as the research tool, Nardin found that the mere possession of threat capability did not function to exacerbate dyadic conflict as compared to a control condition. I n fact, subjects who were given threat capability and