Humor is a quintessentially social phenomenon, since every joke requires botha teller and an audience. Here we ask how humor operates in task-oriented group discussions. We usetheories about thefunctions ofhumor togenerate hypotheses aboutwhojokes, when and in what situations. Then we useeventhistory techniques to analyze humor attemptsand successes in six-person groups. Our results combine to suggest an image of jokingasa status-related activity, with men, highparticipators, frequentinterrupters, and those who arefrequently interrupted allshowing status-related patterns of humor use. Wefind substantial time dependence in humor use, in whichhumor may serve to form a status hierarchy earlyin a group's developmentand to dissipate task-related tension later in the discussion. We usethese results, in conjunction with core insights on status and emotion from thegroup processes literature, to develop a newtheory of humor usein task-oriented groups. The new theory generates predictions about the contentof humor episodes, which we examine with additionaldatafrom ourgroup discussions. Consistent with the theory, we find that a higher proportion of men's humor is differentiating, whilea higher proportion ofwomen's humor is cohesion-building. We find the samegeneral pattern with our otherstatusvariable, participation.allof the fun out of the subject. Accepting that risk, we follow up on a long-dormant thread in the literature on small group interaction that discusses when and how humor is important in task-group discussions (Bales & Slater 1955). We begin by reviewing the wide-ranging literature that has developed on humor. Then we discuss the implications of this literature for group interaction, deriving predictions about the actual distribution of humor use among group participants and the timing of joke-telling during a task-oriented conversation between strangers. We test these predictions by examining the actual distribution and timing of humorous episodes during conversations in 29 task-oriented groups. We use our findings to develop a new theory about the relationship between status, affect and humor. Finally, we take a deeper look at our data to look for patterns that support or contradict our new theoretical predictions.