In considering expressive interaction in two distinct cultural settings, a sheltered workshop in Israel and a section of a lead and zinc mine in Zambia, joking activity was found to be governed by sets of rules which we term joking frames. Two types of joking frames, setting-specific and category-routinized, are distinguished. Setting-specific joking depends primarily on resources derived locally within the setting in which this activity occurs. Such frames are highly fragile and setting-specific joking follows an indeterminate course. Category-routinized joking frames are anchored in more general social conventions, and are more resistant to subversion and to re-transposition to overtly serious activity. Of central concern to our analysis is the isolation of conditions which lead to the establishment, maintenance, and destruction of joking frames during interaction. We indicate that joking activity, and indeed any form of expressive activity, must be understood in terms of the emergent, self-generative, form of the activity itself. So our analysis indicates that while setting-specific and category-routinized frames are established and destroyed by similar mechanisms, the rules included in the latter have the greater time-depth because these rules are better able to integrate subversive elements which emerge from the course of joking activity, and the original definitions of category-routinized frames can be maintained for lengthier periods of time.