This study examined naturally occurring idea-generation in organizational groups completing an extended problem-solving task. Meetings held by 11 groups engaged in a quality improvement process in a governmental agency were analyzed to identify 37 idea-generation episodes. All groups had available a group decision support system (GDSS), although some opted not to use it. Across all groups, idea productivity was low though fairly efficient, but GDSS-supported idea-generation produced significantly fewer ideas. In general, idea-generation as a process appears to be more complex than has been thought; observed deviations from idealized idea-generation norms were not necessarily dysfunctional. The findings encourage reassessment of the assumptions underlying existing normative models of idea-generation.