“Levels of analysis” has been a controversial topic in group research and has often been ignored by IS researchers leading to “cross-level fallacies” and “aggregation biases.” While the predominant practice in IS research is to assess group behavior by computing the “arithmetic mean of individual members’ behaviors,” recent studies have argued that groups should be treated separate from the individuals who constitute them, and group behavior should be measured using “global” constructs. This study examines the applicability of these two divergent practices in the context of technology adoption by groups, a phenomenon of growing importance in organizations that has not been investigated in-depth in IS research. Preliminary results from an experimental study provide some resolution to the debate surrounding the appropriateness of these two approaches, and highlight the contexts in which aggregation of individual-level measures may not be suitable for understanding group behavior.