Researchers often attribute individual characteristics to groups or other collectives. For example, groups are often described as more or less satisfied, efficient, or committed. Of course, there are other characteristics of groups that do not translate as well to the individual level. A group can have a positive climate or high cohesiveness, but it makes little sense to say that about individual group members. Other constructs really don't have group-level correlates; although individual group members may be intelligent, for example, it is nonsensical to describe a group in that way. Alternatively, one way to think of satisfaction and related constructs is that they potentially apply to groups as well as individuals within groups (dual-level constructs). Cohesion, climate, and intelligence are level-specific constructs, as they commonly apply to either groups or individuals but not both.
But here's the catch: Researchers generally collect data from individuals within groups and use those data to make claims about groups (for either dual-level or level-specific constructs). For example, a set of 10 items might be used to measure satisfaction. How is it possible to use satisfaction responses from individuals within groups to claim that groups are more or less satisfied? Although there are several methods for aggregating individual scores (e.g., group average), there are better and more elegant alternatives, including treating the group as a latent variable — a construct that is unobserved but measured by the use of observable indicators. Many variables in the social sciences are latent; the challenge is how to measure them with imperfect mechanisms. So, at the individual level, a person's score on satisfaction is influenced by the unobserved satisfaction factor. At the group level, each member of the group (or, more precisely, each member's score on satisfaction) is an indicator of the group-level construct. The more similar scores within groups, the better the evidence for the group-level construct.
It's possible to test if a dual-level or level-specific construct is supported by the data. Ideally, a level-specific construct would find that most of the variance in responses is accounted for by the desired level (individual or group). For example, cohesion scores within groups should be highly correlated, which means the construct operates mostly at that level. Dual-level constructs, on the other hand, should show that responses are somewhat associated within groups but not so much that there is little variance at the individual level. Another question might be whether the group and individual latent factors for satisfaction have roughly the same effect across group members' individual responses.
An article by Joseph A. Bonito and Joann Keyton in Group Dynamics: Theory, Research, and Practiceopens in new window describes how to do just that. If the test shows isomorphism, or similarity, across levels, then one might argue that the group level merely reflects processes at the individual level. But if the effect of the latent factors at the two levels is different, then the process that operates on individuals is likely different from the one that operates on groups. For its explanation of this principle, this article won the 2019 Group Dynamics Most Valuable Paper Award.
For researchers (both basic and applied) and practitioners, interventions designed to improve or otherwise influence group processes and outcomes should be designed for the appropriate level. For example, if satisfaction displays group-level characteristics, then it makes sense to design interventions to improve satisfaction for groups rather than individuals.
Note: This article is in the Social Psychology and Social Processes topic area. View more articles in the Social Psychology and Social Processes topic area.