Completing tasks in teams of two produces the most satisfaction, but members are most comfortable with 4 or 5 people, found Hackman & Vidmar (1970).
Experimental laboratory. The experiment was run simultaneously at two institutions (Yale University and the University of Illinois, Urbana) using male undergraduates at each institution as subjects.Hackman and Vidmar (1970), p. 40
What is the optimal team size? Research by Slater (1958) suggests one straightforward means of estimating the “optimal” group size — simply determine what size group members prefer. Although both Hare (1952) and Slater find a general increase in dissatisfaction as size increases (a finding replicated by this study), Slater further proposes that five-person groups may be optimal. The reason, Slater suggests, is that smaller groups are too intimate and members may be inhibited from expressing disagreements in them.Hackman and Vidmar (1970), pp. 47-48
-  It should be noted, however, that studies by Ziller (1957) and Miller (reported by Thomas & Fink, 1963) found no consistent relationship between size and member satisfaction.
The present data allow reexamination of Slater’s conclusion. Items one and two of the Member Reaction Questionnaire (“the group is too small”; “the group is too large”) reflect two opposing types of general dissatisfaction with the group size. If item scores are standardized (assuming interval data) and the data is plotted on a graph with size on the abcissa and satisfaction on the ordinate, the intersection of the two items will indicate the point of optimal reported satisfaction with the size of the group. Since neither task type nor laboratory interacted with size, the data were averaged across these variables before converting to standard scores. The results are shown in Figure 1. Consistent with Slater’s finding, optimal satisfaction with size is found between four and five members. Slater’s groups were from a single population and worked only on “human relations” tasks; the conclusion may now be generalized over three types of intellective tasks and two laboratory populations.Hackman and Vidmar (1970), pp. 48-49
Yet the finding that members are most comfortable with groups of size four or five is, in some ways, inconsistent with other data from the same people. According to items on the Member Reaction Questionnaire, members of dyads were clearly the most satisfied, and dissatisfaction increased, in approximately linear fashion, from size three to size seven. Further, the middle-sized groups (i.e., sizes three through five) were less creative than the dyads or the relatively large groups. Why, then, should members of these groups report that they are the most satisfied with the size of their groups?Hackman and Vidmar (1970), pp. 49
It may be that members of smaller groups feel unusually “exposed” and, while in fact they experience few objective difficulties in working together, they are still vaguely uncomfortable. The present data do not, unfortunately, address this possibility. The data are clear for larger groups. In these groups, members are unhappy, and the reasons center around the coordination difficulties they encounter (see Table 2). Yet even if the above line of reasoning is valid, we must explain why performance tended to be less adequate in middle-sized groups than in dyads and in very large groups.
[Editorial note: In Table 2, Group Product Characteristics are: (i) Action orientation; (ii) Length; (iii) Originality; (iv) Optimism; (v) Quality of presentation; (vi) Issue involvement; and (vii) Creativity.]
One possibility is that middle-sized groups are, in a sense, too comfortable for their own good. It appears that members of dyads may have responded to their “exposure” by pouring their energies into task performance. And, since there are very few coordination problems in these groups, the result was relatively good performance.Hackman and Vidmar (1970), pp. 49
Source: Hackman, J. Richard, and Neil Vidmar. “Effects of size and task type on group performance and member reactions.” Sociometry (1970): 37-54. https://www.jstor.org/stable/2786271
Slater, Philip E. “Contrasting correlates of group size.” Sociometry 21, no. 2 (1958): 129-139.