There has been considerable interest in studying how the research output of a group of N researchers depends on the group-size, N. Several workers have studied this, but with conflicting conclusions, ranging from finding constant per-capita output to per-capita output varying linearly as N, and even exponentially with N. The present communication states afresh the author's earlier theory of productive interactions and gives analyses of the outputs of two prolific research groups: one from Dhaka University, Bangladesh, and one from Karachi University, Pakistan, each over nearly two decades. The data, obtained from published bibliographies, are sub-divided into small successive ranges of lab. group size, 1–2, 3–4, 5–6, etc., and analyzed by calculating the relevant publication-rate per person (R) for each range. Plots of the data from each group show evidence of an initial approx, linear rise of per-capita publication rate, R, up to about N=5, followed by a maximum at group-size of 6 to 8 persons. This group size would correspond to the optimum efficiency, as a balance between the benefits of increasing interaction ( N2) and Parkinsonian loss of efficiency. This is in agreement with the first peak in the author's earlier analysis (of recent U.K. and U.S.A. data) published five years ago inScientometrics, as well as his previous work published elsewhere. Possible reasons for the failure of statistical criteria to show up this phenomenon of increasing per-capita output are indicated and further indepth studies on two University research groups are planned.