In view of conflict with the conclusions of several earlier studies, a fresh analysis has been made ofCohen's data on publication-rate for various lab. group sizes in the National Cancer Institute (USA) and the National Institute of Medical Research (U.K.) for 1976–77. The present analysis is based on subdividing the data into successive ranges of lab. group size, 1–3, 4–6, 7–9 ... and calculating the relevant publication rate person (R) for each range.When plotted, this yields graphs with aninitial linear rise of per-capita publication rate, R, followed by well-marked maxima of publication rates (maxima/minima 1.6) at group size of 6,16, and 27±2 persons, which are confirmed through analyses with ranges of two different group sizes. The group size at the peaks presumably correspond to optimum efficiency, in general agreement with the findings of the present author on samples from various countries published several years ago. Our conclusions fall somewhere between the findings ofCohen (publication rate independent of size) and those ofWallmark et al. (research efficiency increases exponentially with size).