When staphylococcal strains from hospital patients are plotted in a frequency curve according to the size of the growth-free zone in a disk test for sensitivity to penicillin-G, this curve is trimodal. This implies that, besides the sensitive, there are two populations of resistant staphylococci: the moderately and the highly resistant, In a hospital-wide survey of coagulase or mannite positive staphylococci isolated in North Carolina Memorial Hospital between October 1964 and October 1965 it was inferred that: (1) there was a statistically highly significant difference between the distributions of the moderately and the highly resistant strains over duration of hospital stay prior to the time of isolation; (2) the highly resistant strains were seldom introduced from outside; and (3) there was no significant association between use of penicillin-G by a patient harbouring a strain and the frequency with which such a strain was found to be highly resistant to this drug. The latter finding is in contrast to significant associations found when the penicillin resistance is moderate, or when the drug used is other than penicillin-G. It is concluded that a fall in the incidence of staphylococcal strains highly resistant to penicillin-G in patients who did not receive other antibiotics may well be a valid index of the efficacy of programmes to control hospital infection. If so, there is a practical merit in adopting the routine of distinguisning between two groups of penicillin-resistant staphylococci, which behave, epidemiologically, as two distinct bacterial populations.