The penetration of various aminoglycosides into uninfected and infected fluids of steel net cages, implanted subcutaneously into rabbits, was studied. The pharmacokinetics of the antibiotics tested in these fluids were characterized by a peak concentration which was delayed in relation to that in serum after both intramuscular and intravenous administration, and by a slower elimination from cage fluids than from serum. Comparing amikacin, gentamicin, netilmicin and tobramycin, the latter seemed to have a somewhat lower penetrability into uninfected cage fluids. Infection of the cage fluids with gram-negative aerobic bacteria resulted in a reduction of the measurable concentrations of amikacin, gentamicin or netilmicin in the cage fluids when compared to those obtained in uninfected fluids in the same rabbits. Elimination of the aminoglycosides from the infected cage fluids was slower than from the uninfected ones. The lower concentrations of the aminoglycosides in infected cage fluids were considered to be primarily due to a penetration barrier created by the infection. The viable counts in infected cage fluids were only marginally affected in cages where the aminoglycoside concentrations were above the minimum inhibitory concentrations (MIC's) of the aminoglycosides against the bacterial strains used for infection when tested in vitro according to standard techniques. In infected cage fluids the pO2 and pH were low, while the pCO2 was high. The number of viable bacteria was high. These factors, which in vitro increased the MIC's of the agents, and the low concentrations achieved in infected cage fluids could explain the inefficacy of aminoglycoside treatment in this experimental model.