Laboratory tests of Metarhizium anisopliae var acridum Driver & Milner, at a dose of 1.3 x 10(6) conidia ml-1, had no adverse effects on nymphs of mayfly, Ulmerophlebia sp or 8-week-old fry of the rainbow fish, Melanotaenia duboulayi Castelnau. This dose was toxic to the cladoceran, Ceriodaphnia dubia Richard, causing 100% mortality in 48 h. When this test was repeated at doses of up to 6.7 x 10(3) conidia ml-1, there was only 5% mortality after 192 h. Spraying of artificial water sources with a very high dose of the fungus as an aqueous spray resulted in 80-130 conidia ml-1 at 15 cm depth in the first 24 h after spraying. The conidia rapidly settled out and were absent from the top 15 cm layer of water after about 50 h. A similar experiment using the oil formulation as used in field control resulted in a 2- to 20-fold lower level of conidia in the water. Finally, sampling actual water sources in spray areas revealed a very low level of contamination of the water, with a maximum mean level of 29 conidia ml-1 in the first 24 h after treatment. Thus the level of conidia likely to enter water during control campaigns is a small fraction of that required to kill cladocerans, the only sensitive non-target organism tested. It is concluded that the biopesticide is very unlikely to pose any hazard to aquatic organisms.