The present study was designed to address whether exercise influences the experimental metastasis of a mouse mammary tumour; a related issue was the impact of timing of exercise onset relative to tumour exposure on NK and LAK mediated cytotoxicity. After 8 weeks of forced treadmill exercise or voluntary wheel running or remaining sedentary, female BALB/c mice received an intravenous (i.v.) injection of female BALB/c mice received an intravenous (i.v.) injection of MMT line 66 tumour cells. Mice were then randomized into continuation of activity (TT, WW), cessation of activity (TS, WS), initiation of activity (ST, SW) and maintenance of sedentary condition (SS) for three weeks. Tumour control (TC) mice, who were matched to the SS mice for age, received an i.v. injection of heat-killed MMT 66 tumour cells. In total there were 8 groups including the tumour control. The average number of lung tumours did not differ by activity condition; however, the mice in the continuous treadmill group (TT) tended to have a higher tumour multiplicity (162 +/- 22) and those in the treadmill x sedentary condition (TS) tended to have lower tumour multiplicity (109 +/- 16) compared with the other groups except the SW group (95 +/- 15). The lymphokine activated killer activity in the spleen was significantly elevated in the TS (49 +/- 3%) and WS (44 +/- 3%) mice compared with the sedentary animals (30 +/- 3%) (p < 0.003 and 0.05, respectively). NK activity was lower in the mice that had stopped exercising (TS and WS) after injection of tumour compared with sedentary animals. These data suggest that although exercise training influences natural immune cytotoxic mechanisms in vitro, this may not translate into clinically significant changes in tumour burden. The dissociation between natural immunity and tumour outcome may reflect the relative resistance of the tumour line to lysis by natural killer cells. It remains to be tested if infusion of IL-2 (to induce LAK activity) in exercise trained animals results in fewer tumour metastases.