Meta-analysis techniques have been used to accumulate data from 94 studies describing the natural killer (NK) cell response of some 900 volunteers to acute and chronic exercise. NK cell numbers have been indicated in terms of CD3-CD16+CD56+, CD16+ or CD56+ phenotypes, and cytolytic activity has been expressed per 10,000 peripheral blood mononuclear cells or in terms of lytic units. Acute exercise has been categorised as sustained moderate (50 to 65% of aerobic power), sustained vigorous (>75% of aerobic power), brief maximal or 'supramaximal', prolonged, eccentric or resistance, and repeated exercise. In general, there was a marked increase in NK cell count at the end of exercise, probably attributable to a catecholamine-mediated demargination of cells. Following exercise, cell counts dropped to less than half of normal levels for a couple of hours but, except in unusual circumstances (e.g. prolonged, intense and stressful exercise), normal resting values are restored within 24 hours. If activity is both prolonged and vigorous, the decrease in NK cell counts and cytolytic activity may begin during the exercise session. Although the usual depression of NK cell count seems too brief to have major practical importance for health, there could be a cumulative adverse effect on immunosurveillance and health experience in athletes who induce such changes several times per week. There is a weak suggestion of an offsetting increase in resting NK cell counts and cytolytic action in trained individuals, and this merits further exploration in studies where effects of recent training sessions are carefully controlled.