In three separate tests in 1987, 1988 and 1989, a total of 436 one-year-old N'Dama cattle were maintained for 12, 18 and 24 weeks under a medium natural tsetse-trypanosome challenge in Gabon, Central Africa. Matching health and performance data were recorded on 4, 10 and 13 occasions respectively, to allow simultaneous evaluation of the effect of different criteria of trypanotolerance on animal performance. Under trypanosome prevalences of 25, 31 and 9%, respectively, ability to control the development of anaemia had a very major effect on daily weight gain, four times that of the ability to control parasitaemia, while previous exposure to trypanosome infection from birth to one year had no effect. Anaemia control, measured by average packed red cell volume percent (PCV) over the test period or by lowest PCV reached, was more closely associated with animal performance than when measured by average PCV when detected as parasitaemic. Above average PCV values in the first two measures resulted in a 44% to 48% superior daily weight gain over below average PCV values. PCV post-test recovery was shown to be rapid following a single trypanocidal drug treatment. In practice, it appeared that a suitable field test would be where natural infection could be effected as early in the test as possible and anaemia control measurements carried out over 6 weeks of detected parasitaemia. A field test would become even more feasible if satisfactory correlation could be obtained between the results of natural infection and those of an experimental alternative.