Teleost fish possess discrete blocks of oxidative red muscle (RM) and glycolytic white muscle, whereas tetrapod skeletal muscles are mixed oxidative/glycolytic. It has been suggested that the anatomy of RM in teleost fish could lead to higher intramuscular O2 partial pressures (PO2) than in mammalian skeletal muscles. This study provides the first direct experimental support for this suggestion by using novel optical fibre sensors to discover a mean (+/- S.E.M., N=6) normoxic steady-state red muscle PO2 (PrmO2) of 61+/-10 mmHg (1 mmHg=133.3 Pa) in free-swimming rainbow trout Oncorhynchus mykiss. This is significantly higher than literature reports for mammalian muscles, where the PO2 never exceeds 40 mmHg. Aerobic RM powers sustained swimming in rainbow trout. During graded incremental exercise, PrmO2 declined from 62+/-5 mmHg at the lowest swim speed down to 45+/-3 mmHg at maximum rates of aerobic work, but then rose again to 51+/-5 mmHg at exhaustion. These measurements of PrmO2 during exercise indicated, therefore, that O2 supply to the RM was not a major limiting factor at exhaustion in trout. The current study found no evidence that teleost haemoglobins with a Root effect cause extremely elevated O2 tensions in aerobic tissues. Under normoxic conditions, PrmO2 was significantly lower than arterial PO2 (119+/-5 mmHg), and remained lower when the arterial to tissue PO2 gradient was reduced by exposure to mild hypoxia. When two sequential levels of mild hypoxia (30 min at a water PO2 of 100 mmHg then 30 min at 75 mmHg) caused PaO2 to fall to 84+/-2 mmHg then 61+/-3 mmHg, respectively, this elicited simultaneous reductions in PrmO2,to 51+/-6 mmHg then 41+/-5 mmHg, respectively. Although these hypoxic reductions in PrmO2 were significantly smaller than those in PaO2, the effect could be attributed to the sigmoid shape of the trout haemoglobin-O2 dissociation curve.