The role of different isoforms of cyclo‐oxygenase (COX) in mediating the acute (0‐6 h) and late (24 h) phases of inflammation was investigated in the rat knee joint following intra‐articular injection of carrageenan. The hyperaemic response was assessed transcutaneously using laser Doppler imaging (LDI). Samples were taken at corresponding time points for detection of synovial COX‐1, COX‐2 and inducible nitric oxide synthase (iNOS) mRNA, and measurement of urinary prostaglandin (PG) and nitric oxide metabolites (NOx). A non‐selective COX inhibitor (indomethacin, 15 mg kg−1i.p.), a selective COX‐2 inhibitor (SC‐236, 16.8 mg kg−1i.p.) or vehicle were administered 1 h prior to carrageenan in the acute phase study. LDI scans were taken hourly for 4 h post‐induction. Inflammatory hyperaemia in the vehicle group was attenuated in the indomethacin‐ (P < 0.001, two‐way ANOVA) and SC‐236‐treated groups (P < 0.0001), with no difference between these treatments. At 24 h, i.v. infusion of indomethacin (0.1 mg min−1), increased vascular resistance (24 ± 7.1 %; P < 0.05) compared to vehicle infusion, whereas SC‐236 (0.11 mg min−1) did not. Resistance changes to indomethacin also differed from SC‐236 (P < 0.05). Knee joint diameter progressively increased over 24 h (P < 0.0001, one‐way ANOVA). Urinary PG levels increased by 6 h (P < 0.05), but returned to baseline by 24 h. COX‐1 mRNA was detectable at all time points; COX‐2 mRNA only at 3 h. Urinary NOx levels increased progressively over 24 h (P < 0.05), paralleled by induction of iNOS in the 3 and 24 h samples. Prostaglandin production via COX‐2 appears to mediate the development of acute inflammatory hyperaemia, but nitrergic mechanisms may supervene subsequently. COX‐1 but not COX‐2 contributes to the maintenance of basal blood flow in the hyperaemic joint at 24 h.