Objectives: The findings that inducible cyclooxygenase (COX)-2, but not constitutive COX-1, is upregulated in the brain of conscious rats ∼1.5 h after intraperitoneal pyrogen administration, that the systemic administration of COX-2 inhibitors abolishes fever, and that COX-2-deficient mice do not develop fever in response to intraperitoneal lipopolysaccharide (LPS) have strongly implicated COX-2 in the mediation of the febrile response. However, the biosynthesis of COX-2 is significantly slower than the onset of the fever produced by intravenously injected LPS. It consequently seems improbable that inducible COX-2 could play a role in the initiation of this febrile response, but a role for COX-1 has not yet been categorically ruled out; or, alternatively, a constitutive isoform of COX-2 could have such a role. We have studied, therefore, the effects of the non-selective COX inhibitor indomethacin, the COX-1-selective inhibitor SC-560, and the COX-2-selective inhibitor nimesulide on the characteristically biphasic fever induced by intravenous LPS in conscious guinea pigs; it has an onset latency of ∼10 min. Methods: We injected the inhibitors 30 min before LPS, in various combinations of doses and routes; their respective vehicles were the control solutions. Core temperatures (T<sub>c</sub>) were monitored continuously, and plasma and brain PGE<sub>2</sub> levels were measured before and at 2-hour intervals after LPS administration. Results: Intraperitoneal indomethacin at 10 mg kg<sup>–1</sup> attenuated both phases of intravenous LPS (2 µg kg<sup>–1</sup>) fever, but the first more so than the second; at 50 mg kg<sup>–1</sup>, it inhibited the febrile response completely. Intraperitoneal SC-560 (5 mg kg<sup>–1</sup>) did not affect the febrile response to intravenous LPS (2 µg kg<sup>–1</sup>). Intraperitoneal nimesulide (0.3, 1.0, and 3.0 mg kg<sup>–1</sup>) dose dependently attenuated intravenous LPS (0.1 and 2 µg kg<sup>–1</sup>) fever; the second phase of the biphasic T<sub>c</sub> rise was affected significantly more than the first. Intraperitoneal nimesulide also prevented the associated rises in plasma and brain PGE<sub>2</sub> levels. Intracerebroventricular LPS (150 ng kg<sup>–1</sup>) evoked a monophasic fever with a long onset latency (∼30 min); it was accompanied by a rise in brain PGE<sub>2</sub> only, implying that the febrigenic PGE<sub>2</sub> was generated directly in the brain. This response, however, was completely abolished by intraperitoneal nimesulide (3 mg kg<sup>–1</sup>), indicating that nimesulide crosses the blood-brain barrier. Intracerebroventricular nimesulide at 0.3 mg kg<sup>–1</sup> prevented the rise in plasma PGE<sub>2</sub> after intravenous LPS (2 µg kg<sup>–1</sup>) and again attenuated the second febrile peak significantly more than the first. Conclusions: COX-1 is not involved in intravenous LPS fever production, and COX-2 appears to play a greater role in the late than in the early phase of intravenous LPS fever in guinea pigs. The involvement of a constitutive COX-2 is inferred in the early phase.