The sensitivity of dengue virus identification by mosquito inoculation and four reverse transcription-polymerase chain reaction (RT-PCR) procedures (Am. J. Trop. Med. Hyg. 45 (1991) 418 (H); J. Clin. Microbiol. 29 (1991) 2107 (M); J. Clin. Microbiol. 30 (1992) 545 (L); and Southeast Asian J. Trop. Med. Public Health 27 (1996) 228 (Y)) were compared using coded clinical specimens derived from areas in Thailand where all four dengue serotypes circulate. The sensitivity of virus detection in serologically confirmed dengue cases was 54, 52, 60, 79, and 80% for mosquito inoculation, procedures H, M, L and Y, respectively. In comparison to clinical specimens which yielded virus isolates by mosquito inoculation, there was relatively low sensitivity in detecting each of the four dengue serotypes by PCR: procedure H-dengue 4 (25%), procedure M-dengue 3 (73%), procedure L-dengue 1 (70%), and procedure Y-dengue 1 (79%). Dengue virus was detectable by RT-PCR for more days of illness and in the presence of dengue-specific antibody when compared to virus isolated in mosquitoes. Procedures L and Y were more sensitive than mosquito inoculation or procedures H and M in detecting all four dengue serotypes in clinical specimens and may be the RT-PCR methods of choice for virus surveillance or research use. © 2002 Elsevier Science B.V. All rights reserved.