It is now clear that almost everybody forms antibodies to HIV within six months of infection, and that in the majority of infections, this happens within six weeks of infection.

Studies which have attempted to define the average length of time before seroconversion have taken two approaches.

One is to select a group of people likely to be at high risk of HIV infection who remain HIV antibody negative (for instance, gay men who have had unprotected anal sex with partners known to be HIV-positive), and to use polymerase chain reaction testing to detect proteins from HIV in the blood before the formation of antibodies. The other is to trace the recipients of HIV infected blood donations, and to use a well–tried method of mathematical projection to estimate how long seroconversion took.

In 1988 Simmonds studied stored blood samples from 18 seropositive haemophiliacs; seroconversions were estimated to have taken place between 30 and 160 days from infection in this group.

In 1989 Horsburgh estimated that from a sample of 39 seropositive men for whom stored serum was available before and after seroconversion, the median time to seroconversion was 2.4 months; in other words, half the sample took longer than 2.4 months, with 95% of the total sample having seroconverted by 10.3 months. The researchers performed a similar analysis of data for other studies and concluded that the median time to seroconversion was 2.1 months, with 95% having seroconverted after 5.8 months.

This formed the basis of advice that anyone wishing to take an HIV antibody test should wait for between 3–6 months after the last occasion they might have been at risk of infection. However, past studies had a number of drawbacks. They used less sensitive antibody tests, and also employed PCR testing at a time when experience with the technology was still very limited. False results were much more common. Researchers also assumed that those exposed to the virus were likely to become infected after a small number of exposures. It's now known that some people may not become infected despite repeated exposures to HIV, and so what is assumed to be a lengthy delay between exposure and the formation of antibodies may be a combined result of erroneous PCR results and infection at a much later stage in the study than assumed.

A very small number of people may take longer than six months to form antibodies, as studies by Horsburgh, Wolinsky and Imagawa have noted. These studies all used polymerase chain reaction testing to detect genetic fragments of HIV, which appeared to be present despite no evidence of seroconversion.

However, in a follow–up to their original study, Imagawa and colleagues retested blood samples from gay men who had been seronegative despite evidence of infection by more advanced PCR methods, and discovered that only one of the sample still showed evidence of infection. They concluded that what they had previously observed were unsuccessful cases of infection, not latent infection. The phenomenon of unsuccessful infection is still poorly understood, but it is interesting to note that in the period between the first and second test all those who had shown earlier evidence of infection ceased having unprotected anal intercourse. This suggests that the earlier PCR result may well have been erroneous.