To examine the relationship between adherence, viral suppression and antiretroviral resistance in HIV-infected homeless and marginally housed people on protease inhibitor (PI) therapy. A cross-sectional analysis of subjects in an observational prospective cohort systematically sampled from free meal lines, homeless shelters and low-income, single-room occupancy (SRO) hotels. Thirty-four HIV-infected people with a median of 12 months of PI therapy. Adherence measured by periodic unannounced pill counts, electronic medication monitoring, and self-report; HIV RNA viral load; and HIV-1 genotypic changes associated with drug resistance. Median adherence was 89, 73, and 67% by self-report, pill count, and electronic medication monitor, respectively. Thirty-eight per cent of the population had over 90% adherence by pill count. Depending on the measure, adherence explained 36-65% of the variation in concurrent HIV RNA levels. The three adherence measures were closely related. Of 20 genotyped patients who received a new reverse transcriptase inhibitor (RTI) when starting a PI, three had primary protease gene substitutions. Of 12 genotyped patients who received a PI without a new RTI, six had primary protease gene substitutions (P < 0.03). A substantial proportion of homeless and marginally housed individuals had good adherence to PI therapy. A strong relationship was found between independent methods of measuring adherence and concurrent viral suppression. PI resistance was more closely related to the failure to change RTI when starting a PI than to the level of adherence.