Almost no HIV survey published in Africa is based on a statistically random sample; therefore, there is selection bias in all data.  The seroprevalence data used in this report generally are drawn from the largest and most recent samples available.  Both time and geographic place were controlled in the comparisons.  All data used had at least two verification tests.  Reliability of the data have been assessed; but given the quality of the original data, the patterns described still should be considered tentative indications.  Detailed information on sample size and population tested for each of the seroprevalence estimates used in this paper are contained in appendix tables.