Sub-optimal pill adherence to antiretroviral treatment among HIV-infected persons may have serious consequences, including a decline in health status, eventually leading to death. We developed an instrument to assess self-reported pill adherence to antiretroviral therapy (ART) among patients attending a public health clinic in South Africa. The instrument, based on previously published guidelines, consisted of five questions and a visual analogue scale assessing pill adherence behaviour and was administered to 101 patients living with HIV who commenced ART at least six months prior to data collection. The scale showed fair but not excellent internal consistency (alpha reliability=0.63). Participants reported generally high levels of pill adherence. One question on the scale was able to distinguish between a detectable and a lower than detectable viral load (p<0.01). Yet, the combined items did not predict biologically measured adherence. Despite good comprehensibility of items, the study may have been vulnerable to recall and social desirability bias and was conducted only among clinic attenders rather than those with varied levels of clinic attendance, which is potentially a matter for further study.