OBJECTIVES: This study compared the reliability and feasibility of three different versions of VAS on socioeconomic heterogeneous groups in a developing country. METHODS: Ten health states were described and applied on forty adults with mean age 33. In the first version, 10 cubes were placed against health states, and respondents were requested to remove the number of cubes equivalent to loss in health associated with the state with ten cubes equivalent to full health and zero cube equivalent to death. In the second version, they had the conventional VAS and were requested to point at the scale. Thirdly, respondents were requested to pick from a collection of cubes the number equivalent to each health state with ten cubes equal to death and no cube being equivalent to full health. Mean explanation and valuation times were 5 and 18 minutes respectively. Thirteen percent of the respondents had problems understanding the first version of the instrument. RESULTS: The standard deviation of the elicited values with the second version was highest for all the disease cases (Disease:Std{1st, 2nd, 3rd version}; Epilepsy: Std{0.25, 0.28, 0.23}, Bronchitis: Std{0.20, 0.27, 0.22}, Dental Pain: Std{0.21, 0.23, 0.18}, Vertigo: Std{0.17, 0.23, 0.17}, Madness: Std{0.29, 0.32, 0.28}, etc). ANOVA procedure showed that mean scores obtained using the different versions were statistically different for eight out of the ten diseases (Epilepsy{<.0003}, Bronchitis {<0.0012}, Dental Pain {<0.0165}, Vertigo {<0.0001}, Madness {<0.8321}, etc). CONCLUSIONS: In comparing the preferences elicited with VAS, caution should be exercised as different modifications of the instrument were demonstrated to yield different values. In the elicitation process, the alternative versions of VAS may have allowed the incorporation of other values which lacked with the conventional VAS.