In order to establish an efficient selection criterion the variability in three asparagus populations was evaluated defining the most important yield components and analysing its evolution along three growing seasons. The yield components, coefficient of variation (CV) and the proportion of plants contributing to 80% of the total yield were estimated. The elite plants were selected by mean of total yield and clusters techniques. Multiple regression showed that spear number (SN) and spear weight (SW) were the most important yield components. In every population, total yield (TY) and SN showed the highest values of CV, independently of sex. 69% of the plants contributed to the 80% of the total yield in the first year while in the second and third year the contribution was 57%. At the end of the third year, 17 plants were selected by the average of the total yield and 43 by clusters. It is suggested to select for SW in the first year, reducing in 68% the experimental material. In the second year, the selection for SN would reduce to 5% the plants to evaluate for total yield in the third year. In this way the selected plants are the same but the number of plants to evaluate is dramatically reduced along the years, therefore facilitating the breeders work.