In developing countries, dietary diversity is usually assessed during a single yearly period and the effects of seasonal variations remain unknown. We studied these variations in women living in a Sahelian rural area (Burkina Faso). A representative sample of 550 women was surveyed at the beginning and at the end of the seasonal cereal shortage in April and September 2003, respectively. For each season, a dietary diversity score (DDS) representing the number of food groups consumed over a 24-h period, was computed and nutritional status was assessed by the BMI. The DDS increased from 3.4 +/- 1.1 to 3.8 +/- 1.5 food groups between the beginning and the end of the shortage season (P < 0.0001), and the proportion of women exhibiting low DDS decreased from 31.6 to 8.1%. This was due to the consumption of foods available during the cereal shortage season and despite the decrease in the consumption of some purchased foods. The increase in DDS was lower in women for whom DDS was already high in April and vice versa. Over the same period, the percentage of underweight women (BMI <18.5 kg/m(2)) increased from 11.1 to 17.1%. The relation between DDS and the women's socioeconomic characteristics or nutritional status was weakened in September. Thus, in April, fewer women were underweight when their DDS was high than when it was medium or low [odds ratio = 0.3 (0.2; 0.6)], but not in September [odds ratio = 0.6 (0.3; 1.0)]. In such a context, it would be useful to measure dietary diversity at the beginning of the cereal shortage season, when many women exhibit low DDS.