The short-term consumption of foods that differed in energy density (ED) and/or NaHCO3 supplementation, on subsequent food intake and diet selection in sheep were measured. Thirty sheep weighing 35.9 (sd 2.89) kg were used. Two foods were formulated: H had 11 and L had 8 MJ metabolizable energy/kg fresh matter. Four further foods were formulated by adding either 40 g NaHCO3/kg or 16.5 g NaCl/kg to foods H and L. NaCl was added to give the same Na concentration as with 40 g NaHCO3/kg to control for any effects of Na per se. In a preliminary test, it was found that a 2 h consumption of food H supplemented with NaHCO3 could buffer potential impact on the rumen environment of subsequent consumption of food H alone (as judged by rumen pH and acid-buffering capacity); however, it was not as effective as the consumption of food L alone in doing so. Each food treatment was offered to one of six groups (n 5) for 2 h following 16 h of food deprivation. Sheep were then offered a choice between H and L for a further 6 h. Supplementing H or L with either NaHCO3 or NaCl had no significant effect on either intake or diet selection. ED significantly (P<0.01) affected intake during the 2 h single feeding period, with sheep offered H or L consuming 540 and 663 (sed 37) g respectively, but had no effect on subsequent intake during the choice period. During the choice period all sheep showed a preference for food H, but sheep previously offered L selected significantly more H (0.873 g/g) than sheep previously offered H (0.544 (sed 0.028) g/g; P<0.001). It is concluded that short-term consumption of foods that differ in ED, and hence in their potential impact on the rumen environment, significantly affects subsequent diet selection. This is in agreement with the hypothesis that ruminant animals select a diet to help maintain the rumen environment within a certain physiological range. Food H with 40 g NaHCO3 added/kg may not have been sufficient to affect subsequent diet selection. It is suggested that larger, rather than smaller, changes in the rumen environment achieved through previous feeding should be expected to alter subsequent diet selection.