Five levels of L-valine [3.3 (basal), 5.5, 7.8, 10.1 and 12.4 g/d] were fed to gilts from one estrus before mating through two pregnancies and for 30 d of a third pregnancy. Five litter-mate outcome groups of five animals each were allocated randomly to the diet treatments. Animals were individually fed 1.82 kg/d premating and during pregnancy. A common diet was fed to all animals during lactation at levels of 4.0 and 4.75 kg/d for first and second farrowings, respectively. Animals fed 3.3 g/d L-valine retained less N (P less than .10) at d 45 and 90 for two pregnancies than did animals fed higher valine levels. Plasma valine concentrations determined after a 24-h fast and 2 h postfeeding increased sharply (P less than .01) among animals fed more than 5.5 g/d L-valine. There was a treatment X bleeding time interaction (P less than .01) for plasma valine, explained by a postfeeding depression compared with a higher fasting concentration for animals fed 3.3 g/d L-valine. At all other levels of valine, the fasting and postfeeding plasma valine concentrations were equal or increased after feeding. Sow weight gain during the experiment increased (linear, P less than .05) as dietary L-valine increased. Estimated milk yield was greatest for animals fed 5.5 g/d, but average pig gain was maximized at 7.8 g/d L-valine. We conclude that 5.5 g/d L-valine met the requirement for pregnancy in this experiment.(ABSTRACT TRUNCATED AT 250 WORDS)