The main aim of this work was to assess factors affecting the secretion of carotenoids in cows’ milk. Our objectives were 1) to determine the kinetics of the decrease in carotenoids in plasma, milk, and adipose tissues following a switch from a high- to a low- carotenoid diet; and 2) to specify whether, during lipomobilization, the restitution of these compounds stored in the adipose tissues is sufficient to modify their secretion in milk. During the preexperimental period, 32 cows in midlactation were fed a grass silage-based diet, and were then assigned to 4 groups; 2 groups were maintained on the grass silage diet and 2 were switched to a late hay diet. For each forage diet, one group was fed according to net energy for lactation and nitrogen requirements, and the other was submitted to an energetic underfeeding, with similar forage and carotenoid intake between groups. Variations in concentration of carotenoids and color index (CI) of plasma and milk were monitored over 8 wk. Other components of nutritional interest; i.e., vitamin E (VE), vitamin A, and fatty acids, were also measured. The switch from grass silage to hay diet induced a rapid decrease in concentration of ßl-carotene (BC) and VE and in the CI of plasma and milk during the first 2 wk. Pools of BC in adipose tissues also decreased by 40%. Concentrations of BC at the end of the experiment for silage and hay groups were 5.10 and 1.71 µg/mL in plasma and 0.17 and 0.07 µg/mL in milk, respectively. The energetic underfeeding did not affect BC concentration in plasma and induced a small increase in milk BC concentration, related to a decreased milk yield. In the silage group, the energetic underfeeding after 3 to 4 wk induced a decrease in CI and VE of plasma, but not of milk. The fatty acid profile in milk was modified by the change from grass silage to hay diet (C10 to C14 and linoleic acid increased; stearic and linolenic acid percentages decreased) and by underfeeding (oleic, vaccenic, and rumenic acid percentages increased). This study shows that BC and VE levels persist in midlactation cows’ plasma and milk for about 2 wk. The results could not confirm a release of BC by bovine adipose tissue, but the level of underfeeding was moderate in this trial. The concentration of BC explained 58 and 40% of variation in CI of plasma and milk, respectively. These CI appear to be valuable tools for diet traceability (i.e., silage vs. hay).