Six rumen-cannulated Holstein cows in early lactation were assigned to 3 treatments: grazing (G), zero-grazing (ZG), and grass silage (GS) harvested from the same perennial rye grass sward in a 3×3 Latin square design with three 21-d periods. The objectives of this study were to investigate the underlying mechanisms for the reported elevation in milk rumenic acid (RA) concentration associated with G compared with ZG and GS, and to identify the important variables contributing to the milk RA response. Grazing animals were offered 20kg of dry matter/cow per day; indoor animals were offered ad libitum grass or silage. A concentrate at a rate of 3kg/d was also offered to all cows. Rumen, plasma, and milk samples were collected in the third week of each period. Data were analyzed by the MIXED procedure of SAS. Dry matter intakes were less for GS with no difference between G and ZG. Milk yield was greater for G than for ZG or GS. Milk fat and protein contents were less for GS with no difference between G and ZG. The combined intake (g/d) of linoleic and linolenic (18:3n-3) acids was different across the treatments (G: 433; ZG: 327; and GS: 164). Rumen pH was less for G with no difference between ZG and GS. Concentrations of volatile fatty acids and ammonia nitrogen in rumens were not different across the treatments. Wet rumen fill was less for G with no difference between ZG and GS. Vaccenic acid concentrations were different across the treatments in rumen (G: 12.30%, ZG: 9.31%, and GS: 4.21%); plasma (G: 2.18%, ZG: 1.47%, and GS: 0.66%) and milk (G: 4.73%, ZG: 3.49%, and GS: 0.99%). Milk RA concentrations were greater for G (2.07%) than for ZG (1.38%) and GS (0.54%). Milk desaturase index based on the ratio cis-9-14:1/14:0 was not different across the treatments. Milk RA yield per 100g of linoleic acid and linolenic acid intake (efficiency) was 2.23, 1.50, and 0.62g in G, ZG, and GS, respectively, suggesting that G cows were more efficient than ZG and GS cows in milk RA production. Stepwise regression analysis of a group of variables revealed that plasma vaccenic acid accounted for 95% of the variation in milk RA production. Milk desaturase index did not enter into the model. Overall findings suggest that substrate intake influenced milk RA production but it was not the only factor involved. There were differences in efficiency of milk RA production, which appears to depend on the factors regulating ruminal vaccenic acid production and its supply to the mammary tissue.