AbstractForty British Friesian cows with a mean calving date of 28 January were used in a randomized block design experiment to evaluate a high‐quality grass silage for milk production. The high‐quality silage was made from two consecutive cuts of a perennial ryegrass sward after regrowth intervals of 37 days, wilted to 51% dry matter, finely chopped and ensiled using 2.2 1 t‐1(0.5 gallons per ton) of formic acid. The resulting silage had a dry matter digestibility of 72.9% and was fedad libitumwith 3.8 kg (8.4 lb) concentrates as a supplement. A medium‐quality grass silage, of lower digestibility and dry matter content (70.0 and 25.4% respectively), was also fedad libitumin addition to either 3.8, 5.7 or 7.6 kg (8.4, 12.5 or 16.7 lb) concentrates. The feeding treatments were commenced immediately after calving and were terminated on 9 April giving a mean period of 72 d on the treatments. Over the experimental period the animals on the high‐quality silage consumed 15% more silage dry matter and produced 8% more milk that those on the medium‐quality silage with the same level of concentrate supplementation. From the results it was estimated that 1.9 kg (4.3 lb) of additional concentrates would be required with the medium‐quality silage to give an equivalent milk output to that achieved with the high‐quality silage. Milk composition, liveweight change, ration digestibility and nitrogen balance data are