Two silages were prepared from the Ðrst-cut sward of timothy-meadow fescue and wilted to a dry matter (DM) content of 300 g kg~1. One was ensiled with the addition of a formic-acid-based additive (4 litres formic acid (FA) per tonne) and the other with the addition of a bacterial inoculant (LAB) at a rate of 5 ] 106 colony forming units g~1. Both silages were well preserved, but the extent of fermentation was greater in LAB-silage than in FA-silage as indicated by a lower concentration of water soluble carbohydrates (68 vs 177 g kg~1 DM) and a higher concentration of lactic acid (147 vs 32 g kg~1 DM). Four Ayrshire cows were used in a 4 ] 4 Latin square experiment with 21-day periods to study the e †ects of silage fermentation and postruminal casein supplementation on silage intake, nutrient supply and milk production. The four treatments were FA-silage without casein (FA-0), LAB-silage without casein (LAB-0), FA-silage with casein (FA-C) and LAB-silage with casein (LAB-C). Both silages were given ad libitum with 8 kg day~1 of barley without or with 400 g day~1 of casein infused into the duodenum. Organic matter digestibility was lower (0É723 vs 0É753 ; P \ 0É01) for FA-silage than for LAB-silage. Cows o †ered FA-silage had a higher molar proportion of acetate and a lower proportion of propionate in ruminal Ñuid than cows o †ered LAB-silage. Microbial protein synthesis estimated from the output of purine derivatives in urine was greater (288 vs 260 g N day~1 ; P \ 0É05) for cows given FA-silage compared with LAB-silage. Feeding LAB-treated silage tended (P \ 0É10) to decrease silage DM intake compared with FA-treated silage (10É61 vs 11É77 kg DM day~1). Silage composition did not a †ect signiÐcantly milk yield or milk composition. Casein infusion increased milk yield (25É1 vs 27É1 kg day~1 ; P \ 0É05), milk protein content (32É4 vs 33É8 g kg~1 ; P \ 0É05) and protein yield (808 vs 905 g day~1 ; P \ 0É01). The responses were similar for both silages.