The effects of an exogenous enzyme preparation, the application method and feed type on ruminal fermentation and microbial protein synthesis were investigated using the rumen simulation technique (Rusitec). Steam-rolled barley grain and chopped alfalfa hay were sprayed with water (control, C), an enzyme preparation with a predominant xylanase activity (EF), or autoclaved enzyme (AEF) 24 h prior to feeding, or the enzyme was supplied in the buffer infused into the Rusitec (EI). Microbial N incorporation was measured using (15NH4)2SO4 in the buffer. Spent feed bags were pummelled mechanically in buffer to segregate the feed particle-associated (FPA) and feed particle-bound (FPB) bacterial fractions. Enzymes applied to feed reduced neutral-detergent fibre content, and increased the concentration of reducing sugars in barley grain, but not alfalfa hay. Ruminal cellulolytic bacteria were more numerous with EF than with C. Disappearance of DM from barley grain was higher with EF than with C, but alfalfa was unaffected by EF. Treatment EF increased incorporation of 15N into FPA and FPB fractions at 24 and 48 h. In contrast, AEF reduced the 24 h values, relative to C; AEF and C were similar at 48 h. Infused enzyme (EI) did not affect 15N incorporation. Xylanase activity in effluent was increased by EF and EI, compared to C, but not by AEF. Xylanase activity in FPA was higher at 48 h than at 24 h with all treatments; it was higher with EF than C at 24 and 48 h, but was not altered by AEF or EI. Applying enzymes onto feeds before feeding was more effective than dosing directly into the artificial rumen for increasing ruminal fibrolytic activity.