The method of Zinn & Owens (1986; Canadian Journal of Animal Science66, 157–166), based on release of purine bases by HClO4 followed by their precipitation with AgNO3, was used to study recovery of purines from lyophilized rumen microbial or Escherichia coli preparations added to matrices such as cellulose, starch and neutral-detergent fibre. The recovery of purines was poor (approximately 50 %). Under the hydrolysis conditions (12 M-HClO4, 90–95° for 1 h) used in the method of Zinn & Owens (1986), the recovery of purines from the rumen microbial preparations added to matrices measured using an HPLC method was 95–102 %, suggesting that the lower recovery of purines in the method of Zinn & Owens (1986) was not due to incomplete hydrolysis of nucleic acids. Using the HPLC method, adenine and allopurinol (an internal standard) were found to be heat-labile as substantial destruction was observed on heating at 121°. On the other hand, another commonly used internal standard, caffeine, was stable at 121°. A complete hydrolysis of nucleic acids from the rumen microbial preparation was observed with 2·5 ml 0·6 M-HClO4 in a total volume of 3 ml (0·5 M-HClO4 during hydrolysis) at 90–95° for 1 h, and under these conditions adenine, guanine, allopurinol and caffeine were stable. Moreover, under these milder hydrolysis conditions, the recovery of purine bases from the rumen microbial or E. coli preparations added to matrices ranged from 92 to 108 % using the method of Zinn & Owens (1986). Based on the results, changes in hydrolysis conditions have been proposed for accurate determination of purine bases using spectrophotometric or HPLC methods.