Two round robins (RRs) of the quantitative phase analysis (QPA) of silicon nitrides (Si 3 N 4 ) using the mean normalized intensity (MNI) method and the Rietveld method were conducted as one of the projects for establishing standard methods of characterizing advanced ceramic materials. Accuracy and precision of three techniques, namely the MNI method using peakheight intensity (MNI+P), the MNI method using integrated intensity (MNI+I) and the Rietveld method (R), were tested. Precision of the methods was found to follow the order R < MNI+I < MNI+P in the ®rst RR and MNI+I < R < MNI+P in the second RR. Resulting accuracy of the methods was ranked R 9 MNI+P < MNI+I in the ®rst RR and MNI+P < R 9 MNI+I in the second. The MNI+P method, which relies upon a simple and routine procedure for measuring peak-height intensities, gave the best precision in both RRs. Both the accuracy and the precision of the Rietveld method were the worst among the three techniques in the ®rst RR. They were, however, signi®cantly improved in the second RR. Although the precision of the MNI+I method was the worst in the second RR, it was better than that in the ®rst, and the accuracy was the best in both the ®rst and the second RR. The degree of improvement from the ®rst to the second RR, in both precision and accuracy, was MNI+P < MNI+I < R, coinciding with the ease of these three techniques in reverse order. This result is largely due to (i) a new protocol for experimental and analytical parameters and (ii) improved skill of the participants in data analysis in the second RR. Magnitudes and signs of the observed errors could be interpreted through results of the theoretical studies.