Abstract
φInterobserver errors in scoring presence or absence were investigated for 31 nonmetric cranial traits based on a sample of 203 almost complete Polynesian skulls using the correlation coefficients of attributes. The φ coefficients for all 31 traits were statistically significant at the 0.01 level. When the φ value of +0.7 was regarded as the critical point for assessing interobserver replicability, 20 traits were selected. Since the MMD on these 20 traits did not reach the significant level of 0.05, these traits were considered to be reliable for comparisons between different observers. Because interobserver error was larger than intraobserver error and some anthropologically important traits did not clear the threshold for assessing consistency, it is recommended that physical anthropologists should investigate nonmetric cranial traits by themselves.
