Intra- and interobserver errors in facial flatness measurements

This study examines intra- and interobserver errors in the facial flatness measurements defined by Yamaguchi (1973) using two-way ANOVA with replication. The sample used consists of 10 recent Japanese skulls from Tohoku District. These skulls are measured twice by each of 6 observers. The interactions between skulls and observers are significant in the frontal chord, the simotic subtense, and the zygomaxillary chord. In the latter 2 items, interobserver errors are not significant, suggesting the nature of differences varies from skull to skull. When the among-skull mean squares over the mean squares of total error are tested, they are significant at <0.1% level in all the items. The variance components expressed as percentages of the sum of variances indicate that intraobserver and total error variances are, on average, the largest in the simotic region, the second largest in the frontal region, and the smallest in the zygomaxillary region. On average, however, the variance components of among-skulls are very large in all the items. These indicate that it may be safe to use the facial flatness measurements obtained by different investigators in analyses of population affinities when the morphological differences among the populations are relatively large. The students employing facial flatness measurements in the study of population affinities should, however, pay attention to the measurement errors, especially those of simotic measurements, in interpretation of the biological significance of their results. The results of two-way ANOVA and nested ANOVA are also discussed in terms of the interaction components.