The assessment and meaning of intraobserver error in population studies based on discontinuous cranial traits.

Intraobserver scoring precision of 50 discontinuous (discrete) cranial traits was assessed on a sample of 125 intact Iroquois crania using the square of the phi coefficient. Scoring consistency was high for most traits although several were subject to low precision (phi 2 less than 0.7). Most errors were random rather than systematic. Taken collectively the problematic traits exceeded their expected contribution to MMD coefficients among four Woodland samples from southern Ontario. A traitlist with the problematic variants excluded produced more meaningful biological relationships among the samples. Further, the % contribution of the problematic traits increased with the time interval between the scoring of the individual samples. It was suggested that intraobserver error could effectively distort the interpretation of biological relationships and that it must in part be responsible for the poor performance reported for discontinuous traits in some studies. Therefore, it was argued that the results support rather than compromise the use of discontinuous skeletal traits in population studies and that the assessment of intraobserver error should be a standard procedure of the research design.