Red cell oval morphology is still the only accepted basis for the clinical or epidemiological diagnosis of ovalocytosis. Therefore it is important to know the errors when detecting and counting morphological ovalocytes. In all previous studies of ovalocytosis there was no assessment of the variation which may have occurred in classification due to smearing and staining techniques or the criteria for the diagnosis of ovalocyte morphology; nor was inter or intraobserver variation assessed. We report how different peripheral blood smear methods influence the diagnosis of ovalocytosis in populations in the Madang and East Sepik Provinces in Papua New Guinea. We also examined within and between observer variation in the quantitative assessment of ovalocytosis at ×40 and ×100 microscopy powers. A modified method of making a thin malaria blood smear gave the best preservation of red cell morphology and was adopted for the quantitative ovalocytosis studies. A special haematology smear is unnecessary. Ovalocyte frequency estimations were similar when ×40 and ×100 lenses were used, but ×40 was preferable for assessing morphology. Two observers were consistent in their findings and produced very similar results for the high-quality smears from the planned Madang survey, and rather different results for the smears from the unplanned routine Sepik survey. We conclude that measurement error for ovalocytosis assessment can be quite small and unimportant, minimized by careful planning and quality control. Otherwise measurement error is substantial and threatens validity of classification and grading of ovalocytosis.