A study was made to determine whether two experienced clinicians elicited and scored primitive reflexes (PR) differently and whether reliability could be improved by standardization. Three studies were carried out, using a protocol for the examination of 14 PR. In the first study with 31 healthy young subjects, two investigators found virtually no difference in the routine neurological examination. However, the interobserver agreement was very poor, indicating the need for a further improvement of the PR protocol. In the second study, 30 neurological patients were examined with an improved, more explicit and standardized protocol, in which the amplitude and the persistence of the reflex were scored separately. Interobserver agreement improved considerably, and was high for amplitude as well as persistence. In the third study, 36 neurological patients were examined twice by one investigator within 2 weeks. Good to excellent intraobserver agreement was found. No pathognomonic or strictly localizing reflex could be distinguished.