Interobserver variation among four observers in evaluation of eight selected neurological signs was investigated. MATERIAL & METHODS--Two hundred and two consecutive unselected inpatients were examined by two senior neurologists and two trainees, all without knowledge of the neurological case history. The signs examined were: anisocoria, jerky eye movements, facial palsy, elbow extension force, finger-nose test, Barré sign, knee jerk, and extensor plantar reflex. Observed agreement rates and kappa coefficients were calculated in order to compare the interobserver variability among neurologists and trainees, and to evaluate differences in the interobserver variability between signs. RESULTS--Observed agreement rates varied from 0.80 to 0.95 for neurologists and from 0.65 to 0.98 for trainees. For neurologists kappa coefficients ranged from 0.40 to 0.67 and for trainees from 0.22 to 0.81. The neurologists had higher kappa values than the trainees in 5 signs, but this difference was only statistically significant for jerky eye movements. For the individual signs the observed agreement rates were between 0.50 and 0.93 for all four examiners combined, and overall kappa values varied from 0.32 to 0.71 with highest agreement for facial palsy and lowest for knee jerk. CONCLUSION--The magnitude of the interobserver and intersign variation indicates that the interpretation of the neurological signs tested, without knowledge of the case history, should be done with some caution.