Two diagnostic schemes for detection of pulmonary embolism by ventilation-perfusion (V-P) scintigraphy were compared for relative accuracy by two groups of observers interpreting 70 V-P scintiscans. Observers in Group B, who used the criteria recently proposed by Biello et al., had a significantly smaller average number of 'indeterminate' interpretations (41%) than did the observers in Group A (55%), who used a simpler scheme (p < 0.05). In addition, Group B showed a slight improvement in positive predictive value without deterioration in the negative predictive value compared with Group A. Along with this improvement in diagnostic performance, Group B achieved a significant reduction in interobserver variability compared with Group A for patients without pulmonary embolism (p < 0.05). There was no significant difference in interobserver variability between the two groups for patients with pulmonary embolism. The diagnostic scheme introduced by Biello et al. represents a useful improvement for the diagnosis of pulmonary embolism by V-P imaging.