We performed a parallel evaluation of 5 automated reticulocyte counters to produce the immature reticulocyte fraction (IRF). We analyzed 225 samples from healthy control subjects, 115 from patients with various diseases, 38 with advanced aplasia, and 22 in early erythropoietic recovery after chemotherapy or bone marrow transplantation. The reference intervals were different for each instrument (ADVIA 120, 0.04-0.25; CELL DYN 4000, 0.15-0.35; GEN-S, 0.20-0.37; SE 9500 RET 0.05-0.21; VEGA RETIC: 0.06-0.23). The imprecision, obtained by 1-way analysis of variance on duplicates, was satisfactory for clinical use for all methods (coefficient of variation, 7.6%-20.5% in healthy subjects), although it was higher than the analytic goal based on biologic variability within subjects. The comparison of different methods shows that agreement is good only between SE 9500 RET CELL DYN 4000, and VEGA RETIC (r2 = 0.72-0.78). The study of diagnostic performance in distinguishing aplasia from early bone marrow recovery shows slightly different results (area under the curve from 0.70 for ADVIA 120 to 0.96 for SE 9500 RET). Even with slight differences, the fluorescence-based methods seem to be more robust than other methods for IRF measurement.