Because most automated platelet counters cannot be relied on in thrombocytopenia, clinicians face a problem when decision making is based on platelet counts. Therefore we evaluated a visual platelet counting method from a blood smear with white blood cells (WBCs) as reference (PCW = platelet count based on WBC). Platelet counting for 74 thrombocytopenic (<120 x 10(9)/L) children was performed with PCW and with an automated counter (impedance principle); both methods were compared with evaluation by phase-contrast microscopy as the standard method. The PCW correlated well with the phase-contrast microscopy evaluation (y = -0.38 + 1.01x, r2 = 0.99). For platelet counts <20 x 10(9)/L the maximal deviation was 2 x 10(9)/L. The correlation between automated counts and the standard method was poor. The regression was y = 9.63 + 0.94x, r2 = 0.86. For platelet counts <20 x 10(9)/L the maximal deviation was 37 x 10(9)/L; on average, 7 x 10(9)/L platelets were counted in excess when compared with the standard method. PCW, in contrast to the automated impedance method, discriminated platelets from nonplatelet particles such as debris, fragments of red blood cells (hemolytic-uremic syndrome [HUS]) and of blast cells, and identified platelets of abnormal size. In addition, the appearance ofplatelets, WBCs, and RBCs gave clues to the etiology of thrombocytopenia, such as leukemia, infection, HUS, familial macrothrombocytopenia, and immune thrombocytopenia. PCW is a fast, reliable platelet counting method requiring less experience than the phase-contrast method. Visual evaluation from a stained smear clearly differentiates platelets and nonplatelet particles in contrast to most automated counters. In addition, the original smear can be preserved and reevaluated.