A scanning electron microscopy was used after in vitro and in vivo tests to investigate any alterations caused by the peristaltic roller pump in erythrocyte morphology. The electron micrographs of samples were examined as follows: 1) by image analyser; 2) by applying Bessis's classification for the qualitative study of crenated red blood cells (RBCs). The in vitro test was repeated four times using blood from healthy donors. Each basal blood sample was divided into 250 ml portions, each of which was recirculated for 12 minutes at different flow rates. In order to verify any persistent erythrocyte damage caused by the peristaltic pump, 15 minutes after recirculation at 450 ml/min, another sample was prepared using the blood remaining from the last test. A statistically significant direct correlation was found between blood flow (Qb) increase and the percentage of morphologically altered RBCs, when either using an image analyser (r = 0.97; p < 0.05) or Bessis's classification (r = 0.95; p < 0.05). However, neither method showed any statistically significant difference between the percentage of deformed RBCs, determined in the basal sample, or in the percentage found at the end of the 450 ml/min test after standing 15 minutes at room temperature. The in vivo test was carried out on 6 patients over 2 dialysis sessions, which differed only for the Qb: 250 versus 400 ml/min. The two dialysis sessions gave comparable results when using both study methods regarding the presence of deformed RBCs.(ABSTRACT TRUNCATED AT 250 WORDS)