The red blood cell distribution width (RDW), which provides a quantitative measure of the heterogeneity of the red cell population (anisocytosis) in the peripheral blood, the mean corpuscular volume (Mcv) and a regression model combining both variables were used to assess their predictive accuracy in differentiating 51 dogs with regenerative anaemia from 92 dogs with non-regenerative anaemia, which had been diagnosed on the basis of the corrected reticulocyte count. A classification tree analysis was constructed to generate an optimum set of diagnostic rules to differentiate between the two types of anaemia. Seventyfour dogs with a normal haemogram were used as controls. An increase of 1 per cent in the RDW and of 1 fl in the mcv increased the odds of an anaemic dog suffering from regenerative anaemia by factors of 1.3 and 1.14, respectively. By the classification tree, 78 per cent of anaemic dogs with a RDW of 16.25 per cent or less would be expected to have non-regenerative anaemia. With a RDW over 16.25 per cent, an McV of 68.2 fl was the cut-off between dogs expected to have regenerative (71 per cent) or non-regenerative (75 per cent) anaemia. The RDW and mcv are measured by most automatic haematology analysers and may give the first indication of the bone marrow response of an anaemic dog. However, different electronic counters give different normal values of the RDW and MCV.