PROGRAMMED cell death (apoptosis) is a prominent feature of the development of the immune and nervous systems^1,2. The identification of the Caenorhabditis elegans cell death gene, ced-3, as a prototype of the interleukin-lβ converting enzyme (ICE) protease family has led to extensive evidence implicating these enzymes in apoptosis^3,4. Among the ten or more members of the ICE protease family, CPP32/yama/apopain^5–7 exhibits the highest similarity to CED-3 in both sequence homology and substrate specificity⁸. To analyse its function in vivo, we generated CPP32-deficient mice by homologous recombination. These mice, born at a frequency lower than expected by mendelian genetics, were smaller than their littermates and died at 1–3 weeks of age. Although their thymocytes retained normal susceptibility to various apoptotic stimuli, brain development in CPP32-deficient mice was profoundly affected, and discernible by embryonic day 12, resulting in a variety of hyperplasias and disorganized cell deployment. These supernumerary cells were postmitotic and terminally differentiated by the postnatal stage. Pyknotic clusters at sites of major morphogenetic change during normal brain development⁹ were not observed in the mutant embryos, indicating decreased apoptosis in the absence of CPP32. Thus CPP32 is shown to play a critical role during morphogenetic cell death^9,10 in the mammalian brain.