With the localisation of the gene for the autosomal recessive forms of proximal spinal muscular atrophies (SMA) to the chromosomal region 5q13 and the later detection of homozygous deletions of the SMN gene located in this region, prenatal prediction of SMA has become feasible and is widely applied now. In our experience with 77 prenatal predictions of SMA, follow-up of the 39 liveborn children from these pregnancies never led to a false-negative result. Application of SMN deletion analysis has consequences for prenatal prediction of SMA. When the index patient has a homozygously deleted exon 7 of the SMN gene, prenatal prediction and interpretation of results are straightforward. In families in which no DNA from the index patient is available, prenatal detection of a homozygous SMN deletion may be considered almost proof of SMA in the fetus. Absence of a deletion, however, will not guarantee an unaffected child. A real problem exists if the index patient does not show a homozygous deletion of SMN exon 7. In such cases with non-homozygous SMN deletions, one cannot be certain of 5q linkage and autosomal recessive inheritance until other SMN mutations are detected. This is an argument to abstain from prenatal diagnosis by linkage analysis in these families.