Three large and comparable series of births were used to test the working hypothesis that if there is a real seasonal variation in the frequency of a given congenital malformation; it would have to be shown by adequate analysis; to be more overt in non-tropical areas; and to be six months out of phase in northern and southern hemispheres. The data set were hospital births from tropical (287,165 births) and non-tropical (582,585 births) South America, and from Italy (508,536 births). Sixteen well-defined malformation types were tested: anencephaly, spina bifida, cephalocoele, hydrocephaly, microtia, cleft palate, cleft lip, oesophageal atresia, anal atresia, hypospadias, pes equino-varus, pes talovalgus, postaxial polydactylyl, pre-axial polydactylyl, diaphragmatic hernia, and Down's syndrome. No seasonal variation was proven (p less than 0.01) for any malformation type in any of the three series of data by means of Walter and Elwood's test, or Hewitt et al's non-parametric test2 applied to seven instances with sample sizes smaller than 50 cases. Variations of borderline significance (p less than 0.05) included oesophageal atresia in tropical South America, none in non-tropical South America, and anencephaly in Italy. It is concluded that seasonal variation in the occurrence of congenital malformations is a rare phenomenon when tests are strictly used within their recommended limitations.