Editor,—We read with interest the paper by Rahimet al examining the relationship between heat shock protein (hsp) gene polymorphisms and sudden infant death syndrome (SIDS).1 The authors described a total of 15 restriction fragment length polymorphisms (RFLP) in three hsp genes in a control population, and in 12 SIDS infants. One RFLP, a 15 kb fragment in the Mspl digest of hsp 60, was found to be present in different frequencies between the control group and the small SIDS group, with a χ² value of 7.37, giving a p value, using one degree of freedom, of 0.00663. However, as 15 statistical comparisons have been made between the two groups, Bonferroni’s correction should be applied,2 raising the p value to 0.0995 (not significant). Further, to choose controls from the general population may not be appropriate, as polymorphism frequency may vary greatly between different ethnic groups. We suggest a further study investigating only the one polymorphism implicated above, in which the parental allele frequency is also ascertained. This would enable non-transmitted allele frequency to be used as an ideal ethnically matched control and statistically analysed by a family based control method.3