A multiple-tubes procedure is described for using PCR to determine the genotype of a very small DNA sample. The procedure involves dividing the sample among several tubes, then amplifying and typing the contents of each tube separately. The results are analyzed by a statistical procedure which determines whether a genotype can be conclusively assigned to the DNA sample. Simulation studies show that this procedure usually gives correct results even when the number of double-stranded fragments in the sample is as small as 30. The procedure remains effective even in the presence of small amounts of laboratory contamination. We find that the multiple-tubes procedure is superior to the standard one-tube procedure, either when the sample is small or when laboratory contamination is a potential problem; and we recommend its use in these situations. Because the procedure is statistical, it allows the degree of certainty in the result to be quantified and may be useful in other PCR applications as well.