Current routine genotyping methods typically do not provide haplotype infor
mation, which is essential for many analyses of fine-scale molecular-geneti
cs data. Haplotypes can be obtained, at considerable cost, experimentally o
r (partially) through genotyping of additional family members. Alternativel
y, a statistical method can be used to infer phase and to reconstruct haplo
types. We present a new statistical method, applicable to genotype data at
linked loci from a population sample, that improves substantially on curren
t algorithms; often, error rates are reduced by >50%, relative to its neare
st competitor. Furthermore, our algorithm performs well in absolute terms,
suggesting that reconstructing haplotypes experimentally or by genotyping a
dditional family members may be an inefficient use of resources.