The likelihood ratio approach for DNA typing in criminal cases is desc
ribed. It is shown how this approach uses both the sizes and pattern o
f discrepancies between the crime scene profile of fragment lengths an
d the suspect profile for quantifying the strength of the evidence. In
contrast to the current match-binning approach, it avoids an initial
decision about whether the two profiles match. Likelihood ratios for p
airs of profiles that meet the published statistical criteria for a ma
tch show a wide range of values including some that indicate the evide
nce is strongly against identity.