The mismatch analysis of PCR-amplified DNA has generally assumed the absence of artificially introduced base substitutions in a significant proportion of the amplification product. This technique, however, differs from the direct sequencing of amplified DNA in that non-specific substitutions will render a molecule useless in analysis. The expected signal-to-noise ratio is heavily influenced by several parameters viz. initial template copy number, number of replication cycles, eventual product yield and the type of experimental system adopted. Mathematical modelling can be used to optimize fragment length with respect to the method applied and suggests as yet undescribed improvements such as partial modification or cleavage to optimize signal detection.