The nonsynonymous synonymous substitution rate ratio (dN/dS) is commonly used to detect selection in coding sequences at both global (entire genes) and local (codon position) levels. Among a large variety of organisms, this estimation is frequently performed from data sets where recombination has likely occurred. However, several studies have shown that recombination may bias selection estimates at local level, in particular by increasing the number of false positively selected sites. Here we describe this bias and provide alternative strategies to estimate dN/dS in the presence of recombination. Several practical examples are also included to illustrate this procedure.