Although genome-wide association (GWA) studies are not worth the effort in crosses between inbred lines, many crosses are actually made up of divergent yet outbred populations. Despite its relevance, however, this experimental setting has not been studied at a time when SNP microarrays are available in many species. To assess whether GWA can be useful in this setting, we performed combined coalescence--gene dropping simulations. We studied the influence of marker density, QTL effect and QTL allele frequency on power, false discovery rate (FDR) and accuracy. Our results suggest that GWA in outbred F(2) crosses is useful, especially in large populations. Under these circumstances, accuracy increased and FDR decreased as compared with classical linkage analysis. However, current SNP densities (in the order of 30-60 K SNPs/genome or equivalent to 10-20 SNPs per cM) may not be much better than linkage analysis and higher SNP densities may be required. SNP ascertainment had an important effect; the best option was to select SNPs as uniformly as possible without setting any restriction on allele frequency. Using only SNPs with fixed alternative alleles in each breed controlled false positive rate but was not useful to detect variability within lines. Finally, the most significant SNP was not necessarily the closest to the causal SNP, although the closest SNPs were usually above the significance threshold; thus, it is prudent to follow-up significant signals located in regions of interest even if they do not correspond to absolute maxima.