The molecular signature of selection depends strongly on whether new mutations are immediately favorable and sweep to fixation (hard sweeps) as opposed to when selection acts on segregating variation (soft sweeps). The prediction of reduced sequence variation around selected polymorphisms is much stronger for hard than soft sweeps, particularly when considering quantitative traits where sweeps are likely to be incomplete. Here, we directly investigate the genomic signal of soft sweeps within an artificial selection experiment on Mimulus guttatus. We first develop a statistical method based on Fisher’s angular transformation of allele frequencies to identify selected loci. Application of this method identifies about 400 significant windows, but no fixed differences between phenotypically divergent populations. With two notable exceptions, we find a modest average effect of partial sweeps on the amount of molecular variation. The first exception is a polymorphic inversion on chromosome 6. The increase of the derived haplotype has a broad genomic effect due to recombination suppression coupled with substantial initial haplotype structure within the population. Second, we found significant increases in nucleotide variation around selected loci in the population evolving larger flowers. This suggests that “high” alleles for flower size were initially less frequent than “low” alleles. This result is consistent with prior studies of M. guttatus and illustrates how molecular evolution can depend on the allele frequency spectrum at quantitative trait loci.