Results of a fractional creep-rupture experiment using seam specimens fabricated from an ethylene-propylene-diene terpolymer (EPDM) roofing membrane material and a butyl-based adhesive are reported. Three variables were included: adhesive thickness (thin and thick), EPDM-surface cleanness (well-cleaned and particulate-contaminated), and test specimen configuration (peel and shear). Only a few of the shear specimens failed and the failures were for specimens having thin adhesive and contaminated rubber. In contrast, the majority of the peel specimens failed. The ranking, in decreasing order of the median creep-rupture lifetime of the peel specimens, was: thick, clean > thick, contaminated > thin, clean > thin, contaminated. The effect of adhesive thickness on lifetime was greater than the effect of surface cleanness. When the early-failure behavior of the peel specimens was examined, somewhat different results were found. The thick, clean specimens displayed significantly longer times to early failure, by a factor of at least 100, than those fabricated using the other three combinations of pplication conditions. However, the predicted early times-to-failure for the sets of specimens fabricated using the other three application conditions were similar. The results emphasize that, for the system investigated, proper application of seams using solvent-based adhesives requires both thick adhesive layers and clean rubber surfaces. It is recommended that creep-rupture testing, particularly in peel, be an integral part of any methodology that evaluates the performance of seams.