Many studies of the elegant architectures of orb webs are conducted in controlled laboratory environments that remove environmental variability. The degree to which spider behavior in these circumstances resembles that of spiders in the wild is largely unknown. We compared web architecture and silk investment of furrowed orb weavers Larinioides cornutus (Clerck 1757) building webs in laboratory cages and spinning webs on fences in the field and found significant differences. The volume of major ampullate silk in radii was 53% lower in cage webs, primarily because the silk was 50% thinner, but also because spiders tended to spin 14% fewer radii than in fence webs. Cage spiders also invested about 40% less flagelliform silk and aggregate glue in the capture spiral, although the difference was not statistically significant, a trend primarily driven by a decrease in the length of the glue-coated capture spiral. These patterns were consistent with spiders reducing silk investment when building at new web sites while they assessed insect abundance. Differences in the type of substrate for web attachment, amount of available space, and condition may also have influenced web architecture. Cage webs were more symmetrical than fence webs, which displayed an unusual horizontal asymmetry that may have maximized their capture areas within the constraints of the available fence-railing attachment sites. Our findings suggest using caution when generalizing the properties of laboratory-spun webs to more natural conditions. More importantly, they demonstrate that orb spiders actively modify their behaviors when spinning webs under different conditions.