The Environmental Protection Agency evaluates the risk of granular pesticides to birds using an LD50s/ft2-risk index. The index does not include avian response to granular characteristics, pesticide load per granule, or weather conditions, and it assumes that avian risk increases as granule availability increases. Fensulfothion was formulated on granular pesticide carriers to evaluate the effects of carrier type (silica, clay, and corncob), pesticide load per granule (1/2, 1/8, and 1/32 of the LD50), granule color (blue, white, and uncolored), and granule size (large and small) on captive house sparrows (Passer domesticus). In the initial experiments, silica and clay granules posed a greater risk to the birds than did corncob granules, and the risk from the granular pesticide decreased as the pesticide load per granule decreased. Findings regarding granular size and color were inconclusive. Carrier type was further evaluated by offering formulations on the three granular carriers to house sparrows that were given less food than the birds in the first evaluations. In this instance, corncob granules posed a greater risk to the birds than did clay or silica granules. When house sparrows were offered silica, clay, and corncob granules (formulated with fensulfothion) on dry or wet soil, the granular carriers on dry soil posed the greater risk to birds. Wet soil significantly reduced the risk to the birds from corncob and silica granules, whereas the risk from clay granules was not significantly reduced on wet soil. Captive house sparrows were offered silica granules formulated with fensulfothion at 1/2 the LD50 per granule at application rates of 50, 100, 200, 400, 800, 1200, 1600, or 2000 LD50s/ft2. A significant negative correlation between brain ChE activity and the natural logarithm of application rate indicated that the relationship between avian risk and granule availability was asymptotic rather than linear. The information now available regarding avian response to pesticide granules as a source of grit should be included in risk assessment models to increase the accuracy of predicting avian exposure. This information also could be used to design granular pesticide formulations less hazardous to birds.