Marker methods are needed for estimating fecal output by grazing animals in studies of nutrition and environmental impact. In addition, estimates of prefecal mass and turnover time are relevant to exercise performance and certain digestive disorders. As a first step in developing marker methods for field use, a chromic oxide model of fecal kinetics was developed and tested in the context of a digestion balance experiment with stall-fed horses. The model consists of removal of feces at a constant rate from a single compartment, the prefecal mass. Four horses were fed hay, and another four were fed hay and concentrate. Balance-marker experiments were conducted for 10 d, following 7 d of adaptation. A dose of chromic oxide mixed in chopped hay and molasses was administered from a nose bag at 0700 daily for 10 d. Dry matter and Cr were measured in feeds and feces. Fecal Cr concentration (C, mg/kg DM) varied during the day, so data from total daily collections were used for model development. These fecal Cr data (Ct) at time t (days) were fitted to a single exponential, with one rate constant (k), rising to an asymptote (Ca): Ct = Ca - Ca x e(-kt). Superior fits were obtained when a delay (d) was introduced between the pulse oral dose and the entry of marker into the prefecal pool: Ct = Ca - Ca x e(-k(t-d)). Using pooled data, delays of 2.7 and 2.0 h gave best fits (highest estimates of R2) for pooled data from horses fed hay or hay plus concentrate, respectively. The model generated estimates of 3.4 and 3.8 kg/d of DM for fecal outputs (dose/Ca) of horses fed hay or hay and concentrate, respectively. The rate constants yielded turnover times (1/k) of 33 and 18 h, and prefecal masses [(dose/Ca)/k] of 4.6 and 2.9 kg of DM for hay or hay and concentrate groups, respectively. Using data from individual horses, mean estimates for each diet were similar to corresponding values for the pooled data. In balance experiments, feces collected were 3.7 and 4.4 kg/d, and Cr recoveries were 108 and 115% dose for the hay and hay plus concentrate diets, respectively. Marker estimates (M) were correlated with total collection estimates (T) of fecal output [M = T(.890 +/- .045); r = .70, P = .041]. Adjusting for recovery improved the regression coefficient to 1.009 +/- .028 (r = .87, P = .002). The findings suggest that if Cr doses are more frequent than daily and if Cr inputs other than dose can be eliminated this method should give accurate and precise estimates of fecal output.