means of the lognormal distributions used to generate the synthetic samples (Equation 6) but only about 1.5 percent less than the estimate, 565 nCi/g, based on the fistulated steer data (Smith, this volume). DISCUSSION The results described above indicate that the simulation model, Equation l, does a fair job of duplicating the results of the fistulated steer study (D. D. Smith, this volume). Considering the large error term indicated by simulation, 557 + 526 nCi/day, the differences among various estimates of I, are negligible; but both the arithmetic (557 nCi/day) and the geometric (571 pCi/day) estimates based on simulation are closer to the estimate of I based on fistulated steer data (565 nCi/day, Smith et al., 1976) than were previous independent estimates based on theory (585 nCi/day, Martin and Bloom, 1977) or assumed dietary composition (620 nCi/day, Gilbert et al., 1977). When an earlier but quite similar version of this paper was presented at the San Diego meeting of NAEG, two questions were raised which merit consideration here. One had to do with the “apparently inevitable outcome" of the simulation "given the manner in which it was carried out.'' The second was an expression of skepticism concerning the evidence cited for assuming that all factors of the simulation model, Equation 1, are lognormally distributed. These questions are discussed below. Except for the parameter D in Equation 2, which was adjusted to meet the apparent requirement that a 410-kg cow should be able to obtain enough digestible energy for maintenance on a daily ration of about 6 kg of vegetation, the simulation model and the fistulated steer study were independent. Because of the "adjustment" of D, a critical parameter, the two estimates of I, have been characterized as “almost independent." The only development required to make them truly independent is an independent and site-specific method of estimating D, the digestibility of vegetation available to grazing animals. If the simulation model yields estimates of I, which are essentially the same as estimates based on the rumen conténts of fistulated steers allowed to graze a fenced area (and it does), this outcome can be charac- terized as "inevitable" if, and only if, the model correctly simulates ' the grazing process. This means that the assumption incorporated in the model design and the input data used to implement the simulation must be essentially correct. The principal assumptions incorporated in the simulation model are (1) that grazing is a random process which, given sufficient time, results in the ingestion of a composite random sample of vegetation and soil and (2) that the factors of the simulation model, Equation 1, can be represented as independent random variables having lognormal distributions. The best evidence that these assumptions and 301