1. A mixed sward was divided into two areas receiving high (H) or low (L) levels of fertilizer nitrogen, and on two occasions, in spring and summer, herbage cut from each area was given to three steers in a continuous digestibility trial lasting 45 days. Digestibility and faecal nitrogen concentration were measured for each animal for each 24 hr. period, and equations relating these variables were calculated for each trial from the unit 24 hr. observations and from averages over three animals and over successive 2-, 3-, 5- and 9-day periods.

2. In the two spring trials herbage digestibility remained constant for the first 12 days and then declined by about 0·4 units per day. In the summer trials digestibility declined throughout by 0—25 units per day. The nitrogen content of the herbage was consistently 25% higher for treatment H, and declined steadily throughout each trial.

3. Both season of cutting and fertilizer treatment had a significant effect on the faecal index regressions obtained. Equations calculated from average values for three animals and 3 days predicted digestibility coefficients from 3% faecal nitrogen as follows: equation (1) spring H, 76·5; (2) spring L, 77·5; (3)summer H, 71·8; (4) summer L, 72·8.

4. The joint residual standard deviation for equations (1) to (4) had the low value of ± 0·90 units of digestibility, indicating that ‘single-sward’ regressions are considerably more accurate than those embracing a great variety of herbages. Even greater accuracy was obtained when digestibility and faecal nitrogen concentrations were measured for periods longer than 3 days; in general, residual variance varied inversely with period length. Using values for individual animals in the regressions increased the residual variance slightly, but because differences between animals in the digestibility-faecal nitrogen relationship were small and generally nonsignificant, the increase was by a factor of the order of 1·6, not 3.

5. Standard errors of prediction were calculated, although there is evidence to suggest that these underestimate the error of digestibility coefficients predicted for free-grazing animals. It is concluded that faecal index regressions are accurate enough for measuring digestibility per se in grazing experiments. In the measurement of herbage consumption, however, the experimental error is likely to be high, partly because indigestibility is the factor required to be predicted, but mainly because of large differences in appetite among similarly treated animals.