1. Data from twenty-four short-term trials on the nutrition of dairy heifers have been used in a study of the effects of management, method of estimation of live-weight gain, length of experimental period, covariance analysis and the use of monozygous twins upon experimental error.2. The value of rigid adherence to a routine and allowing ample time for the animals to settle down before the trial have been demonstrated.3. Error standard deviations were associated with length of experimental period, those for 42-day periods being about half those per 21-day periods, but were not appreciably correlated with initial live weight.4. No advantage could be demonstrated for the use of monozygous twins, or covariance analysis with initial weight or rate of gain, indicating that genetic effects contributed little to the error in experiments lasting 3–6 weeks.5. There was no significant difference between errors in live-weight gains calculated from the regression of live weight on time, from single initial and final weights, or from two initial and final live weights, indicating that deviations from linear regression are not solely in the nature of random day to day fluctuations in live weight.