Using the tibial nerves of diabetic rats, alternative sampling schemes have been compared for estimating the sizes of fibres in nerve trunks of mixed fascicularity. The merits of each scheme were evaluated by comparing their reliability, precision, cost in time, and efficiency with 'absolute' values obtained by first measuring every fibre. The external diameter of all myelinated fibres was measured in each of six nerves (c. 2900 fibres/nerve). Total measurement time was about 29 hours. All sampling schemes produced group means within +/-4% of the absolute value of 5.52 micron. The most efficient schemes were those in which only 6% of all fibres were selected for measurement. For these the measurement time was 2 hours or less. Results are discussed in the general context of measurement of the sizes of nerve fibres. It is concluded that future studies should place more emphasis on sampling fewer fibres from more animals rather than on measuring all fibres very precisely. These considerations are likely to be of special concern to those wanting to analyse specimens with large fibre complements and those screening large numbers of specimens.