The present investigation presents a discussion of two methods used to define a datum line with respect to which the error of straightness is calculated, namely the least squares line method and the end points line method. The paper investigates uncertainties of both methods. Adopting statistical accuracy, the analysis of experimental data using the two methods demonstrated that both methods give almost the same value of error of straightness of the surface under test with the same uncertainty. Nevertheless, for practical application the end point line method is more favourable for not only does it offer a suitable mathematically defined datum but also it is simple, straightforward and independent of personal judgement.