With the use of a generalisation of the theory of single-stage line intersect sampling, estimators for population total per unit area and variance are derived for the case of line intersect three-stage sampling. In this case line intersect sampling is followed by two further stages of sampling: a sample from the intersecting first-order elements and a sample from the second-order elements that constitute the latter. Estimators for two types of two-stage sampling follow as special cases. The techniques developed may reduce costs or make sampling feasible in cases where single-stage line intersect sampling is impractical. Forest Sci. 20:129-133.