An automated method was developed for the quantification of dislocations in natural fibres, and the method was tested successfully on spruce pulps and hemp fibres. The method employs polarized light microscopy and image analysis based on two images per fibre. One image is optimised for the identification of the fibre edges, the other for the identification of dislocations. Identification of the fibre implies automatic threshold determination followed by correction of obvious errors and smoothing of the fibre edges. Identification of the dislocations implies differentiation of the dislocation image followed by a step where dislocations are separated from the background and finally smoothing of the dislocations using a series of erosion and dilation steps. From these values the relative dislocation area may be calculated, i.e. the area of the dislocations in percent of the area of the fibre. For this parameter, the new method may give large errors for single fibres, but since the errors are not systematic, average values for batches of fibres are estimated with reasonable accuracy. For two different spruce pulp batches the estimated mean was respectively 14.9 % compared to the manually determined value of 14.3 % (52 fibres) and 17.7 % compared to 16.4 % (53 fibres). For a hemp batch comprising 23 fibres the estimated mean was 21.6 % compared to the manually determined value of 20.9 %. For mill and laboratory pulp fibres produced from the same batch of spruce wood, the amount of dislocations was analysed both according to the Thygesen/Ander method and a previous manual method (Ander/Marklund). The results of the two methods are not directly comparable, since the Ander/Marklund method gives an estimate of the number of dislocations per mm fibre. The Ander/Marklund method did not show any significant difference between the two different spruce pulps in contrast to the Thygesen/Ander method, which surprisingly showed that the mill pulp fibres on average had a lower relative dislocation area than the laboratory pulp. Possible reasons for this result are discussed.