Ten flexible characters for two collections of a pulmonate land slug that differed in their method of preservation (ethanol or freezer/ethanol), were measured four times - twice with calipers and twice with a stage micrometer - by two different people. Repeated measurements were used to estimate the measurement error (ME) associated with the ten characters. ME ranged from 2\% to 90\% and differed significantly between the characters. Characters with low mean values and a high flexibility showed the largest ME. With the stage micrometer, one measurer obtained significantly higher ME, but both measurers obtained the same ME with calipers. There was no detectable effect of preservation method on the size and shape of characters, yet, results differed when characters were measured with either calipers or a stage micrometer. One measurer obtained significantly larger mean values for three characters with calipers. Additionally, presumed species differences between three Carinarian species (Arion fasciatus, A. silvaticus and A circumscriptus; subgenus Carinarian Hesse 1926) were tested using principal component analysis and canonical discriminant analysis. Despite possible measurer biases and high ME, the different datasets yielded highly similar results, indicating that biometrical data of soft, flexible structures may yield valuable and reliable data which can be examined statistically. Our results indicate that A. fasciatus is larger than the other two species, but it is hard to distinguish from both of the other species when size is not considered. Arion silvaticus and A. circumscriptus can be separated only when colour characters are used. © 2002 The Linnean Society of London.