Results presented in this paper demonstrate the relationship between two data sets from Mubwindi Swamp, south-west Uganda. The first data set is from a survey of dryland montane forest within the immediate catchment and the second is in the form of pollen spectra recovered from surface samples of sediment. Numerical analysis grouped pollen types encountered into five categories, according to their level of representivity when compared with abundances of the same taxa in the sampled vegetation: under representative, moderately under representative, representative, moderately over representative, and over representative. The northern part of Mubwindi Swamp also yielded a 5-m-long sequence of organic-rich sediments (core MB6) from which sub-fossil pollen and radiocarbon data have been previously obtained. This provides an opportunity to apply calibration factors derived from modern pollen–vegetation relationships to sub-fossil pollen data that were obtained from the same site. These calibration factors effectively down-weighted the influence of over representative pollen types and permitted the discrimination of a phase of forest disturbance that had not previously been so clearly visible in sub- fossil pollen data.