The use of surrogates for biodiversity is a practical tool to improve the cost effectiveness of regional conservation planning. However, there is still much uncertainty about the biological representativeness of surrogates. Using a biophysical classification system known as the Mitchell Landscapes, we compare plant species composition in contiguous versus disjunct units of nine Landscape types and hence the ability of this surrogate to capture patterns of plant species composition. We found that plant species homogeneity was higher within a contiguous Landscape than between non-contiguous units of the same Landscape. Overall, the dissimilarity between non-contiguous units and their contiguous counterparts was significant (P=0.004). Biophysical classes with very high dissimilarities between non-contiguous units of the same region may be of limited utility.