Elephants confined to protected areas could affect co-occurring species. We expected measurable deviations in the density, species number, turnover and composition, abundance-incidence and rank-abundance patterns of woody plant communities where elephants browse. We examined these variables in the presence (inside South Africa's Tembe Elephant Park) and absence (uninhabited communal land adjacent to the park) of elephants for sand forests, closed and open woodlands. Woodland type influenced the apparent effect of elephants on tree densities and number of species. In sand forests, the presence of elephants was not associated with lowered tree densities but this did not hold for mixed woodlands. The rates of species turnover in the park were consistently higher than on communal land for all three woodland types. Species composition also differed significantly between these areas, as did the number of favoured food trees of elephants. Elephants had a clear influence on vegetation at the species level. However, at the community level they had no apparent affect on woodland-specific abundance-incidences and rank-abundance relationships. Elephants were confined to Tembe for only some 20 years before this study and still occur at relatively low densities. This may explain why we could not detect an impact on the bivariate relationships.