Phytolith analysis is conventionally an archaeo-botanical tool used to study past human activity using material from excavations or soil pits. However, phytolith analysis also has potential as a palaeoecological tool, to reconstruct vegetation changes through periods of climatic change and human influence. To study phytoliths from lake sediment alongside pollen requires an understanding of phytolith taphonomy in lakes. Theoretical models suggest phytoliths represent more local vegetation at smaller spatial scales than pollen from lake sediments, but this has not been tested empirically in the Neotropics. This paper compares pollen and phytolith assemblages from the same lake sediment surface sample, from a suite of lakes of different sizes across different vegetation types of lowland tropical Bolivia. We find three factors driving phytolith composition in lakes: taphonomy, lake size and phytolith productivity. By comparing phytolith assemblages with pollen assemblages, we find that they provide different taxonomic information and generally complement each other as palaeo-vegetation proxies. We also demonstrate empirically that pollen assemblages in lake samples represent a larger catchment area than phytolith assemblages. Our findings suggest that phytoliths can be particularly useful in providing local-scale vegetation histories from large lakes, to complement the regional-scale vegetation histories provided by pollen data.