The development of weighted averaging models for diatoms and lake-water temperature is reviewed in relation to other factors controlling diatom abundance in lakes. It is argued that diatom–temperature models are weaker than those developed for salinity, pH and phosphorus. Moreover, even where a statistically significant model is developed for water temperature, its application over long time periods requires caution. For example, over long time periods (such as the Holocene), soil and vegetation will have changed considerably, altering lake-catchment hydrochemistry and hence the lake biota. Applications of diatom-temperature models in these instances may result in inferred temperatures that are merely artefacts of changing lake-water chemistry. However, palaeolimnological studies using variance partitioning analysis of diatom profiles in sediment cores and independent climate data show that climate signals are recorded in lake sediments. It is suggested, therefore, that the link between changing climate and diatoms be addressed via an alternative method than that of temperature-inference models. The situation demands a more complete understanding of the interactions between lake hydrodynamics and nutrient inputs from catchments over a variety of timescales.