AbstractAimThe well‐proven positive correlation between the increased proportion of entire‐leaved woody dicotyledonous species and increased mean annual temperature has been commonly used to estimate temperature in the past. However, in regions of colder climates, this relationship is not straightforward, questioning the accuracy of temperature estimation.LocationFinland, Poland, Germany.TaxonDicotyledons.MethodsThe floristic composition of 10 × 10 km squares in 20 km wide transects through Finland, Poland and Germany was analysed.ResultsAt higher temperatures, deciduous woody plants appeared to show the expected positive relationship between mean annual temperature and the proportion of entire leaf margins. However, we found a negative correlation within woody deciduous plants at higher latitudes with mean annual temperature values from approximately −2.5°C to +2–4°C and at all temperatures when all woody plants were included. Herbaceous species showed a weak relationship between morphology and temperature.Main ConclusionsThe hypothesis that the phenomenon was caused by a large percentage of entire‐leaved evergreen species that winter under snow cover was rejected. These results indicate that using the leaf margin analysis method for past temperature estimation is increasingly inaccurate at colder temperatures. Consequently, we recommend avoiding this method at locations where the mean annual temperature falls below 5°C.