Sieving samples for chironomid analysis with a 150 lm mesh was shown to greatly reduce sample preparation time, and use of only larger specimens did not affect chironomid-inferred salinities in African lakes. Here, we tested if this method is suitable for temperature reconstruction in colder lakes at higher latitudes. Removal of specimens\150 lm in two training sets, one from Canada and one from Sweden, had little impact on the performance statistics of the calibration models. Chironomid abundance, however, decreased greatly because more than half of the head capsules in assemblages were\150 lm. This had major impacts on the temperature reconstructions. Inferences were on average 2°C warmer with the modified models (all specimens [150 lm) than those obtained with the full model (all specimens[100 lm). General patterns of temperature change were also altered. For Lake 7 on Southampton Island, Canada, a cooling trend was reconstructed with the full Canadian model while the modified Canadian model yielded a warming trend. When only specimens[150 lm were used, two to three times more wet sediment was needed to obtain a sufficient number of head capsules. These results indicate that, in cold lakes (mean July/ August air temperature B11°C), large proportions of head capsules are \150 lm, and sieving the samples in a 150 lm mesh leads to altered temperature reconstructions. Keywords Chironomid-inferred temperatures Á Alternative sampling methods Á Temperature reconstruction Á Cold lakes