Calibration data sets give a unique opportunity to establish patterns of biological existence and their statistical associations with environmental variables. By use of calibration data sets, environmental variables can be inferred quantitatively. The resulting long time-series may assist in distinguishing natural environmental variability from human-induced variability, both in terms of climate change and biotic turnover. However, the validity of the palaeoenvironmental reconstructions depends on their accuracy, precision and sensibility. Before performing palaeoenvironmental inferences, key mechanisms controlling contemporary species’ distribution, abundances and dynamics should be identified and understood. An inference model is developed to produce reconstructions. A major challenge lies in validating and interpreting the reconstructions. Calibration data sets involving midges (Diptera: Chironomidae) suggest that climate has a broad-scale, regional control over midge existence and abundance, often over-riding the influence of local within-lake variables. In recent years, the use of midges as quantitative indicators of past temperatures has greatly expanded. As the number of reconstructions increase, especially in Fennoscandia and North America, it seems the among-site variability is so large that it is unlikely to be due only to local differences in climate. Hence, we question whether the long climate gradients in calibration data sets can accurately be used to calibrate local variables, when most local gradients in time and space are short. Ten Holocene chironomid-inferred temperature curves from Fennoscandia are compared. We illustrate some general principles in palaeoecology by identifying factors that may cause bias. Especially, we consider how calibration data sets simplify the complexity of the real world by maximizing single ecological gradients and by not taking into account co-varying variables. We give some recommendations and criteria that chironomid analysis should meet in order to improve the reliability of the temperature inferences. Finally, we discuss how the complex interactions between species and environment may have implications when we aim at predicting future biodiversity.