Abstract According to current knowledge, the non-proportionality of the light yield of scintillators appears to be a fundamental limitation of energy resolution. A good energy resolution is of great importance for most applications of scintillation detectors. Thus, its limitations are discussed below; which arise from the non-proportional response of scintillators to gamma rays and electrons, being of crucial importance to the intrinsic energy resolution of crystals. The important influence of Landau fluctuations and the scattering of secondary electrons (δ-rays) on intrinsic resolution is pointed out here. The study on undoped NaI and CsI at liquid nitrogen temperature with a light readout by avalanche photodiodes strongly suggests that the non-proportionality of many crystals is not their intrinsic property and may be improved by selective co-doping. Finally, several observations that have been collected in the last 15 years on the influence of the slow components of light pulses on energy resolution suggest that more complex processes are taking place in the scintillators. This was observed with CsI(Tl), CsI(Na), ZnSe(Te), and undoped NaI at liquid nitrogen temperature and, finally, for NaI(Tl) at temperatures reduced below 0 °C. A common conclusion of these observations is that the highest energy resolution, and particularly intrinsic resolution measured with the scintillators, characterized by two or more components of the light pulse decay, is obtainable when the spectrometry equipment integrates the whole light of the components. In contrast, the slow components observed in many other crystals degrade the intrinsic resolution. In the limiting case, afterglow could also be considered as a very slow component that spoils the energy resolution. The aim of this work is to summarize all of the above observations by looking for their origin.