A new theoretical framework breaks down the overdispersion observed in luminescence dates from single grains and aliquots, proposing a generative variance budget that separates the different sources of uncertainty. The approach, presented in a paper titled «Deconstructing overdispersion: a generative variance budget for luminescence dating from single grains to aliquots», aims to explain why measured dispersions often far exceed estimates based on conventional experimental errors.

Luminescence dating —used in archaeology and geology to date sediments and artefacts from the radiation dose accumulated in minerals such as quartz and feldspar— has long grappled with overdispersion: the real variability between grains or between aliquots is systematically larger than expected from known error sources. This discrepancy directly affects the precision of calculated ages and the interpretation of sedimentary contexts.

The proposed budget treats the total variance as a sum of terms attributable to distinct processes —intrinsic grain-to-grain variability, heterogeneity in the environmental dose, microdosimetric effects and experimental contributions— so that each source can be identified and quantified separately. Moving from single grains to aliquots, the model also allows researchers to track how these contributions combine and average, which is key to correctly interpreting laboratory results.

The framework's usefulness lies in turning an ad hoc fitting parameter —overdispersion— into a quantity with physical meaning, making it easier to compare measurement protocols, assess sample quality and fit dose models more realistically. The work paves the way for applying the same budget to other luminescence dating methods and for revisiting previous dates with more transparent uncertainty estimates.