Fig. 2

a–f Comparative End Boundary distributions generated from single-phase MERLF radiocarbon data associated with a recent study of Achanduin Castle [49] highlighting the effect of different outlier models on standalone and multidisciplinary constructional estimate precision. In standalone model 1a, an exponential prior was applied, and all three available radiocarbon determinations were tagged with a 5% Outlier Probability in the default OxCal General Outlier Model (subfigure a); in standalone model 2a, a uniform prior was applied, and all three available radiocarbon determinations were tagged with a 100% Outlier Probability in a Charcoal Outlier Model with a time-constant modified to 100 years (subfigure c); and in standalone model 3a, a uniform prior was applied, and all three radiocarbon determinations were tagged with a 100% Outlier Probability in the default Charcoal Outlier Model (subfigure e). Multidisciplinary models 1b, 2b and 3b are a development from 1a, 2a and 3a, and each includes a 1310 AD documentary TAQ (subfigures b, d and f). All models have been updated using OxCal v4.4 [51] and the IntCal20 atmospheric curve [53], with results rounded out to 5 years. X axis scales in these plots have been standardised to facilitate visual comparison, but note also the different y-axis probability density scales, as well as the contrasting End Boundary distributions and medians