Over the years, the effects of scale-up on the performance of pulverised-coal burners has been the subject of much controversy. This is due in part to the fact that not all physical and thermo-chemical processes scale in the same way when the scale of the burner is changed, even if the geometric similarity of burners and the momentum ratios between fuel and air streams are maintained constant. The problem is made even more complex by differences in flame surface-to-volume ratios; by scale effects on the turbulent mixing process; and by differences in two-phase interactions between coal particles and the main flow field. The latter has an influence on particle distribution within the flame, and consequently it affects the location and environment of devolatilisation; it also affects ignition behaviour, char burnout within the volatile flame envelope, and consequently the entire thermochemical and aerodynamic structure of the flame. For reasons of economics or ease of in-flame diagnostics and flexibility of operation, burner developers tend to evaluate smaller-scale versions of the burners than will ultimately be applied. This paper describes the results of a series of experiments where a 12 MW(t) large-scale burner is compared with two 2.5 MW(t) geometrically similar burners designed around a constant-velocity and constant-residence-time scaling criterion. High- and low-NOx flames were produced from burners and were extensively probed to determine in-flame thermo-chemical structures. Results show that complex differences exist between flames. The results detailed in the text are interpreted with reference to the effect of scale on turbulent mixing and two-phase interactions; they confirm that, regardless of the scaling criterion selected, similarity of the in-flame thermochemical features is almost impossible to maintain as burner scale is changed.