The hygroscopicity parameter κ is important for assessing biomass-burning aerosol impacts on visibility, direct radiative forcing, and cloud-related indirect effects. Discrepancies among κ values derived from different methods have been reported, but often separately across fuel types, burn phases, and experimental systems, limiting assessment of their consistency and controlling factors. Here, we compare κ derived from a cloud condensation nuclei counter (CCNC), a hygroscopicity tandem differential mobility analyzer (HTDMA), and AMS–SP2 composition-based predictions within a single laboratory framework. Fresh and aged biomass-burning particles were produced from four representative fuels: hardwood, softwood, peat, and leaves, burned under  smouldering or flaming conditions. CCNC-derived κ systematically exceeded HTDMA-derived κ, with discrepancy magnitudes strongly dependent on fuel type and burn phase. Composition-based κ generally fell between CCNC and HTDMA values. Coupled effects of κ size dependence and size representation, together with externally mixed non-hygroscopic black carbon, affected discrepancies among the three approaches but only partly explained them. Using O:C-informed κ_org values instead of a constant κ_org = 0.1 consistently improved AMS–SP2 prediction accuracy relative to HTDMA values in peat experiments only, by 16–59 %, while substantial residual discrepancies remained. These results suggest the presence of additional controlling factors not tested here, motivating measurements of surface-active species, particle phase state, and co-condensation of soluble gases to potentially explain the remaining discrepancies in future work. They further highlight the need for regime-aware κ parameterizations in atmospheric and climate models, alongside careful treatment of size representation, black-carbon inclusion, and κ_org assumptions in composition-based hygroscopicity predictions.