Size effect on bending strength and failure modes of finger-jointed timber

• Bayesian analysis of 1100 QC bending tests on finger-jointed boards, 20-40 mm thick. • Thinner boards showed higher strength, 43.7 MPa at 20 mm and 38.3 MPa at 40 mm. • Width showed no credible effect on strength, with slope about 0.02 MPa per mm. • Dirichlet regression revealed outside-joint failure rises with thickness. • Uniform characteristic strength is conservative for 20-40 mm range with 40% margin. Finger-jointed timber boards for cross-laminated timber production are typically assigned the same characteristic bending strength ( f m,j,k ) if produced from the same strength class, regardless of differences in their cross-sectional dimensions. To validate the relevance of this approach, this study investigated the effects of cross-sectional dimensions on the bending performance of finger-jointed timber produced from spruce (strength class: C24). A large industrial dataset of 1100 specimens, with seven thicknesses (ranging 20 to 40 mm) and variable widths, were statistically analyzed. The bending tests were performed with a constant span-to-depth ratio ( l / h = 18), meaning thinner specimens had a shorter test span. Bending strength was modeled with Bayesian multilevel linear model, and the proportions of three quality-control failure modes (joint interface failure, joint base failure, and outside-joint failure) were analyzed with a zero-one-inflated Dirichlet regression. Based on the results, all groups with n ≥ 100 exceeded the declared f m,j,k of 27.6 MPa, with characteristic strengths of 43.7 MPa (+58.3%), 40.0 MPa (+45.0%), and 38.3 MPa (+38.8%) for the 20-, 30-, and 40-mm thickness groups, respectively. Thinner specimens demonstrated higher bending strength with convincing evidence (pairwise contrasts PD = 100% with 95% HDPIs entirely below zero), while width had no credible effect (PD < 95%). Dirichlet regression revealed shifts in failure mode proportions with varying strength. Higher bending strengths were associated with a higher proportion of joint interface failure. Outside‑joint failure was observed with a higher proportion in lower-strength and thicker specimens. Overall, assigning uniform f m,j,k to various cross-sectional dimensions proved to give adequate safety margins. Beyond the uniform f m,j,k , however, statistical evidence of a size effect on both bending strength and failure modes was observed. The magnitude of the observed size effect reflects the combined influence of increasing thickness and test span under the current quality control bending test regime, which means a coupled change in stressed volume and geometry rather than a pure cross-section scaling effect. These findings are relevant to flatwise four-point bending tests on finger-jointed boards from industrial production made with visually graded C24 spruce with thickness ranging 20 to 40 mm.