A study was performed to investigate the reproducibility of haemagglutinin-inhibition (HI) and virus neutralising (VN) assays for detection of anti-influenza antibody. Participants in 11 laboratories from eight countries measured antibody to egg-grown A/Japan/434/2003, cell-grown A/Japan/434/2003 and A/Panama/2007/99 (H3N2) viruses in 18 human and two post-infection ferret sera. There was significant intra-laboratory assay variability for VN compared to HI. For replicate assays within laboratories, 14/410 (3%) and 130/631 (21%) titres differed by >2-fold (p < 0.0001), and 0/410 (0%) and 35/631 (6%) titres differed by >5-fold (p < 0.0001) by HI and VN, respectively. Although both assays showed inter-laboratory variation, VN assays were significantly more variable than HI. Median geometric coefficients of variation (GCV) for VN assays with each virus were 256%, 323% and 359% compared to 138%, 155% and 261% with HI. A serum standard improved inter-laboratory agreement and reduced median GCVs. This study raises concern about comparability of serology results from H5N1 vaccine trials and it is proposed that an International Standard for influenza H5N1 antibody is developed. Crown Copyright © 2007.