A shipboard analytical intercomparison of dissolved (<0.2 μm) iron in the surface waters of the Atlantic Ocean was undertaken during October 2000. A single underway surface (1–2 m) seawater sampling and filtration protocol was used, in order to minimise differences from possible sample contamination. Over 200 samples (1/h) were collected over 12 days and analysed immediately using four different analytical methods, based on three variants of flow injection with luminol chemiluminescence (FI–CL) and cathodic stripping voltammetry (CSV). Dissolved iron concentrations varied between 0.02 and 1.61 nM during the intercomparison. On average, CSV [Electroanalysis 12 (2000) 565] measured 0.08 nM higher iron concentrations than one FI–CL method [Anal. Chim. Acta 361 (1998) 189], which measured 0.13 nM higher iron values than the other two [Anal. Chem. 65 (1993) 1524; Anal. Chim. Acta 377 (1998) 113]. Statistical analyses (paired two-tailed t-test) showed that each analytical method gave significantly different dissolved iron concentrations at the 95% confidence interval. These data however, represent a significant improvement over earlier intercomparison exercises for iron. The data have been evaluated with respect to accuracy and overall inter-laboratory replicate precision, which was generally better than the 95% confidence intervals reported for the NASS Certified Reference Materials. Systematic differences between analytical methods were probably due to the extraction of different physico-chemical forms of iron during preconcentration, either on the micro-column resin (in the FI methods) or with competing ligand equilibration (in the CSV method). Small systematic concentration differences may also have resulted from protocols used for quantification of the analytical blank and instrument calibration.