El Niño–Southern Oscillation (ENSO) is one of the most important coupled ocean–atmospheric phenomena to cause global climate variability on interannual timescales. Efforts to understand recent, apparently anomalous ENSO behaviour are hampered by the phenomenon's unstable (non-stationary) nature and the limitations inherent in palaeoclimate records. In this paper, the complexities associated with isolating ENSO signals in observational and palaeoclimate records are reviewed. The utility and limitations of high-resolution (tree-ring, coral, speleothems, ice and documentary) proxy data for ENSO reconstruction are discussed. To overcome the regional biases contained within each palaeoclimatic source, it is necessary to compare complementary signals derived from multiple proxy climate records. To date, there have been limited attempts to reconstruct large-scale ENSO using these ‘multiproxy’ methodologies. A critique of the complexities associated with previous approaches of reconstructing discrete ENSO events and atmospheric/oceanic indices is provided. Abundant potential remains to better characterise teleconnection patterns, propagation signatures and non-stationary features of large-scale ENSO behaviour. If key uncertainties in ENSO dynamics (such as the response of extreme events to natural/human forcing) are to be adequately assessed, then complementary attempts must be made to model the historic synoptic conditions with apparent changes in reconstructed indices. Copyright © 2006 John Wiley & Sons, Ltd.