In order to understand natural sea-level variability, and to enhance future predictions, accurate and precise estimates of Holocene tidal levels are required. Although the application of diatom-based transfer functions can yield such data, these work best when underpinned by local training sets. Urbanized estuaries offer little prospect of obtaining local training sets and, instead, the reliability of regional transfer functions has to be assessed. The performance of a published regional (UK) diatom-based tidal-level transfer function applied to fossil assemblages from two contrasting core sites in the Mersey Estuary, UK, is assessed using modern analogue techniques and sediment isotope data. We find that, although estimated tidal levels coincide with changes in organic matter source, the frequent lack of modern analogues mean that palaeotide estimates are without basis. This is likely a consequence of the site-specific nature of diatom assemblages in higher intertidal and supratidal areas in particular, where local factors are expected to exert a greater control on their ecology. This situation may be partly resolved by constructing and applying much larger regional training sets from multiple higher intertidal and supratidal sites (where intact). Otherwise the application of alternative techniques, such as regional foraminiferal tidal-level transfer functions, may be more appropriate.