In order to ensure that patterns detected in field samples reflect real ecological processes rather than methodological idiosyncrasies, it is important that researchers attempt to understand the consequences of the sampling and analytical designs that they select. This is especially true for sandy beach ecology, which has lagged somewhat behind ecological studies of other intertidal habitats. This paper investigates the performance of routine estimators of macrofaunal abundance and species richness, which are variables that have been widely used to infer predictable patterns of biodiversity across a gradient of beach types. To do this, a total of six shore-normal strip transects were sampled on three exposed, oceanic sandy beaches in the Eastern Cape, South Africa. These transects comprised contiguous quadrats arranged linearly between the spring high and low water marks. Using simple Monte Carlo simulation techniques, data collected from the strip transects were used to assess the accuracy of parameter estimates from different sampling strategies relative to their true values (macrofaunal abundance ranged 595-1369 individuals transect(-1); species richness ranged 12-21 species transect(-1)). Results indicated that estimates from the various transect methods performed in a similar manner both within beaches and among beaches. Estimates for macrofaunal abundance tended to be negatively biased, especially at levels of sampling effort most commonly reported in the literature, and accuracy decreased with decreasing sampling effort. By the same token, estimates for species richness were always negatively biased and were also characterised by low precision. Furthermore, triplicate transects comprising a sampled area in the region of 4 m 2 (as has been previously recommended) are expected to miss more than 30% of the species that occur on the transect. Surprisingly, for both macrofaunal abundance and species richness, estimates based on data from transects sampling quadrats at 3 m intervals across the intertidal outperformed those based on data from transects sampled at either 2 or 4 m intervals, instead being more similar to those from transects sampled at I m intervals (the highest level of sampling effort considered). This suggests that across-shore patches of organisms might be distributed in response to some process acting at spatial scales close to 3 m. Together these results provide a compelling argument against combining data from surveys employing different sampling regimes. They also indicate a need for increased sampling effort, improved estimation techniques or both. Considering the limitations imposed on sandy beach ecology, improved estimation seems most reasonable, with emerging techniques such as geostatistics, extrapolative estimates of species richness and general additive models representing areas of potential progress. (C) 2003 Elsevier Ltd. All rights reserved.