Geochemical soil samples were taken along two traverses over shear-hosted gold mineralisation in Area '223' at the Tunkillia gold prospect, located 550 km north-west of Adelaide. The objective was to assess the effectiveness of near-surface soil sampling for locating mineralisation in areas where the bedrock is deeply weathered and is covered by an aeolian dune field. One sampling traverse lay along the crest of a longitudinal dune that cross-cuts the mineralised zone, while the other lay along the adjacent swale. Samples were collected from 5-25 cm depth, sieved to <200 microns, and gold and other trace element concentrations were determined on partial extracts obtained by leaching using sodium pyrophosphate with cyanide. The partial leach soil analytical preparatory process was chosen so as to increase the proportion of loosely bound ions detectable in the sample that can potentially reflect the presence of underlying mineralisation, as compared to using a process that digests the entire mineral grain, thus acting to dilute the effect of such ions present in the sample. The specific leach chemistry was chosen based on previous work done at this location (Fabris et al., 2009). The cyanide is an effective extractant for gold and silver, and sodium pyrophosphate is regarded as effective at liberating metal ions from organic matter, plus any associated with aluminium oxides and some clays. A hand-held X-ray fluorescence analyser was used to determine iron, calcium and its proxy strontium. Gold assay results for both of the traverses had their highest values over the mineralised zone. Samples from the dune traverse, however, yielded significantly reduced levels of extractable gold, and this produced an Au anomaly-to-background ratio that was much less than for the samples collected along the swale traverse. It is concluded that near-surface soil sampling and partial leach analysis could potentially be a viable and cost-effective geochemical exploration strategy to utilise in the region, provided that regolith, landform and vegetation factors are taken into consideration in determining what are the appropriate background values against which individual sample data will be evaluated.