Mercury is a potent neurotoxin with which food and beverages may be contaminated from a number of sources, both natural and anthropogenic. The determination of mercury at concentrations close to instrumental detection limits suffers from problems related to memory effects and loss, both during sample preparation and within sample introduction systems. L-cysteine (1%) was added to rice samples, standards, and rinse solutions in order to keep the mercury in solution and decrease the memory effect. Gold (1 μg L⁻¹) was added online as an internal standard to improve accuracy and precision, while further decreasing the memory effect. A comparison of methods involving microwave digestion or acid extraction showed that both were capable of detecting single-digit μg kg⁻¹ concentrations of mercury in rice. The microwave digestion ICP-MS procedure was further validated by a comparison of results with those obtained with a solid-sampling mercury analyser, based on CV-AAS, for which no significant differences were found. Both instrumental techniques were also validated by recoveries of spikes at various stages of the procedures and by the analysis of rice flour CRMs (NIST SRM 1568a and NIST SRM 1568, containing 5.8 μg kg⁻¹ Hg and 6.0 μg kg⁻¹, respectively). Recoveries between 80 and 120% were obtained and the concentrations measured in the CRM 1568a were not significantly different from the certified value.