We examined maintenance effects following training in a self-generated strategy versus a mnemonic strategy, designed to improve recall of four-digit numbers in older adults. This was accomplished by reassessing participants from a previous study eight months after completion of training. A control group that received no training was also included. The results showed that both training groups maintained training-related gains in number recall performance eight months later in a criterion task providing cognitive support. However, when no support was provided, number recall performance for the mnemonic group dropped slightly whereas performance for the self-generated strategy group improved. The control group maintained performance across assessments. These data indicate the numerical memory training may result in long-term effects in older adults and that, under some conditions, practising one's own memory strategies may have advantages over mnemonic training in old age.