The theoretical distinction between arithmetic facts and procedures was first made by Groen and Parkman (1972). This was confirmed with a neuropsychological single case described by Warrington (1982) who had impaired arithmetical facts but well preserved arithmetical procedures. Since this time there have been several patients described who showed a selective impairment of arithmetic facts. There have also been reports of cases with impaired arithmetical procedures. However, there has not yet been a case reported with the selective impairment of procedures in the context of intact arithmetic facts. This paper describes a patient, SR, with probable Alzheimer's dementia who had well preserved addition, multiplication and subtraction facts but who nevertheless had severe difficulties with a range of arithmetical procedures such as multidigit sums, decimals and fractions. The implications of this case for current theoretical models are discussed.