Functional MRI is increasingly recognized for its potential as a powerful new tool in clinical neuropsychology. This is likely due to the fact that, with some degree of innovation, it is possible to convert practically any familiar cognitive test into one that can be performed in the MRI scanning environment. However, like any assessment approach, meaningful interpretation of fMRI data for the purpose of patient evaluation crucially requires normative data derived from a sample of unimpaired persons, against which individual patients may be compared. Currently, no such normative data are available for any fMRI-based cognitive testing protocol. In this paper, we report the first of a series of fMRI-compatible cognitive assessment protocols, a matrix reasoning test (f-MRT), for which normative samples of functional activation have been collected from unimpaired control subjects and structured in a manner that makes individual patient evaluation possible in terms of familiar z-score distributions. Practical application of the f-MRT is demonstrated via a contrastive case-study of two individuals with cognitive impairment in which fMRI data identifies subtleties in patient deficits otherwise missed by conventional measures of performance.