Training in interventional radiology (IR) relies on a traditional apprenticeship; to protect patients, expert supervision is mandatory until knowledge, attitudes and practical skills have been certified as satisfactory. However, the current quality of IR training is threatened by reduced time for trainees to learn, as well as a loss of basic diagnostic, training cases to non-invasive imaging. At the same time, IR techniques are becoming a focus of interest to a range of other clinical specialities. To address this training shortfall there is a need to develop novel training alternatives such as simulator models. Few simulator models in any medical field have been successfully validated to show improved clinical skills in treating patients. To date no endovascular simulator has met this standard. A good simulator must be based around key performance measures (metrics) derived from careful analysis of the procedure to be replicated. Metrics can be determined by trained psychologists from a direct analysis of the content of the job or task to be tested. The identification of these critical measures of performance is a complex process which must be tailored to a training curriculum to be effective. Simulators based on flawed metrics will invariably lead to unsatisfactory assessment. It follows that simulator development must involve the statutory licensing authorities. Equally it is essential that we do not assume that training on a particular simulator will correlate with the ability to perform the task in the real world. This "transfer of training" must be rigorously proven by validation studies.