There is a need for valid objective tests of neurological improvement or deterioration to more accurately define response or progression in phase II studies of malignant glioma. The Edinburgh Functional Impairment Tests (EFIT) incorporate objective measures of upper and lower limb function, memory and a rating scale for dysphasia. We examined the intra-observer repeatability of the (EFIT) 24 hours apart in 55 patients with brain tumors and stable neurological disease and the inter-rater repeatability in 33 patients in the perioperative period (54 dual assessments). Intra-observer studies of the four subtests, failed to demonstrate any learning effect and showed close agreement. Inter-rater studies were affected by a treatment effect (steroids) and identified slight inter-rater bias for the ten meter walk. Altman-Bland plots showed that the level of agreement was less good in patients with more severe impairment. Correction for the severity of handicap was possible using a simple formulae: (timed tests: [rater 1 - rater 2]/[rater 1 + rater 2], Williams Delayed Recall Test [WDRT] (rater 1 - 2/81). Using this correction, all intra- and inter-rater variance of patients tested within 12 hours were < 0.2. A change of > or = 0.2 for the timed tests and WDRT, and a change in dysphasia score of > or = 2, represent a significant change in impairment using the EFIT. The EFIT should be a useful addition in phase II studies where objectively recording response or time to progression is important.