The results of multi-gated blood pool imaging (MBPI) for determination of left ventricular ejection fraction (LVEF) in 48 patients at rest and during exercise were analysed in order to evaluate observer variance and interstudy variance. Twenty-three MBPI studies carried out with the patient at rest and 37 studies carried out during graded ergometer exercise were analysed by two independent observers in order to determine intra- and inter-observer variance. On the intra-observer level regression analysis yielded excellent correlation between rest and exercise studies (r = 0,998 at rest, r = 0,964 during exercise), and showed no statistically significant difference (P less than 0,01); this was also true of the inter-observer measurements (r = 0,999 at rest, r = 0,988 during exercise). In order to evaluate interstudy variance, MBPI was repeated in 15 patients at rest. Interstudy variation of the LVEF was minimal as indicated by regression analysis (r = 0,98), and showed no statistically significant difference (P less than 0,01). During the repeat study a mean increase in the LVEF of 2,1% was obtained, but this difference was not statistically significant.