In 28 patients with various cardiac diseases we compared ejection fractions obtained by magnetic resonance imaging in a single oblique slice with monoplane ventriculography in the right anterior oblique projection, the latter serving as the standard. Also, results were evaluated for clinical relevance and relation to image quality. The correlation between the two techniques was moderate (r = 0.65). According to our standardized limits for clinical relevance, insufficiently correlating ejection fractions were obtained in 14 patients. In 8 of these patients this was attributed to poor endocardial edge detection. Edge detection problems were more frequently encountered by imaging with echo-time 20 msec than with echo-time 32 msec. Other causes for mismatching of the obtained ejection fractions are discussed. It is concluded that determination of ejection fraction by single slice magnetic resonance imaging should not be used for clinical application. Improvement can be expected by using a contiguous slicing technique, a longer echo-time in the spin-echo pulse sequence, or in due course by application of newly developed fast-imaging pulse sequences.