It has previously been shown that - in theory - magnification variations can occur in an imaging system as a function of defocus, depending on the field curvature of the illuminating system. We here present the results of practical experiments to verify this effect in the transmission electron microscope. We find that with illumination settings typically used in the electron microscopy of biological macromolecules, systematic variations in magnification of ~ 0.5% per microm defocus can easily occur. This work highlights the need for a magnification-invariant imaging mode to eliminate or to compensate for this effect.