Research has previously shown that adding consistent stereoscopic information to self-motion displays can improve the vection in depth induced in physically stationary observers. In some past studies, the simulated eye-separation was always close to the observer's actual eye-separation, as the aim was to examine vection under ecological viewing conditions that provided consistent binocular and monocular self-motion information. The present study investigated whether large discrepancies between the observer's simulated and physical eye-separations would alter the vection-inducing potential of stereoscopic optic flow (either helping, hindering, or preventing the induction of vection). Our self-motion displays simulated eye-separations of 0 cm (the non-stereoscopic control), 3.25 cm (reduced from normal), 6.5 cm (approximately normal), and 13 cm (exaggerated relative to normal). The rated strength of vection in depth was found to increase systematically with the simulated eye-separation. While vection was the strongest in the 13-cm condition (stronger than even the 6.5-cm condition), the 3.25-cm condition still produced superior vection to the 0-cm control (i.e., it had significantly stronger vection ratings and shorter onset latencies). Perceptions of scene depth and object motion-in-depth speed were also found to increase with the simulated eye-separation. As expected based on the findings of previous studies, correlational analyses suggested that the stereoscopic advantage for vection (found for all of our non-zero eye-separation conditions) was due to the increase in perceived motion-in-depth.