Ames' trapezoidal window is a well-known demonstration of the effect of monocular information about surface slant on other visual processes, such as motion perception (Ames, 1951). While performing informal observations of a hand-held reproduction of Ames' display, we noted some peculiar effects that may be due to the interaction of vision with proprioception. Specifically, during monocular observation while holding the trapezoid with the arms fully extended, several observers experience a striking proprioceptive distortion (e.g., one hand appears farther from the other, or one arm appears longer than the other). Furthermore, in bimodal judgments of slant several observers are strongly biased by vision despite clear, unambiguous proprioception. In this sense, the effect may appear as an instance of visual dominance in the multimodal perception of one's position in space. However, our data suggest that the proprioceptive misalignment of the hands is, in fact, rather less than the apparent slant of the window when this is not held. This finding argues against a “visual capture” account, supports an explanation in terms of bimodal integrative processes, and underscores the importance of supplementing phenomenological observation with objective measures. Implications of the visual-proprioceptive Ames' window for current models of bimodal integration are discussed.