The induced size effect is an apparent rotation about a vertical axis that results from binocularly viewing a target in which one half-image is vertically magnified. A previous paper (Arditi et al., 1981, Vision Res. 21, 755-764) described a theory of this effect in terms of horizontal disparities that exist between vertically magnified images of oblique features and their unmagnified counterparts. The present studies test two aspects of that theory: the requirement of oblique features in stimuli eliciting the induced effect, and the assumption that binocular associations (inputs to disparity computations) are made across horizontal meridians. The former aspect was confirmed in a stereo discrimination experiment in which the direction of rotation (tilt) for crossed line patterns of varying orientation was judged, for a fixed vertical magnification of one half-image. The latter aspect was rejected on the basis of the results of that experiment, and of two experiments in which observers matched the apparent tilt of the lines with a horizontal adjustment line which could be stereoscopically rotated in depth. The data and some associated demonstrations suggest that stereoacuity and apparent depth of oblique lines vertically magnified in one half-image are determined by the horizontal separation between binocular points which are nearest in a fixed binocular coordinate map, rather than by purely horizontal point-matchings. This "nearest neighbor hypothesis" seems to be operative in classic measures of stereoacuity as well as in the induced effect.