The document discusses quality prediction of asymmetrically distorted stereoscopic images. Existing methods generate substantial prediction bias when applied to asymmetrically distorted images. The paper builds a database of single-view and symmetrically/asymmetrically distorted stereoscopic images. A subjective test found prediction bias could be overestimation or underestimation depending on distortion type/level. Eye dominance was found to not strongly impact quality decisions. A new model using information content and divisive normalization pooling, inspired by binocular rivalry, successfully eliminates prediction bias, improving quality prediction of stereoscopic images.