The effect of labeling and numbering of response scales on the likelihood of response bias