Pertinent to the question of projective error, two methodological factors that threaten the descriptive validity of projection measures are considered in conjunction with the effects of item type (abilities vs. opinions). Study 1 examined the effect of experimental paradigm. Compared to the assumed similarity method, projection indices from the consensus method were significantly lower resulting in reduced accuracy. Study 2 focused on measurement sequence effects. Measuring group estimates before self-ratings produced higher, but more valid projection indices. The ability-opinion distinction was a factor common to both studies. The effect of item type on accuracy was mediated by projection: accuracy was greater on opinions because projection was stronger. For both item types, measures of projective error implied under-projection. Whilst the degree of error rests ultimately on paradigm validity, the heuristic value of projection should remain unrealised. Copyright © 2004 John Wiley & Sons, Ltd.