Some writers suggest that retrieval may be improved in certain contexts by indexing only particular characteristics in a text. For example, Oh (1998) suggests that, in psychology, the indexing only of “empirical facts” (variable names, correlation values, and significance level information) would improve the retrieval situation. While highly specialized indexing of this type may be justified in a few unusual situations, it is unlikely to be a majority requirement and will probably be much more expensive than a more conventional approach.