In response to Dr. Humphrey's comments , I find it necessary to clarify that the article (Humphery, 1987) I was referring to in my paper is indeed the correct one, but the context was made less than clear, which I take full responsibility. The passage in question states two different but related meanings in the context of that paragraph. The first one, "More recent research on indexing vocabularies in the context of MEDLINE. . .," was intended to mean the performance of free-text and controlled vocabulary in representing the same concepts or documents regardless of the human indexer's role. While difficulties exist in evaluating the performance of both controlled and natural language in representing concepts, Humphrey (1987) demonstrates how a knowledge-based indexing system can aid human indexing using MeSH in the indexing process, so that human errors and inconsistencies can be minimized. The second meaning is about the use of search performance in measuring the performance of controlled and natural language in representing document concepts. My citation to Hersh and Hickam (1995) merely states the fact and their opinion, and it would not be appropriate for me to judge whether or not the 10% benefit reflects a significant increase in searching performance. Though it was not my intent in my paper to discuss to what extent an increment in search performance can be considered as significant, the addition of Humphrey's (1996) correspondence would have given audience a balanced view on the contribution of controlled vocabulary to the improvement of search performance.