Although a large part of a document retrieval system's resources are devoted to indexing, the question of how people do subject indexing has been the subject of much conjecture and only a little experimentation. This dissertation examines the relationships between a document being indexed and the index terms assigned to that document in an attempt to quantify the extent of "machine-like" indexing occurring when librarians and scientists index technical text. A number of possible relationships between the text and the index assignments are predicted and tested with two models: a multiple linear regression model and a Boolean combinatorial model. It is concluded that indexers in general do not index technical text in a "machine-like" fashion and that neither model is useful as a general predictor of human indexing. (Author/NH)