The use of text classification as a strategy in the training of translators is discussed. Frequently, translator trainees are required to complete a text profile classifying the text on dimensions such as purpose, formality, difficulty, and emotional tone. It is argued that the empirical basis for such classification is very shaky, but could be made more secure by three types of experiments: (1) verification of validity and interrater reliability on the categories used; (2) tests of statistical and logical (or psychological) interdependence of the various dimensions; and (3) examination of the extent to which translators at various levels actually use text analysis when making decisions about translation. (Contains 13 references.) (Author/MSE)