PERHAPS THE MAJOR PROBLEM IN TASK ANALYSIS FOR INDUSTRIAL TRAINING IS TO DETERMINE WHAT TO DESCRIBE AND ON WHAT LEVEL OF DETAIL. MANY DIFFERENT LEVELS OF DESCRIPTION MAY BE NEEDED TO ESTIMATE THE COST OF INADEQUATE PERFORMANCE TO A SYSTEM AND THE PROBABILITY OF ADEQUATE PERFORMANCE WITHOUT TRAINING--THE PROBLEM OF IDENTIFYING DIFFICULT COMPONENTS OF A JOB. IN THE ABSENCE OF DIRECT EMPIRICAL MEASURES OF THESE FACTORS, WORKING ESTIMATES CAN BE MADE BY APPEALING TO EXISTING METHODS AND CONCEPTS. SINCE SOME DIVISION OF TASKS INTO PERFORMANCE UNITS WILL BE NEEDED FOR VARIOUS PURPOSES, TRAINING TAXONOMIES ARE REQUIRED. ALTHOUGH TAXONOMIES SHOULD INCLUDE A HIERARCHY OF EXHAUSTIVE, MUTUALLY EXCLUSIVE CATEGORIES, EACH WITH A SPECIFIC TRAINING REQUIREMENT, THE RELATIVE POSITION OF SUCH CATEGORIES CAN BE EXPECTED TO VARY. IN RESPECT TO ACTUAL EVALUATION OF TRAINING TECHNIQUES, EVIDENCE ON SPECIFIC TRAINING CONDITIONS AND THEIR APPLICABILITY IS STILL FAR FROM COMPLETE. MOREOVER, TASK ANALYSIS MUST TAKE INTO ACCOUNT THE ENVIRONMENT AS WELL AS THE CONTENT OF TRAINING. (THE DOCUMENT INCLUDES 59 REFERENCES.) (LY)