TRIO is an expert system for training F-14 interceptor radar operators in the basic tactics of high speed air intercepts. It introduces artificial intelligence methods into real-time training. The TRIO task environment supports simulations of airborne radars, interceptor and target aircraft, and weapons models. It provides dynamic displays of heading, bearing, and displacement vectors, radar screens, flight instruments, intercept parameters, missile envelopes and interceptor/target aircraft ground tracks. It incorporates real-time speech recognition and synthesis subsystems including highly advanced capabilities for recognition of naturally articulated, extended utterances. TRIO supports three instructional modes: pre-flight demonstrations, in-flight monitoring and guidance, and post-flight debriefing. The instruction employs an articulate expert to demonstrate and explain intercept tactics, a set of daemons for real-time performance monitoring, and a knowledge-based performance analysis program to detect and diagnose student errors. TRIO is currently implemented on a personal LISP machine, the BBN Jericho computer.