THIS STUDY FOCUSED ON THE GENERAL PROBLEM OF HOW THE TOTAL EFFECT OF A MULTIDIMENSIONAL STIMULUS IS COMPOUNDED FROM THE SIMPLE EFFECT OF ITS SEPARATE COMPONENTS IN THE CONTEXT OF STIMULUS GENERALIZATION. FROM A THEORETICAL LEVEL, THE PROBLEM WAS STUDIED IN TERMS OF THE GEOMETRY OF MINKOWSKI. AN EXPERIMENT WAS CARRIED OUT WITH 20 SUBJECTS WHO WERE TRAINED ON SUCCESSIVE DISCRIMINATION INVOLVING 3 FREQUENCY MODULATED SINUSOIDAL TONES THAT DIFFERED IN CENTER FREQUENCY AND/OR THE RATE OF MODULATION. THE PATTERN OF REINFORCEMENT FOR THE EMISSION OF EITHER OF TWO MUTUALLY EXCLUSIVE RESPONSES WAS CONTINGENT ON THE PRESENCE OF ONE OR TWO VISUAL CUES. IN THIS WAY, THE STIMULUS PROPERTY ATTENDED WAS BROUGHT UNDER EXPERIMENTAL CONTROL. THE STRUCTURE INDUCED ON THE EXTENDED STIMULUS SET WAS EXAMINED USING GENERALIZATION TESTING PROCEDURES WITH THE TWO VISUAL CUES PRESENTED SINGLY FOR 10 SUBJECTS, AND WITH CUES PRESENTED IN COMBINATION FOR THE OTHER 10 SUBJECTS. MEASURES OF THE RESPONSE PROBABILITY AND RESPONSE LATENCY WERE OBTAINED. MEASURES OF DISCRIMINATION LATENCY WERE RELATED TO MEASURES OF RESPONSE PROBABILITY. IN THE CHOICE OF A MODE OF EXPRESSION FOR RESPONSE LATENCY, THE MOST REASONABLE CANDIDATES WERE THE RECIPROCAL AND LOGARITHMIC TRANSFORMATIONS. (JC)