ASSUMING THAT SPEECH IS THE RESULT OF A NUMBER OF DISCRETE NEUROMUSCULAR EVENTS AND THAT THE BRAIN CAN STORE ONLY A LIMITED NUMBER OF MOTOR COMMANDS WITH WHICH TO CONTROL THESE EVENTS, THE RESEARCH REPORTED IN THIS PAPER WAS DIRECTED TO A DETERMINATION OF THE SIZE AND NATURE OF THE STORED ITEMS AND AN EXPLANATION OF HOW SPEAKERS ENCODE A SEQUENCE OF THESE DISCRETE LINGUISTIC UNITS INTO A CONTINUOUSLY CHANGING SIGNAL. THE TECHNIQUE OF ELECTROMYOGRAPHY WAS USED TO STUDY THE ACTIONS OF SOME OF THE MUSCLES IN THE MOUTH IN ARTICULATING "B,P,D" COMBINED WITH 12 AMERICAN ENGLISH VOWELS. RESULTS SHOWED THAT NO ONE-TO-ONE CORRESPONDENCE EXISTS BETWEEN PHONEMES AND MOTOR COMMANDS. TWO MOTOR COMMANDS DIRECTED SIMULTANEOUSLY TO THE SAME MUSCLE RESULT IN MUSCULAR ACTIVITY EQUAL TO THAT NECESSARY FOR THE PRODUCTION OF THE SOUND REQUIRING THE GREATEST TENSION. IF MOTOR COMMANDS CALL FOR A SEQUENCE OF GESTURES AND THE FIRST GESTURE REQUIRES MORE MUSCLE ACTION THAN THE SECOND, THE ACTION FOR THE SECOND GESTURE IS DECREASED. HOWEVER, WHEN THE SECOND GESTURE REQUIRES A GREATER OR EQUAL AMOUNT OF ACTIVITY, A MORE OR LESS SIMPLE CONCATENATION OF GESTURES RESULTS. ALTERNATIVE HYPOTHESES ARE DISCUSSED TO ACCOUNT FOR THESE DATA AND ARE SUGGESTED AS TOPICS FOR FURTHER EXPERIMENTS. THIS PAPER WAS PRESENTED AT A MEETING OF THE ACOUSTICAL SOC. OF AMER., BOSTON, JUNE 1, 1966. (JD)