THE 71 VARIABLES UNCOVERED IN THE LITERATURE SEARCH FELL INTO FOUR CATEGORIES: GENERAL RESEARCH CHARACTERISTICS, SAMPLING AND SAMPLE CHARACTERISTICS, QUESTIONNAIRE CHARACTERISTICS, AND RESEARCH PROCEDURES. THE MEAN FINAL RESPONSE RATE FOR ALL 98 STUDIES WAS 60.6 PERCENT WITH SLIGHTLY OVER 2 CONTACTS REQUIRED. THE NUMBER OF CONTACTS MADE BY THE RESEARCHER AND THE RECIPIENTS' PERCEPTION OF THE SALIENCE OF THE QUESTIONNAIRE EXPLAINED 51 PERCENT OF THE VARIANCE IN RESPONSE RATES. CONTACTS WHICH ENHANCED THE IMPORTANCE OF THE QUESTIONNAIRE (SPECIAL DELIVERY, CERTIFIED MAIL, TELEPHONE CALLS, AND PERSONAL CONTACT) INCREASED THE RESPONSE RATES. RESPONSE RATES WERE HIGHEST FOR GOVERNMENT-SPONSORED SURVEYS. SIMILARLY, THE USE OF METERED OR FRANKED POSTAGE ON THE OUTER ENVELOPE SEEMED TO BE ASSOCIATED WITH GOVERNMENTAL OR 'OFFICIAL' RESEARCH SPONSORSHIP. LONGER SURVEY FORMS ARE PERCEIVED AS MORE IMPORTANT THAN SHORTER AND, WHEN OTHER FACTORS ARE CONTROLLED, ACTUALLY ARE ASSOCIATED WITH GREATER RESPONSE. PAYING THE RESPONDENT TO FILL OUT THE QUESTIONNAIRE AND PROVIDING RETURN POSTAGE ALSO INCREASED THE RESPONSE RATE. A CAUSAL MODEL IS PRESENTED WHICH INCLUDES THESE AND OTHER FACTORS (POPULATION SURVEYED, TYPE OF QUESTIONS ASKED, PROVISIONS FOR ANONYMITY). THESE FACTORS EXPLAIN 90 PERCENT OF THE REPORTED VARIATIONS. A REGRESSION EQUATION PREDICTING FINAL RESPONSE RATES IS PRESENTED TO SHOW THAT HIGH RESPONSE LEVELS ARE ACHIEVABLE BY MANIPULATING THE COSTS OF RESPONDING AND THE PERCEIVED IMPORTANCE OF BOTH THE RESEARCH AND THE INDIVIDUAL RESPONSE. A BIBLIOGRAPHY IS APPENDED. (GLR)