on the left margin of the instrument. As the screening process 

 begins, i.e., as you encounter questions which refer only to a 

 segment of the respondents, indentations (normally 5 to 10 spaces) 

 from the margin are used. This alerts the coder, keypuncher, and 

 researcher to the fact that only a segment of the respondents are 

 providing information to these questions. Such a procedure allows 

 for a smoother transition from section to section. It also helps to 

 alleviate some of the problems which normally are encountered 

 when asking recall or historical information by providing the 

 interviewer with handy "prompts" or reminders should the re- 

 spondent indicate confusion over the question. Such a procedure 

 would be of invaluable assistance in the current CETA Follow-Up 

 Instrument. 



One final suggestion for future surveys is offered. We believe 

 that all instruments (mail, telephone, and/or person-to-person 

 questionnaires and schedules) should be thoroughly pretested prior 

 to the onset of the survey instrument. It is a means of assessing how 

 the questionnaire works and whether any changes are necessary 

 prior to the start of the full-scale study. A good pretest is designed to 

 test not only the questionnaire, but the questions as well. If such a 

 pretest had been conducted prior to the SNHS sponsored research, 

 we believe many of the problems herein reported could have been 

 avoided. 



SUMMARY 



This report addresses common methodological problems facing 

 survey researchers. The design of a valid and reliable instrument 

 for the collection of social science data is a necessary first step to 

 survey research. Unfortunately, not enough attention is usually 

 given to issues of experimental design. 



The use of a control group for comparisons with an experimental 

 or treatment group greatly enhances the utility of a survey 

 instrument. In the usual case, that is, a One-Group Pretest-Post-Test 

 Design, several major threats to internal validity arise. These 

 include the effects of history, maturation, testing, instrument decay, 

 and statistical regression. By introducing a control group and using 

 a randomization procedure for the assignment of study groups, 

 these threats are controlled. Further, the introduction of the control 

 group allows for a more meaningful analysis of relationships. 



Additionally, the survey scientist needs to focus his attention on 

 issues which relate directly to questionnaire construction. A 

 procedure for locating respondents must be developed prior to the 

 start of a longitudinal survey. Endorsements from relevant groups 

 should be gathered prior to the collection of data to help insure a 

 high response rate. The instrument itself needs to be worded so as to 

 not confuse the respondents, especially when the data gathering 



