APPENDIX 



SURVEY METHODS 



Que»tlonnalra 



The dfiveloiment phase of the study 

 consisted of 57 depth interviews conducted 

 by specialists in this type of interviewing. 

 Respondents were chosen for these inter- 

 views in a non- systematic, but also non- 

 random method. These interviews consisted 

 of informal and casual discussions covering 

 aspects of household consumers' preferences 

 for canned fish. The respondent was al- 

 lowed to take whatever direction she wanted 

 to in these discussions, following her own 

 natural inclinations. No attempt was made 

 to limit or restrict the discussion to 

 predetermined sa*eas of interest. Only when 

 the respondent had exhausted some topic did 

 the interviewer attempt to give some fur- 

 ther direction to the interviews by asking 

 a very genereLl and open "why" or "how" 

 question. 



In addition, a number of specific 

 techniques were used in these initial inter- 

 views as further aids in eliciting consumer 

 attitudes and motivations. Such techniques 

 as word association, sentence completion, 

 response projection, role taking, and 

 cartoon tests were used. 



After the first few of these 57 inter- 

 views were completed, discussions were held 

 by the staff and the interviewers. Sug- 

 gestions were made regarding procedural 

 changes in order to increase the prospects 

 for more complete emd detailed infonnation. 



Using these 57 preliminary inteirviews 

 as a basis, a list of associations was 

 developed showing all of the releveint areas 

 to be explored in the full scale study. 

 This "item list ' formed the basis for de- 

 velopment of the "guided association" 

 portion of the questionnaire. 



Several drafts of the pre-test ques- 

 tionnaire were then developed, in consul- 

 tation with staff members of the U. S. Fish 

 and Wildlife Service, with each draft re- 

 ceiving limited field tests by a specialist 

 in interviewing techniques. An improved 

 draft of the questionnaire was also for- 

 warded to members of the fish canning 

 industry for their ccnmients and suggestions. 



A full scale pre-test of the ques- 

 tionnaire was carried out in the three 

 urbanized areas selected for the full 

 scale survey, Boston, Massachusetts, Blr- 

 mlDgham, Alabama and Detroit, Michlgein. A 

 total of 6l pre-test interviews were com- 

 pleted. These were distributed approxi- 

 mately equally among the three urbanized 

 areas. A coniplete review of all questions 

 included in the pre-test was carried out 

 with differences in local interpretation 

 noted especially. Based on this review 

 final revisions in the questlonnsdre were 

 made and specific instructions to the in- 

 terviewers prepared. A copy of the ques- 

 tionnaire is included in this Appendix. 



Sample Design 



The sample design for this study was 

 constructed with two basic requirements in 

 mind. First, the sampling techniques em- 

 ployed must be consistent with the demands 

 of soiind research methodology; they must be 

 techniques by irtiich valid inferences may be 

 drawn from the sample for the population 

 group under investigation. The only known 

 way to meet this requirement is through 

 probability sampling. With probability 

 samples, the chance of observing a given 

 individual or element of the population of 

 interest is known. It permits the re- 

 seeurcher to not only control the sampling 

 areas, but also to measxire them. It is 

 this property, the measurability of area, 

 which lends validity to the conclusions 

 drawn from probability samples. 



Second, the sample design must be 

 economically and statistically efficient; 

 that is, it should, for the budget allotted 

 and resources available, provide the most 

 accurate estimates of the characteristics 

 studied. The estimates derived from the 

 sample must be of sufficient accuracy to 

 be used with confidence. Selection of the 

 most efficient design implies knowledge of 

 the sources of variation effecting a set 

 of sample observations or measurements. 

 The problem of sample design is to make 

 that Judicious selection among the many 

 techniques available for controlling these 

 sources of variation, and hence the 

 event\uil sampling area, irtilch will achieve 

 €ua appropriate balance between adminis- 

 trative efficiency and statistical 



31 



