Bill --
  Sounds like one of your problems is with the form of the questionnaire.
  Instead of having a paper and pencil questionnaire, have you thought
  about using an interviewer-administered one?  Although the up-front
  costs are higher (you have to hire interviewers), the data quality
  is also better, since you can insure random selection and need fewer
  interviews.  BTW, most self-administered interviewing situations
  do not produce random samples.  Even if visitors are selected at random
  participants tend to be self-selected.  The big problem is in determining
  the extent of the bias due to self-selection.  At least with the interviewer
  administered  study you can measure the extent of bias (through a logistic
  regression on the probability of refusal) and adjust your sample accordingly.

  These may not help you with your Feb. study, but you might want to
  consider it.  E-mail me privately if you want to see some examples.
  --Adam

Adam Bickford, Smithsonian Institution,
Institutional Studies Office:: [log in to unmask]
(202) 786-2289