Know-How Exchange : Research/Metrics

What exactly was the problem with the first version of these 3 questions?

Were they simply unclear to the respondents or did they mess up the responses to the full questionnaire, e.g. because they were biasing answers to your key research question?

For instance, I often find that questionnaires on the adoption of a new product/technology are strongly suggesting this adoption. In those cases, many respondents indicate they will try the new product on the survey, but won't in real life....

Depending on answering the above, your options are:

a) keep only these three questions seperate for the first group of 277 and the last group of 160. Compare the answers across groups: if they are similar, you may still want to add them together in your final report, thus benefiting from the full sample size. You can add all other questions across groups.

b) if you suspect rephrasing the three questions has changed respondents' answers to (some of) the other questions as well, analyze each group seperately. Again, you can add them later if the key insights are similar. However, if they are not and you believe the answers of the last group are more valid, I would only report the results of these 160 respondents.

Koen is exactly right. You need to analyze the surveys separately and see if the results are the same. If so, then you can probably combine them. Otherwise, you need to report them separately.

Any questions that precede the new questions can probably be combined, since the new questions would not have biased the respondents (any differently than those in the original survey were biased). But anything after those new questions is suspect -- until you can verify that there are no significant differences between responses to the first survey and those in the second.

In general, it's not a great idea to make changes in a survey mid-stream (i.e., once it's fielded). Strange things can happen, and then you're stuck with the very dilemma you're facing.