Numerous examples of questions which have not worked as they were intended to work have been presented throughout the previous chapters. Reference has also been made to a number of studies in which questions that had supposedly been piloted in order to iron out any faults were misunderstood by large percentages of respondents. We will take another look at these studies, now, because they set the stage for the rest of this chapter.

Nuckols (1953) took nine questions from ‘The Quarters Polls’ that varied in difficulty and presented them to forty-eight randomly chosen, middle-income respondents. The respondents were asked to restate the questions ‘in their own words’. One in six of the interpretations given was either partly or completely wrong. An unexpected finding was that there was little, if any, relationship between the error rate and the difficulty ratings for the questions.

Belson (1981) advances similar findings. He and his colleagues analysed over 2000 questions that had been designed by twenty-four different market researchers over a two-year period and identified questions that might be expected to cause respondents some difficulty. They then selected for in-depth study four examples of each of the six most frequently occurring kinds of problem questions. The kinds of questions selected for investigation were questions that required more than one answer; contained too many meaningful words; included qualifying clauses or phrases; included reference to two or more subjects; included difficult vocabulary; or included instructions.