A certain type of questionnaire has the purpose of ascertaining the subject's opinion on some issue. One kind of items seeks to find out about the extent to which subjects agree with some proposition. Such items are designed as follows:

  1. The items are not formulated as questions to be answered, but as statements to be agreed or disagreed with.
  2. The statements are positive. Not only do they not contain a grammatical negation. The predication itself is semantically positive. For instance, the example item below says ‘politicians are generally trustworthy’ rather than ‘politicians are generally untrustworthy/malicious’. The purpose of this principle is twofold:
    • Positive formulations are less biased (carry less presuppositions) than negative statements.
    • The processing of negative statements is prone to polarity conflicts.
  3. Instead of forcing a yes/no reaction, it is often more interesting to allow for shaded opinions. These use a scale on which a subject may mark the degree of his agreement. Generally, it will be sufficient to distinguish between full agreement and partial agreement, and similarly for disagreement.
  4. Depending on the kind of question and the research interest, provision may be made for one or more types of undecided opinion:
    • There may be a middle position for salomonic attitudes declaring the statement half true and half false.
    • The subject may not know or may not care. Apart from the scale, there is then a separate box for that answer.

Here is a section of such a questionnaire; the numbers in the ‘value’ row need not be shown.

Please mark the extent to which you agree with each of the following statements.
value -1 -0.5 0 0.5 1 I don't know./
I don't care.
I disagree strongly. I tend to disagree. This is partly true and partly false. I tend to agree. I agree strongly.
French is the most beautiful language in the world.
Politicians are generally trustworthy people.

The results are computed as follows:

  1. The positions on the scale are assigned numerical values. The middle value is zero; a positive reaction is symbolized by a positive number, a negative reaction by a negative number. This is, again, in order to avoid polarity conflicts, this time on the part of the investigator.1
  2. The numbers are real numbers between 0 and 1/-1, for ease of statistical computing.
  3. If the questionnaire made a distinction between a neutral answer and ignorance/abstention, it may or may not be reasonable to assign the value zero to the latter, too, and to merge them with the neutral answers.

1 Alternatively, one may use positive numbers only. ‘1’ then signifies ‘yes’, and ‘0’ ‘no’, with ‘0.5’ standing for the undecided middle position. While this is closer to boolean logic, it is not quite as intuitive.