An Italian proverb says: Tra il dire e il fare, c'è in mezzo il mare.¹ If one wants to find out how much sport people do on average, one has to accompany a sample of them through their daily life, make observations in sports institutions or collect data on their actual behavior in some similar way. If, on the other hand, one presents a set of subjects with a questionnaire asking about their sports activities, one may, at best, find out what they think they do in terms of sports, but more probably, what they want the investigator to think about it.

Questionnaires are appreciated in many sciences as an easily applied method of data collection. Not seldom are they used as a makeshift substitute for the observation of actual behavior. That is a basic methodological mistake.

There are many private reasons why somebody would prefer to appear behaving differently from the way he actually behaves. Apart from these, there is commonly a norm in the society which its members are expected to observe. Its role is particularly important in the case of language. In many societies, observation of the linguistic norm is an essential part of formal education from early linguistic socialization on. What is more, almost all reflection on language has been shaped by formal education referring to the norm. Unless they have an academic formation in empirical linguistics, people are simply unable to observe their own actual linguistic behavior and to reflect neutrally on it.

From this it follows that if the epistemic interest is in actual linguistic behavior, viz. in questions such as the following:

• what speakers actually say,
• how they actually code certain concepts and operations in their language,
• whether a certain sentence is grammatical,
• which language they choose to speak in a certain situation,
• whether they are more fluent in L_i than in L_j,
• and so forth,

then it is useless to ask them about it because they will be either unable to answer such questions or it will be against their self-image to answer them sincerely. Questionnaires about the linguistic behavior of subjects in any such respect are a useful means if the investigator wants to know about the relevant self-image of subjects, i.e. about what they think should be the case.

¹ Between saying and doing there is in the middle the sea.