If the object of research is some temporary or permanent cognitive or emotional state or process of human beings, then the researcher may himself be in such a state. Instead of trying to get data from other people, he might then try to observe himself. Since what is at stake are inner states and processes, this procedure is called introspection. For a while, it was thought in some strands of linguistics that the linguist could obtain data by introspection. Its products commonly take the form of example sentences created by the linguist. Meanwhile, it has not only become empirically obvious that such products are often unreliable, but there is also a consensus that, in view of the methodological standards set out in the section on sources of data, they are not data in the sense of any scientific discipline.

This does not, of course, imply that the linguist is not allowed to generate example sentences on the basis of introspection, i.e. by simply putting his language competence to work. This happens frequently, and with complete justification, in the composition of textbooks. An example is not a piece of data.

What remains is the intuition of the researcher concerning his object. It has the same methodological status as the intuition of a chemist about which molecular compound may have which physical properties: none. Its value lies exclusively in the field of heuristics, helping the scientist in conceiving ingenious and fruitful hypotheses and thus abridging long series of mechanical experiments by the trial-and-error method. Such hypotheses, however, are not methodologically justified (falsified or verified) by an appeal to the intuition that generated them, but by established methods. Introspection is at the same heuristic level that the linguist may produce example sentences or may employ his intuition in the understanding of his data or in figuring out what kind of data he would need in order to prove a certain hypothesis. He may then try and obtain such data in one of the methodologically accepted ways.

What holds for introspection on the side of the researcher is also valid, mutatis mutandis, for introspection on the part of experiment subjects or linguistic informants. This has been known for a long time:

Report on a report by Wilhelm Wundt (end of 19th century) on the heterogeneous results of an experiment where subjects received a stimulus and reported on their subsequent introspection:

Such disagreements could not be settled in any scientific fashion owing to the inherently private nature of internal events. In more technical terms, introspection failed as a bona fide scientific method because it violated a fundamental rule concerning scientific investigation: that of independent access to both causes and effects. Although the cause (i.e., stimulus) was open to public observation, the effect (i.e., internal sensation) was not. Without such independent observation of the internal sensation, it was impossible to tell which of two conflicting introspective reports was the correct one. (Dellarosa 1988:5)

It is of especial relevance in informant work. An informant's introspection is no more reliable than a linguist's introspection. The researcher therefore must be aware of what it makes sense to ask a native speaker and what not. S. the section on informant judgements.