Thomas Wasow & Jennifer Arnold, "Intuitions in linguistic argumentation," Lingua 115 (2005) 1481–1496.
Generative grammarians have relied on introspective intuitions of well-formedness as their primary source of data. The overreliance on this one type of data and the unsystematic manner in which they are collected cast doubt on the empirical basis of a great deal of syntactic theorizing. These concerns are illustrated with examples and one more detailed case study, concerning the English verb-particle construction.
Some people raise an eyebrow at linguists' practice of treating their own sentence judgments as objective empirical data. The danger is that a linguist's pet theory could unconsciously warp his or her judgments. It's a legitimate worry, but in practice linguistic judgments can go a long way. One of the perquisites of research on basic cognitive processes is that you always have easy access to a specimen of the species you study, namely, yourself. When I was a student in a perception lab I asked my advisor when we would stop generating tones to listen to and start doing the research. He corrected me: listening to the tones /was/ research, as far as he was concerned, since he was confident that if a sequence sounded a certain way to him, it would sound that way to every other normal member of the species. As a sanity check (and to satisfy journal referees) we would eventually pay students to listen to the sounds and press buttons according to what they heard, but the results always ratified what we could hear with our own ears. I've followed the same strategy in psycholinguistics, and in dozens of studies I've found that the average ratings from volunteers have always lined up with the original subjective judgments of the linguists.
（Steven Pinker, The Stuff of Thought, Penguin, 2007, p.34；強調は引用者によるもの；see also this）