Gradedness and Consistency in Grammaticality Judgments

The methodological issue of the unreliability of certain introspective data circulating in syntactic literature has already been mentioned by several authors (e.g. Schutze, 1996; Adli, 2003). One particularly problematic phenomenon is that questionable judgments are sometimes quoted in theoretical studies without prior critical empirical verification contributing thus to the formation of “myths” in the literature. I shall argue in this paper that one case in point is the widely-quoted que qui ‘rule’ in French. It will be demonstrated that this rule, which has been introduced into the literature solely on the basis of un-controlled introspective data, is not confirmed by experimental studies in which a controlled process of data-collection is applied to a whole sample of test persons and which essentially make use of a graded concept of grammaticality. Furthermore, it will be shown that the use of quantitative data on graded grammaticality does not only allow the interpretation of the (mean) grammaticality value, but also the consideration of another measure, namely the (intraindividual) judgment consistency observed for a given construction. This newly introduced measure has the advantage to complement the information about the exact grammaticality value with the information on the difficulty to give stable judgments allowing a more complete evaluation of the grammatical quality of a structure.