It is now widely recognized among educators that explicit attention to language is necessary in order to optimize both language and content learning in situations, such as CLIL, where learners, teachers or both operate in a second language. However, the requirement of attention to language sits uneasily with the fact that content-subject specialist teachers frequently feel unprepared to think and operate in linguistic dimensions. In an attempt to create a conceptualisation that would speak to subject teachers in terms that are meaningful to them from within their own subjects, a Construct of Cognitive Discourse Functions (henceforth CDFs) has been proposed. This construct is theoretically founded in both educational curriculum theory and linguistic pragmatics and consists of a seven-fold categorization of verbalizations which express acts of thinking about subject matter in the classroom (CLASSIFY, DEFINE, DESCRIBE, EVALUATE, EXPLAIN, EXPLORE, REPORT). As the theoretical background of the CDF Construct has been discussed at length elsewhere (Dalton-Puffer 2013, 2016), it will be presented only briefly at the outset of this article. The main purpose of this contribution is to report on steps taken towards an empirical validation of the CDF Construct. A total of four smaller-scale studies each focusing on the classroom discourse in one subject (biology, physics, economics, history) will be surveyed in order to find answers to the questions of whether CDFs actually occur in classroom interaction and if they do, which and to what extent. Comparisons to a small complementary study on a set of EFL lessons will be made where appropriate. Results show that CDFs are indeed a staple of teaching and learning in classroom-based education but are in no way equally distributed. Also, they are almost never the object of conscious attention. As a number of further questions remain yet to be answered on the way towards empirical validation of a fully articulated CDF Construct, these will be discussed in the conclusion of the article.