Instructional process management (encompassing instructional design and classroom management) is known to be very complex, mainly due to its context and the large and diverse amount of knowledge driving it: content knowledge, pedagogical knowledge (PK), curriculum knowledge, knowledge of learners and their characteristics, knowledge of educational contexts, pedagogical content knowledge, and knowledge of educational ends (Shulman, 1987). This complexity makes researchers unable to detect well-defined practices, and a knowledge base leading to efficient teaching with low training cost is missing (Koedinger, Booth, & Klahr, 2013). The attempts so far to investigate these pieces of knowledge can be categorised in two paths. The first path considers novice vs. expert knowledge comparisons to model knowledge growth across experience (Hogan, Rabinowitz, & Craven, 2003). The second path considers knowledge bases every teacher needs in order to work efficiently. These two paths both have some concerns. Teachers' expertise is not so clear-cut and the way teachers develop it through experience is difficult to model and diagnose. Further, specifying a comprehensive knowledge base about teaching is often externally-driven (Hattie, 2009; Wang, Haertel, & Walberg, 1993), listing a superficial knowledge base of 'what works', often unrelated to teachers' beliefs and knowledge and/or their cognitive abilities. Pedagogical Content Knowledge (PCK), as Shulman (1987, p. 8) argued, is " that special amalgam of content and pedagogy that is uniquely the province of teachers, their own special form of special understanding ". This proposition, as well as Shulman's professional knowledge development model, has become widespread in the educational sciences research field. However, some problems can be raised, notwithstanding its vagueness (" special amalgam " , also see Kind, 2015). First, teaching is mainly seen as a " learned profession " (see Shulman, 1987, p. 9), while the social and informal facet of this activity, relying on mainly innate abilities, remains unaddressed. Second, we refer to van Driel et al.'s (2001) definition of PCK: " the knowledge the teachers must have in order to teach science " (emphasis added) to highlight that the orientation of such a knowledge base is highly prescriptive in nature: it states appropriate knowledge so that anyone having it can be a good teacher, and many research lines have been oriented to differentiating novices from experts with regard to PCK. Third, this model emphasises a unidirectional way in which teachers' beliefs and knowledge 2 influence his or her social behaviour, whereas bidirectional ways are more likely to occur (Rimm-Kaufman & Hamre, 2010). The aim of this paper is to explore a cognitive way to define teachers' professional knowledge (TPK), arguing that some 'natural' knowledge, stemming from several human social abilities – and, for many of them, animal – is thus engaged in teaching as well. The actions grounded on such knowledge are undertaken automatically or at a low cognitive load due to the nature of the latter. Some theoretical views on teaching include such an assumption (Csibra, 2007; Csibra & Gergely, 2011; Strauss, 2005; Strauss & Ziv, 2012), but so far, little research has investigated teachers' cognitive processes in relation to both natural cognition and Cognitive Load Theory (CLT) (see however Feldon, 2007; Moos & Pitton, 2013). The remainder of this paper seeks firstly to consider teachers' actions through the lens of natural cognition and pedagogy, then to set up a framework for teacher cognition and knowledge, showing that several social abilities and knowledge can be used for teaching purposes, and with a low cognitive load. Then, we describe the abilities for teaching as primary vs. secondary knowledge. Eventually, we use this framework to assess or predict which cognitive load is in relation with teachers' performances according to the CLASS, a renowned classroom observation system.