Analyzing coherence for conceptual learning in a Grade 2 numeracy lesson

National and international evaluations of numeracy knowledge have shown that numeracy levels in South Africa are below the competencies specified in the curriculum. Within the broader context of ongoing poor performance in numeracy, a longitudinal research and development project - the Wits Maths Connect - Primary project (WMC-P) has been launched aimed at improving the teaching and learning of primary mathematics. As part of the baseline data collected for this project, the project team observed and videotaped a single numeracy lesson across the Grade 2 classes in the ten project schools, with a view to gaining insights about the nature of teaching and learning, and the classroom contexts of these activities. The analysis has been framed by the Systemic Functional Linguistics concept of coherence and the Variation Theory concept of structured variation. This paper presents evidence to illustrate, firstly, poor coherence in and across pedagogic communication and activities and secondly, random selection and sequencing of exercises that militate against meaning making. The ambiguous use of co-references, co-classifications and co-extensions coupled with random selection and sequencing of tasks give rise to weak coherence within practices that exhibit "extreme localization" and 'ahistoricity'. Such practices work to severely impair possibilities for connected learning of number concepts.