Critical decisions in iron ore project evaluation and production are based on global and local spatial estimates of primary mineral deposit properties (such as in situ geochemical grades) and geometallurgical responses (such as screened, washed or concentrated product yields and grades), which are derived from limited sample data. These spatial estimates are typically made using classical geostatistical techniques, which fundamentally assume that the attributes of interest behave additively. This assumption is generally appropriate for primary' variables but response' variables may behave non-additively and therefore their direct linear estimates may be biased.In this paper an existing additivity testing mechanism is used to quantify the bias resulting from direct linear estimation of a non-additive iron ore attribute. Data from a magnetite project are used to determine the circumstances under which such biases may be of material impact. The implications for iron ore resource estimation, geometallurgical modelling and grade control are outlined and discussed.CITATION:Cornah, A J, 2013. Assessment of the impact of non-additivity in the estimation of iron ore attributes, in Proceedings Iron Ore 2013 , pp 129-136 (The Australasian Institute of Mining and Metallurgy: Melbourne).