The Theory of Sampling (TOS) has contributed significantly to chemometrics by allowing data analysts to become more aware of total measurement uncertainty as total sampling error + total analytical error. This has allowed a new view on, for example, validation of multivariate data models and their performance. This keynote focuses mostly on the reverse perspective: how can chemometric multivariate data modelling be of help in an augmented sampling context? The basic principles of multivariate data analysis are presented and application examples are outlined from several areas in the geosciences to facilitate answers to the question posed, including a geochemical multivariate anomaly concept defined by a multivariate background model. Applications from geochemical prospecting, geochemical mapping, environmental chemistry and anthropogenic outlier recognition are outlined, as are the interdependent relationships between geostatistical modelling (variographics) of singular variables versus linear combinations of a multitude of variables, in chemometrics called scores. The potential synergy between chemometrics and TOS is far from exhausted at this time; much work remains.CITATION:Esbensen, K H, 2014. Theory of sampling in a multivariate perspective - what can chemometrics contribute?, in Proceedings Sampling 2014 , pp 7-10 (The Australasian Institute of Mining and Metallurgy: Melbourne).