A multivariate statistical technique which aims to develop spatial structure from numerical data by estimating the differences and similarities between analytical units. Points or items are distributed in a hyperspace, whose dimensions are a large number of variables, and can be similarly distributed in a space of fewer dimensions. The points, originally randomly distributed, are moved about in the new space until the distances between points are similar in proportion to those between points in the original hyperspace. For example, a group of artifacts, about which a large number of characteristics and measurements have been recorded, can be represented by a two-dimensional plot. The reverse is also possible: distributions in a space of few dimensions can be 'unfolded' into space of many more dimensions. I.e. a statistical technique that distorts multidimensional distances/dissimilarities so that they can be portrayed on a two-dimensional map.