This talk presents some recent and current developments in the area of multidimensional scaling, an exploratory method that represents proximities among objects as distances among points in a low-dimensional space. After a brief introduction to state-of-the-art MDS techniques and some computational discussions, the focus of the talk shifts to MDS goodness-of-fit assessment with special emphasis on permutation tests. Subsequently, structural components (e.g., clustering) are integrated into the MDS target function (we call this “cluster-optimized proximity scaling”). The final segment concerns the use of sup-port vector machines within an MDS context, which aims to separate the MDS space into facets in a (non-)linear way ac-cording to a supervised classification problem.