1.IntroductionIn chemometrics, the most common approach for exploratory data visualisation is using Principal Components Analysis1,2(PCA). When there are many groups of samples, PCA can be used to explore whether there are patterns in the data. The best way to do this if there are many samples is to colour code each sample according to its group. However if there are a large number of samples, in the order of hundreds and if there are many groups often requiring different symbols, different colours, and overlapping points, traditional PC scores plots can become crowded and often contain an uninformative cloud of overlapping points.Self Organising (or Kohonen) Maps3,4(SOMs) are an alternative approach for data visualisation. There have been relatively few descriptions of the applications of SOMs in analytical chemistry5–14compared to PCA, probably due to the lack of well established packages. Yet SOMs can be a powerful means of visualisation. In addition to maps that provide information similar to scores plots about the relationship between samples, component planes,3,4which have some analogy to loadings plots, can be used to visualise characteristic variables, specific samples or groups of samples.In this paper we report a new application to the characterisation of polymers using Dynamic Mechanical Analysis (DMA). Previously we have reported the use of PCA, cluster analysis, multivariate classification techniques and Learning Vector Quantization15–18to the analysis and grouping of polymers. DMA involves measuring the response of a material to an applied external force as a function of temperature. As the temperature increases the physical properties of the polymers change as transition occurs from a solid to glass to liquid, yielding curves that are characteristic of the material concerned. The polymer dataset reported in this paper, is distinguished by the fact that it can be visualised at different levels. At the top level we can ask whether a polymer is amorphous or semi-crystalline, at the next level whether it belongs to a specific group (e.g. type of material), and at the third level what grade it comes from. This allows the data to be presented in different ways according to the objective of the visualisation.There are many tools associated with SOMs, which can be employed to provide detailed insight into a dataset. There are numerous approaches for displaying maps, some of which are not well represented in the analytical chemistry literature: the most common approach used by analytical chemists being the U-Matrix.32In this paper we develop our own Matlab routines for SOMs, which overcome some of the limitations of various public domain approaches and which extend the capabilities of visualising data especially in the graphical representation. SOMs are an important alternative to PCA that can be employed in all areas of analytical chemistry and are particularly effective where there are several classes: most multivariate classification and visualisation methods perform most effectively when there are 2 or a small number of classes characterising a dataset.