Seismicity is a distributed process of great spatial and temporal variability and complexity. Efforts to characterise and describe the evolution of seismicity patterns have a long history. Today, the detection of changes in the spatial distribution of seismicity is still regarded as one of the most important approaches in monitoring and understanding seismicity. The problem of how to best describe these spatio-temporal changes remains, also in view of the detection of possible precursors for large earthquakes. In particular, it is difficult to separate the superimposed effects of different origin and to unveil the subtle (precursory) effects in the presence of stronger but irrelevant constituents. I present an approach to the latter two problems which relies on the Principal Components Analysis (PCA), a method based on eigen-structure analysis, by taking a time series approach and separating the seismicity rate patterns into a background component and components of change. I show a sample application to the Southern California area and discuss the promising results in view of their implications, potential applications and with respect to their possible precursory qualities.