The extensive spatial coverage of seismic data provides a unique source of information widely used for surface and facies modelling, faults detection, and more generally, static heterogeneities modelling. The advent of time-lapse (4D) seismic surveys opens a new dimension-the possibility of monitoring fluid flow, and hence, production. Discounting the favourable cases in which fluid displacement can be seen directly from time-lapse seismic data, the preliminary challenge is to establish a correlation between saturation changes and seismic data changes. Because of the low resolution of seismic data, one should not expect any useful point-to-point correlation. Instead, in general one might expect correlations between spatial patterns of seismic time differences and corresponding spatial patterns of saturation changes. Spatial patterns involve multiple locations within a fixed template window and are summarized by the principal/canonical components (PC/CC) of the within-template variability of each variable. This preliminary study aims at developing a general methodology to establish such pattern correlations, thereby preparing the way for a systematic utilization of time-lapse seismic surveys beyond trivial visual observation. The very well-known synthetic Stanford V 3D elastic reservoir, for which both flow simulation results and repeated synthetic seismic surveys are available, is used to explore such correlations. It is found that the first few PCs or CCs of the 4D seismic and water saturation variables do capture the spatial patterns, and they account for most of the total variance. Through the principal/canonical analysis, significant correlation is found between multiple-point spatial patterns of the 4D seismic variable and spatial patterns of the water saturation time-lapse variable. The 4D seismic patterns are then put to use to predict changes in saturation.