Two-dimensional images of (Plane) Laser Induced Fluorescence (PLIF) have been used to study the turbulent mixing process in a model stirred tank. A calibration procedure is presented and discussed in terms of its accuracy. Data from the literature are used for comparison. A pattern-recognition algorithm has been designed to identify and quantitatively describe large-scale structures in the flow. This methodology, called "structural analysis", is based on a conditional analysis of the PLIF data and requires the definition of an appropriate structure-detector function which is calculated locally. The mathematical tools developed have been used to study the mixing in a Rushton turbine-agitated reactor. Particular attention is paid to two specific regions of the tank; namely the bulk and the impeller stream regions, at two measured power input (0.3 and 0.7 W kg(-1)). The averaged concentration fields show a common two-dimensional steady circulation pattern. Concentration probability density functions reflect well the instability of the flow in the two regions investigated. The data reveal the non-isotropic distribution of these instabilities around a reference point when the feed port is situated in the bulk region only. In this case, the structural analysis quantitatively shows the presence of a folding of the concentration field. It was found that this phenomenon can last several seconds.