Diagnosis of the surface chemical factors playing a part in flotation separation of a value sulfide phase requires measurement of the hydrophobic and hydrophilic species that are statistically different between the concentrate and tail streams. Statistical methods, based on the monolayer-sensitive time of flight secondary ion mass spectrometry (ToF-SIMS) technique, have been developed towards this ultimate aim by measuring hydrophobic species (collector ions, dimers and metal complexes, polysulfides) as well as hydrophobic metal ions, precipitates and added depressant species. Reliable identification of specific mineral particles is central to this statistical analysis. A chalcopyrite/pyrite/sphalerite mineral mixture conditioned at pH9 for 20 min to study transfer of Cu from chalcopyrite via solution to the other two mineral surfaces, since this mechanism can be responsible for their inadvertent flotation in copper recovery, showed no statistical difference in the copper intensities on pyrite and sphalerite (selected from Fe and Zn images) after this conditioning. Principal component analysis (PCA) identifies combinations of factors strongly correlated (positively or negatively) in images or spectra from sets of data. In images, PCA selects these correlations from the mass spectra recorded at each of 256 x 256 pixels in a selected area of particles. In the image mode, PCA has proved to be a much better method of selecting particles by mineral phase with clearer definition of particle boundaries due to multi-variable recognition. It has clearly separated a statistical difference in copper intensities between the sphalerite and pyrite phases. The PCA method has been applied to concentrate and tails samples collected from the Inco Matte Concentrator demonstrating extensive CuOH and NiOH transfer between the chalcocite (Cc) and heazlewoodite (Hz) minerals. Statistical differences illustrate the important discriminating depressant action of NiOH in flotation despite the activation of Hz by Cu transfer. The adsorption of the collector at specifically-identified Cu sites has been elucidated by the study. Importantly, the statistical analysis has been able to confirm some mechanisms and deny others proposed to control recovery and selectivity giving more focus on the control factors. (C) 2005 Elsevier Ltd. All rights reserved.