To characterize the flow instability in mechanically agitated vessels, digital particle image velocimetry (DPIV) was used to measure the near-instantaneous flow fields in a baffled, stirred vessel, equipped with a Rushton turbine impeller of large diameter ratio of impeller to vessel. From the measured near-instantaneous flow fields, distinct large-scale flow patterns were identified and the formation mechanisms of the patterns were explored via multi-scale analysis. Further investigations show that the flow patterns are rather complex. Macro-instability (MI) appears as a swith among those flow patterns. Spectral distribution function P(ƒ), equivalent to probability distribution function from a purely mathematical point of view, was introduced to extract information on MI fron the time series of the vorticity of the fluctuation velocity field. The two-dimensional distribution of MI and its dependence on Reynolds number were studied. Results show that probability analysis via spectral distribution function can be successfully used to quantify the relative intensity of macro-instabilities. Low-frequency, large-scale fluctuations of flow patterns were much more intensive in the lower and upper corner of the vessel than in the rest parts of the vessel. The fluctuations of flow patterns also had an offset to the core of the vessel due to the influence of baffles. Compared with visual observation technique, this new analysis method via DPIVis more objective and can be used to give deep understanding to the spatio-temporal properties of macro-instability.