This work addresses some fundamental questions in the area of partially premixed combustion—what parameters control the fraction of flamelets that are premixed (versus non-premixed), and what are the locations of high probability of premixed (versus non-premixed) combustion? To answer these questions there is a need to measure the flame index (ξ) and its statistical properties, and this information previously has not been available. Flame index is+ 1 where a premixed flamelet exists and is-1 at the location of a non-premixed flamelet. A new method to measure flame index was developed that adds NO 2 to the air; acetone is used as one component of the fuel. Laser-induced fluorescence images indicate the locations of flamelets and whether the gradients of the fuel and O 2 are in the same direction or not. Flame index was measured in a gas turbine model combustor that was designed at DLR that is a good example of partially premixed combustion. Measurements show how the mean flame index varies in space; near the fuel injector the combustion is 50% non-premixed and 50% premixed while downstream the flamelets are mostly premixed. This trend is consistent with two numerical simulations of swirl flames; however for simple lifted jet flames the premixed flamelets do not extend so far downstream. It was found that one parameter that controls the fraction of flamelets that are premixed is the ratio of the fuel injection velocity to the air velocity. Increasing this ratio increases the fraction of flamelets that are premixed because it increases the distance that the fuel stream penetrates into the more intense mixing region. Good signal-to-noise ratios of 24 (for acetone) and 13 (for NO 2) were achieved and an uncertainty analysis is presented that is based on calibration experiments.