The Eulerian, joint-scalar probability density function (pdf) approach, at the single point and time level, was applied to predict the evolution of a piloted methane-air turbulent jet diffusion flame. At a certain fuel inlet velocity, the measurements indicate that the flame exhibits local extinction at 20 diameters from the fuel nozzle and subsequent downstream relighting. The objective was to investigate whether the current pdf approach can treat extinction/reignition phenomena and accurately predict the concentrations of species such as CO and H-2. Two different mixing models-coalescence-dispersion and linear mean square estimation (LMSE) - were tested, together with two turbulence models (k-epsilon and full Reynolds stress). The solution method involved a coupled finite-volume solution of the velocity held and a Monte Carlo simulation of the pdf transport equation. A global hydrocarbon scheme (no radical species present) was employed to represent the chemistry. The predictions demonstrated that in general the mixing and fuel consumption rates were well represented. Other scalars such as H2O and temperature were also satisfactorily predicted for the flame not subjected to local extinction; furthermore, in this particular case, the results obtained were very similar regardless of the mixing models and turbulence closures applied. However, with regard to the intermediate species, H-2 and CO, some overprediction was observed. For the flame where extinction was observed experimentally, the predictions showed that with LMSE, extinction with no subsequent relighting was predicted whereas with the coalescence-dispersion model stable burning was reproduced though the low temperatures observed at x/D = 20 were not predicted. In addition the results were sensitive to the predicted spreading rate and hence to the turbulence model.