Limitations of the stochastic approach, a method widely used in simulating droplet dispersion in turbulence, are examined using two well-defined two-phase, turbulent flow problems. It is found that the stochastic approach is restricted to flows in which the continuous and dispersed phases can meet each other's correlative condition. Also, use of the steady-state formula of the drag coefficient in the Lagrangian equation of motion for models based on the stochastic approach could cause significant errors in the prediction of complete two-phase turbulent structures, particularly for large droplets.