The spatial and temporal development of a spray strongly depends on the local characteristics of turbulence. The turbulence-droplet coupling gives rise to droplet dispersion, which is the underlaying physical phenomenon of interest in this study. Large eddy simulations (LES) provide details of the instantaneous flow field and anisotropy of the larger scales. Hence, LES has the potential of improved spray simulations in flows that are highly nonisotropic/nonstationary. A numerical study on the effect of droplet diameter (d) on spray shape is described by carefully varying d. The droplets are assumed to be non-interacting with each other. They are also assumed to maintain their shape and diameter. The droplet Stokes numbers are within the range 0.07 <= St(p) < 2.56, corresponding to diameters 2 < d < 12 am for a common liquid fuel. In order to emulate a fuel spray, a droplet-laden jet at Re = 10,000 and Ma = 0.3 is considered as a model problem that avoids the dense spray regime. A novel technique to visualize the simulated sprays in a realistic manner is presented, and a qualitative comparison to a diesel spray experiments is made. It is shown that the spray-cloud shape depends strongly on droplet Stokes number. A spray penetration correlation formula is suggested. The nonlinear character of the droplet-eddy interaction and its dependence on droplet size is studied by visualization of droplet trajectories. We show that the spray behavior can be coherently explained by considering the statistical properties of the droplet cloud. The results show that the instantaneous/short-time-averaged probability density functions (PDFs) of droplet statistics explain very coherently the St(p) dependency of the spray shape. The PDFs of the axial and radial components of droplet-gas slip velocity (u(g) - u(p)) are used to explain the visual observations on the spray cloud evolution.