In virtual auditory display, the accurate simulation of early reflection is helpful to guarantee audio fidelity and enhance immersion. However, the early reflection may not be easily distinguished from the direct sound due to the masking effect. This work investigated the spatial resolution of early reflection for speech and white noise under different conditions, in which three-down-one-up adaptive strategy with three-interval-three-alternative forced-choice (3I-3AFC) was employed. Results show that, for both speech and white noise, the spatial resolution of early reflection decreases with the increasing deviation of reflection orientation relative to the direct sound, and has no relationship with the time delay; Moreover, the spatial resolution of early reflection for speech is always lower than that for white noise under the same condition.