Humans can easily recognize materials, such as metal or plastic, under various illuminations. It has been suggested that the visual system utilizes some statistical properties of natural illumination to recover the surface reflectance. The purpose of this study is to understand which components of illumination determine the appearance of surface material. For this purpose, we examined surface material perception under various illuminations in which their spatiotemporal frequency components are manipulated. It is known that surfaces in computer-graphics images rendered with unrealistic illuminations appear unrealistic. To examine the surface material appearance of real objects in the real world, we made a room that is uniformly illuminated from all directions in 3-D space. The room is large enough (3.5 × 3 ×3 m) to serve as realistic environment for observation. To examine the spatiotemporal characteristics of illumination on material perception, we projected various patterns, such as sinusoidal gratings having various spatiotemporal frequencies or band-pass-filtered random-dots, on a wall of the room. The angle of the object (boards having different surface materials) was adjusted to make the projected patterns function as illumination. Subjects judged how strongly they were able to perceive the surface material of the object. We found that when the illumination was uniform (zero frequency), metal, plastic, and porcelain completely lost their material impression. Surface material perception of those objects showed low-pass characteristics: low-spatial frequency illumination induced more veridical perception than illumination having high- spatial frequency components. The temporal frequency of illumination also showed low-pass characteristics. These results imply that the visual system can recover surface properties under some artificial illuminations, such as one having a low spatiotemporal frequency component.