Abstract
When a textured surface is modulated in depth, the level of illumination varies across the surface, producing modulations of local mean luminance (LM). The shading also produces modulations in the luminance amplitude of the visual texture (AM). The AM co-varies with LM such that local contrast remains constant. It is well known that human vision can use LM cues to infer the shape of the underlying surface, but the role of AM has not been explored. We conjecture that the phase relation between LM and AM is an ecologically valid depth cue that might be used by human vision. To examine this, we used a 2ifc task in which both intervals contained a textured plaid: binary noise with oblique (+/−45 deg, 0.5 c/deg) LM and AM components. In one interval both obliques consisted of an LM+AM pair, but in the other interval one oblique was an anti-phase pair (LM-AM). We measured LM thresholds for discriminating the two plaid types at fixed levels of AM and found that observers always required more LM to distinguish the plaid types than they did to detect LM without AM. We also measured AM thresholds for plaid discrimination at fixed levels of LM and found that when LM was at its own detection threshold no amount of AM would enable observers to do the task. However, when LM was 8 times its detection threshold the amount of AM required to discriminate the plaids fell below the AM detection threshold. This facilitation of AM by LM suggests some integration of the two cues, perhaps at the level of surface depth coding. The 3-D appearance of the two plaids was very different: the plaid with the LM-AM component appeared corrugated only in the LM+AM direction, while the other plaid appeared doubly corrugated. The data thus suggest that supra-threshold LM may be necessary for shape from shading. Conversely, supra-threshold AM is not necessary for shape from shading but when present its phase relationship with the LM component may be a key determinant of 3-D surface interpretation.