Brightness and lightness

Experiments in perceived illumination

Alessandro Soranzo, Stephen Ivory and Alan Gilchrist

Classic theorists, like Helmholtz, Hering, and Katz, suggested that perceived illumination within a field is determined by average luminance. Zdravković et al. (2011; J Exp Psychol Hum Percept Perform, Dec 5) reported evidence that illumination is tied to highest luminance. We tested perceived illumination using a new technique. Looking into two square windows in the far wall of a vision tunnel, observers could see a patch of the far wall in each of two chambers. Each patch contained two shades of grey. They turned a knob to adjust the illumination level in one chamber to match that of the other. The stimuli placed in the chambers varied in luminance range, spatial frequency, and relative area. Illumination was matched for highest luminance, not average. Spatial frequency made no difference. Area effects were also found. If an aperture contained a large dark grey region and a small light grey region, illumination level appeared lower and surface lightness appeared higher than if the aperture contained equal light and dark regions. These results, which are consistent with Koffka’s invariance theorem and with the role of highest luminance in anchoring theory, provide a foundation for integrating the perception of illumination with perception of surface lightness.

Relationship between constancy and memory for stimulus hue and lightness

Maria Olkkonen and Sarah Allred

Both constancy and memory have important roles in color-based object identification: constancy aids in scene-invariance of perceived color, and memory links perception across time. Here, we characterize color constancy and memory for stimuli varying in lightness or hue. Observers compared the lightness or hue of two briefly presented 2-degree patches in a 2AFC experiment. Patches were presented 1) with a 2.5-second inter-stimulus interval on a uniform surround (memory condition), 2) simultaneously with a luminance or hue difference in stimulus surround (constancy condition), or 3) with both the delay and the surround difference (combined condition). We fit psychometric functions to the proportion-lighter or proportion-bluer data in each condition and estimated the bias (the point of subjective equality). We developed a probabilistic response model with noise and bias parameters fit to all data from the single-change conditions and predicted the bias in the combined condition. Every manipulation elicited a bias in perceived lightness or hue. Pure memory bias was generally well-described as a tendency towards mean stimulus value. Surround biases were consistent with constancy predictions. The response model predicted the combined bias based on the single-change fits with reasonable accuracy. We conclude that memory and constancy act roughly additively in color perception.

Can a veiling luminance make a black room look white?

Alan Gilchrist and Stephen Ivory

What happens when observers view a miniature room painted black and filled with black objects through the reflection of a homogeneous sheet of light (veiling luminance)? Prior work leads to conflicting predictions. Because all-white rooms produce a low contrast image due to mutual illumination, while all-black rooms produce a high contrast image (Gilchrist & Jacobsen, 1984), a black room seen through a veil should look white due to the reduced contrast produced by the veil. But because lightness constancy is excellent when a 3D image is viewed through a veil (Gilchrist & Jacobsen, 1983), but completely absent when a 2D Mondrian is seen through a veil (unless the edges of the veil are visible), a black room seen through a veil should appear as a black room covered by a veil. Fifteen observers judged the room as white (Munsell 9.0) and none saw a veiling luminance. The fact that the room appeared white even though our stimulus did not duplicate the luminance gradients in a white room shows that the precise shape of the gradients is not important, only the gross level of contrast. A separate group of 15 viewed the room without the veil saw the room as gray (Munsell 6.2).

Testing luminance polarity and spatial grouping principles of three recent lightness theories

Michael Rudd

Will varying the luminance of a non-contiguous background affect target lightness? Gilchrist's Anchoring Theory (AT) says “yes” only if the background is the highest luminance. Bressan's Lightness Grouping (LG) and Rudd's Edge Integration (EI) theories predict effects for all backgrounds. Incremental or decremental squares were presented on two sides of a monitor, surrounded by frames of equal luminance but different widths (right narrower). The background was varied from lowest to highest display luminance. Observers adjusted the left square to match the right in lightness. The background was predicted to affect the right square lightness more than the left due to spatial proximity (LG) or edge weight decay (EI), yielding a background effect on matches. This effect was seen at all backgrounds, contrary to AT. But the effect was strongest with highest luminance backgrounds, which can be explained by LG (EI) only if there is an interaction between proximity (edge weight) and frame/background polarity. Incremental target matches were non-monotonic for backgrounds intermediate between the square and frame luminance. Non-monotonicity has been seen in other lightness studies and indicates assimilation (Rudd, Journal of Vision, 2010, 10(14):40, 1-36). I propose a set of principles based on proximity and luminance polarity that any lightness theory must explain.

Lightness contrast at the leading edge of motion

Hiroshi Ashida and Nick E Scott-Samuel

Uniform backgrounds appear lighter or darker when elements containing luminance gradients move across them (first presented by Hiroshi Nakamura at the 2010 Illusion Contest in Japan). This effect is intriguing because it integrates the processing of brightness/lightness, gradients, and motion. We psychophysically investigated the phenomenon by measuring the apparent lightness of a grey background overlaid with moving square patches of vertically oriented luminance gradient. For grey-to-black gradients, the background appeared lighter when the black edges were leading than when they were trailing. For grey-to-white gradients, the background appeared darker when the white edges were leading than when they were trailing. For white-to-black gradients, the background appeared darker with a white edge leading and lighter with a dark edge leading, whereas the effects were weaker. These results demonstrate that lightness contrast can be modulated by motion direction of the inducing patterns. This effect might be partially explained by stimulus enhancement at the leading edge and/or suppression at the trailing edge (e.g. Arnold et al, 2007, Vision Research, 47, 2403-2410), but it is also likely to be related to asymmetry in processing spatiotemporal gradients (Murakami et al, 2006, Vision Research, 46, 2421-2431). [Supported by Global COE program (D07 Kyoto University) by MEXT, Japan, and JSPS Grant-in Aid for Scientific Research (A22243044) for HA]

Highest luminance modulates search in complex tasks

Suncica Zdravkovic, Ivana Jakovljev and Ian M. Thornton

In a visual scene, surfaces with highest luminance (HL) are considered “special” because they determine the lightness assignment of all other surfaces. This implies that the visual system has implicit access to the relative intensity of the HL. Here, we ask whether the relative location of the HL is also implicitly encoded. We used the MILO task (Thornton & Horowitz, 2004 Perception & Psychophysics 66 38-50) to explore whether the location of the HL influences behavior during sequential search. Each display consisted of nine gray squares, ranging from black to white. Participants clicked on each square in a specific order, either from black to white (BTW) or from white to black (WTB). A located target would either vanish (Experiment 1) or remain (Experiment 2) on the screen. Finally, all the targets were either brighter than the background (increments, Experiment 3), or darker than the background (decrements, Experiment 4). Physically removing items did not modify search behavior, confirming the existence of efficient inhibitory tagging previously demonstrated with alphanumeric sequences. WTB search was always more efficient than BTW search except when all targets were decrements. These search benefits suggest that we use implicit knowledge about the position of HL during complex tasks. This research was supported by Ministry of Education and Science, Grants No. 179033 and III47020.