LIGHTNESS SYMPOSIUM

Historically, two aspects of colors were considered: chromatic (red, blue, yellow etc.) and achromatic (black, white, and various shades of gray). Although lightness perception (perception of achromatic colors) may seem like a trivial visual task, it is in fact a difficult scientific problem that has challenged vision researchers for more than 150 years. In this symposium, we will present various lightness phenomena, describe challenges that confront researchers in their attempts to develop theories and construct models of such effects, and lastly introduce a novel explanatory framework that builds on previous theories in this field.

In the first talk, Dejan Todorović will provide an introduction to the topic of lightness perception. He will first briefly review several basic relevant physical aspects of light, such as illumination, luminance, and reflectance, as well as their perceptual counterparts, such as perceived illumination, brightness, and lightness. He will then describe and demonstrate a number of lightness phenomena, including lightness constancy and various compelling and intriguing lightness illusions.

Elias Economou will introduce the major elements of the Anchoring Theory of lightness perception and its Gestalt approach to lightness encoding. Since its publication in 1999, this theory has influenced the field, has been widely tested, and has proven to be a useful tool for generating new experiments. At the time of its publication, it claimed to account for a wide range of the most relevant phenomena and the majority of published experimental findings in the field.

In the following talk, Sunčica Zdravković will review the empirical data acquired over the past 25 years that now pose serious challenges to some of the core propositions of Anchoring Theory. These challenges concern failures of constancy that arise both from changes or differences in illumination and from variations in background and surface arrangements within a scene.

The last two talks will outline a new unified approach, consisting of two complementary components, that we have been developing with the previously mentioned challenges in mind. This approach builds on the foundations of Anchoring Theory and the general Gestalt framework, while also introducing novel and improved formulas that could account for phenomena that the original theory could not explain.

In his talk, Predrag Nedimović will introduce the new computational formula and the main assumptions of the new model, and will demonstrate how it explains illumination-related errors. Anna Riga will then show how the model accounts for lightness illusions arising from different backgrounds and surface arrangements.

The symposium will conclude with a summary of the model’s strengths and weaknesses. The new approach accounts for both illumination and background-dependent failures of lightness constancy, and phenomena that posed a challenge to the Anchoring Theory (Hypercompression and Maniatis variation of Simultaneous Lightness Contrast illusion). Current weaknesses of the model concern the prediction of failures of constancy under different shadow intensities and certain illusions, such as Helson-type assimilation.

TALKS

Basics of lightness perception: physics and phenomenology

Dejan Todorović¹,* Anna Riga² & Predrag Nedimović¹

¹Laboratory for Experimental Psychology, Faculty of Philosophy, University of Belgrade, Serbia

²Department of Cognitive Science, University of Malta, Msida, Malta

*dtodorov@f.bg.ac.rs

In this introductory talk to the symposium, several notions basic for the understanding of lightness perception (and presupposed in subsequent talks) will be introduced, and a number of illustrations of lightness phenomena will be presented and discussed.

The topic of lightness perception is the study of perception of achromatic surfaces. Physically, such surfaces have uniform reflectance, that is, they reflect approximately equal percentages of light of all wavelengths; phenomenally, they are perceived as having various shades of gray, including black and white. However, the relationship between the relevant physics (reflectance percentage, which usually varies between 3% and 90%) and corresponding phenomenology (lightness or perceived reflectance or shade of gray) is surprisingly complex and still not fully understood.

There are two main factors that complicate our understanding of this field. One is that the luminance of a surface (the amount of light reflected from it and entering our eyes) does not depend only on its reflectance but also on the amount of illumination falling on it, according to a simple formula: luminance = illumination * reflectance. Thus, surfaces of equal reflectance, when illuminated differently, can have different luminances and vice versa. Nevertheless, our perception in such cases is usually more in accord with reflectance than with luminance; that is, we tend to perceive that surfaces of equal reflectance have approximately equal gray levels, even when they are under different illumination. This is the phenomenon of lightness constancy, which is still under investigation.

The other complicating factor is that there are cases in which surfaces of equal reflectance, which are under equal illumination and thus have equal luminance, are perceived to have different shades of gray. As a rule, such cases involve equal surfaces embedded in different contexts. There is a large and bewildering variety of such phenomena, which are usually called ‘lightness illusions’. They constitute a source of diverse challenges for any theory whose goal is a broad and general account of lightness. In this talk, many examples of such effects will be presented and briefly discussed. They include versions of the following phenomena: simultaneous contrast, remote contrast, gradient effects, assimilation, induced glare and mist, White’s effect, Cornsweet’s illusion, Adelson’s checker shadow, and others.

keywords: lightness; lightness constancy; lightness illusions

A review of the Anchoring Theory of Lightness

Elias Economou^1,* & Alan Gilchrist²

¹Psychology Department, Lab of Experimental Psychology, University of Crete

²Psychology Department, Rutgers The State University of New Jersey, NJ USA

*elias@psy.soc.uoc.gr

Since its publication 25 years ago, the Anchoring Theory of Lightness (1999) has provided a very successful framework for understanding the computations involved in the perception of white, black, and gray surfaces. At the core of Anchoring Theory, two key concepts can be identified: a luminance ratios and not absolute luminance values constitute the input of lightness encoding processes, and b. surfaces in the scene are grouped together based on Gestalt organizational principles. These lead us to propose that an Anchoring Rule is needed in order to convert luminance ratio into lightness values and that the final lightness of any single surface in the visual field is “Co-determined” by its relationship to the several perceptual groups to which it belongs.

In this talk, we will explain the major computational components of the Theory. With respect to the Anchoring Problem, we will review both the theoretical and empirical work that led us to identify the rule that the system uses: the Highest Luminance Rule (i.e. Highest luminance in the framework is assigned the value of White). With respect to the Co-determination problem, we will describe the way we think the system parses the image into groups of surfaces that we call “Frames of reference”, the subsequent computations that take place within those Frames or Frameworks, and the way these are combined to provide the final lightness value of a surface.

In the final part of the presentation, we will apply Anchoring Theory to some very well-known lightness illusions, and we will also review how the theory explains major errors that occur when large illumination differences are present in the visual field at the same time.

keywords: lightness; Anchoring theory

Empirical Challenges to the Anchoring Theory

Sunčica Zdravković^1,2,*, Elias Economou³ & Dejan Todorović²

¹Laboratory for Experimental Psychology, Department of Psychology, Faculty of Philosophy, University of Novi Sad, Serbia

²Laboratory for Experimental Psychology, Department of Psychology, Faculty of Philosophy, University of Belgrade, Serbia

³Psychology Department, Lab of Experimental Psychology, University of Crete

*suncica.zdravkovic@ff.uns.ac.rs

The previous talk presented the Anchoring Theory (AT) and its main concepts. The theory was able to account for a very wide range of lightness phenomena. But the original publication (1999) already noted several lightness effects that were not explained.

One of them was the compression of the perceived range under intense illumination. When a group of grey surfaces is suspended in midair and illuminated by a spotlight, the perceived range of greys is compressed relative to the presented physical range. The theory predicted that the compression should increase as the intensity of the spotlight increases, but stopped when the spotlight was 30 times higher than the prevailing illumination. Empirical results, however, showed that the compression increases beyond this point, and all the shades start to appear white and very light grey. Authors were immediately aware that such hypercompression of perceived range cannot be explained by AT.

Another problem concerned the variation of the classical simultaneous contrast illusion (SLC). In SLC, the two equal grey targets appear different due to their black and white backgrounds. The theory attributed the effect to the difference in the highest luminance (HL) in each half of the display, i.e., framework; white as HL on one side and middle grey on the other. But Maniatis (2015) noted that the illusion persists even when a white patch is added to the black background. According to AT, this should have eradicated the illusion, since there is a real white on both sides of the SLC. Our own experiments showed that the illusory effect is only weakened in Maniatis’ variation (Cohen d_SLC=1.057, Cohen d_MANIATIS=0.899), and the theoretical consequences of such findings will be discussed.

Further, the model was practically structure-blind, which means that it could not account for the change in the organization of the visual scene, or the impact of the elements differently positioned in relation to the target. Finally, several core concepts (such as belongingness, grouping, framework, and framework strength) needed to be made more precise.

keywords: lightness; lightness constancy; lightness illusions; Anchoring theory

New model explains illumination-dependent errors in lightness perception

Predrag Nedimović^1,*, Anna Riga², Elias Economou³ & Alan Gilchrist⁴

¹Laboratory for Experimental Psychology, Faculty of Philosophy, University of Belgrade, Serbia

 ²Department of Cognitive Science, University of Malta, Msida, Malta

³Psychology Department, Lab of Experimental Psychology, University of Crete

⁴Psychology Department, Rutgers The State University of New Jersey, NJ USA

*predrag.nedimovic@f.bg.ac.rs

Lightness, which refers to our perception of the percentage of light that an object reflects, remains relatively unchanged when the object’s luminance changes due to a change in illumination. This phenomenon is known as lightness constancy. However, experiments show that the constancy is not perfect. When objects are brightly illuminated (i.e., under a spotlight), they appear somewhat lighter. Moreover, objects with low reflectance increase in lightness more than objects with higher reflectance, and thus the perceived range is compressed. Additionally, the amount of compression is positively related to the intensity of the spotlight. Our previous studies show that the amount of compression is predicted by the ratio between the highest luminance in the spotlight and the highest luminance in the prevailing illumination (R² = .94, p < .001).

The New model accounts for this compression by segmenting the scene into different illumination frameworks and by calculating the lightness of all surfaces with regard to each framework. First, the visual scene is segmented into frameworks that are defined by the boundaries of illumination. The framework of illumination that contains the target is called the Relevant framework, while the rest of the scene is the Foreign framework. Lightness values are then determined with regard to each framework by linear regression with fixed point (Anchoring) and slope (Scaling). Anchoring translates relative luminance values into specific lightness values, and it assumes that the highest luminance within each framework appears white. Lightness of other surfaces within a framework is determined by a Scaling component, which produces a slope that varies with the ratio between the highest luminance in the Relevant framework and the highest luminance in the foreign framework.

This approach predicts constancy when there is only one illumination in the visual scene, by assuming a Slope of 1. It also accounts for and predicts increasing amounts of compression as a spotlight becomes brighter or a shadow becomes darker.

keywords: lightness; lightness constancy

Lightness Errors and Perceptual Grouping

Anna Riga^1,*, Predrag Nedimović², Dejan Todorović², & Alan Gilchrist³

¹Department of Cognitive Science, University of Malta, Msida, Malta

²Laboratory for Experimental Psychology, Faculty of Philosophy, University of Belgrade, Serbia

³Psychology Department, Rutgers University, Newark, NJ, USA

*anna.riga1357@gmail.com

Errors in lightness perception can arise from changes in illumination (illumination-dependent) or from changes in the background (background-dependent). The most renowned example of the latter is the Simultaneous Lightness Contrast illusion (SLC): a gray square placed on a black background is perceived to be lighter than a square of equal luminance on a white background. This difference in the appearance of the shade of gray of the two squares demonstrates how perceptual grouping influences lightness.

We propose that grouping affects lightness through anchoring. The highest luminance within a perceptual group is assigned the value of white in that group, while other surfaces within that group are assigned values relative to their luminance ratio with the anchor. In SLC, the gray square belonging to the white background appears veridical, as the white background serves as the anchor. In the case of the square that belongs to the black background, the gray square itself becomes the highest luminance and gets a value of white in that framework. Since all four surfaces belong to the same single-illumination framework, they will be assigned veridical values in that framework. Through codetermination, however, the final value for the part belonging to the black background will be slightly higher than that belonging to the white background. Lightness computations within perceptual groups can also account for the Maniatis variation of SLC, reverse contrast, and various other illusions.

Grouping can also mediate errors caused by illumination changes (i.e., illumination-dependent errors). In the 5 square paradigm, the error pattern produced by the spotlight placed on the display will vary depending on the spatial arrangement of the squares (orderly vs scrambled versions). Despite the robust effects of perceptual grouping on lightness perception, important computational questions remain open.

keywords: lightness; lightness illusions