The Colour Size Effect Revisited

After briefly introducing the colorimetric basics required to understand what knowledge is beyond when we talk about colour measurements I would like to turn back to the size effect phenomenon how and since when it has appeared in technical literature and what the first observations and statements of the problem were.

1.5.1. Small Field Phenomena

As visual angle becomes excessively small (1'-5'), due to the absence of the S-cones in the foveola (it is often said to be “blue colour blind”), colour stimuli seem to loose some of their colour components¹⁶, and, at approximately 1', only intensity perception remains, and the reflected light from a coloured object seems completely achromatic if viewed foveally.

With the regress of visual angle, first, yellow tones tend to appear white, and then, at approximately 3' blue seems to be black. Blue-green and red tones keep their original appearance the furthermost with the reduction of size.^17,18 At about 3', colour discrimination ability is in complete accordance with that of observers being diagnosed of colour deficiency (tritanopic dichromacy). As Burnham¹⁹ reported, a related fact of foveal tritanopia was first observed by König.²⁰

1.5.2. Medium Field Investigations

Another phenomenon reported in connection with the field size of a colour stimulus was first observed by Hering,²¹ in 1893 (also reported by Burnham¹⁹). When comparing monochromatic yellow light with a mixture of red and green lights, matching was no longer present when the subtended visual angle decreased, as the mixed light appeared to exhibit more red colour. Later, during anomaloscope matchings, these findings were confirmed^22-24 and the reverse effect²⁵ was also demonstrated, namely, by increasing the

viewing distance of constant size stimuli, more red component was needed to achieve a match with monochromatic yellow. This effect takes place between 1°-25° of visual angle.

In the same domain of visual angles (1°-25°) possibly the most valuable, yet anecdotic, phenomenon from the viewpoint of this work was first investigated by Marshall and Guilford.²⁶ Their goal was to investigate a common remark of their observers to their simultaneous colour matching experiments of larger paint samples. Their subjects claimed that they experience no difference if the colour of larger samples is assessed employing the standard size mid-grey mask, provided with the Munsell Book of Colour, or discarding the use of it. After conducting the experiments Guilford concluded that only black and white changed in Munsell value as size was varied. But this phenomenon remained an open question because of the inconsistent results. A possible reason of unsuccessfully investigating the chromatic changes of colours of the different sizes may have been the use of colours with the highest chroma at the gamut boundaries of the Munsell Book of Color where there was no choice to detect any shift towards colours of e.g. greater chroma. After these early works, colour size effect findings were generally quantified by the colorimetric terms (chromaticity, excitation purity) as well as perceived colour attributes (hue, lightness, and chroma) but spread in literature really seldom later.

Actually, this is the domain (1°-25°) that contains the critical 4° frontier and where colour measurements should be split according to the CIE recommendations to a ‘small’

and a ‘large field’ region. Note that, in this Section, introducing a comprehensive grouping of the perceptual derivation of colours along the variation of the stimulus size, field size categories do not coincide with the expressions ‘small’ and ‘large’ field domains often used by the CIE referring to the use of the 2° or the 10° colorimetric observers.

1.5.3. Large Field Effects

Perceived properties coloured surfaces of field sizes as large as 22° and 77° were matched with that of a 2° field, and, as the result of modifying the field size, primarily a change in excitation purity was identified.¹⁹ Burnham concluded that the change is of greater magnitude if field size is increased to 22° from 2° than to 77° from 22°, and, for certain colours, the experienced excitation purity rise stopped or even a drop was observed. It must be noted that neither chromatic adaptation nor separation of the large stimulus was controlled. These factors may be important if unlimited time exposure is allowed for the observers to the large stimulus, nevertheless, their unimportance for short term durations (up to 8 s) in terms of the size phenomenon will be reflected later in this work. In another

work of the same author the size effect was found to be similar to that of increasing the illumination over the colour samples,²⁷ namely, colours seem brighter and more colourful if their size is enlarged.

1.5.4. Peripheral Colour Vision

When studying the size effect, it is important to review the results achieved in the extensive research of human peripheral vision. If objects are not seen centrally then their projected image on the retina will be off-foveal. It is well-known that the perception of colours at the periphery differs from that in the centre of vision (i.e. in the fovea).^28,29 Here, besides of the desaturation of colours, hue shifts also take place with the increase of eccentricity (due to the decreased activity of the L−M channel).^30,31 With the increase of the physical size of the stimuli, it is possible to achieve a fovea-like colour perception, up to eccentricities of 20°. However, even the larger stimuli fail to produce fully saturated hues at 40°.³² Also the surround conditions proved to have an effect on the colour perception in the periphery.³³ A vital phenomenon regarding the appearance of extra-foveal colours is rod intrusion that is the contribution of the rod photoreceptors in the evolving colour sensation.^34-39 The major difference between the colour size effect and peripheral colour appearance is that, though, regarding the size effect, colour stimuli are projected to off-axis retinal regions, too, basically they are observed centrally. In this case an

„integration” of the regionally different perception within the uniform patch may occur – to „finalize” the perceived colour of the patch. Sometimes, the large patch is perceived yet to be inhomogeneous. But, despite the major difference between the two phenomena the contribution of the peripheral parts of the retina for colours viewed immersed suggests that the overall perceptual changes may show some similarities with the peripheral colour vision phenomena.

1.5.5. Recent Research

Concerning the possible physiological reasons causing the colour size effect, some of them were verified (e.g. the absence of S-cones in the foveola) and others were disproved (e.g.

the theory of independent yellow receptors in the retina²⁵) in the past.

Recently, with the advance and the widespread use of new experimental techniques, such as colour matching on computer-controlled colour monitors or the magnitude estimation technique and the new imaging devices, such as calibrated colour displays providing large size self-luminous stimuli, it has been possible to begin a more extensive

research on the colour size effect in terms of more sophisticated and powerful tools in colour science, such as CIELAB and particularly the colour appearance models (CAMs).

Both the small field and the large field effect have been revisited.^40-47 As colour is more and more instrumental in transmitting information, colour naming under different illumination levels for small visual fields have also been investigated.⁴⁸ On the one hand, authors concluded for the case of the small field (smaller than 1°) that the colour of small objects became more and more achromatic with the decrease of size⁴⁰ and, on the other hand, authors agreed that for larger samples (i.e. larger than 2°), the perceived colour seemed to exhibit more lightness, and, the perceptual correlates related to chromaticness i.e. chroma and saturation also increased.^41,43

The revision came basically from the architectural field⁴⁹ and aimed to find out what hides behind the anecdotic phenomenon that the interior and exterior wall surfaces of houses do not match the expectations after the paintwork – if the paint was previously selected from a swatch collection of the manufacturers. In Anter’s work,⁴⁹ thousands of outdoor observations of traditional Swedish houses gave a large dataset matching their perception with NCS cards (on the Natural Colour System see e.g. Ref. No. 2). Results showed some consistent variation patterns in the relationship between the inherent colour (the “real” colour matched with directly laying the NCS sample onto the surface of the wall) and the “approximate perceived colour” (viewed from a distance) of painted facades observed under specific conditions such as “full daylight”. The most obvious and pervasive tendency was that the perceived colour of the facade had less blackness than the inherent colour (average difference: 11 NCS units, statistically significant). The perceived colour of yellowish brown, red, yellowish green, and blue facades had more chromaticness than the inherent colour (average difference: 4 NCS units, statistically significant). The perceived colour had more whiteness than the inherent colour (average difference: 7 NCS units, statistically significant) except red. The often significant hue shifts between the perceived colour of the facade and its inherent colour, where also a consistent variation pattern was found, depended much on the kind of the inherent colour. Above findings have been explained by several different factors related to the nature of these outdoor observations.

The illumination of the facades (blue sky, overcast sky, or different phases of daylight, with or without haze) may differ significantly from that of the colour sample (daylight simulator). The level of illumination may be different (e. g. 1000 lx for the sample and 20000 lx for the outdoor facade observation). Chromatic adaptation (or even colour constancy) may be incomplete. The different levels of illumination may cause differences

in perceived colourfulness and hue shifts may come into play (Bezold-Brücke effect). The proximal field and the surround may also be different for the facades (soil, paving, vegetation, and other buildings) and small colour samples (white paper). Though it was not stated explicitly that the differences between the physical size of the colour sample and the façade may be attributed to the perceptual mismatch, it can be assumed that this was a typical practical instance of the size effect of colours in the medium angular domain (viewing distances were 4 m, 50 m, and for longer distances up to 1300 m).

The colour appearance changes of a real room⁴² and the colour perception of different visual fields^43,44 (2°-50°) of surface colours were also investigated and modelled in comparison to standard size samples. Observers assessed the colour appearance of several fully painted rooms using small (2°) matching samples (both reflective and CRT matching) and different size painted fields on walls, too. A correction method based on CIELAB space and later an LMS cone signal based formula⁴⁴ was suggested embedding the size of the stimulus as a factor and another one for predicting the appearance of the completely painted rooms.⁴² Mainly lightness and chroma changes were found to significantly alter with size. Hue was usually found to be unchanged. The overall average colour difference between the colorimetric and the perceived colour of the patches reached values up to 18 CIELAB units depending on size and were decreased to around 9 units by the use of the formulae. Indeed, with the increase of the visual angle, both CIELAB and the LMS model performances decreased and the CIELAB model did not converge to the room appearance model at all, which would have been expectable as a transient behaviour. It is susceptible that for larger visual angles the applied experimental techniques employing surface colours to asses their appearance owing to their perceptual complexity may not comprise enough control.