Classical painting styles are remarkably different between Europe and East Asia. Classic, post-Renaissance European paintings are approaching photorealism with its rich expressions of shading and highlights whereas paintings from northeast Asia consist of motifs drawn in faint shades and dark contours. Given recent findings that artwork follows the statistical regularities of natural scenes, it is sensible to hypothesize that western and eastern painting styles reflect the visual environment of the respective province. Here, we propose that the different climates of Europe and Asia produced different natural light environments that changed the visual appearance of objects, which in turn influenced painting style. Analysis of meteorological data and optical simulations show that directional lightfields in Mediterranean climates produce object images with variegated shading and sharp highlights. Cloudy and diffused monsoon-like lightfields, in comparison, produce line-shaped shading only around the deepest concavities and remove highlights as well as cast shadows. Image statistics analysis suggests that western and eastern painting styles mimic such differences in visual appearance. The style of classical artworks that have been appreciated in a particular cultural realm could partially mirror the implicit structure of images as constrained by the natural light environment of the corresponding habitat.
There is a remarkable difference between the styles of classical painting in Europe and Asia (Fig. 1) while we can also see small but clear variations within each region (e.g., Italian vs Dutch paintings). It has been pointed out that classical European paintings from the Renaissance period through to the romanticism period often consist of bright, richly-shaded objects on dark backgrounds. In contrast, classical paintings in northeastern Asia, including China, Korea, and Japan, tend to consist of motifs drawn in dark, blurry shades and/or sharp contours on a bright background. Notably, historically speaking, northeastern Asian paintings have been virtually devoid of the cast shadows and specular highlights that are so central to European photorealism. There are significant variations in style within each region, but such variations are far smaller than those between regions.
This huge discrepancy in painting style has been ascribed to the diversity in various factors ‒ e.g., philosophy, aesthetics, and technology ‒ in Europe and East Asia that have remained culturally separate until the last few centuries. For example, it has been suggested that the style of Chinese traditional paintings has long been under the deep influence of an aesthetical theory originating in the 6th century (cf. Wang, 1996). It has also been pointed out that three-dimensional representations, including linear perspectives in Renaissance paintings, are closely associated with an appreciation for the sciences of geometry and optics that were being developed in Europe. In the preceding study, Graham and Field (2008) have suggested that painting materials ‒ oil vs water ‒ and the uniformity of the background are important factors. Since these factors are interdependent, it would be generally difficult to attribute differences to a single source. Yet, one can still consider one fundamental origin, which is the difference in the natural environment between Europe and East Asia.
Increasing evidence from computational aesthetics provides the intriguing hypothesis that image artwork follows the statistical regularity of natural scenes (Brachmann and Redies, 2017; Graham and Redies, 2010; Mather, 2018; Tyler, 2020). For example, many paintings, including abstract works such as Pollock’s, tend to have a fractal structure ‒ or the so-called 1/f spatial-frequency spectrum ‒ (Taylor et al., 1999) which is observed ubiquitously in natural images (e.g., Field, 1987; Simoncelli and Olshausen, 2001). To be sure, other classes of image statistics in paintings found in museums often mimic those of natural scenes (Brachmann and Redies, 2017; Graham and Redies, 2010). Humans, then, tend to appreciate paintings with the implicit image structures to which they have been exposed. Recent psychophysical experiments also show that human observers tend to prefer visual patterns that follow natural image statistics (Spehar et al., 2003), and violations of natural-scene statistics tend to produce visual discomfort or unpleasantness (Fernandez and Wilkins, 2008; Juricevic et al., 2010; Ogawa and Motoyoshi, 2020, 2021; O’Hare and Hibbard, 2011).
According to these findings, it is sensible that a specific painting style in a cultural region reflects the statistical properties of the visual environment in which it was developed. Here, we propose the tentative hypothesis that climate, via its effect on natural-scene image statistics, plays a significant role in the development of these painting styles (Motoyoshi, 2011).
2. Natural Illumination Depends on Climate
The canonical style of classical European painting after the Renaissance was originally developed in cities of Italy. The East-Asian style, in comparison, has one of its origins at southern estates along the Jiang River (and around northern capitals) in China, although other areas within each region also influenced each other. Italy enjoys sunny skies in the Mediterranean climatic zone whereas China experiences frequently cloud cover in the monsoon (and yellow-sand phenomena in the northern capitals). As illustrated in Fig. 2a, this difference in sky condition inescapably produces a large difference in the spatial pattern of natural illumination (i.e., lightfield) between the two regions. Under the sunny sky, the light field is highly directional. Under the cloudy sky, however, the illumination is often heavily diffused due to light scattering through the clouds. Such a diversity of light environments is also found to some extent within each region, although arguably to a lesser extent than the differences between Europe and Asia. These differences and their possible influence on the development of painting styles within each region will be discussed later.
To confirm the theoretical difference in light scattering quantitatively, we analyzed the relationship between weather and the light field’s spatial structure by using a large meteorological database containing all-sky radiance maps for various weather conditions (standardized from the data collected every 30 min over 10 years; Takeda et al., 2005). The all-sky radiance map consisted of 145 intensity data points, each measuring average local intensity within a cubic angular division of the whole 180-degree sky range. Figure 2b shows the RMS contrast (left panel) and skewness (right panel) of the all-sky radiance map plotted as a function of the direct illuminance of the sun (the inverse of the cloud ratio). The plots reveal that, as the sky is less cloudy, the radiance map’s intensity histogram tends toward a higher contrast (correlation coefficient, r = 0.86, p < 0.001) and skewness (r = 0.85, p < 0.001). These empirical data confirm a general law that the sky’s light distribution is more spatially homogeneous and less skewed for cloudy or rainy weathers, which are far more frequent in the monsoon than for typically more sunny Mediterranean weather. In accordance with the optics of the atmosphere, these low-level illumination statistics are likely to be determined primarily by cloud ratio (weather) rather than region (climate). It is unlikely that the low-level illumination statistics in a sky with the same cloud ratio would be largely different between Italy and China.
3. Natural-Image Appearance Depends on Illumination
Importantly, differences in the spatial structure of illumination are mirrored in the visual appearance of natural objects in the natural environment, as suggested by recent studies on material/color perception (Fleming et al., 2003; Motoyoshi et al., 2007; Pont, 2019). As shown in Fig. 3a, directional illumination on a bumpy surface produces strong shadings and cast shadows in the image, while diffuse illumination only produces shadings in the depths because the light comes from various directions. This law is so robust that it is almost impossible for the opposite effect to occur. Figure 3b shows computer-generated images of the same object (i.e., Happy Buddha; Curless and Levoy, 1996) with identical reflectance properties. The inset in the upper left shows the ‘lightfield’ that was used to illuminate the object. The object on the left is illuminated by the lightfield having a high contrast and skewness (Beach lightfield; Debevec, 2006). The object on the right is illuminated by the same lightfield whose contrast and skewness were lowered. As expected, the images have remarkably different visual appearances. Under the ‘sunny’ lightfield (high contrast, high skew), a straight ray from the dominant light source, the sun, makes the object involve richly variegated shading, deep cast shadows, and sharp strong highlights. Alternatively, under the cloudy lightfield (low contrast, low skew), incident light from various directions brings out shadings only in the deepest concavities, and renders highlights extensively blurred and faint.
Psychophysical evidence shows that the human visual system cannot overcome these apparent effects to see through physical shape or the reflectance properties of the object (Fleming et al., 2003; Ho et al., 2008; Kartashova et al., 2016; Pont, 2019). The perceived glossiness, lightness, and bumps necessarily depend on the statistical structure of the lightfield ‒ that is, diffuse illumination makes an object look almost matte, lighter, and less bumpy (Fleming et al., 2003; Ho et al., 2008; Motoyoshi and Matoba, 2012). As a consequence, objects under directional illumination appear to consist of masses with clear three-dimensional shapes and materials whereas objects under diffuse illumination appear flat, matte, and somehow contour-based. These apparent differences in the image due to illumination appear to be in parallel with the apparent difference in the European and East-Asian painting styles shown in Fig. 1.
4. Parallelism in Image Statistics between Objects and Paintings
To quantitatively examine the above differences between the images of objects under sunny vs cloudy light environments, we compared the image statistics of 3D objects under sunny vs cloudy lightfields. We generated images (256 × 256 pixels) of six different objects (happy buddha, dragon, and lion, viewed from two different directions). Each of these 3D models consisted of a highly complex and diverse set of directional and scaled bumpy surfaces based on tens of thousands of polygons. Thus, the image statistics obtained for an entire object can be regarded as a summary of a number of local statistics produced by individual shape. Each object was rendered under three different clear outdoor lighting fields (Eucalyptus, Campus, and Beach; Debevec, 2006) and the same lightfields that mimicked cloudy and diffuse weather (i.e., reducing the contrast by a factor of 1/4 and reducing the skewness by a power of 1/4). All objects had a bluish surface with a diffuse reflectance of 70% and a specular reflectance of 3%. Considering that the natural illumination is mainly from the top, the objects were placed on a very large area of flat floor of 50% reflectance to prevent the effect of illumination from below. Each image was converted to the CIE La*b* color space, and the following image statistics were computed: log RMS contrast, skewness, log kurtosis of the luminance image, and log chromatic contrast. All image statistics were computed only on image pixels falling on the object.
Red and blue squares in Fig. 4 shows the average image statistics for objects under sunny lightfield (W) and cloudy lightfields with reduced contrast and skewness (E), respectively. Each panel shows the results for contrast, skewness, kurtosis, and chroma. Results show that object images under sunny lightfields have higher contrast [t(44) = 15.6, p < 0.00001], larger skewness [t(46) = 3.88, p < 0.001], lower kurtosis [t(45) = –2.02, p < 0.05], and higher chromatic contrast [t(44) = 13.6, p < 0.00001], than images generated under diffuse cloudy lightfields. The results are consistent with the casual observation of Fig. 3 as well as with the previous psychophysical data on surface perception (Fleming et al., 2003; Ho et al., 2008; Motoyoshi and Matoba, 2012).
Next, we applied the same analysis to 120 images of classical paintings considered to be masterpieces in the history of European (60) and northeast Asian (60) art (Graham and Field, 2008). Images were collected from the Internet and trimmed to a square area (256 × 256 pixels) centered on the main motif (Fig. 1). In the analysis, the image was converted to luminance data by assuming a gamma of 0.5, and the four image statistics described above ‒ log RMS contrast, skewness, log kurtosis, and chromatic contrast ‒ were computed. Red and blue circles show the average of European and Asian paintings respectively. Data reveal that European paintings have a higher contrast [t(118) = 16.4, p < 0.00001], a larger skewness [t(118) = 10.5, p < 0.00001], a lower kurtosis [t(118) = –2.99, p < 0.01], and a higher chromatic contrast [t(118) = 12.1, p < 0.00001] than Asian paintings. The results obtained for luminance contrast and skewness replicate the results of the previous study (Graham and Field, 2008), but not those for kurtosis (~sparseness). This may be due to the fact that the previous study used a wide range of paintings from the Eastern hemisphere, including non-classical ones (from a single museum), while the present study used masterpieces of classical painting from northeast Asia. A similar pattern of the results was obtained for the painting images with no gamma or with a gamma of 2.5. The interregional differences in image statistics between painting style were obvious, as described above, and so we did not use a larger number of images.
It should be noted that here we focused only on the low-level statistics of the lightfield as a factor that alters the image features in a weather-dependent manner. There would be other factors that vary with weather. One is the higher-order illumination variables such as ‘diffuseness’ and ‘brilliance’, which have recently been shown to influence the perception of materials and shapes (Pont, 2019; Zhang et al., 2019). The other is haze, which is known to significantly reduce the contrast and blur the image of distant objects (Narasimhan and Nayar, 2002). The homogeneous background, which has been suggested in the previous study as one of the important properties that characterize East-Asian paintings (Graham and Field, 2008), may also be considered as a partial reflection of haze in natural scenes. In general, haze scatters not only the illumination from above but also from below, so that the effective lightfield for the target object is diffused in all directions. Therefore, it is possible that the difference in the appearance of object between sunny and cloudy conditions is even larger than shown in Figs 3 and 4.
In the present study, we proposed a tentative, naturalistic explanation of the origin of style differences between classical paintings of Europe and East Asia. We argued that western and eastern climates create different light environments and that images of natural objects have correspondingly different characteristics. And so, simply, each region ‒ Europe and East Asia ‒ has developed a painting style that mirrors those characteristics. The idea that people in a given region will adopt a style of painting that reflects the habitat’s environment is consistent with experimental evidence showing that humans prefer images that follow the statistical regularities of natural images (Spehar et al., 2003; Fernandez and Wilkins, 2008; Juricevic et al., 2010; Ogawa and Motoyoshi, 2020, 2021; O’Hare and Hibbard, 2011). It should also be noted that this ‘environmental’ explanation is limited to groups of work dating from a time when cultures of East Asia and western Europe had little interactions. Our environmental explanation is not necessarily applicable to more modern periods during which interregional exchanges have been more prevalent.
Experimental studies have shown that humans tend to prefer visual stimuli that have been shown repeatedly (Zajonc, 2001). This mere exposure effect is known to occur not only for artificial figures but also for natural images ‒ including faces ‒ even when the exposure does not involve conscious awareness (e.g., Monahan et al., 2000). Continual exposure to an implicit structure of sensory inputs, then, may induce a preference for that structure. In line with this idea, implicit image statistics for light environments specific to Europe and East Asia may have promoted a particular style of painting. Moreover, there is also a possibility that each painting style emphasizes the distinctive light environment. Everyday scenes in Europe cannot always have extremely dark shadows as in paintings by Caravaggio or the tenebrists, and everyday scenes in China cannot always have the misty atmosphere that characterizes ink paintings. The appearance of paintings may be exaggerated compared to their real-world scene counterparts. Considering that many European painters before the Impressionists, as well as many Asian classical painters, did not sketch what they saw, but created works based on their memories, such an emphasis seems quite possible.
Whereas this exaggeration is not strictly consistent with the general principle of familiarity whereby images obeying the statistical regularities of natural scenes are favored, it would demonstrate the peak shift ‒ an effect in which humans prefer stimuli that are not only close to familiar stimuli but also far from unfamiliar stimuli (Ramachandran and Hirstein, 1999).
As suggested by Graham and Field (2008), painting materials also play a significant role in determining painting style. In Europe, paintings were mainly oil-based whereas paintings in in East Asia used watercolors ‒ pigments fixed to the surface of the painting by water-soluble mediums such as glue. Clearly, oil paintings tend to have higher contrast and greater skew in their luminance distribution. Watercolors, by comparison, have greater difficulty representing dark blacks and therefore naturally produce images with lower contrast. It is possible, then, that differences in painting styles between Europe and East Asia can be attributed to differences in painting materials. We compared the statistics of oil paintings and frescoes of several famous European artists (e.g., Tiepolo). Results showed that the contrast and skewness of the frescoes were as low as in East Asia, but the kurtosis was nearer to that of European oil paintings and still different from that of East-Asian paintings. Therefore, it would seem difficult to attribute differences in image statistics between the two styles to painting materials alone. Moreover, even if painting materials had a large impact on the style-related image statistics, it is also possible that painting materials were chosen for their suitability to depict landscapes and objects specific to a particular climatic region. This raises the chicken-vs-the-egg question of whether style influenced choice of painting material or vice versa. As mentioned earlier, painting styles are the product of a complex interaction of various natural, technological, and social factors.
As mentioned earlier, the relationship between painting style and climate may be more complicated if we restrict ourselves to a smaller geographical area. For example, our hypothesis does not seem to explain the fact that the light environment in England (cloudy) is very different from that in Italy, and yet painting styles in the two regions are similar. However, one can also point out that classical paintings in England, especially after the 16th century, were under the influence of continental painting. The canonical style of painting at the English court was greatly influenced by the work of Anthony van Dyck, a Flemish painter, colleague of Rubens, and frequent visitor to Italy. Interestingly, before such continental influences became prominent, a famous painter, Nicholas Hilliard, left the following message regarding the unique climate and lighting conditions of England: “seeing that best to show oneself needeth no shadow of place but rather the open light … Her Majesty … chose her place to sit for that purpose in the open alley of a goodly garden, where no tree was near, nor any shadow at all …” (Strong, 1975). It is also noteworthy that this policy of emphasizing open light rather than shading was revived in the style of J.M.W. Turner ‒ a prominent painter in the mid-nineteenth century ‒ at a time when British painting was considered to have shed its continental influences.
One might also point out that the development of colorful paintings, such as ukiyo-e in Japan’s monsoon climate zone, poses a challenge to the climate theory. However, we believe the argument is weak for the following reasons: (1) ukiyo-e was one of the popular media among citizens during the period of Japan’s cultural isolation, unlike the traditional arts that were prized by the ruling class; (2) unlike Western paintings, no ukiyo-e depicted cast shadows or highlights except for the few that were influenced by the style imported from the Netherlands ‒ the only European country that traded with Japan; and (3) Japan has more sunny days than the Jiangnan region of China. The fact that ukiyo-e with vivid color and contrast was created at a time when this region was culturally isolated rather indirectly supports the climate theory.
Although small compared to the obvious differences between East Asia and Europe, it is true that there is a wide variety of painting styles within Europe. In particular, the differences in painting styles between northern and southern Europe are well known. It would be interesting to look at the differences in painting styles within Europe from the perspective of differences in light environment due to climate and lifestyle. The climate is not uniform even in Europe, so the light environment may differ from region to region. While we focused only on low-level illumination statistics, these regional differences may be related to higher-order lighting characteristics (cf. de Kroon and de Kroon, 2003). By introducing variables such as ‘diffuseness’ and ‘brilliance’ as mentioned above (Pont, 2019; Zhang et al., 2019), we may be able to gain insights for why Italian and Netherlandish paintings are so different. Although systematic historical research is less developed than for European paintings, a similar analysis could be applied for variations in Asian paintings.
The present study proposed that the statistical regularity of natural images in the light environment and of the neural image processing in the visual brain could partly constrain painting style. The evidence is still limited and our hypothesis (in its original form) would be difficult to generalize over all styles. One of the weaknesses of the present study is that the range of the paintings we dealt with was limited in both time and space. Like many other studies dealing with Western paintings, we did not deal with the Lascaux murals or medieval paintings. However, we would also like to note that the ancient Roman style, as seen in the Pompeian murals and Enkaustik works, is similar to the classical European style after the Renaissance in terms of variegated shading and highlighting. We expect that such a naturalistic approach can be useful to uncover the ecological origins of visual arts, especially when combined with careful historical considerations.
This study was supported by JSPS KAKENHI 15K12042, 18K19801, and 20K21803. Part of the study had been presented at the Vision Science Annual Meeting 2011 and the Visual Science of Art Conference 2012. The author thanks the anonymous reviewers for their thoughtful comments on the manuscript.
Supplementary material is available online at: https://doi.org/10.6084/m9.figshare.19107845
Brachmann, A. and Redies, C. (2017). Computational and experimental approaches to visual aesthetics, Front. Comput. Neurosci. 11, 102. doi.org/10.3389/fncom.2017.00102.
Curless, B. and Levoy, M. (1996). A volumetric method for building complex models from range images, in: SIGGRAPH ‘96: Proceedings of the 23rd Annual Conference on Computer Graphics and Interactive Techniques, pp. 303–312. doi.org/10.1145/237170.237269.
Debevec, P. (2006). Image-based lighting, in: SIGGRAPH ‘06: ACM SIGGRAPH 2006 Courses, pp. 4–es. doi.org/10.1145/1185657.1185686.
Fernandez, D. and Wilkins, A. J. (2008). Uncomfortable images in art and nature, Perception 37, 1098–1113. doi.org/10.1068/p5814.
Field, D. J. (1987). Relations between the statistics of natural images and the response properties of cortical cells, J. Opt. Soc. Am. A, 4, 2379–2394. doi.org/10.1364/JOSAA.4.002379.
Fleming, R. W., Dror, R. O. and Adelson, E. H. (2003). Real-world illumination and the perception of surface reflectance properties, J. Vis. 3, 3. doi.org/10.1167/3.5.3.
Graham, D. J. and Field, D. J. (2008). Variations in intensity statistics for representational and abstract art, and for art from the Eastern and Western hemispheres, Perception 37, 1341–1352. doi.org/10.1068/p5971.
Graham, D. J. and Redies, C. (2010). Statistical regularities in art: Relations with visual coding and perception, Vis. Res. 50, 1503–1509. 10.1016/j.visres.2010.05.002.
Ho, Y.-X., Landy, M. S. and Maloney, L. T. (2008). Conjoint measurement of gloss and surface texture, Psychol. Sci. 19, 196–204. doi.org/10.1111/j.1467-9280.2008.02067.x.
Juricevic, I., Land, L., Wilkins, A. and Webster, M. A. (2010). Visual discomfort and natural image statistics, Perception 39, 884–899. doi.org/10.1068/p6656.
Kartashova, T., Sekulovski, D., de Ridder, H., te Pas, S. F. and Pont, S. C. (2016). The global structure of the visual light field and its relation to the physical light field, J. Vis. 16, 9. doi.org/10.1167/16.10.9.
Mather, G. (2018). Visual image statistics in the history of western art, Art Percept. 6, 97–115. doi: 10.1163/22134913-20181092.
Monahan, J. L., Murphy, S. T. and Zajonc, R. B. (2000). Subliminal mere exposure: Specific, general, and diffuse effects, Psychol. Sci. 11, 462–466. doi.org/10.1111/1467–9280.00289.
Motoyoshi, I. (2011). Ecological-optics origin of the style of European and East-Asian classical painting, J. Vis. 11, 1188. doi.org/10.1167/11.11.1188.
Motoyoshi, I. and Matoba, H. (2012). Variability in constancy of the perceived surface reflectance across different illumination statistics, Vis. Res. 53, 30–39. doi.org/10.1016/j.visres.2011.11.010.
Motoyoshi, I., Nishida, S., Sharan, L. and Adelson, E. H. (2007). Image statistics and the perception of surface qualities, Nature 447, 206–209. doi.org/10.1038/nature05724.
Narasimhan, S. G. and Nayar, S. K. (2002). Vision and the atmosphere, Int. J. Comput. Vis. 48, 233–254. doi.org/10.1023/A:1016328200723.
Ogawa, N. and Motoyoshi, I. (2020). Differential effects of orientation and spatial-frequency spectra on visual unpleasantness, Front. Psychol. 11, 1342. doi.org/10.3389/fpsyg.2020.01342.
Ogawa, N. and Motoyoshi, I. (2021). Spatiotemporal frequency characteristics of the visual unpleasantness of dynamic bandpass noise, Vision Res. 184, 37–42. doi.org/10.1016/j.visres.2021.03.001.
O’Hare, L. and Hibbard, P. B. (2011). Spatial frequency and visual discomfort, Vision Res. 51, 1767–1777. doi: 10.1016/j.visres.2011.06.002.
Pont, S. C. (2019). Light: toward a transdisciplinary science of appearance and atmosphere, Annu. Rev. Vis. Sci. 5, 503–527. doi.org/10.1146/annurev-vision-091718-014934.
Ramachandran, V. S. and Hirstein, W. (1999). The science of art: A neurological theory of aesthetic experience, J. Consc. Stud. 6, 15–51.
Simoncelli, E. P. and Olshausen, B. A. (2001). Natural image statistics and neural representation, Annu. Rev. Neurosci. 24, 1193–1216. doi.org/10.1146/annurev.neuro.24.1.1193.
Spehar, B., Clifford, C. W. G., Newell, B. R. and Taylor, R. P. (2003). Universal aesthetic of fractals, Comput. Graph. 27, 813–820. doi.org/10.1016/S0097-8493(03)00154-7.
Takeda, G. Inanuma, M., Yoshizawa, N. and Isozaki, K. (2005). Standard weather data and heat load calculation program LESCOM. Inoue Shoin.
Taylor, R. P., Micolich, A. P. and Jonas, D. (1999). Fractal analysis of Pollock’s drip paintings, Nature 399, 422. doi.org/10.1038/20833.
Tyler, C. W. (2020). The intersection of visual science and art in Renaissance Italy, Perception 49, 1265–1282. doi.org/10.1177/0301006620974973.
Zajonc, R. B. (2001). Mere exposure: A gateway to the subliminal, Curr. Dir. Psychol. Sci. 10, 224–228. doi.org/10.1111/1467–8721.00154.
Zhang, F., de Ridder, H., Barla, P. and Pont, S. (2019). A systematic approach to testing and predicting light-material interactions, J. Vis. 19, 11. doi: https://doi.org/10.1167/19.4.11.