Culture, Perception, and Artistic Visualization: A Comparative Study of Children's Drawings in Three Siberian Cultural Groups



In a study of three indigenous and non-indigenous cultural groups in northwestern and northeastern Siberia, framed line tests and a landscape drawing task were used to examine the hypotheses that test-based assessments of context sensitivity and independence are correlated with the amount of contextual information contained in drawings, and with the order in which the focal and background objects are drawn. The results supported these hypotheses, and inspection of the regression relationships suggested that the intergroup variations in test performance were likely to result from differences in the attention accorded to contextual information, as revealed by the drawings. Social and environmental explanations for the group differences in context sensitivity are also discussed. The conclusions support the argument that cultural differences in artistic styles and perceptual tests reflect the same underlying perceptual tendencies, and they are consistent with the argument that these tendencies reflect corresponding differences in patterns of social and environmental interaction.

1. Introduction

1.1. Cross-cultural differences in object/context perception

Research on cross-cultural differences in perception and processing of (mostly visual) stimuli and their relation to attitudes and cognition dates back to seminal works by William Rivers (e.g., Rivers, 1905) in the first years of the 20th century and by Alexandr Luria (e.g., Luria, 1979) in the 1930s. Since the second half of the 1960s, much of this research has been focused on how representatives of different cultures differ in the degree to which they attend to focal objects in their perceptual field as opposed to attending to the context within which the objects reside. The first studies of this topic were undertaken by Herman Witkin and his colleagues (particularly John Berry) as part of their research into psychological differentiation, a theoretical concept that was initially developed to account for differences between persons in their educational success (Witkin, Dyk, Paterson, Goodenough, & Karp, 1962), but later applied also for explaining cross-cultural differences in perception and thinking (e.g., Witkin, 1967; Witkin & Goodenough, 1981). The researchers asserted that “in the perceptual domain greater differentiation shows itself in the tendency for parts of the field to be experienced as discrete from the field as a whole rather than as fused with the field, or experienced as global, which is indicative of lesser differentiation” (Witkin & Berry, 1975: 6; see also Witkin et al., 1962; Witkin & Goodenough, 1981). They initially suggested that progress toward greater differentiation (including a shift from field dependency toward field independency) was related to individual psychological development (Witkin & Goodenough, 1981; Witkin et al., 1962).

Witkin and his colleagues developed two basic research tools, the Frame-and-Rod (FRT) test and the Embedded Figure Test (EFT), that measured the subject's ability to abstract a part of the perceptual field from its context and so could be used to assess his or her progress toward field independency (Witkin, 1967; Witkin & Berry, 1975; Witkin et al., 1962). Although the connection between differentiation and psychological development was reformulated in comparative non-evaluative perspective in the context of cross-cultural research (Berry, 1976; Witkin & Berry, 1975), the same tests were used to assess the differences in field dependency/independency across cultural groups. A number of comparative studies of field dependency/independency of perception among representatives of aboriginal and non-aboriginal cultural groups living in North and South Americas, Africa, and Australia confirmed the existence of these differences (see Berry, 1976; Witkin & Berry, 1975 for review and discussion).

Generalizing from their results, John Berry (1976) suggested that the degree of field dependence/independence of perception among representatives of any given culture—as well as their level of psychological differentiation in general—is determined by aspects of socialization practice, most notably socialization toward assertion as opposed to socialization toward conformity, that vary greatly across cultures (e.g., Barry, Child, & Bacon, 1959). The relevant aspects of socialization practices were thought to be determined, in their turn, by a set of closely interwoven eco-cultural and societal factors. The former included the type of subsistence economy and the way of life (nomadic vs. sedentary), while the later included such parameters as the degree of social role differentiation and “tight” versus “loose” social authority and pressure. In particular, mobile hunters and nomadic reindeer herders, who could not be expected to have significant role differentiation or tight social authority, were predicted to be more field independent in comparison to sedentary societies (Berry, 1976).

After a brief lull in the 1980s and early 1990s, research interest in cross-cultural differences in object/context perception has increased again during the last decade and half. The main inspiration of these new studies has been the recently proposed theory of independent and interdependent cognitive styles (Markus & Kitayama, 1991, 2003; Nisbett, 2003; Nisbett, Peng, Choi, & Norenzayan, 2001), which, in contrast to the psychological differentiation theory, was developed specifically to address cross-cultural differences in perception and cognition. According to this theory, the underlying cause of these differences should be sought in cultural models of the relationship between Self and Other—and in particular the degree to which this relationship is understood in terms of independence or of interdependence. Each of these cultural options is thought to have its own cognitive and perceptual implications.

As with the earlier research, these recent studies have focused on the distinction between context sensitivity and context independence, now re-christened as a contrast between “holistic” and “analytic” perception. Many of them, however, employ a new research instrument, the Framed Line Test (FLT, discussed in detail in section 'Prediction 1' of this article), developed by Kitayama and colleagues in 2003. In contrast to the FRT and EFT used in the previous studies, which tested entirely for abilities related to context independence, the FLT measures the subjects' performance both in a task that demands abstraction of an object from its context (the absolute task) and in a task that involves judging an object in relation to its context (the relative task). Empirically, the research fields have also changed, with a tendency for work in the new theoretical framework to concentrate on the contrast between Western society (particularly the United States) and East Asian societies (Japan, Korea, and China). It is argued that perceptual holism—and the cognitive style of interdependency in general—is more likely to be found in Asian communitarian cultures. Conversely, perceptual analytism—and the cognitive style of independency—is thought to be typical for American society, in which recognition of individual aspirations and self-agency prevail (Kitayama, Duffy, & Uchida, 2007; Markus & Kitayama, 1991, 2003; Nisbett, 2003).

Three kinds of empirical evidence have been presented to support this view. First of all, several studies employing the FLT have demonstrated that Americans perform more accurately in the absolute task and less accurately in the relative task in comparison to Eastern Asians (Duffy, Toriyama, Itakura, & Kitayama, 2009; Ji, Peng, & Nisbett, 2000; Kitayama, Duffy, Kawamura, & Larsen, 2003; Vasilyeva, Duffy, & Huttenlocher, 2007). It has also been demonstrated that Americans have greater activation of attention-related brain regions while performing the relative task, while East Asians have greater activation of these regions in the absolute task (Hedden, Ketay, Aron, Rose Markus, & Gabrieli, 2008; Ketay, Aron, & Hedden, 2009). This suggests that the relative task demanded greater attention (due to its perceived difficulty) than the absolute task in the case of Americans, while for East Asians the opposite was the case. However, at least one extensive study that employed the same test (Zhou, Gotch, Zhou, & Liu, 2008) demonstrated that both Americans and East Asians (Chinese) performed better in the relative task than in the absolute task, although the mean error made by the East Asians in the relative task was smaller than that of Americans. One explanation for this, which is consistent with the theory of independent/interdependent cognitive styles, is that although an intercultural difference in holism/analytism will manifest itself in differences in the absolute errors that are made in each kind of task, these differences may not always be great enough to change the rank order of the tasks themselves.

Second, several studies demonstrated that East Asians both notice and remember changes in contextual relations of objects in the visual field better than Americans (Masuda & Nisbett, 2001, 2006). They also have more difficulties then Westerners in recalling a previously seen object embedded in a novel context (Chua, Boland, & Nisbett, 2005; Masuda & Nisbett, 2001).

Finally, one study (Chua et al., 2005) demonstrated a difference in eye movements between Americans and Chinese while attending to a visual scene: During initial presentation of a scene, Chinese made more fixations on the background and looked longer before making the first fixation on the focal object than Americans, who, on the other hand, fixated on the focal object for a longer time than Chinese. However, at least two later studies (Evans, Rotello, Li, & Rayner, 2009; Rayner, Li, Williams, Cave, & Well, 2007) have failed to replicate these results, while one of them (Evans et al., 2009) also failed to replicate the difference between East Asians and Americans in recalling previously seen objects embedded in novel contexts. More research seems to be needed in order to clarify the possible effects of perceptual holism/analytism on visual memory and ocular locomotion.

The cross-cultural differences in the cultural models of self and the corresponding cognitive styles are usually explained by referring to social practices that prevail in a given society, among which the design of social institutions, religion, and ideology have been particularly mentioned (Kitayama et al., 2007; Nisbett, 2003; Nisbett et al., 2001). However, as in the earlier theoretical tradition, there has also been some interest in the impact of eco-cultural factors. One recent study identified differences in perceptual holism/analytism among nomadic pastoralists, fishermen, and sedentary agricultural groups in the Eastern Black Sea region of Turkey—explaining them on similar lines to those proposed by Berry (Uskul, Kitayama, & Nisbett, 2008). A third view is that differences in perception might be linked to differences in the perceptual affordances provided by the physical environments in which the corresponding cultures exist. Thus, one recent study (Miyamoto, Nisbett, & Masuda, 2006) demonstrated that both Americans and Japanese noticed more changes in the background of a culturally neutral visual scene if, before the experiment, they had been asked to examine a set of photographs of street scenes taken in Japanese cities. Conversely, both Americans and Japanese noticed fewer background changes if they had been primed by comparable photographs taken in American cities.

It is important to note that the recent studies on holism/analytism have paid relatively little attention to differences in holism/analytism between individuals as opposed to groups, and they have been criticized for this (Berry, Poortinga, Breugelmans, Chasiotis, & Sam, 2011;; van de Vijver, Chasiotis, & Breugelmans, 2011;). However, the existence of individual differences within cultural groups has been acknowledged by the recent researchers, and several of their theoretical works have stressed the need to study them (e.g., Kitayama et al., 2007; Markus & Kitayama, 2003).

1.2. Perceptual holism/analytism and artistic visualization

As can be seen from this short review, although researchers differ in their views concerning the factors and mechanisms that cause cross-cultural differences in object/context perception, the existence of these differences in perceptual style is established beyond any reasonable doubt. As emphasized both by Witkin and Berry (Berry, 1976; Witkin & Berry, 1975) and by Kitayama and Nisbett (Kitayama et al., 2007; Nisbett, 2003), these perceptual findings are likely to correlate with wider cognitive and behavioral differences between the cultures concerned. One important sphere in which a link between society and culture, on the one hand, and perceptual stimuli processing, on the other, may be observed, is the production of visual images. Masuda, Gonzalez, Kwan, and Nisbett (2008) have recently suggested that cultural differences in artistic styles might reflect the same tendencies toward greater or lesser context sensitivity as are measured by psychological tests of visual perception. This is a fascinating idea because, if correct, it would provide psychological backing for art-historical ideas about the connections between visual styles and other aspects of sociocultural development.

To test this idea Masuda and colleagues made certain assumptions about the connection between picture contents and perceptual style. They predicted that the distance between the bottom of the picture and the horizon line (which marks the upper border of the background space in a landscape drawing) would be greater in images produced by holistic (context-sensitive) individuals because they would need more space for background contextualization. Conversely, they predicted that analytic (context-independent) individuals would produce or prefer images in which the focus of attention—whether a person, a face, or an object—was comparatively large, allowing it to be represented in greater detail. In order to test these conjectures, they compared masterpieces of East Asian and Western art created between the 15th and the end of the 19th centuries, photographs and non-professional landscape drawings produced by contemporary Japanese and Americans, and their preferences for different photographic images. The test results confirmed that the Asian artists, and test participants, were more likely to produce or prefer the kinds of image predicted for holistic individuals, while the Western artists and participants gravitated toward the kinds of image predicted for analytic individuals.

Stimulating though it is, Masuda et al.'s argument is open to criticism. While previous research has indeed shown that East Asians (including Japanese) and Westerners (including Americans) do differ in perceptual processing, the differences in artistic styles observed by Masuda et al. might well be related to some yet unspecified aspects of the two societies and their cultures rather than to the differences in perceptual holism/individualism of their members. Some relevant aspects might be the different conventions of drawing learned during formal school education, early exposure to different visual styles, differences in the roles which visual images normally play (or played in the past) in the societies in question (e.g., a utilitarian role as interior decoration vs. a more esthetic role as works of art), different esthetic norms, etc. Differences in physical environments may also have an effect (as suggested by Miyamoto et al., 2006). Miyamoto and colleagues demonstrated an effect on perceptual processes, but it also seems possible that environmental factors might affect visual representations directly, without affecting perception along the way. For example, differences in landscape (flat vs. mountainous) can regularly expose members of two cultures to qualitatively different types of vista, characterized, among other things, by different perceived heights of the horizon. These differences might get reflected in the images they produce even if perceptual processing is the same in both groups.

One obvious way of showing that there really is a direct connection between perceptual processing and artistic style would be to demonstrate that the link between the parameters of drawing and the level of perceptual holism/analytism also holds at individual level and that the individual-level associations are strong enough to account for the existence of parallel intercultural differences. It should be possible to demonstrate a correlation between the level of perceptual holism/analytism and the position of the horizon/size of objects in drawings across individuals as well as across groups, and to prove that this correlation across individuals remains even after controlling for their group membership. This would provide strong evidence that the differences in drawing are due to individual characteristics (the stimuli processing style) that are unevenly distributed between the groups rather than due to group-level factors such as peculiarities of school education.

We can also conjecture that differing levels of perceptual holism/analytism will have implications for the very process of image production—that is, for the ways in which pictures are drawn. Thus, individuals with more holistic perception can be expected to prefer drawing focal objects embedded in a previously depicted context, while individuals with more analytic perception can be expected to start their drawings with detailed representation of the focal objects and add context later if at all. This prediction follows directly from Masuda's theory of the connection between perceptual holism/analytism and artistic style, and it can be tested empirically.

1.3. Aims and tasks of this study

This study aims to explore the relations between culture, perceptual holism/individualism and artistic visualization taking into account the theoretical and methodological suggestions discussed above. In particular, our aim is to test the theoretical claim (derived from Masuda et al., 2008) that certain differences in freely drawn landscape pictures—namely those in the position of the horizon and the size of objects—as well as in drawing strategies—namely the order in which foreground and background elements are drawn—a) can be observed between cultural groups and b) can be related to the differences in perceptual processing—namely the relative levels of perceptual holism/analytism—that exist between the members of these groups. We also wish to make a new test of the theoretical claims by Berry (Berry, 1976; Witkin & Berry, 1975) and Uskul et al. (2008) that the differences in perceptual processing that exist between cultural groups can be related to ecocultural characteristics of these groups, particularly to their economy and way of life.

To test these claims, we compare the individual levels of holism/individualism and the landscape drawings of teenagers originating from two separate regions and three cultural groups:

  1. The indigenous and predominantly nomadic Nenets living in the Tazovsky district, Yamal-Nenets Autonomous Province of the Russian Federation (northwestern Siberia).
  2. The incomer population (composed of ethnic Russians and others) in the same district.
  3. The indigenous, predominantly settled sea hunting and reindeer herding population (Chukchis and Yup'ik Eskimos) in several settlements in Chukotski Autonomous Province of the Russian Federation (northeastern Siberia).

These groups differ in their mother tongue, economic activities, and the way of life (nomadic, sedentary, or semi-sedentary—see the next section). However, their members share the same national identity (Russian Federation), are subjected to the same political and administrative system and attend the same system of formal school education. Besides that, the non-indigenous and the Nenets children in our study were born and have grown up in similar ecological conditions. Therefore, in contrast to many previous studies, the present inquiry fits better to the “just a minimal difference” approach (Cohen, 2007) and allows more precise testing of the causal role of eco-cultural (e.g., economic activity and way of life) and purely ecological factors in the formation of cross-cultural differences in perceptual holism/analytism. If the predictions are fulfilled, our experimental design would also demonstrate that the cultural association between picture characteristics and psychological measures is not limited to the familiar North American/East Asian comparison but applies to other intercultural comparisons as well.

We use our comparison to test the following four predictions that, as the discussion above suggests, should be true if theoretical claims under question are to be rigorously supported:

  1. The amount of contextual information in the drawings, as reflected in the position of the horizon and the size of objects, correlates with individual levels of children's perceptual holism/analytism: the higher the level of perceptual holism, the smaller the size of objects and the higher the horizon.
  2. The group-level differences in drawing scores follow the same pattern as those for the visual-processing scores, and this fact can be explained by the individual-level relationship between the two sets of scores.
  3. The members of more sedentary groups have higher levels of perceptual holism in comparison to the nomadic pastoralists. They are also expected to include more contextual information in their drawings.
  4. Individuals from groups with higher mean levels of perceptual holism (lower mean levels of perceptual analytism) are more likely to draw the background elements first and add the foreground elements later. In contrast, individuals from the groups with lower mean levels of perceptual holism (higher mean levels of perceptual analytism) are more likely to draw foreground elements first and add background later.

In the following sections, we first provide details of our study regions and groups—focusing on ecological and cultural features (e.g., ecological conditions, type of economy, type of authority, differentiation of social roles) that previous studies have used to explain cross-cultural differences in object/context sensitivity. Next, we describe our study, which collected drawings from all three groups and correlated their characteristics with the results of a test of visual perception, completed by the same individuals. Finally, we discuss our findings in order to show how far group differences—in both test results and drawing styles—can be explained by differences in the economic and ecological circumstances of the three study populations.

2. Study regions and study groups

This study was performed in two geographic regions: (a) Tazovsky district, Yamal-Nenets Autonomous Province, northwestern Siberia (the study site 1) and (b) Chukotski Autonomous Province (Chukotka), northeastern Siberia (the study site 2) of the Russian Federation. Two study samples—those of non-aboriginal and Nenets teenagers—were collected in the study site 1, while the third study sample—that of Chukchi/Yup'ik teenagers—was collected in the study site 2.

The study site 1 (Tazovsky district) represents a typical flatland tundra. The region has a very flat and monotonous landscape with no natural elevations higher than 10 m, but with many natural depressions formed by a dense system of rivers and streams. Vegetation consists mainly of lichens, grasses, and sedges as well as of bushy willow stands on the southern slopes of hills and natural depressions. Trees are absent except in several small “islands” of coniferous forest on the banks of large rivers. Therefore, the environment normally provides vast vistas in all directions, objects can be seen from large distances, and the line of horizon is always present.

The indigenous population of the region—Nenets—are predominantly nomadic reindeer herders and fishermen. Most of them live in small and predominantly kin-based groups (nomadic camps) that migrate through the tundra during all or most of the year. Although each camp has an informal leader, no traditional authority exists beyond that level. No differentiation of social roles apart from those based on gender is evident (see Donahoe & Istomin, 2007; Dwyer & Istomin, 2008; Istomin & Dwyer, 2010 for further details on Nenets economy and way of life). Our observations suggest, however, that nearly all Nenets who are currently younger than 30–35 years have been subjected to at least 6 years of compulsory education in a boarding school. Correspondingly, all younger Nenets speak fluent Russian and seem to be able to read and write in this language. At the moment, a growing number of Nenets households settle in several permanent settlements that exist in the region and find employment as hired labor, usually low paid. However, as yet this process of sedentarization has involved only a minority of the Nenets.

The non-indigenous population of the district consists mostly of recent newcomers who have settled there during the last 35 years, since the beginning of work to extract the region's gas deposits. The non-indigenous population resides mostly in permanent settlements (where it constitutes the absolute majority of inhabitants) and, partly, in gas drilling bases. While the majority of newcomers are ethnic Russians, this group also includes representatives of ethnic minorities living mainly in the central and southern parts of the country, such as Ukrainians, Tatars, and Bashkirs. All the newcomers use the Russian language for daily communication. Many of them are employed in the gas drilling industry, while the rest work in the rapidly growing sectors of services and small-scale commerce, pre-school and school education, and administration.

The study site 2 (Providensky and Chukotski districts of the Chukotski Autonomous Province) is a rather mountainous region with relatively diverse landscape and vegetation. The province represents a mosaic of seashore, coastal marshes, basins, and mountain ridges of different sizes. Areas densely overgrown with mid-size bushes of alder are intermixed with areas of mountain tundra covered by grasses and sedges. Everywhere, mountains form an essential part of the vista sometimes closing it and often making up the line of the horizon.

The native population of the region consists of Chukchis and Yup'ik Eskimos. The majority of the participants in the present study were drawn from three coastal communities: Novoe Chaplino (about 400 inhabitants), Yanrakynnot (with population of ca. 350 inhabitants), and Lorino (around 1,400 residents). About 5% of the overall population in these communities still practice the traditional economic activities that include hunting sea mammals (whales, seals, and walrus) and certain species of seabirds, small-scale inland hunting as well as sea and lake fishing. Besides that, reindeer herding is present in two of the communities (Lorino and Yanrakynnot). In contrast to the aboriginal population of Tazovsky district, the hunters, fishermen, and reindeer herders in this study site are settled or semi-settled, working in the tundra on seasonal duty. The absolute majority of indigenous locals are either employed in the non-customary state-owned economy (local school, administration, Housing Management) or officially unemployed. Kinship ties continue to play an important role in social organization. Kin group elders have informal authority over the members of their groups and some of them (e.g., an old experienced hunter or a wise woman—a visionary) can be highly regarded by the whole community. Division of roles is essential among hunters; here, the criteria are age, experience, and current standing in terms of successful catch. In other spheres, acquired social position prevails over the ascribed one. Gender and age affect social roles, space division, behavior patterns, choices, and taboos. All the children and youngsters in the hamlets receive formal education according to the Russian Federal standards and speak fluent Russian. Indeed, Russian is now the mother tongue of Yup'ik children of school age.

3. Procedure, methods, and results of the study

3.1. Procedure and methods

The data for our main study were collected in February and March 2011 in Tazovsky (study site 1), and in April and May 2011, in Chukotka (study site 2). The study sample consisted of teenagers attending two public schools (Tazovsky boarding school and Tazovsky secondary school) in study site 1 and three public schools belonging to the three communities described in the previous section in study site 2. The Tazovsky boarding school represents the main facility providing obligatory basic school education for the children of nomads. These children are transported there from the tundra each autumn, once the school year begins, and live there at the state's expense till the end of the school year. Since the school provides classes of Nenets as mother language and has a curriculum adapted to the needs of children from aboriginal families, it is attended also by many Nenets children whose parents live permanently in the town. Nearly, all children from non-aboriginal families, in contrast, attend the Tazovsky secondary school with Russian as the only language of instruction. In Chukotka, the three schools involved in the research are attended by the children from the corresponding communities. Since the communities themselves are predominantly indigenous, most of these children are from indigenous families. After all the cases of incomplete data were excluded, the sample consisted of 183 cases: 81 cases of Nenets children (27 boys and 54 girls) attending the Tazovsky boarding school (referred to below as the Nenets sample); 80 cases of non-Nenets children (37 boys and 43 girls) attending the Tazovsky secondary school (referred to below as the Non-indigenous sample); and 22 indigenous children (14 boys and 8 girls) attending schools in Chukotka (referred to below as the Chukchi–Yup'ik sample). The children in the Tazovsky samples were aged between 11 and 14 (i.e., 15 was the upper limit), while the Chukchi sample was aged between 10 and 16. The mean age for all three samples was between 12 and 13.

In the course of the study, the children were asked to complete a questionnaire containing background questions about the age and gender of the child, his/her ethnic background, and the profession of his/her parents as well as two experimental tests. In Tazovsky, the children completed these questionnaires at school during one of their regular classes. In study site 2, the children were tested in their home environment after the classes. The experimenter was always present during the testing, gave necessary instructions and explanations, and observed the process of performing the tests.

The tests offered to the children included a simplified version of the FLT developed by Kitayama et al. (2003) and a landscape drawing. The FLT was used to assess the level of perceptual holism of the children by testing their performance in two experimental tasks. In both tasks, the children were shown a square stimulus frame with the side length of 150 mm, within which a 50 mm vertical line was printed. The line was extended downward from the center of the upper edge of the square. The children were also given a set of eight smaller response frames representing squares with the side lengths of 60 mm. In the first of the tasks—the absolute task—the participants were instructed to draw lines in each of the response frames that were of the same absolute length as the line in the stimulus frame. In the second, relative task, the children were instructed to draw lines whose proportions to the sizes of the response frames were the same as the proportion of the stimulus line to the size of the stimulus frame. The children were instructed to use only pencils while performing these tasks and to remove all the other tools, including rulers and erasers, from their desks.

Note that in contrast to the original FLT test as developed by Kitayama and colleagues, we did not vary the relative sizes of the stimulus and response frames and the children had the stimulus frame in front of their eyes while drawing lines in the response frames. These deviations from the original variant of the test were introduced to enable us to give the test as a part of the questionnaire and to cope with the rigid time limitations that we faced: In the study site 1, we were allowed by the school administrations to use only one regular class hour—lasting 40 min—for testing children in each of the study units. We believe that these modifications in the standard test procedure would not affect the general tendencies in responses, although they might make these tendencies less visible. In each test, only the lines drawn in the last four response frames were assessed as the response, while the rest of the response frames were ignored as training ones.

The main assumption behind the FLT is that, in comparison to the individuals with analytic processing, the individuals with holistic processing can be expected to have more difficulty in abstracting the line from the frame as needed to perform the absolute task, but less difficulty in considering the line in relation to the frame as needed to perform the relative task (Kitayama et al., 2003). So, the FLT produces two scores for each individual: the error in the absolute task, which measures difficulty with context-independent (analytic) judgment; and the error in the relative task, which measures difficulty with context-sensitive (holistic) judgment (note that in both cases the error itself is measured in absolute terms). In our analysis, we used the difference between these two FLT scores (the error in the absolute FLT task minus the error in the relative FLT task) as a single continuous variable that would assess and meaningfully compare the levels of perceptual holism/analytism of individuals and could be correlated with picture-based measurements. Note that the value of this variable increases with the absolute task error and decreases with the relative task error and, therefore, individuals with holistic perceptual styles will score more highly.

The second task consisted of drawing landscape scenery on regular paper (A4 paper format 210 mm × 297 mm). The children were asked to draw a tundra landscape, which would contain four compulsory elements—a human figure, a traditional dwelling (nomadic or hunting tent), a water area (such as river, lake, or sea), and hills or mountains as well as additional elements children themselves would find necessary. Note that the children were not directly instructed to draw the horizon line. Some examples of pictures produced by children from different groups are present on Fig. 1.

Figure 1.

Examples of pictures produced by Chukchi/Yup'ik (A1-A4), Nenets (B1-B4), and non-indigenous (C1-C4) children.

The Russian text in the upper part of each picture formulates the drawing task for the children. (“Please draw a tundra landscape keeping the sheet horizontal. Please include four obligatory objects into the drawing: a traditional tent, waters, sopki (this can mean both hills and mountains in Russian), and a human being. You can include any other objects if you like.”)

To process the data, we measured the distance between the bottom of a drawing and the horizon line (defined as the line separating the ground from everything else that lies above it) on the left and on the right sides of each picture and took the average of these two measurements as the mean height of the horizon. After these measurements were taken, all the pictures were digitalized using the same size (4,665 × 3,292 pixels) and resolution (400 dpi). Then the sizes of human figures on the pictures were measured. To do this, the number of pixels inside the contour of the human figures was calculated by means of the Adobe Photoshop tool “Histogram.” If the drawing contained several human figures, the figure with the greatest number of pixels—that is, the largest figure—was used for the further analysis. This choice is justified, since the smaller figures could arguably represent a part of background against which the bigger figure was drawn rather than foreground elements.

In 26 drawings (18 of them drawn by non-indigenous and 8 by indigenous children), the horizon line was missing. In these cases, we took its height as being zero and concluded, therefore, that the picture did not contain any contextual information. Two reasons accounted for this methodological decision. First of all, the horizon line in the treeless tundra environment represents such a pervasive element of the visual background that the failure to depict it represents very likely the failure to include any contextual information. Second, our assessment of the pictures with the missing horizon line revealed that 12 of them did not contain any other elements except those claimed by the task as obligatory. These elements were typically drawn next to each other in the lower part of the picture or “hanging in the air” in various parts of the paper—indicating that no attempt had been made to relate the elements to a background. In three other pictures, the only optional element included was the sun.

As with the results of the FLT, the measurements of the horizon and the size of the human figure in each drawing have been collapsed into a single index on which the object/context relations can be meaningfully compared across drawings. To do this, the horizon and human figure measurements were first divided each by its own standard deviation to bring them to the same scale. Then, the human figure score was subtracted from the horizon score to form what we have called the Masuda Pictorial Index (MPI). Note that positive values of the MPI reflect a comparatively context-based orientation, and negative values a more object-oriented approach.

Finally, to obtain comparative data on the use of the alternative drawing strategies, two groups of children from study site 1 were selected. The groups represented separate study units (classes) of the Tazovsky secondary school and the Tazovsky boarding school, respectively. The group from the secondary school included 14 children, while the group from the boarding school included 12 children. The reason for choosing exactly these two groups was that the groups, as a pre-test survey of their members showed, were mono-ethnic, that is, all the children in one of the groups were of indigenous background, while all the children from the other group were of non-indigenous background. It should be noted, however, that the non-indigenous children from the secondary school group were about 1 year older then the indigenous children from the boarding school group (Mages 14 and 13, respectively). Both groups were administered the two tests described above. As the children performed the drawing task, the researcher carefully observed the drawing strategies they adopted and counted the number of individuals who started their drawing from background elements (the line of the horizon, surface objects far away, sun, and clouds) and proceeded to foreground objects later as well as the number of individuals who started from the foreground elements (the human figure, the house, or the river/lake) and proceeded to background elements later. In each group, the order of drawing of three randomly selected individuals was recorded by means of a regular automatic photo camera. Fig. 2 shows examples of (a) “object first” strategies and (b) “background first” strategies redrawn from the photographs taken during the task.

Figure 2.

Examples of the “object first” (A1-A4) and “background first” (B1-B4) drawing strategies.

3.2. Results

The data obtained during the study were used to test the four predictions outlined in section 'Aims and tasks of this study' of this article. The following results were obtained:

3.2.1. Prediction 1

The amount of contextual information in the drawings, as reflected in the position of the horizon and the size of objects, correlates with individual levels of children's perceptual holism/analytism: the higher the level of perceptual holism, the smaller the size of objects and the higher the horizon.

To test this prediction, we investigated the correlation between the results of the FLT (the difference between errors in the absolute and the relative tasks) and the picture-based scores (size of the human figure, mean height of the horizon, and the MPI) across the whole sample as well as in each of the eco-cultural groups. Table 1 reports the correlation coefficients and their statistical significance obtained for the whole sample.

Table 1. Correlations between the difference of the two FLT measures and the pictorial measures
 Size of the Human FigureMean Height of the HorizonMasuda Pictorial Index (Horizon−Human Figure)
  1. Note. *Correlation is significant at 0.05 level, **correlation is significant at 0.01 level.

Difference of FLT measures (absolute−relative)−.170*.208**.238**

As can be seen from this table, the results of the FLT correlate significantly (although admittedly weakly) with all the three picture-based scores and the directions of the correlations are consistent with our prediction. The correlation between the FLT results and the MPI has the highest coefficient, showing that the MPI that we constructed not only summarizes but also slightly improves the explanatory power of the picture-based measures. Overall, these results support our prediction.

3.2.2. Prediction 2

The group-level differences in drawing scores follow the same pattern as those for the visual-processing scores, and this fact can be explained by the individual-level relationship between the two sets of scores.

To test this prediction, we investigated how well the results of the FLT across our sample are explained by the between-group distribution of the MPI and the within-group correlations between the MPI and the FLT results. We chose this strategy rather than investigating, as might seem more logical, how well the results of FLT and their between- and within-group distributions explain the MPI, because the exploratory analysis of our data suggested that the pictorial variables showed greater differences in standard deviation terms than did the FLT results (see Table 3 below). If, in a linear model, two variables exhibit similar intercategory differences but the differences in variable A are greater in standard deviation terms than those of variable B—then the differences in variable A might possibly provide a full explanation of the differences in variable B, but variable B could only provide a partial explanation, at best, of the differences in variable A. For this reason, we decided to reverse our analysis and to see whether we could use the differences in picture measurements (summarized by the MPI) to explain the differences in the FLT scores.

To test how well the between-group distribution of the FLT scores is explained by the between-group distribution of the MPI (combined with the within-group MPI-FLT correlations), we compared two general linear models. In the first model, the difference between FLT task errors is predicted by group membership alone, while in the second model the MPI is added as a covariate. If this addition removes the between-group effect, we can conclude that the original effect was indeed due to unmeasured intergroup differences in the MPI.

Table 2 presents and interrelates the relevant findings. The first column sets out the intergroup comparisons for the MPI —with the Non-indigenous group as the base category. The Chukchi–Yup'ik show the highest value for context-orientation, but the difference between Nenets and Non-indigenous scores is also significant at the .001 level. The FLT models themselves are presented in the next two columns—starting with the statistics for the “groups-only” model. Both the overall group effect in this model and the contrasts between the groups are significant at the .05 level. Note that the pattern of the between-group contrasts is the same as for the MPI.

Table 2. Statistics of the intergroup comparison for the MPI and of the two general linear models that predict the difference between FLT task errors on the basis of the group membership alone (left column) and the group membership plus MPI (right column)
ModelMPIDifference of the FLT Error Measures
Groups OnlyGroups OnlyGroups and MPI
Base number183183183
Intergroup F-statistic (significance)33.879 (<.001)3.685 (.027)0.729 (.484)
Regression parameters:   
Masuda index (significance)1.540 (.031)
Chukchi–Yup'ik (significance)2.261 (<.001)7.443 (.019)3.961 (.258)
Nenets (significance)1.395 (<.001)4.167 (.044)2.018 (.373)

The third column sets out the results for the model that also includes the Masuda index as an explanatory covariate. The parameter value for the MPI effect, and more particularly its significance level, is perhaps the most important result here—since it derives from pooled estimates of the within-group association between the pictorial and test variables. The fact that the association holds within, as well as between, the ethnic groups is consistent with our prediction and fills the main gap in the earlier argument by Masuda and colleagues (Masuda et al., 2008) by demonstrating a direct association at individual level between pictorial style and test-based measures of cognitive orientation.

The final step is to see whether including the regression relationship with the MPI enables the model to explain the intergroup differences in FLT scores. The results confirm that it goes a long way toward doing so. The overall intergroup F-statistic is reduced by over 80%, moving from significance (p = .027) to well below significance level (p = .484). The estimated differences between the group values are also substantially reduced and cease to be statistically significant. Readers can check that these changes really are due to the regression relationship with the MPI scores, by multiplying the MPI score for the group concerned by the Masuda regression coefficient and subtracting the product from the area value in the second column. In each case the result is equal to the residual group effect shown in the third column. These findings are consistent with our prediction.

However, despite being reduced by about half, the estimated intergroup differences do not completely disappear. So, it is reasonable to ask why the regression relationship does not explain all of the association. One technical factor, which may play a role here, concerns measurement error in the MPI itself. There is bound to be some measurement error in the picture scores, in the sense that different drawings produced by a given individual would not have exactly the same characteristics every time—and would thus produce variable estimates of the psychological preferences underlying the choice of style. The effect of this variation will be to flatten the regression slope connecting the MPI with the FLT score, thus leading to some under-estimation of the contribution that it makes to the area-level FLT score. We return to this point in the overall discussion below.

3.2.3. Prediction 3

The members of more sedentary groups have higher levels of perceptual holism in comparison to the nomadic pastoralists. They are also expected to include more contextual information in their drawings.

To test this prediction, we used one-way anovas to investigate the differences between the three eco-cultural groups in our sample in their mean FLT results and picture-based scores. Table 3 summarizes the results of these tests by reporting the group means and the overall sample mean for each score, the overall standard deviation of each score, the corresponding F-statistics and their significance levels. The final rows of the table give significance levels for the pairwise comparisons between the three eco-cultural groups, using two different tests. The t-test is the same as that used in the regression models shown in Table 2, while Tukey's HSD (Honestly Significant Difference) test applies a stricter criterion that corrects for the risk of over-reporting significant differences in the context of multiple comparisons. (Note that some of the results for the MPI and for the difference of FLT scores are also reported in Table 2 above).

Table 3. Statistics of intergroup comparisons for the pictorial measures and the difference between FLT task errors
Sample N Size of the Human Figure (in 1000s of Pixels)Mean Height of the HorizonMPIDifference of the FLT Measures (Absolute−Relative)
Overall M 183192.395.5.792.07
Overall SD 203.855.21.5913.22
F-statistic (2/180 d.f.) 19.69318.64733.8793.685
Significance of F-statistic <.001<.001<.001.027
Significance of specific intergroup comparisons: Chukchi–Yup'ik/Nenets
 t-test .
  Tukey-HSD .458.018.024.549
 t-test <.001<.001<.001.019
  Tukey-HSD <.001<.001<.001.049
 t-test <.001<.001<.001.044
 Tukey-HSD <.001<.001<.001.108

As can be seen from this table, all the variables place the three groups in the same order—with the Chukchi–Yup'ik children being most holistic in their perception and context-oriented in their drawings and the non-indigenous children being the most analytic in their perception and object-oriented in their drawings. Most, but not all, of the pairwise comparisons are significant in their own right. The F-statistics confirm that the overall intergroup differences are all statistically significant—at the .05 level for FLT score, and at the .001 level for the three pictorial measures.

Although the differences between the group means on all the variables consistently put the groups in the same order, this order does not completely coincide with our prediction. Though the children of the partially nomadic Nenets do indeed score more highly for analytic perception, and include less context in their drawings, than do the more settled Chukchi–Yup'ik group (and this difference is statistically significant in the case of the drawing scores), the non-aboriginal children, who are arguably the most settled as far as their way of life is concerned, demonstrate the highest levels of both perceptual analytism and object-orientation in drawing in the whole sample.

3.2.4. Prediction 4

Individuals from groups with higher mean levels of perceptual holism (lower mean levels of perceptual analytism) are more likely to draw the background elements first and add the foreground elements later. In contrast, individuals from the groups with lower mean levels of perceptual holism (higher mean levels of perceptual analytism) are more likely to draw foreground elements first and add background later.

To test this prediction, we compared the observations of drawing strategies made in the two groups of non-indigenous and Nenets children in Tazovsky district. These observations showed that 8 of the 12 Nenets children (67%) in the boarding school group used the “background first” strategy and only four children used the “objects first” strategy. In contrast, only 4 of the 14 non-aboriginal children in the secondary school group used the background first strategy, while 10 children (71%) used the “object first” strategy. This suggests a difference between the groups in the preferred strategy of drawing (Pearson chi-square test is significant at p = .05). Since, as has been demonstrated above, the Nenets children as a group consistently demonstrated higher perceptual holism in comparison to the non-indigenous children, these results are fully consistent with our prediction.

4. Discussion

The data obtained in our study confirm most of our hypotheses—but not entirely in the ways expected. The surprises have been

  1. the low to moderate correlations between the FLT score and the pictorial measures,
  2. the fact that the intergroup differences in the picture-based MPI scores were far larger (in standard deviation terms) than those for the FLT score, and
  3. the falsification of our prediction that the partially nomadic Nenets group would have the highest average score for analytism (context independence).

Before we can assess the overall significance of our findings, we need to account for these surprises. We will deal with points A and B together, and then go on to consider point C.

The weak-to-moderate correlation between the MPI and the FLT score suggests that the MPI-picture scores are indeed associated with levels of holism/analytism—but that either (a) the association is weak or (b) our version of the FLT is an imprecise measure of holism/analytism. There are two reasons for considering that (b) might be correct. The first is that any measure of a psychological variable is only an indicator of the underlying quality, not the quality itself. The second reason is that, as explained earlier, our own version of the FLT was a simplification of the original FLT, designed to fit easily into the questionnaire booklets used in this study. This modified version is likely to provide less precise results than would be obtained using the full test procedure.

In this situation, the hypothesized relationship between artistic styles and holism/analytism is empirically equivalent to stating that the MPI and FLT scores are both approximate indicators of the unobservable underlying values of holism/analytism. If this hypothesis is correct, what relationships would we expect to find between the two observable variables? There are in fact four relevant points, which include three testable predictions.

  1. There will be some individual-level correlation between the two variables. As we have seen, this prediction is confirmed by the findings in Tables 1 and 2 above.
  2. Both variables will show similar intergroup differences (since both variables reflect the underlying intergroup differences in holism/analytism values). This prediction is also confirmed—by the patterns of intergroup differences shown in Tables 2 and 3.
  3. The intergroup differences will be greater (in standard deviation terms) in the case of the indicator which best represents the underlying values of holism/analytism. As the figures in Table 3 show, the MPI measure emerges as the stronger indicator according to this criterion.
  4. The individual-level relationship between the two indicators, combined with the intergroup differences in the stronger indicator, should be able to explain (in a statistical sense) much—but not necessarily all—of the intergroup variation in the weaker indicator. This prediction is also confirmed: As the results in Table 2 show, the individual-level relationship between the MPI and FLT scores explains around 50 percent of the intergroup variation in FLT scores. (If the stronger indicator corresponded exactly to the underlying holism/analytism values, it would explain all of the intergroup variation in the weaker indicator. The fact that the MPI explains about half of this variation suggests that it provides a good, but not perfect, approximation to the underlying holism/analytism values)

The hypothesized relationship between artistic style and holism/analytism levels is thus confirmed as strongly as it could be—given that neither indicator corresponds exactly to the underlying values of holism/analytism.

This leaves us with the problem of explaining why the MPI provides a better indication of underlying holism/analytism levels than is given by the FLT results. We have already identified one possible reason—namely that the simplified version of the FLT which we used in our study could not be expected to produce a very precise assessment of perceptual holism/individualism. This explanation would be equally compatible with the psychological differentiation theory and the theory of independent/interdependent cognitive styles. However, another potential explanation, which is specific to the cognitive styles view, would be that cultural preferences for in- or inter-dependence lead to different perceptual priorities (identified by the picture-based MPI), which in turn lead to different patterns of perceptual processing skills (assessed by the FLT). Since the picture-based measures are closer in the causal chain to the underlying cultural distinction, it is not surprising that they provide a sharper distinction between the different cultural groups. If this explanation is correct, the superior performance of the MPI would remain, even if we repeated the experiment with a more precise version of the FLT. We cannot of course know what the results of this hypothetical experiment would be—but since the prediction depends on the cognitive style view, the results would provide an empirical assessment of the otherwise rather elusive distinction between that view and the earlier psychological differentiation theory.

The fact that our prediction 3—namely that the children of nomads would have higher analytism scores than those of sedentary populations—was refuted by our observations (point C above) is not particularly surprising given that, in our study, the distinction between nomadic and sedentary groups overlaps with the distinction between aboriginal and non-aboriginal populations. Our study shows that the non-aboriginal children have more analytic perception than the children from both aboriginal groups, which replicates the results previously obtained by Berry (1976) and others (see Witkin & Berry, 1975 for a review) on aboriginal and non-aboriginal populations in Australia and Northern America. According to Witkin and Berry (1975), this difference can be related to a higher degree of psychological differentiation among the non-aboriginals, which, they argue, is due to the effects of formal school education and a tendency of industrial economies to foster socialization toward assertion. In the terms favored by Kitayama one could argue, consistently with much ethnographic evidence, that this effect is due to the high levels of informal cooperation found in many indigenous societies.

An alternative explanation of this difference would involve the fact that the non-aboriginal group in our study consists mostly of recent migrants. As was mentioned, although in a different context, by Kitayama et al. (2007), people who leave their homeland to try their luck in frontier regions with a harsh environment and a fairly undeveloped infrastructure can be expected to have a rather distinctive set of personality traits. They are likely to be less dependent psychologically on the established network of social relationships, more ambitious, and more willing to take risks in comparison to the majority of those whom they leave behind. In other words, migrants to frontier regions are likely to be self-selected for an unusually independent and assertive sense of the self. This would in turn affect the socialization practices they employ, leading their children also to acquire a distinctly analytic (field independent) perceptual style—an explanation which would be consistent with both the psychological differentiation and cognitive style theories.

Turning to the comparison between the two indigenous groups, although the difference between the Chukchi–Yup'ik and Nenets scores was consistent with our third prediction, it too raises some theoretical problems. Though currently sedentary, the Chukchi–Yup'ik group has a history of herding and hunting and some individuals still hunt. The arguments for high analyticity among nomads also apply to hunters (Uskul et al., 2008; Witkin & Berry, 1975), and high levels of context independence have indeed been reported for Yup'ik of St. Lawrence Island and Inuit subgroups in Alaska, Canada, and Greenland (Berry, 1966, 1971). For this reason, we might have expected the Chukchi–Yup'ik children in our study to have similar analytism levels to the children of the nomadic Nenets—instead of which they recorded substantially more holistic scores on both the FLT and MPI measures.

We should not make too much of this result since the number of individuals in the Chukchi–Yup'ik sample is small—just 22 school children. What is more, there are at least two possible explanations: the impact of sedentarization on the transmission of hunting-related attitudes and skills, and the visual affordances of the Chukotka landscape, which, as the earlier descriptions suggest, offers richer, more diverse, and arguably more chaotic perceptual scenes in comparison to that of the Nenets—and so would be expected to foster more holistic visual perception among the Chukchi–Yupik group (Miyamoto et al., 2006).

Nevertheless, since the existing literature offers conflicting predictions concerning the respective levels of holism/analytism of the groups in our study, as well as several alternative explanations for the differences in perceptual processing that were actually observed, it is clear that further research is needed in order to choose between the competing theories.

5. Conclusions

The primary aim of the research presented in this article has been to investigate how culture and society can influence both the production of visual images and test-based measures of perceptual processing—and then to use the results to help explain cultural differences in artistic style. To do this, we attempted to investigate directly the relation between the levels of perceptual holism/individualism measured in frame-line tests and particular aspects of free drawing, namely the height of the horizon line and the size of objects, both across individuals and across groups. This approach, in our opinion, has important advantages over the previously used approach of simply juxtaposing intergroup differences in perceptual tests with intergroup differences in drawing style. Our approach allows us to assess the extent to which interpersonal differences in artistic style (captured by the characteristics of free drawing) are associated with differences in visual judgment (captured by the FLT) and to demonstrate that this association is capable of producing parallel sets of inter-area differences in both kinds of measure.

Consistently with the previous research (Masuda et al., 2008), our findings show that certain characteristics of free drawings—the height of the horizon and the size of focal figures—as well as drawing strategies (the order in which objects and background are drawn) differ between cultural groups and that these differences are related to differences in perceptual processing—namely the relative levels of perceptual holism/analytism that exist between the members of these groups. Furthermore, our findings also suggest an interpretation of the association—according to which artistic style reflects underlying cognitive orientations more precisely than do the measures of visual judgment captured by our version of the FLT.

These results are consistent with theoretical views expressed by Nisbett, Kitayama, and colleagues that the uneven distribution of individual test scores across groups are ultimately the product of different patterns of attention, associated with cultural differences in cognitive style. Several alternative explanations for these cultural differences—focusing on such factors as the way of life, type of subsistence economy, characteristics of the natural environment, and aspects of social structure—are consistent with our data. Additional research is needed to choose between these explanations. Generally, our results demonstrate that an inquiry into relation between perceptual processing and artistic style has a great potential.


The study presented in this article was conducted for the Collaborative Research Centre (SFB) 586 “Difference and Integration” of the German Research Foundation (project E86 coordinated by Prof. Dr. Günther Schlee) and the framework project “Conditions and Limitations of Lifestyle Plurality in Siberia” designed by Dr. Joachim Otto Habeck at the Siberian Studies Centre at the Max Planck Institute for Social Anthropology in Halle, Germany (Directors—Prof. Dr. Chris Hann and Prof. Dr. Günther Schlee, Coordinator—Dr. Joachim Otto Habeck). The authors thank their numerous Russian, Nenets, Chukchi, and Yup'ik collaborators, in particular Alexandr Liakhov, the Tyseda extended family, Antonina Yatta, Dmitriy Sigunylik, and Vitaly Rannav. We also express our gratitude to the Department of Education in Tazovsky District, Yamal-Nenets Autonomous Province (Director Larisa Salomatina), and to the administrations of the Tazovsky secondary school and Tazovsky boarding school. We owe thanks to our colleagues Prof. Dr. Chris Hann, Prof. Dr. Günther Schlee, and Dr. Joachim Otto Habeck from the Max Planck Institute for Social Anthropology in Halle.