(Reprinted from: Proceedings for the 32nd IFLA World Congress, Bankok, Tailand. 273-76,1995) INTRODUCTION The dramatic development of national and international tourism in China beginning in the early 1980s has had strong impacts on the landscapes as well as on the profession of landscape architecture. "Destructive construction" (Zhu 1982; Chen 1988; Sun 1985; Gan 1988) has become a serious problem for national scenic and historical landscapes. Many factors may contribute to this "destructive construction" in the process of tourist development. It is hypothesized that some of the "destructive constructions" is caused by cognitive problems, namely the constructors may be unaware of the destructive nature of their work, or even look upon these constructions as improvements in the landscape. They might believe that what they perceive as the most attractive constructions must be equally attractive to the tourists who travel a long distance to see these parks. In most cases national parks are located in remote areas, and tourist development in these areas is usually carried out by local people who directly manage the landscape. It is thus reasonable to hypothesize that the local people who develop and make changes in the landscapes might have different preference and change models than the tourists who actually use the landscapes. Testing this hypothesis is the main objective of this research. PROCEDURES Preference Rating In total 50 different scenes (color slides) were selected for this study, randomly coded from 1 to 50. Scenes are from the Red Stone National Park in South China's Guangdong Province. The slides were expected to be most representative of the diverse landscape features and spatial dimensions seen in this park. Twenty nine groups, in total 573 subjects, were invited to participate in the experiment (Yu 1995); one of the groups is composed of graduate students at Harvard Graduate School of Design (GSD, n=24), all other twenty eight groups were composed of Chinese subjects. These participants have very different cultural backgrounds in terms of macro-cultural influence (Western vs. Chinese), landscape expertise (experts vs. general public), living environment (rural vs. urban), educational level (high vs. low) and gender (male vs. female) Using the SBE procedure (Daniel and Boster. 1976), members from each of twenty-nine groups were asked to rate the 50 scenes (slides) on 1 to 10 point scale. Each slide was shown for 10 seconds. The SBE procedure has been widely recognized as an effective and reliable procedure (Hull, Buhyoff et al. 1984; Schroeder 1984; Hull 1986).. Experiments on the Chinese groups were carried out in Beijing, China; the experiment on the GSD(24) was carried out in Cambridge, MA, USA. Correlation and factor analyses were carried out to investigate the cultural and sub-cultural variations in landscape preference (Yu 1995). Landscape Categorization Based on Preference Ratings In my case, an empirical approach to landscape categorization was used. This approach uses observers' preference ratings as indicators to categorize landscapes. This approach has become more and more recognized and proven to be meaningful and practically more relevant than others (Kaplan 1985; Amedeo, Pitt et al. 1989) Underlying this approach is the presumption that some physical features of landscapes and their combinations significantly influence the scenic value (preference level) discrimination; furthermore, observers have internally consistent responses to certain landscape features and their combinations. The commonly used method in this approach to landscape categorization is factor and cluster analysis. Using the preference ratings for 50 individual slides by twenty-nine subject groups,. a principle component factor analysis model using a varimax rotation method was used to interpret and categorize the landscapes. Preference Models: Regression with Dumming Variables On the basis of landscape categorization using factor analysis and clustering, landscape compositions of individual scenes are analyzed by using the principle factors as binary (dummy) variables (Jobson 1992; Hardy 1993). Regressional models were built for landscape preference prediction based on these dumming variables.. For each of the factors (variables), any single landscape can be represented in a combination of the following possible values: one or zero for each of the factors: 1 If factor i is existent Xi = 0 If factor i is not existent Using the dummy variables as regressors and the preference ratings as dependent variables (predicted values), regression models of landscape preference were constructed (see Hardy 1993). The advantages of this approach of analysis and regressional analysis are: (1) It recognizes the fact that existence/absence (1 or 0) of certain contents is an overriding indictor of perceptual categorization. (2) It avoids a misleading result by imposing unrealistic measurement assumptions on the more or less quantitative and categorical variables. (3) It facilitates the process of categorization. (4) It can be most easily adapted by electronic media, so that the analysis can easily fit into a GIS format, and be integrated into the planning process. RESULTS AND DISCUSSION: TWO MODELS OF LANDSCAPE PREFERENCE AND CHANGE Correlation and factor analyses based on preference scales of twenty nine subject groups show that subjects living in a rural environment having a lower education, and subjects living in an urban environment having a higher education are significantly different in their landscape preferences. The influences of macro-cultural differences (Western vs. Chinese), of expertise (landscape architects vs. general public), and of gender (male vs. female) are not significant . These findings indicate that two models of landscape preference and landscape changes exist: the rural model representing the local farmers, and the urban model representing the tourists (detailed reports see Yu 1995.) Further inquiry as to what might affect viewers' landscape preferences depends on landscape categorization analysis. Factor analysis on the fifty scenes is also based on preference scales of all twenty-nine groups. The principle factors emerging from the calculations allow the author to analysis the landscape composition in the categories listed in Table 1. The following function is used in a regressional analysis with dummy variables (Hardy 1993): yi = B0 + (公式见打印稿) Here, yi is the preference rating for scene No.i. Bik is weight for category k of item j (in this case any of the three zones, and weather items), m is the total number of items (it is 4 in this case), rj is the number of categories in item j (in this case kj variesm from 2 to 8). Ei is the error of rating for the scene. Bo is costant. Two preference models are built (Table 1): the local farmers' model (rural model) and the tourists' model (urban model). Table 1 Two models of landscape preference Table 2 is a comparison on the weights of he categories that reflect the preference variations suggested in the two preference models. Table 2 Tourists vs. local farmers in landscape preference Generally speaking, the tourists (highly educated urban dwellers) have a much higher preference for natural landscapes (especially water and rock features) than cultural landscapes (with the exceptions of historic features and to some degree traditional settlements). The farmers have a much higher preference for cultural landscapes, especially landscapes associated with tourist development and agricultural production. The forested landscapes are perceived much more beautiful than the deforested barren hills by both tourists and farmers. Urban subjects join rural subjects in highly preferring historic features, one of three most preferred categories together with the natural features of water and red rock. Some best or worst combinations are listed in Table 3. Table 3 The best and worst combination of landscape categories for the two models Cultural and sub-cultural variations in landscape perception have been amply recorded in the past years. Most of the findings were, however, based on the observations of Westerners; only a few have addressed the cross-cultural and sub-cultural influences in Asians (Tips and Savasdisara 1986a, 1986b; Yu 1988,1990; Yu and Ji. 1990) and cross-cultural comparisons between Westerners and Asians (Nasar 1984; Hull and Revell 1989; Yang and Kaplan 1990). Research to date seems to yield two opposing conclusions. Many researchers have found that observers with different expertise, nationality and origin, etc., show a statistically strong agreement in landscape preference (Shafer and Tooby. 1973; Zube, Pitt et al. 1975; Daniel and Boster. 1976; Ulrich 1977; Brush 1979; Wellman and Buhyoff 1980; Ulrich 1983; Nasar 1984; Yang and Kaplan 1990). On the other hand, some findings seem to suggest landscape preference and perception vary from culture to culture ( Zube, Pitt et al. 1983; Hull and Revell 1989). The influence of cultural and sub-cultural factors on landscape preference is far from clear; many more comparative investigations are needed, especially of non-western subjects. Results of this research provide some comparable information about cultural and sub-cultural variations in landscape perception and preference. It is suggested that the living environment (urban vs. rural) together with the general educational level (high vs. low) are the major predictors of variations in landscape preference. By contrast, macro-cultural backgrounds (Western vs. Chinese), landscape expertise (landscape architects vs. general public), and gender (male vs. female) do not significantly influence the observers' preferences for different kinds of landscapes. The Findings from this research suggest that lower educated rural people and higher educated urban residents do have very different models of landscape preference and change. The meaning of these findings for the issue of "destructive construction" in Chinese scenic and historical landscapes should be understood in the following context. Currently, there are about 120 national parks (National Scenic and Historical Areas) in China, and thousands of regional parks. A considerable amount of these protected landscapes are located in remote areas. Tourist development in these areas is greatly encouraged by the central government and is considered means leading to local economic improvement. Constructions of tourist services are usually undertaken by local people, at least at the beginning of tourist development, without careful planning and design. The rural model (local farmers' model) of landscape preference, and also landscape change, becomes dominant in tourist development and landscape management in remote landscapes. The most striking finding from this research is that lower educated rural people and higher educated urban residents in general, and landscape architects in particular, have extremely different attitudes toward modern tourist services. To the rural people, modern tourist services are the most attractive features, while tourists have just the opposite preference. This finding, together with others, has proven the hypothesis that differences in landscape cognition are responsible for the "destructive construction" in national parks in China. This research suggests that local farmers do believe that the glorious hotels and bituminous concrete paths can beautify their landscapes. They also believe that these modern constructions are more attractive than their traditional farmsteads, even more than the rocky surroundings, and might also believe that the tourists from the urban areas or from foreign countries must have the same taste as they themselves. Therefore, local farmers change their landscape in a way that results in the construction of modern glazed-tiled hotels, colorful decorated service buildings, smoothly paved roads, etc. This is, of course, is a misconception, which leads to what landscape architects and well educated tourists called "destructive" development. On the other hand, landscape architects will sooner or later find that their designs, given the opportunity to carry them out, will hardly be fully accepted by the local clients. The professional design of a modest construction that is aimed at lower environmental impacts can easily be misused by the local people. By taking some of the ideas, changing the materials (e.g. from clay brick to concrete block, from plain local tiles to glazed tiles) and dimensions or location, the local developers can turn a landscape architect's "improving design" into a "destructive construction". One of the author's personally experienced examples in the Red Stone Park is that a humble gate designed by the author, suggesting the use of local red sand rocks, is now transformed into a glorious tall building, with luxurious rooms and glazed tiles. Findings from this research also suggest that there is some common ground between the two models of landscape preference and change. Among others, evergreen forests strongly preferred by both lower educated rural and higher educated urban observers, this indicates that reforestation in this area is a widely acceptable strategy of landscape improvement. The findings suggested that education is one of the most significant factors influencing landscape preference. This fact indicates that the negative impacts of tourist development caused by rural people's misconception about tourist service construction can be most effectively mitigated through education. CONCLUSION It is concluded that the cognitive difference between the lower educated rural people and the higher educated tourists and especially landscape architects is at least partly responsible for the "destructive constructions" in the scenic areas in China. These "destructive constructions" are mainly tourist services that have significant impacts on the natural and historical landscapes. These luxurious tourist services have been constructed and managed by the local people, and are highly visible. The local people believe such constructions to be the most attractive to tourists as well as to themselves, and believe they improve the landscapes. Tourists and especially the landscape architects have the opposite attitudes towards modern tourist services, their preference and change models are, however, hardly acceptable to the lower educated local people. Education is perhaps the most effective solution to the problem of "destructive construction" stemming from the misconception among the local people. ACKNOWLEDGMENT Thanks are due to Carl Steinitz and Ruth Weil for their help and reading of the draft of this paper.