Sustainable Development from Homogenization to Inclusivity: Optimization Strategies for Rural Landscape Design Based on Visual Behaviors and Landscape Preferences for Different Demographic Characteristics

Abstract

The rural landscape, as the core constituent unit of rural space, is vital for promoting sustainable rural development and achieving rural revitalization goals. However, current research is often limited to single groups, neglecting differences in visual behaviors (VBs) and landscape preferences (LPs) among diverse groups. Thus, this study employed eye-tracking technology combined with a landscape preference questionnaire to investigate the VBs and LPs of 160 participants when viewing rural landscapes. The results indicate that there is a notable correlation between VBs and LPs (p < 0.01), and the two aspects mutually influence each other. Moreover, groups with different demographic characteristics exhibit significant differences in both VBs and LPs. Among them, the score for LPs in the group of farmers, service sector workers, and retirees is significantly higher than that in the group of students, while the mean fixation duration is significantly lower than that in the group of students. Based on these research findings, this study suggests that, during the process of rural landscape design, full consideration should be given to the VBs and LPs of diverse groups with different demographic characteristics to enhance the inclusivity of rural landscape design and facilitate the realization of diversified and sustainable rural development.

1. Introduction

The rural landscape is a unique landscape system that is shaped by the interaction of rural natural resources and socio-cultural factors [1]. However, at present, rural landscapes are confronted with the severe challenge of landscape homogenization during their development [2], which hinders the sustainable development of rural areas. Driven by the dual strategies of rural revitalization and sustainable development [3], the functional scope of rural landscapes is no longer limited to traditional residential [4] and production functions [5], but has become a multidimensional integration that balances rural tourism [6], ecological protection [7], cultural heritage [8], and economic vitality [9], which indicates that rural landscapes urgently need to respond to the demands of the times through scientific planning. Existing studies have confirmed that significant variations in visual behaviors and landscape preferences across demographic groups exist during rural tourism experiences [10]. Meanwhile, with the transformation of rural tourism from a “resource–dependent” [11] model to an “experience–driven” [12] one, rural landscape spaces have gradually become composite scenarios that carry the aesthetic needs of multiple stakeholders [13]. This transformation places higher demands on the inclusivity and precision of rural landscape design. Therefore, studying the visual behaviors of groups with different demographic characteristics when exploring rural landscapes and revealing their preferences for rural landscapes can assist designers in accurately grasping the landscape needs of different groups. This, in turn, enables better coordination of the supply and demand relationship between rural landscape provision and the diverse needs of the population, avoids landscape homogenization, creates inclusive rural landscapes, optimizes rural landscape design, and ultimately enhances the quality of rural tourism experiences and achieves sustainable development in rural areas.

1.1. The Use of Eye-Tracking Technology in Rural Landscape Research and Its Sustainable Impacts

Human information processing heavily relies on vision, and eye movements serve as a crucial tool for the human visual system to explore the outside world [14]. With the deepening of rural landscape research, scholars have gradually broken through the traditional paradigm of qualitative analysis and systematically introduced a variety of quantitative research methods. Eye-tracking technology has emerged as a pivotal tool in this shift. The application of eye-tracking in rural landscape research provides a powerful means to uncover the perceptions and preferences of different groups towards rural landscapes, thereby generating multiple positive impacts on the sustainable development of rural landscapes. For instance, Wang et al. used eye-tracking technology to explore the visual behaviors of 72 tourists and villagers towards rural landscapes in Wangshang Village, Shaanxi Province, China. Their research results showed that the differences in visual behaviors between villagers and tourists towards rural landscapes are mainly reflected in the visual observation methods and key focus elements of the participants [2]. When viewing rural landscapes, villagers had a more dispersed field of view, mainly using scanning observation, while tourists had a more concentrated field of view, mainly using staring observation [2]. This discovery is beneficial for rural landscape designers to fully consider the visual needs of different groups during the planning process. Additionally, Su et al. took Huangshandian Village as an example and used eye-tracking technology to explore the preferences of 20 villagers and tourists for the rural landscape. Their research showed that the cognition of the street space landscape of the participants was the most laborious [15]. The landscape features of public green spaces were more prominent, and both tourists and villagers were more interested in public green spaces [15]. Compared with tourists, villagers searched more for landscape information on public green spaces and waterfront spaces, and the frequency of information searches was also higher [15]. These studies not only applied eye-tracking technology to reveal the visual behavioral differences and landscape preference characteristics of participants when observing rural landscapes, but also delivered empirical evidence supporting rural landscape optimization through objective data as well. From a sustainable perspective, such evidence contributes to creating more attractive and suitable rural landscapes. Therefore, compared with traditional qualitative research methods, eye-tracking technology offers more objective data support and effectively enhances the scientific nature of academic research. This supports the idea that the present study can also use eye-tracking experiments to explore the visual behaviors and landscape preferences of subjects when viewing rural landscapes. By gaining an in-depth understanding of the visual behaviors and preferences of different groups, targeted design strategies and practical guidance can be provided for the sustainable development of rural landscapes.

1.2. Research Hypothesis and Theoretical Model Construction: Demographic Characteristics Affecting Rural Landscape VBs and LPs

Demographic characteristics are closely related to individual landscape preferences [10,16]. Users will have different views when using visual quality to evaluate landscape spaces [17]. When focusing on the evaluation of the visual quality of landscape spaces, individuals in the same group tend to form differentiated landscape preferences and evaluation standards based on their subjective experience [18], cultural background [19], and functional needs [20]. It is worth noting that there are also significant differences in the interpretation of landscape visual quality between different groups [21]. For example, in a study on the impact of demography on environmental aesthetics, Stamps showed that the level of consensus on aesthetic preference within the same subject group is high, and the level of consensus on aesthetic preference among different groups is significantly different [22]. Laurie further deepened this point in the rural visual quality research and proposed that group division should be based on familiarity with the landscape (such as tourists and villagers), which can more accurately capture the functional needs and emotional connections of different groups in landscape cognition [23]. In addition, several studies have shown that the reasons for the differences in landscape preferences and visual behaviors are not only due to different group identities, but also affected by the gender, age, and educational background of observers. For example, Zhang et al. found that the visual behavior evaluation of participants in a forest leisure landscape space will be affected by their educational background [24]. Similarly, Wang and Zhao’s research on urban green space vegetation landscapes showed that the landscape preferences of participants were significantly affected by gender and education level: the landscape preference prediction factors for the participants with secondary education and university education were the same [17]. For men, “naturalness”, “plant growth state”, and “elements other than plants” are reliable predictors of landscape preference, but for women, the significant predictors are “plant maturity” and “number of colors” [17]. Svobodova et al. found that participant preferences in mine site landscapes are significantly affected by their own gender and education level [25]. Moreover, Zhang et al. found a significant correlation between the differences in usual residence and the visual behaviors of participants in their study of rural terraced landscapes [26]. The participants with usual residence in rural areas showed a higher frequency of eye twitching in visual behaviors and were more interested in terraced landscapes. However, this study also found no correlation between gender, ethnicity, and age and the visual behaviors of the participants [26]. In contrast, Leite et al. showed in their study that a significant correlation between age and participants’ landscape preferences [27]. Additionally, compared to the elderly population, the younger population tended to prefer natural rural landscapes and prefer the previous (before the change) versions of rural landscapes [27]. Although the above research based on demographic characteristics has achieved certain results in the field of visual behavior or landscape preferences, these studies have limitations in the following three aspects.

Firstly, correlation analysis can only reveal whether there is a correlation between variables [28] but cannot determine the causal direction and mechanism. The existing research overly relies on correlation analysis frameworks at the methodological level, which limits the explanatory power of causal mechanisms. Secondly, most studies focus on landscape types, such as forests, cities, and terraced fields, which exhibit type bias at the research object level and lack systematic research on rural landscape visual behaviors and landscape preferences. Thirdly, existing research has a structural deficiency in exploring the dimensions of different demographic characteristics, neglecting the key role of occupational type and monthly income level as explanatory variables in the formation of group differences. In addition, existing research has not delved deeply into the impact of demographic variables such as group identity, gender, age, occupation, monthly income level, usual residence, and education level on VB and LP, which has resulted in a lack of theoretical underpinnings for rural landscape design, making it difficult to cater to the viewing needs of different groups and achieve the goal of sustainable development. Thus, in response to these research gaps mentioned above, this study attempts to adopt a research method combining eye-tracking experiments and landscape preference questionnaire surveys to systematically investigate the influential mechanism of seven demographic characteristics on visual behaviors and landscape preferences in rural landscapes. The purpose of this approach is to provide scientific strategies and practical guidance for optimizing rural landscape design and enhancing the quality of rural tourism experiences, thereby increasing the attractiveness of rural tourism and promoting the sustainable development of the rural tourism industry. Based on this, the study proposes five research hypotheses, as detailed in Figure 1.

H1. 

There are significant differences in visual behaviors among groups with different demographic characteristics when viewing rural landscapes.

H2. 

Demographic characteristics have significant impacts on rural landscape preferences.

H3. 

A robust correlation exists between visual behaviors and landscape preferences.

H4. 

Landscape preferences significantly affect visual behaviors.

H5. 

Visual behaviors significantly affect landscape preferences.

2. Materials and Methods

2.1. Research Site

Shaanxi Province is located in the inland hinterland of China, at the intersection of the Loess Plateau and the Qinling Mountains. Due to complex terrain and a diverse climate, it has unique rural landscapes, agricultural production, and rural culture. As an outstanding representative of rural revitalization in Shaanxi Province, Tianxi Village has achieved significant progress in rural development and revitalization in recent years, such as becoming a Shaanxi Province Rural Tourism Demonstration Village [29]; inclusion in the second batch of national rural governance demonstration villages [30]; and inclusion in the ninth batch of national democratic and rule of law demonstration villages [31]. Moreover, Tianxi Village has beautiful scenery, abundant tourism resources, complete living facilities, and rural landscape imagery elements, meeting the various standards of this experiment. Therefore, this study selected Tianxi Village as the research site (see Figure 2 for details).

2.2. Eye-Tracking Experiment

2.2.1. Selection of Experimental Photos

Environmental psychology research shows that when influenced or constrained by the surrounding environment, people develop direct or indirect experiential cognition and form subjective environmental cognition [32]. Based on this, and referring mainly to Kevin Lynch’s theory of urban imagery [33], this study introduces the concept of rural landscape imagery, categorizing rural landscapes into five imageable elements: Landmark, Edge, District, Node, and Path [34]. Eye-tracking experiments are easily affected by external factors, such as light changes, noise interference, background clutter, etc. [2], and it has been confirmed that there is no significant difference between the results of eye-tracking experiments based on static pictures and the results of on-site measurement in core indicators [35]. Therefore, this study used control experiments, that is, static rural landscape pictures as visual stimulus materials, and organized participants to watch these photos in an indoor laboratory to obtain eye-tracking experiment data. In the process of experimental photo collection, in order to control the impact of unrelated variables on visual behavior results, this study strictly abided by the variable control norms of environmental behavior research, ensuring that all photos were collected in the same season and the same weather conditions on the same day, and at the same time period (10:00–12:00, 26 April 2024), to maximize the consistency and reliability of the photos.

During the selection of the experimental photos, the research team initially screened 130 photos and classified them according to five types of elements. Secondly, 20 landscape design postgraduates categorically screened the experimental photos again, according to the types of rural landscape image elements. To ensure the objectivity and accuracy of the results, this study further invited 5 administrative leaders from Tianxi Village to carry out the third screening, according to the local characteristics, and invited 5 professors in the field of landscape design to carry out the final academic review of the experimental photos. Finally, 25 photos were selected, and the size ratio was unified as the static stimulus materials for the eye-tracking experiment (Figure 3 for details).

2.2.2. Participants

This study strictly adhered to the principle of anonymity during the process of experimental data collection. The collection of all experimental data used in this study was conducted on a voluntary basis, after the participants had fully understood the entire procedure of the experiment and signed the informed consent form. Explicit consent from the participants was obtained prior to data collection, ensuring that the research complies with ethical standards. To ensure that the subjects could cover different demographic characteristics, this study conducted targeted screening to ensure that the subjects met the requirements of the population structure and actual situation in terms of gender, age, occupation, income level, usual residence, and education level. A total of 160 participants were recruited for this experiment, as shown in

Table 1. Analysis of the demographic characteristics of the participants.

Sample Size	Demographic Characteristics	Category	Quantity	Percentage
N = 160	Group Identity	Villager	80	50%
		Tourist	80	50%
	Gender	Man	80	50%
		Woman	80	50%
		Other	0	0
	Age	18–24	34	21.25%
		25–34	32	20%
		35–44	30	18.75%
		45–64	35	21.875%
		≧65	29	18.125%
	Occupation	Student	31	19.375%
		Professional Technician	24	15%
		Government Staff	19	11.875%
		Service Sector Worker	28	17.5%
		Farmer	32	20%
		Retiree	26	16.25%
	Monthly Income Level	CNY 0–2000	30	18.75%
		CNY 2000–5000	56	35%
		CNY 5000–10,000	43	26.875%
		CNY > 10,000	31	19.375%
	Usual Residence	City	44	27.5%
		Town	62	38.75%
		Village	54	33.75%
	Education Level	Elementary school and below	13	8.125%
		Junior high school	29	18.125%
		Regular/vocational high school	36	22.5%
		Undergraduate/associate degree	52	32.5%
		Graduate and above	30	18.75%

. The vision or corrected vision of the participants is normal, with a male-to-female ratio of 1:1, average age = 41.85, and standard deviation = 16.731.

2.2.3. Process of Eye-Tracking Experiment

This study used E-Prime 3.0.3.9 software to set up the eye-tracking experiment process to ensure the scientific and rigorous nature of the experiment (see Figure 4 for details), which mainly consists of the following five steps: (1) setting the experimental starting instructions to ensure that the participants complete the eye-tracking experiment correctly according to the process; (2) setting a warm-up phase to help the participants to quickly enter the experimental state and master specific experimental operations; (3) encoding 25 formal experimental photos based on 5 types of rural landscape cognitive elements, and setting the order of photo playback to random to avoid subjective interference. Each photo is played for 10 s with a 3 s gray screen for vision calibration [2,33], and the total playback time for the entire formal experimental phase is approximately 6 min; and (4) setting up instructions for terminating the experiment to standardize the ending process and convey the information at the end of the experiment to the participants.

The whole eye-tracking experiment is mainly divided into 3 parts: debugging equipment, the preheating stage, and the formal experiment. Each subject takes about 10 min. Before the beginning of the experiment, experimental personnel 1 led the subject into the laboratory, who sat about 50 cm in front of the computer screen [36], and assisted the subject in wearing the eye tracker. Afterwards, experimenter 2 calibrated each parameter, based on the physiological characteristics of the participants, to ensure that they were ready. After that, they guided the participants by playing warm-up pictures to ensure that they could proficiently operate and complete the warm-up exercises on their own. After the preheating stage was successfully completed, all experimental personnel remained quiet, and the participants completed the formal experiment independently.

2.2.4. Equipment and Environment of Eye-Tracking Experiment

This experiment used a Dikablis Glasses 3.0 Wearable Eye Tracker as the eye-tracking recognition instrument, and a Lenovo Savior R9000P Laptop as the playback device for static stimulus photos. To prevent the impact of the primacy effect on the experimental outcomes, the participants were prohibited from previewing the experimental photographs prior to the experiment. The experiment was conducted in a serene [2], tidy indoor setting with soft lighting. The indoor temperature was maintained at approximately 25 °C, with air humidity ranging between 55% and 60%. The background of the entire laboratory was white, and the tabletop where the instruments were placed was lined with light gray fabric to control variables and reduce the impact of irrelevant factors on the experiment (see Figure 5 for details).

2.3. Landscape Preference Questionnaire

After completing the eye-tracking experiment, the participants were led by experimenter 3 into an adjacent room to independently answer a paper landscape preference questionnaire to avoid interfering with the next participant’s eye-tracking experiment. The questionnaire is mainly divided into two sections: demographic basic information filling and landscape preference scoring (Appendix A for details). The Likert 5-point scale is a psychological measurement tool proposed by Rensis Likert, an American social psychologist, in 1932, which is used to measure the attitude, opinion, or opinion of a statement of a subject. Among them, 1 is the most inconsistent; 2 is not consistent; 3 is general; 4 is consistent; and 5 is particularly consistent [37,38]. The scale is recognized by the academic community because of its good systematization and operability. Therefore, this study used the scale to collect the landscape preference data of the subjects. Specifically, the landscape preference questionnaire used in this study is mainly divided into two parts: demographic basic information filling and landscape preference scoring. Among them, the landscape preference scoring sheet includes color pictures consistent with those from the previous eye movement experiment as visual prompts, so that the subjects can systematically score the complexity, coherence, mystery, and legibility of 25 pictures by using the Likert 5-point scale. The purpose of this setup is to address the following two aspects: (1) to control cognitive interference by using time sequence isolation through spatially and temporally separating the eye-tracking and questionnaire tasks to avoid the expectancy effect [39], so as to ensure the authenticity of eye-tracking experimental results; and (2) to use the working memory activation effect [40] to ensure that, after the participants completed the immersion eye-tracking experiment, their visual cortex retained a clear impression of the spatial layout, color characteristics, and cultural symbols of the rural landscape, providing a cognitive basis for the subsequent preference evaluation.

2.4. Data Analysis Method

A minimum fixation duration threshold of 300 ms was set in the D-LAB 3.72 software, accompanied by the Dikablis Glasses 3.0 wearable eye tracker, and the experimental data were processed to obtain the Mean Fixation Count (MFC), Mean Fixation Duration (MFD), Mean Saccade Count (MSC), and Mean Saccade Duration (MSD) for each participant when viewing every photograph. Among them, the MFC refers to the average number of fixation times made by the participants for each photo; the MFD refers to the average stay time of the participants at each fixation point; the MSC refers to the average number of times a participant’s eyes move from one fixation to another; and the MSD refers to the average duration of each saccade behavior (eye movement between two fixation points) [2,35,41]. This study first used Microsoft Excel 2021 to classify and sort out the data and used SPSS 27.0 software to analyze the correlation between visual behaviors and landscape preferences, aiming to explore whether there is a linear correlation between the two, as well as the degree and direction of correlation. The evaluation criteria are as follows: an absolute value of the Pearson correlation coefficient between 0.1 and 0.3 represents weak correlation, between 0.3 and 0.5 represents medium correlation, and above 0.5 represents strong correlation [42]. Then, in addition to group identity and gender, the other 5 classified variables, including age, occupation, monthly income level, usual residence, and education level, were processed as dummy variables. Finally, 7 demographic variables in this study were taken as independent variables, and linear regression analysis was conducted with visual behaviors and landscape preferences, respectively. Through a series of regression analyses, the variable relationship, internal correlation, and positive and negative impact between different demographic characteristics and rural landscape visual behaviors and landscape preferences were quantitatively explored, providing data support and theoretical reference for the follow-up optimization of rural landscape design and improvement of rural tourism experience quality.

3. Results

3.1. Analysis of the Correlation Between Visual Behaviors and Landscape Preference

As shown in Figure 6, there is a significant correlation between visual behaviors and landscape preferences (p < 0.01), among which, landscape preferences (complexity, coherence, mystery, and legibility) have a significant negative correlation with MFC and MFD, while there is a significant positive correlation with MSC and MSD, which provides strong support for H3.

3.2. Linear Regression Analysis Results Between Visual Behaviors and Landscape Preferences

Table 2. Linear regression analysis results between visual behaviors and landscape preferences.

Independent Variables	Dependent Variables	Results of Linear Regression Analysis
		Adjusted R2	F	B	t	p
MFC MFD MSC MSD	Complexity	0.981	2025.079 ***	−0.491	−12.471	<0.001 ***
				−1.661	−4.978	0.025 *
				0.083	2.674	0.008 **
				0.974	0.850	0.396
	Coherence	0.974	1517.515 ***	−0.533	−11.373	<0.001 ***
				−5.238	−13.182	<0.001 ***
				−0.311	−8.430	<0.001 ***
				9.914	7.267	<0.001 ***
	Mystery	0.852	229.820 ***	0.002	0.022	0.982
				6.975	7.752	<0.001 ***
				1.239	14.825	<0.001 ***
				−24.286	−7.862	<0.001 ***
	Legibility	0.853	232.159 ***	−1.206	−7.334	<0.001 ***
				−18.026	−12.935	<0.001 ***
				−1.957	−15.113	<0.001 ***
				41.239	8.619	<0.001 ***
Complexity Coherence Mystery Legibility	MFC	0.969	1237.781 ***	−1.635	−5.695	<0.001 ***
				−0.548	−2.287	0.025 *
				0.664	2.833	0.005 **
				0.432	2.584	0.011 *
	MFD	0.948	731.399 ***	0.264	5.835	<0.001 ***
				−0.069	−1.821	0.070
				−0.232	−6.278	<0.001 ***
				−0.122	−4.602	<0.001 ***
	MSC	0.985	1261.976 ***	0.055	0.109	0.913
				2.696	6.469	<0.001 ***
				0.069	0.170	0.865
				−0.720	−2.472	0.015 *
	MSD	0.914	423.239 ***	0.096	7.377	<0.001 ***
				0.070	6.414	<0.001 ***
				−0.083	−7.810	<0.001 ***
				−0.059	−7.783	<0.001 ***

Note: *, **, and *** indicate p < 0.05, p < 0.01, and p < 0.001, respectively.

shows the linear regression analysis results of visual behaviors and landscape preferences as independent and dependent variables, respectively: landscape preferences significantly affect visual behaviors, and coherence has the most significant impact on visual behavior. At the same time, visual behaviors also significantly affect landscape preferences, and MSD has the most significant impact on landscape preferences.

Specifically, the regression coefficients of MFC/MFD for complexity, coherence, and legibility are negative and have large absolute values, especially the regression coefficient of average fixation duration on readability (−18.026). This indicates that the lower the MFC of participants, the higher their rating of landscape preferences, while the longer the MFD of the participants, the lower their rating of landscape preferences. Moreover, the MSC of the participants has a positive impact on complexity and mystery, and MSD also has a positive impact on complexity, coherence, and legibility. This indicates that the higher the MSC the participants have, the higher their ratings for complexity and mystery, and the longer the MSD, the higher their ratings for complexity, coherence, and legibility. In addition, the regression coefficient of the MSD of the participants on legibility is 41.239, which is the largest positive value in the table, indicating that MSD has a very significant impact on readability. Additionally, complexity has the greatest impact on MFC, and the regression coefficient is negative (−1.635); coherence also has the greatest impact on MSC, but the regression coefficient is positive (2.696). This indicates that the higher the complexity score of the participants, the lower their MFC, and the higher the coherence score, the higher their MSC. Mystery has a negative impact on MFD and MSD, while it has a positive impact on MFC and MSC, with a large regression coefficient. This indicates that the higher the mystery, the shorter the MFD and MSD, and the higher the MFC and MSC. In addition, legibility only has a positive impact on MFC, indicating that higher legibility leads to higher MFC.

3.3. Linear Regression Analysis Between Different Demographic Characteristics and Visual Behaviors and Landscape Preferences

Table 3. Linear regression analysis results between group identity and visual behaviors and landscape preferences, respectively.

Independent Variables	Dependent Variables	Results of Linear Regression Analysis
		Adjusted R2	F	B	t	p
Villager (reference) Tourist	MFC	0.086	16.013 ***	1.140	4.002	<0.001 ***
	MFD	0.066	12.149 ***	0.123	3.486	0.001 **
	MSC	0.021	4.448 ***	−1.091	−2.109	0.037 *
	MSD	0.107	19.963 ***	−0.035	−4.468	<0.001 ***
	Complexity	0.062	11.495 ***	−0.843	−3.390	0.001 **
	Coherence	0.186	37.310 ***	−1.463	−6.108	<0.001 ***
	Mystery	0.096	17.935 ***	1.005	4.235	<0.001 ***
	Legibility	0.645	289.285 ***	−3.937	−17.008	<0.001 ***

Note: *, **, and *** indicate p < 0.05, p < 0.01, and p < 0.001, respectively.

shows the linear regression analysis results between group identity and visual behaviors and landscape preferences, respectively: when different groups (villagers and tourists) enjoy rural landscapes, there are not only differences in visual behaviors, but also differences in landscape preferences. Despite the relatively low adjusted R² values across the regression analysis models, the influence of group identity on both visual behaviors and landscape preferences is statistically significant (p < 0.001). Consequently, group identity emerges as one of the factors accounting for variations in visual behaviors and landscape preferences.

Table 4. Linear regression analysis results between gender and visual behaviors and landscape preferences, respectively.

Independent Variables	Dependent Variables	Results of Linear Regression Analysis
		Adjusted R2	F	B	t	p
Man (reference) Woman	MFC	0.000	0.001	−0.011	−0.003	0.971
	MFD	0.001	0.194	−0.016	−0.440	0.660
	MSC	0.001	0.130	0.189	0.361	0.718
	MSD	0.000	0.061	0.002	0.247	0.805
	Complexity	0.001	0.079	0.072	0.282	0.778
	Coherence	0.000	0.032	0.048	0.178	0.859
	Mystery	0.005	0.776	0.220	0.881	0.380
	Legibility	0.001	0.174	−0.163	−0.417	0.667

reveals the results of the linear regression analysis between gender and visual behaviors and landscape preferences. The impact of gender on visual behaviors and landscape references is not significant (p > 0.05), and the adjusted R² values of each regression analysis model are very small, almost close to 0. This shows that the influence of gender on landscape preference is very limited; thus, gender is not an important factor affecting visual behavior differences and landscape preference differentiation.

Table 5. Linear regression analysis results between age and visual behaviors and landscape preferences, respectively.

Independent Variables	Dependent Variables	Results of Linear Regression Analysis
		Adjusted R2	F	B	t	p
18–24 (reference) 25–34 35–44 45–64 ≧65	MFC	0.911	406.917 ***	−1.170	−8.442	<0.001 ***
				−2.200	−15.604	<0.001 ***
				−3.519	−25.960	<0.001 ***
				−5.197	−36.522	<0.001 ***
	MFD	0.898	350.485 ***	−0.236	−12.993	<0.001 ***
				−0.409	−22.107	<0.001 ***
				−0.507	−28.518	<0.001 ***
				−0.619	−33.166	<0.001 ***
	MSC	0.885	305.857 ***	3.087	11.154	<0.001 ***
				4.963	17.631	<0.001 ***
				6.767	25.007	<0.001 ***
				9.091	32.005	<0.001 ***
	MSD	0.802	162.502 ***	0.029	5.193	<0.001 ***
				0.053	9.183	<0.001 ***
				0.080	14.520	<0.001 ***
				0.138	23.714	<0.001 ***
	Complexity	0.949	738.093 ***	1.174	12.989	<0.001 ***
				2.145	23.331	<0.001 ***
				3.223	36.468	<0.001 ***
				4.570	49.260	<0.001 ***
	Coherence	0.902	368.031 ***	1.407	10.891	<0.001 ***
				2.369	18.027	<0.001 ***
				3.379	26.751	<0.001 ***
				4.641	34.995	<0.001 ***
	Mystery	0.667	80.697 ***	0.804	3.583	<0.001 ***
				1.715	7.515	<0.001 ***
				2.815	12.838	<0.001 ***
				3.562	15.472	<0.001 ***
	Legibility	0.564	52.485 ***	1.833	4.592	<0.001 ***
				2.911	7.168	<0.001 ***
				3.773	9.668	<0.001 ***
				5.567	13.587	<0.001 ***

Note: *** indicates p < 0.001.

presents the results of the linear regression analysis between age and visual behaviors and landscape preferences, respectively: the influence of age on visual behaviors and landscape preferences is strongly significant (p < 0.001). Specifically: (1) in terms of visual behaviors, MFC and MFD showed a negative increasing trend with an increase in age, while MSC and MSD showed a positive increasing trend with an increase in age. (2) In terms of landscape preferences, the scores of the participants on the complexity, coherence, mystery, and legibility of the rural landscape also showed a positive increasing trend with an increase in age, which indicates that age is an important factor in the differences in visual behaviors and landscape preferences.

According to

Table 6. Linear regression analysis results between occupation and visual behaviors and landscape preferences, respectively.

Independent Variables	Dependent Variables	Results of Linear Regression Analysis
		Adjusted R2	F	B	t	p
Student (reference) Professional Technician Government Staff Service Sector Workers Farmer Retiree	MFC	0.312	15.435 ***	−0.343	−0.808	0.421
				−0.769	−1.688	0.093
				−1.245	−3.054	0.003 **
				−3.131	−7.948	<0.001 ***
				−0.779	−1.874	0.063
	MFD	0.209	9.421 ***	−0.097	−1.743	0.083
				−0.114	−1.91	0.058
				−0.166	−3.103	0.002 **
				−0.342	−6.608	<0.001 ***
				−0.127	−2.324	0.021 *
	MSC	0.276	13.103 ***	0.990	1.294	0.198
				1.411	1.720	0.088
				2.232	3.040	0.003 **
				5.344	7.531	<0.001 ***
				1.555	2.077	0.039 *
	MSD	0.336	17.124 ***	0.007	0.648	0.518
				0.015	1.247	0.214
				0.026	2.360	0.020 *
				0.087	8.185	<0.001 ***
				0.017	1.492	0.138
	Complexity	0.308	15.184 ***	0.459	1.251	0.213
				0.684	1.739	0.084
				1.191	3.385	0.001 **
				2.740	8.058	<0.001 ***
				0.838	2.334	0.021 *
	Coherence	0.280	13.369 ***	0.508	1.312	0.191
				0.784	1.890	0.061
				1.215	3.273	0.001 **
				2.746	7.650	<0.001 ***
				0.927	2.449	0.015 *
	Mystery	0.184	8.184 ***	0.327	0.843	0.400
				0.520	1.251	0.213
				1.015	2.730	0.007 **
				2.094	5.826	<0.001 ***
				0.547	1.444	0.151
	Legibility	0.175	7.759 ***	0.683	1.127	0.262
				0.953	1.466	0.145
				1.401	2.409	0.017 *
				3.308	5.886	<0.001 ***
				1.304	2.199	0.029 *

Note: *, **, and *** indicate p < 0.05, p < 0.01, and p < 0.001, respectively.

, the influence of occupation on visual behaviors and landscape preferences has varying degrees of significance (p < 0.05/0.01/0.001). Specifically, in terms of visual behaviors, participants in the farmer and service sector worker groups had lower MFC and MFD than those in the student group, while the MSC and MSD were significantly higher than those in the student group. In terms of landscape preferences, participants from the farmer, service sector worker, and retiree groups rated the complexity, coherence, and legibility of rural landscapes significantly higher than those from the student group. This indicates that occupation is a significant factor influencing visual behaviors and landscape preferences.

As evident from the linear regression analysis results presented in

Table 7. Linear regression analysis results between monthly income level and visual behaviors and landscape preferences, respectively.

Independent Variables	Dependent Variables	Results of Linear Regression Analysis
		Adjusted R2	F	B	t	p
CNY0–2000 (reference) CNY 2000–5000 CNY5000–10,000 CNY > 10,000	MFC	0.019	0.986	−0.104	−0.244	0.807
				0.473	1.054	0.293
				−0.165	−0.341	0.733
	MFD	0.005	0.284	−0.027	−0.515	0.537
				−0.003	−0.048	0.562
				−0.039	−0.657	0.610
	MSC	0.010	0.531	0.626	0.833	0.406
				0.001	0.001	0.999
				0.728	0.855	0.394
	MSD	0.024	1.276	0.010	0.853	0.395
				−0.007	−0.569	0.570
				0.013	0.987	0.325
	Complexity	0.010	0.534	0.164	0.444	0.658
				−0.233	−0.601	0.549
				0.125	0.298	0.610
	Coherence	0.011	0.526	0.166	0.435	0.664
				−0.197	−0.490	0.625
				0.233	0.540	0.590
	Mystery	0.007	0.354	0.288	0.802	0.424
				0.002	0.005	0.996
				0.075	0.183	0.855
	Legibility	0.014	0.563	−0.021	−0.037	0.970
				−0.467	−0.796	0.427
				0.253	0.400	0.689

, the impact of monthly income level on visual behaviors and landscape preferences is not significant (p > 0.05), and the adjusted R² values of each regression analysis model are all less than 0.03, almost close to 0. This indicates that the explanatory power of monthly income level on landscape preferences is very limited; that is, monthly income level is not an important factor influencing visual behaviors and landscape preferences.

As shown in

Table 8. Linear regression analysis results between usual residence and visual behaviors and landscape preferences, respectively.

Independent Variables	Dependent Variables	Results of Linear Regression Analysis
		Adjusted R2	F	B	t	p
City (reference) Town Village	MFC	0.817	357.026 ***	−1.770	−11.149	<0.001 ***
				−4.305	−26.321	<0.001 ***
	MFD	0.824	374.127 ***	−0.334	−17.521	<0.001 ***
				−0.535	−27.240	<0.001 ***
	MSC	0.812	344.943 ***	4.103	14.520	<0.001 ***
				7.646	26.261	<0.001 ***
	MSD	0.680	169.934 ***	0.043	7.495	<0.001 ***
				0.108	18.121	<0.001 ***
	Complexity	0.851	454.049 ***	1.757	14.221	<0.001 ***
				3.813	29.945	<0.001 ***
	Coherence	0.789	297.935 ***	1.856	12.208	<0.001 ***
				3.811	24.332	<0.001 ***
	Mystery	0.676	166.836 ***	1.616	9.124	<0.001 ***
				3.323	18.207	<0.001 ***
	Legibility	0.435	62.141 ***	1.990	5.469	<0.001 ***
				4.163	11.102	<0.001 ***

Note: *** indicates p < 0.001.

, the influence of usual residence on visual behaviors and landscape preferences was significant (p < 0.001). Specifically, in terms of visual behaviors, compared with the reference group (city), the participants in the town group and the village group had a lower MFC and shorter MFD when watching rural landscape image elements, while the MSC was higher and the MSD was longer. In terms of landscape preferences, the participants in the town group and the village group scored significantly higher on the complexity, coherence, mystery, and legibility of rural landscape image elements than the participants in the city group. This shows that the participants with different usual residences not only have differences in visual behaviors, but also have differences in landscape preferences, which shows that usual residence is indeed a significant factor affecting visual behaviors and landscape preferences in rural landscapes.

Table 9. Linear regression analysis results between education level and visual behaviors and landscape preferences, respectively.

Independent Variables	Dependent Variables	Results of Linear Regression Analysis
		Adjusted R2	F	B	t	p
Elementary School and Below (reference) Junior High School Regular/Vocational High School Undergraduate/Associate Degree Graduate and Above	MFC	0.817	172.945 ***	1.392	5.104	<0.001 ***
				3.124	11.821	<0.001 ***
				5.199	20.527	<0.001 ***
				4.287	15.808	<0.001 ***
	MFD	0.628	65.504 ***	0.088	1.857	0.065
				0.227	4.923	<0.001 ***
				0.514	11.621	<0.001 ***
				0.401	8.463	<0.001 ***
	MSC	0.702	91.199 ***	−3.013	−4.932	<0.001 ***
				−5.083	−8.585	<0.001 ***
				−8.779	−15.471	<0.001 ***
				−7.455	−12.269	<0.001 ***
	MSD	0.793	153.406 ***	−0.077	−9.767	<0.001 ***
				−0.112	−14.747	<0.001 ***
				−0.160	−21.906	<0.001 ***
				−0.142	−18.159	<0.001 ***
	Complexity	0.802	156.537 ***	−1.083	−4.431	<0.001 ***
				−2.444	−10.319	<0.001 ***
				−4.323	−19.044	<0.001 ***
				−3.680	−15.139	<0.001 ***
	Coherence	0.775	133.277 ***	−0.997	−3.702	<0.001 ***
				−2.264	−8.671	<0.001 ***
				−4.392	−17.557	<0.001 ***
				−3.349	−12.502	<0.001 ***
	Mystery	0520	42.022 ***	−0.697	−1.886	0.061
				−1.805	−5.037	<0.001 ***
				−3.008	−8.758	<0.001 ***
				−3.338	−9.078	<0.001 ***
	Legibility	0.590	55.833 ***	−1.117	−2.102	0.037 **
				−2.550	−4.951	<0.001 ***
				−5.569	−11.282	<0.001 ***
				−3.195	−6.044	<0.001 ***

Note: ** and *** indicate p < 0.01 and p < 0.001, respectively.

reports that education level has varying degrees of influence on both visual behaviors and landscape preferences (p < 0.05/0.001). As education level increases, the undergraduate/associate degree group of participants shows the greatest changes in visual behaviors and landscape preferences. Specifically, in terms of visual behaviors, compared with the reference group (elementary school and below), the MFC and MFD in the other groups of participants showed varying degrees of increase, while the MSC and MSD showed varying degrees of decrease, indicating that participants with different education levels have different visual observation styles. In terms of landscape preferences, compared with the reference group, the scores for complexity, coherence, mystery, and legibility in the other groups of participants showed a negative increasing trend, indicating that education level is an important factor affecting rural landscape visual behaviors and landscape preferences.

In summary, when viewing rural landscapes, participants with different demographic characteristics exhibit significant differences in visual behaviors and landscape preferences. Landscape preferences significantly affect visual behaviors, and coherence has the most significant impact on visual behaviors. Meanwhile, visual behaviors also significantly affect landscape preferences, and MSD has the most significant impact on landscape preferences. Thus, these research results above provide strong support for H1, H2, H4, and H5. In addition, these research results above also indicate that demographic characteristics not only indirectly affect landscape preferences by influencing visual behaviors, but also have an indirect impact on visual behaviors by influencing landscape preferences.

4. Discussion

4.1. The Influence of Multidimensional Demographic Characteristics on Visual Behaviors and Preferences Toward Rural Landscapes

As the core resource of rural tourism and the main carrier of rural culture, the rural landscape plays an extremely significant role in the sustainable development of rural areas. With the acceleration of the urbanization process, this fast-paced and high-density social environment has aroused the yearning of people for natural life and promoted the vigorous development of rural tourism. Previous studies have shown that people with different demographic characteristics will have different visual behaviors and landscape preferences when they appreciate the rural landscape in the process of rural tourism [10,16,21]. In order to precisely understand the landscape demands of diverse groups, thereby better harmonizing the supply and demand relationship between rural landscape provision and the multifaceted needs of the population, optimizing rural landscape design, and ultimately enhancing the quality of rural tourism experiences, this study takes demographic characteristic variables as the starting point. By utilizing a research approach that combines eye-tracking experiments with outcome preference questionnaires, this study explores the positive and negative influence relationships between visual behaviors and landscape preferences and reveals the main demographic variables that affect differences in visual behaviors and landscape preferences. The core findings are as follows:

(1)

When viewing rural landscapes, participants with different demographic characteristics not only have significant differences in visual behaviors but also have significant differences in landscape preferences.

(2)

A robust correlation (p < 0.01) was observed between visual behaviors and landscape preferences, characterized by a significant negative association between landscape preferences and MFC and MFD, contrasted with a strong positive association with MSC and MSD.

(3)

Landscape preferences significantly affect visual behaviors, with coherence having the most significant impact on visual behaviors. At the same time, visual behaviors also significantly affect landscape preferences, and MSD has the most significant impact on landscape preference.

(4)

Demographic characteristics not only indirectly affect landscape preferences by influencing visual behaviors, but also have an indirect impact on visual behaviors via affecting landscape preferences.

(5)

Group identity, age, occupation, usual residence, and education level are the main demographic characteristics that influence rural landscape visual behaviors and landscape preferences. Among them, villagers and tourists have significant differences in visual behaviors and landscape preferences. The farmer and service sector worker groups were significantly lower than the student group in MFC and MFD, but significantly higher than the student group in MSC and MSD. Compared with the city group, the participants in the town and village groups showed a negative increasing trend in the average number of fixations and the average length of fixations, and a positive increasing trend in MSC and MSD. With the increase in education level, the range of change in visual behaviors and rural landscape preferences of the undergraduate/associate degree group was the largest. With an increase in age, the MFC and MFD of participants when watching rural landscapes showed a negative increasing trend, while the MSC, MSD, and rural landscape preferences score showed a positive increasing trend.

This research not only facilitates the optimization of design but also offers theoretical and methodological support for enhancing the rural tourism experience and achieving sustainable rural development. The above results provide strong support for the research hypothesis model of the impact of demographic characteristics on visual behaviors and landscape preferences, verifying H1, H2, H3, H4, and H5. Therefore, the hypothesis model should be revised, as shown in Figure 7.

The conclusion in this study that group identity has an impact on the visual behaviors of participants is consistent with the research of Wang et al. [2], which also found a significant difference in MFD between tourists and villagers. At the same time, the research findings on the impact of occupation and education level on landscape preferences have reached a consensus with the views of Wu et al. [43] and Rao et al. [44]. In addition, the research findings that gender affects landscape preferences are consistent with Yao et al. [45], but opposite to the research findings of Svobodova et al. [25] and Abello et al. [46], which showed that gender significantly affects the landscape preferences of participants. However, it is worth noting that Svobodova et al. showed in their research results that age has a significant impact on subjects’ landscape preferences [25], which reached a consensus with our research results. Although the results of this study on the impact of usual residence on landscape preferences differ from Keane’s conclusion [47], which suggested that usual residence (urban or rural) does not seem to have a significant effect on participants’ landscape preferences, we have reached a consensus with Van den Berg & Velks’ [48] and Yu’s [49] research results, both of which found that usual residence has an impact on participants’ landscape preferences. In addition, our research results are similar to Zhang et al.’s [26] research results, which also found that usual residence can affect the visual behavior differences among participants. However, they believed that age does not affect the visual behaviors of participants [26], which contradicts our research findings. Of particular note is that this study conducted regression analysis on seven demographic variables with four types of visual behaviors and four types of landscape preferences, and found that age is the demographic feature that has the greatest impact on visual behaviors and landscape preferences. Based on this, we propose the following speculations and hypotheses.

Firstly, in the linear regression analysis model between age and visual behaviors, as age increases, the MFC (B = −5.197~−1.170, p < 0.001) and MFD (B = −0.619~−0.236, p < 0.001) of the participants show a negative increasing trend. MSC (B = 3.087 and 9.091, p < 0.001) and MSD (B = 0.029 and 0.138, p < 0.001) showed a positive increasing trend. This age-related visual behavior change may be due to the dual effects of cognitive processing ability decline and experience accumulation [50]. Specifically, on the one hand, as age increases, the attention span of the elderly population gradually decreases, resulting in a reduction in the amount of information obtained from a single gaze. They are compelled to compensate for the reduction in information acquisition by increasing MSC, thereby achieving the goal of efficiently screening key information. On the other hand, as the elderly population is mostly villagers from Tianxi Village, they have a high level of familiarity with the rural environment. Therefore, when viewing the rural landscape, they mainly adopt a scanning observation method, as shown in Figure 8. The red region in heatmaps is used to represent the key focus areas of the subjects, which are the parts that have attracted the attention of subjects for a long time and have a high level of interest. Conversely, the yellow region represents moderate attention, while the green region signifies lesser attention [2]. In addition, it is worth noting that the increasing trend of MSD is weak, which further supports the conclusion that the elderly population mainly completes visual search behaviors through rapid eye movement [51].

Secondly, the linear regression analysis model between age and landscape preferences reveals more complex and profound results, with the participants scoring of the complexity (B = 1.174~4.570, p < 0.001), coherence (B = 1.407~4.641, p < 0.001), mystery (B = 0.804~3.562, p < 0.001), and legibility (B = 1.833~5.567, p < 0.001) of rural landscapes all increasing with age, and landscape preferences strengthening with age. This phenomenon may be due to the phased evolution of environmental cognitive patterns under the framework of the life course theory, which suggests that individuals develop cognitive patterns that adapt to their environment at different stages of life [52]. Specifically, the increasing preference of the elderly for landscape complexity may represent a gradual shift in cognitive patterns from exploratory orientation in youth to reflective orientation in old age or may reveal specific cognitive–emotional connection mechanisms formed by the elderly in their long-term adaptation to rural landscapes. In addition, although the rating of rural landscape mystique by the elderly group has reached significance, the effect size is relatively small, indicating that the influence of age on mystique rating is weak, or there may be other moderating variables at other levels, such as the moderating effect of cultural background, or the inference that mystique preference may have a non-linear relationship with individual exploration motivation. This needs to be further verified by introducing motivation scales for multiple regression analysis.

4.2. Suggestions for Optimizing Rural Landscape Design and Improving the Quality of Rural Tourism Experience

To construct rural landscapes that are suitable for different groups and improve the quality of the rural tourism experience, we propose the following design optimization suggestions for rural landscape design based on the experimental results of this study:

(1)

Based on the significant differences in visual behaviors between villagers and tourists, as well as the landscape preferences caused by different usual residences, a “host–guest sharing visual landscape” [53] is constructed to create differentiated visual experiences and achieve landscape gradient translation [54], from urban to rural areas.

(2)

Establish modular landscape units to address the differences in attention patterns among different occupational groups. For example, ecological education nodes that require in-depth observation could be set up for student groups, and open activity spaces that reflect agricultural culture could be reserved for farmer groups.

(3)

Concrete narrative design techniques could be used to create simple and easy-to-understand rural landscape scenes for groups with lower levels of education, such as sculptures with the theme of farming activities and folk performances. For highly educated groups, the abstract design technique could be used to create rural landscape nodes with multi-level visual information and improve the quality of the rural tourism experience of participants.

(4)

A design strategy for strengthening spatial cognition was implemented. Aiming at a reduction in fixation duration and the improvement of sweeping activity brought about by increasing age, a “cognitive buffer zone” was designed, such as transitional spaces with cultural display and rest functions, within visually transformative rural landscape spaces. Based on this cognitive map, rural landscape image elements could be designed, rural roads could be optimized, and landscape markers with cognitive function could be designed as key turning points. Through these landscape markers, the spatial narrative is strengthened and recognition is improved.

4.3. Limitations and Future Research Direction

Although eye-tracking technology has advantages in revealing the correlation between visual behaviors and landscape preferences, eye-tracking technology can not directly explain the physiological response mechanisms of subjects to landscape information processing [2]. Additionally, due to the limitations of the experimental conditions and costs, the sample size of the eye-tracking experiment is usually small and only suitable for a single scene. In addition, a large number of studies have shown that environmental experience is the result of multi-sensory integration [55,56,57,58]. Therefore, future research could combine large-scale social media data to carry out emotional analysis and spatial feature extraction [59,60], capture the dynamic coupling characteristics of the multi-sensory stimulation of subjects in real environments, focus on exploring the rural landscape in central and Western China, further expand the spatial scale and population diversity of this research to obtain more universal and representative research results, comprehensively reveal the core factors affecting the visual behaviors and landscape preferences in rural landscapes, provide a more scientific theoretical basis for rural landscape design, and ultimately help the sustainable development of rural areas.

5. Conclusions

In order to break through the limitations of focusing on a single group in current rural landscape research, this study took Tianxi Village in central and Western China as an example and systematically explored the differences between visual behavior patterns and landscape preferences among 160 subjects by using eye-tracking technology and a landscape preference questionnaire. The results show that groups with different demographic characteristics (group identity, age, occupation, usual residence, and education level) have significant differences in visual behaviors and landscape preferences when viewing rural landscapes. Based on the experimental results, this study suggests that the visual behaviors and landscape preferences of different demographic groups should be fully considered in the process of rural landscape design, so as to achieve inclusive and diversified sustainable development. This study, once again, deepens the research system of rural landscape and promotes the exploration and practice of eye-tracking technology in the field of rural landscape research, not only providing important enlightenment for optimizing rural landscape design, but also providing theoretical guidance and methodological support for improving the quality of the rural tourism experience to achieve sustainable development in rural areas.