Memorable tourism experiences and critical outcomes among nature-based visitors: a fuzzy-set qualitative comparative analysis approach

ABSTRACT

Intense global competition for natural resources, manifesting in negative environmental externalities, has forced hospitality and tourism stakeholders to adopt strategies that may result in economic growth in tandem with environmental conservation. One such strategy is to cultivate travellers’ loyalty to the tourism destinations and encourage them to participate in environmentally friendly activities. Using the affective theory of social exchange, social identity, stimulus–organism–response, tourism consumption system and attachment theories, this study proposes and tests a configurational model that predicts the antecedents of visitors’ pro-environmental behavioural intentions (PEBIs) regarding their desire to revisit (REVI) and recommend (RECI) the services that they experienced. Fuzzy-set qualitative comparative analysis was applied to assess the effect of memorable tourism experiences, place attachment, and demographics on these outcomes. The findings revealed that multiple configurations can predict visitors’ intentions at tourist destinations. This study’s implications for theory, practice, and future research directions are also discussed.

KEYWORDS:

• Pro-environmental behaviour intention
• revisit intention
• recommendation intention
• memorable experiences
• place attachment

Introduction

The globalization of tourism has resulted in dramatic competition to commodify natural resources at the cost of economic growth. However, negative externalities, such as an increased strain on vulnerable ecosystems and the degradation of sites, are only a few of the unfavourable outcomes of this unsustainable development. Despite these negatives, tourism is still prescribed as a chief strategy to provide alternative livelihood possibilities for communities, especially those adjacent to marine protected areas (Pham-Do & Pham, 2020). These areas have the vitally important responsibility of maintaining biodiversity and ensuring wildlife conservation, which is usually funded by non-consumptive activities such as tourism (Maldonado-Oré & Custodio, 2020).

Nevertheless, increased demand fuelled by tourists’ growing desire to encounter endangered species before they disappear forever can pose an enormous risk to the biodiversity of these destinations. The most recent findings detail a tragic decline of almost 70% in the animal population since 1970 (WWF, 2020). Furthermore, the devastating impact of human behaviour on nature was discussed in a compelling UN report, which noted that one million species are threatened with extinction (UN, n.d.).

These reasons drive the growing pressure on destination management and tourism organizations to adopt eco-friendly strategies, especially in ecologically vulnerable tourist destinations, such as protected areas. As the economic resources generated by tourism are necessary to cover the likely expenses of protecting these geographical spaces and the habitats they contain, effective environmental management strategies are vital (Rezapouraghdam et al., 2021). Accordingly, destination managers are continuously searching for mechanisms through which they may achieve the two apparently opposing objectives of customer retention and environmental protection. One of the main objectives of the management efforts in ecologically sensitive places is to decrease visitors’ footprints by encouraging pro-environmental behaviours (PEBs; Rezapouraghdam et al., 2021).

Accordingly, it is important to understand the factors that may determine nature-based travellers’ desirable intentions, such as their pro-environmental behaviour intentions (PEBIs) and revisit (REVI) and recommendation (RECI) intentions towards the destination. This can facilitate sustainable destination development and management (Ramkissoon et al., 2018).

Therefore, our purpose with this study is to take an outcome-focused approach to unveil the factors that best explain the occurrence of combinations of the desired intentional outcomes. To do so, the links among memorable tourism experiences (MTEs), place attachment, and behavioural intentions will be observed for insight (Tsai, 2016). Experience plays such a unique role in the tourism industry that scholars call it the heart of the entertainment business (Loureiro, 2014) and the key to success and competitiveness for destinations (Ellis & Rossman, 2008). Tourism has traditionally been concerned with tourist experiences (Buonincontri et al., 2017; Stamboulis & Skayannis, 2003), such that each transaction associated with visitors at travel destinations can be understood as an experience (Oh et al., 2007).

A memorable experience is formed from individuals’ emotional assessment of real experiences during their travel. Recently, MTEs that are strong enough to enter the long-term memory (Oh et al., 2007) have been viewed favourably due to their significant role in tourism marketing and management (Sthapit et al., 2019). Such memorable experiences influence visitors’ intentions and determine their future behaviours (Kim et al., 2012).

Moreover, individuals’ behaviours and intentions are substantially influenced by the emotional bonds they make with places (Shaykh-Baygloo, 2020). Place attachment is a multidimensional construct comprised of place identity, place dependence, place affect, and developing a social bond with a place, which manifests in visitors’ emotional thoughts about it (Ramkissoon et al., 2013). According to Hidalgo and Hernandez (2001, p. 274), place attachment can be defined as ‘a positive affective bond between an individual and a specific place, the main characteristic of which is the tendency of the individual to maintain closeness to such a place’.

Previous study suggested scholars to extend the concept of MTEs to domains such as nature-based activities (Huang et al., 2019). Furthermore, the study of visitors’ behaviour in protected areas and among human-endangered species in contexts outside of Western countries requires more research (Shi et al., 2019). Additionally, considering a combination of various intentions (i.e. PEBI, RECI, and REVI) as outcome variables offer unprecedented implications to the prescription of strategies that focus on ideal visitor types (Lee et al., 2019).

Nature-based travel is an ecologically sensitive phenomenon (Akhshik et al., 2021), which, because of its numerous beneficial goals and practical significance, has received increasing attention in recent years. Meanwhile, protected natural areas are essential for preserving biodiversity, tourism, ecosystem services, and economic benefits for local communities (Maldonado-Oré & Custodio, 2020). As human activity is considered a major threat to the natural environment in these areas (Kim et al., 2020), the study of visitors’ sustainable behaviours and the observation of the factors that influence their future intentions is highly significant. Furthermore, the financial contribution of nature-based tourism is crucial to the preservation of biodiversity.

Nevertheless, despite the globally increased demand for nature-based tourism (Kim et al., 2020), the study of the influential combinations of factors that may influence travellers’ PEBI, REVI, and RECI in protected areas using non-linear approaches is limited. Therefore, this study’s findings will answer the following question: ‘The presence of which factors together can concurrently predict nature-based tourists’ PEBI, REVI, and RECI?’ Our study focuses on a shift from a linear to a non-linear approach in research design, data analysis, and interpretation and crafts a configural model using fuzzy-set qualitative comparative analysis (fsQCA). We believe intentions may be best evoked when their generating antecedents interact and reach a certain tipping point from which profound, intentional changes in the visitor’s belief system may occur (Gladwell, 2006).

Theoretical background and research model

Some scholars argue that theoretical and managerial techniques for explaining visitors’ behaviours in the complex social milieu of protected areas are poorly understood (Brown et al., 2010). This is because visitors’ impacts and the challenges of managing the human–nature relationship through conventional coexistence are characterized by complexity (Shi et al., 2019). In other words, the intentions and psychological processes that govern these behaviours are complex (Brown et al., 2010). That is why many researchers question the sufficiency of a single theory to comprehensively explain individuals’ PEBs or intentions (Bamberg & Möser; Onwezen et al., 2013). In response, researchers have started to integrate different theories to predict and understand individuals’ intentions and behaviours more accurately. For example, Han (2015) integrated the theory of planned behaviour (TPB) and value–belief–norm (VBN) theory into one theoretical framework to understand hotel guests’ PEB decision-making. Nevertheless, the author indicated that although the integration of these theories provided a better understanding of the intended outcomes (e.g. PEBs), the roles of other factors and variables should not be overlooked. Chuang et al. (2018) integrated sustainability values, TPB, the unified theory of acceptance, and the use of technology into their framework to predict travellers’ PEBs. They also concluded that consumers’ overall decision-making processes can be influenced by many internal and external factors, such as socio-demographic variables. Accordingly, in this study, we used several theories including affective theory of social exchange (Lawler, 2001), attachment theory (Bowlby, 1982), social identity theory (Brown, 2000), tourism consumption system theory (Woodside & Dubelaar, 2002) and stimulus–organism–response theory (Mehrabian & Russell, 1974) to explain how individuals’ various intentions develop based on the outcomes of their memorable tourism experiences, place attachment, and demographic backgrounds (see Figure 1).

Figure 1. Asymmetrical and configurational model of the study.

Memorable experiences

The outcome of a tourism product or service is generally psychological and reflects a variety of experiences (Pine & Gilmore, 1998). The guest experience in a tourism context has been recognized as a crucial element to enhance tourists’ memories of places (Wang, 1999). Lawler’s (2001) affective theory of social exchange can be used in this context to explain how MTEs can affect individuals’ attitudes and behaviours. This theory postulates that positive experiences can create positive emotions and result in attachment to people or places (Hosany et al., 2017). In other words, visitors’ various experiences from their interactions with the people and places they visit and the exchange of knowledge, gratitude, and affection with them can generate positive or negative emotions. MTEs must convey positivity when recalled after the event (Kim et al., 2012), and can lead to emotional attachment to a place.

Kim et al. (2012) developed the most comprehensive measure of MTEs, which includes seven salient dimensions and has been extensively used in the literature (e.g. Sthapit & Coudounaris, 2018). The dimensions of this scale are: 1) hedonism, which refers to the experience of a pleasurable feeling that engenders excitement; 2) refreshment, which is described as experiencing relaxation and being refreshed; 3) local culture, which is the experience of social interaction with local people and developing a positive impression of them; 4) meaningfulness, described as the experience of physical, emotional, or spiritual fulfilment through tourism; 5) knowledge, which is experiencing new facts and information about a new culture; 6) involvement, which is the experience of participating in on-site activities and planning travel; and 7) novelty, which is the experience of partaking in something distinctive and different from past experiences. Social judgement theory and the concept of involvement, which is provoked by specific stimuli in the environment (Havitz & Dimanche, 1997; Zaichkowsky, 1985), can explain how tourists’ participation in leisure activities that are characterized by self-expression, enjoyment, importance, and freedom can attach them to a destination (Xu & Zhang, 2016).

Kim and Ritchie (2014) empirically showed that MTEs affect behavioural intentions among Taiwanese respondents, as other studies have posited. However, more detailed observations are required to unpack the mechanism through which a complex phenomenon such as MTEs may influence different behavioural intentions in a nature-based tourism setting.

Place attachment

Building on attachment theory (Bowlby, 1982), humans have an innate psychological tendency to establish and maintain security via their close links. Many scholars argue that a bond with a meaningful space is a universal affective tie (Relph, 1976). Development theory (Morgan, 2010) also asserts that positive emotions towards a destination can induce attachment through person–place interactions. Therefore, it is reasonable to presume that people who perceive their tourism experiences at a destination positively would tend to become emotionally attached.

Drawing on Ramkissoon et al.’s (2013) conceptualization of place attachment, place dependence is tourists’ awareness of the uniqueness of a setting and its ability to provide their desired travel experiences. Place identity refers to one’s beliefs and perceptions that the self is invested in a specific place. Place affect, alternatively, is the emotional bond that people create with places. Social bonding reflects the experiences that individuals have through social interactions in a place.

Loureiro (2014) observed the role that visitors’ experiences play in place attachment and behavioural intentions in a rural context in Portugal. Importantly, in protected areas, destination managers and scientists agree that the major role of place attachment in decreasing human environmental impacts can have significant implications for PEBIs (Buta et al., 2014; Ramkissoon et al., 2018). A study conducted among visitors to Italy found that tourists’ motivations positively affected their place attachment and that, consequently, place attachment predicted their RECI (Prayag et al., 2018).

Place attachment’s contribution to supporting tourism development was also studied in Kavala, Greece (Stylidis, 2018). Additionally, Vada et al. (2019) found that MTEs significantly influence tourists’ place attachment through the mediation of hedonic and eudaimonic well-being, and residents’ place attachment reportedly affected their positive word-of-mouth behaviour in Sydney and Shanghai (Chen et al., 2018). Moreover, social identity theory (SIT) also supports the linkage between place attachment and individuals’ environmentally friendly behaviours, such that those who identify with a place are more inclined to behave in the interests of that place (Brown, 2000).

Travellers’ place attachment is related to their experiences and engagement with a place (Gross & Brown, 2008). Environmental and personal factors, like positive memories and experiences, are the primary motivators that facilitate the development of place attachment (Loureiro, 2014). Cognitive appraisal theory argues that people’s emotions are the outcome of what they experience in a place, which then influence their intentions (Io, 2018). Previous studies have asserted that the type and level of attachment might lead to different relationships with visitors’ intentions. Scholars have encouraged further research to understand how place attachment affects PEBI (Ramkissoon et al., 2013).

Future intentions

The term ‘behavioural intentions’ refers to a person’s intentions to perform or not perform a particular behaviour and are predictive of actual behaviours (Brown et al., 2010). The drivers of travel intentions (such as MTEs) are regarded as outputs of people’s previous tourism activities, which influence their future intentions (Sthapit & Coudounaris, 2018). MTEs have been asserted as determinant variables that affect individuals’ future intentions (RECI and REVI) after visiting a tourist destination (Bujisic et al., 2015). Nonetheless, a previous study found that the relationships are complex and not all dimensions of a tourism experience influence tourists in the same ways in the post-experience stage (Triantafillidou & Petala, 2016).

Additionally, the relationship between individuals’ place attachment and their environmental behavioural intentions has been thoroughly documented (Buta et al., 2014; Qu et al., 2019). Lee et al. (2015) found a linear structural relationship between recreation experiences and tourists’ environmentally friendly behaviour in Taiwan. People’s interactions in natural environments and their recreational experiences significantly contribute to the development of human–nature bonds (Buta et al., 2014).

Tourists’ behavioural intentions have been studied in various contexts throughout the literature. For example, the effect of street food on consumer behavioural intentions was observed in India (Gupta et al., 2018), and travellers’ REVI was determined to be influenced by destination and country image, mediated by MTEs, among international tourists in China (Zhang et al., 2018). Hasan et al. observed coastal tourists in Bangladesh and showed that visitors’ satisfaction and destination images had considerable effects on their REVI (Hasan et al., 2019). In the context of wine tourism in Greece, a study observed the effect of tourists’ experiences on their emotions, satisfaction, REVI, and RECI (Leri & Theodoridis, 2019). A previous study in the context of sports event tourism asserted the significant role place attachment exerted on individuals’ behavioural intentions in South Korea (Jeong et al., 2020).

Tourism consumption system theory (Woodside & Dubelaar, 2002), argues that travellers’ evaluation of the destination experience affects their overall perception of the destination as well as their behaviour (Sharma & Nayak, 2019). According to this theory, the link between MTEs and their outcomes can be supported as MTEs enhance visitors’ desirable behavioural intentions (e.g. PEB, REVI, and RECI) directly and indirectly by satisfying the innate human need for attachment (Relph, 1976).

Stimulus–organism–response theory (SOR), developed by Mehrabian and Russell (1974), explains that individuals are influenced both physically and emotionally by their surroundings (S), which per se nurture their internal states (O) and result in reactions (R). MTEs can nurture visitors’ emotional attachment and motivate them to have positive intentions towards the destination.

Given the above-discussed theories, this study observes the links among MTEs, place attachment, and PEBI, REVI, RECI, and their combinations. The attributes of MTEs and the components of place attachment were considered antecedent configurations of visitors’ behavioural intentions.

Methodology

Measurement scales

Several established constructs were extracted from the relevant literature. Six items from Kiatkawsin and Han (2017) were used to measure PEBI; for example, ‘I would prefer to buy local products’. REVI and RECI were measured with two items adapted from Toyama and Yamada (2012): ‘Would you like to visit this location again in the future?’ and ‘Would you recommend this location to others?’ Place attachment was measured with 10 items from the categories of place affect (three items), place dependence (three items), and place identity (four items; Kulczycki & Halpenny, 2014). Finally, 24 items from Kim and Ritchie (2014) were used to measure visitors’ memorable experiences in seven dimensions. Most of the constructs were measured on a 5-point Likert scale (anchored by strongly disagree = 1 and strongly agree = 5). REVI and RECI were measured on a 7-point Likert scale (strongly disagree = 1; strongly agree = 7).

Data collection procedure

The study applied a systematic approach to collect empirical data. For the first step, a four-part survey questionnaire (the details of which are provided below) was prepared. A pilot study of 20 participants confirmed the quality of the questionnaire (e.g. the clarity of the questions, readability, and timing). A series of procedural remedies, such as a cover letter explaining the study purpose and outlining the steps that would be taken to ensure anonymity and confidentiality, were employed to avoid common method bias during this phase (Podsakoff et al., 2003). Moreover, to avoid yea-saying and nay-saying, reverse-coded items were embedded into the questionnaire.

In the second step, the Marine Protected Area Administration (a society for the protection of turtles) granted permission for the researchers to observe the turtles and their habitat. The researchers were based at the primary turtle nesting site of Alagadi Beach, Northern Cyprus, during August and September 2019. Guided tours of the area were booked in advance for small groups of no more than 20 people. Activities included participation in a video-based educational programme, a detailed introduction to the species by a specialist, and an interpretive activity in which baby turtles were released into the sea.

Data analysis procedure

In the first step, the composition of the measurement items was explored with a set of factor analyses using a principal component analysis (PCA) method with varimax rotation. The above process was confirmed with confirmatory factor analysis (CFA) through a maximum likelihood estimator (Fornell & Larcker, 1981).

The reliability was tested with Cronbach’s alpha (α), and composite reliability (CR) validated the internal consistency of the variables. Then, the convergent and discriminant validity were evaluated. Harman’s single-factor test was used to evaluate the potential common method variance (Podsakoff et al., 2003).

In the second step, cross-tabulation analyses of two correlated variables (using Cramér’s V test) validated the existence of cases that opposed the main net effect. This step affirmed the complexity of the relationships between variables, further suggesting an asymmetrical approach to data analysis. Fiss (2011) encouraged researchers to perform fsQCA when their findings were subject to asymmetry.

The interpretation of fsQCA results is an elaborate process as it produces substantial exploratory information that goes beyond quantitative and qualitative analysis (Ragin, 2009). Therefore, unlike in conventional symmetric methods (Armstrong, 2011), an asymmetric approach such as fsQCA allows complex relationships of antecedents that generate the same combination of outputs to be assessed. In asymmetrical modelling, the terms ‘antecedent’ and ‘ingredient’ correspond to the ‘independent variable’ in symmetrical modelling, and instead of ‘net effect’, as in symmetrical modelling, the term ‘recipe’ is used in asymmetrical models.

In the third step, fsQCA was performed using Tosmana version 1.6 software to retrieve the causal recipes for the desired outcome and identify the antecedents that were necessary to achieve that outcome using necessary condition analysis (NCA; Dul, 2016). In this step, a three-phase fsQCA was performed using Ragin’s (2014) guidelines. First, the data were calibrated and transformed into fuzzy membership scores. Then, a truth table was crafted with possible low and high outcome scores, and counterfactual analysis refined the conditions to provide the recipes. The results that Tosmana generated were then compared with the results produced by fsQCA (Ragin, 2014). No difference was observed in the generated results or configural models.

Finally, predictive validity was provided and the generated configural models were assessed according to the premise of complexity theory (Woodside, 2017). From almost 500 visitors to Alagadi Beach during the study period, 406 valid questionnaires were extracted for further analysis. The sample included 175 men (43.1%) and 231 women (56.9%). Additional demographic data are listed in Table 1.

Table 1. Respondents’ profile (n = 406).

Characteristics	Frequency	Percentage
Gender
Male	175	43.1
Female	231	56.9
Age
18–29	30	7.4
30–49	126	31.0
50–64	156	38.4
Over 65	94	23.2
Education level
No schooling completed	47	10.6
Some high school	118	29.1
Associate degree/diploma	116	28.6
Trade/technical/Vocational training	77	19.0
Bachelor’s degree	36	8.9
Graduate and higher degree	12	3.0
Marital status
Married	254	62.6
Single	152	37.4
Income
Less than 1000 USD	91	22.4
1000 – 2999 USD	103	25.4
3000 – 5999 USD	184	45.3
More than 6000 USD	28	6.9
Nationality
British	146	35.9
Cypriot	37	9.1
Russian	34	8.3
German	33	8.1
Swedish	24	5.9
Turkey	26	6.4
Others	106	26.1
Total	406	100%

Results and discussion

Results of the preliminary tests

After data screening, several questionnaires were discarded because they had missing values of greater than 5%, leaving 406 valid questionnaires. The mean replacement technique was used to overcome missing values of less than 5%, and this approach did not affect the mean of the variables. In addition, early and subsequent versions of the collected questionnaires were compared to test for nonresponse bias (Armstrong & Overton, 1977). No significant differences (p > .05) were identified, providing further evidence that nonresponse bias was not a concern in this study. The normality of the data was judged by skewness and kurtosis values, which were in the acceptable range of ±1 and ±3 respectively.

The results of the PCA and descriptive statistics are presented in Table A1. All items were loaded accurately and at an acceptable level (λ > 0.69) under the desired measuring constructs. The eigenvalue of each construct was greater than 1.00 (see Table A1). The results of calculating Cronbach’s alpha coefficient and the average variance extracted (AVE) affirm the construct reliability, with alphas higher than the suggested cut-off level (α > 0.70 and AVE > 0.5; Cortina, 1993). The results provided in Table 2 also validate the internal consistency of the measurement items through CR that meets the threshold of 0.70 (Fornell & Larcker, 1981).

Table 2. Result of CFA, CR, AVE, MSV and α.

Scale Items		β	CR	AVE	MSV	α
Pro-environmental behavior intention (PEBI)			.858	.522	.015	.722
Pebi3	I would try to save water and electricity i.e. turning off the tap while washing/brushing teeth, turning off the lights if I leave the room for more than 10 min, walking up the stairs if only need to go 1 floor up, and using hotel towels more than once.	.984
Pebi6	I would buy products in eco-friendly packaging when possible i.e. avoid plastic shopping bags, plastic bottles and try to reuse bottles and bags.	.966
Pebi2	I would try to learn about the recycling facilities and actions of the locals.	.531
Pebi4	I would try to protect local resources as much as I could i.e. I would voluntarily stop visiting a famous spot if it needed to recover from environmental damage and I would not disturb any creatures and vegetation, for example, feeding fish and birds or picking flowers.	.530
Pebi1	I would prefer to buy local products.	.538
Pebi5	I would try to dispose garbage properly if possible i.e. sort my garbage into separate containers for paper, plastic, glass, etc.	.613
Hedonism (hed)			.876	.641	.402	.800
Hed3	Really enjoyed this tourism experience	.814
Hed4	Exciting	.699
Hed2	Indulged in the activities	.901
Hed1	Thrilled about having a new experience	.774
Place Identity (ide)			.871	.628	.009	.793
Ide3	I identify strongly with Alagadi beach	.836
Ide2	Visiting Alagadi beach says a lot about who I am	.838
Ide1	I feel like Alagadi beach is part of me	.788
Ide4	When I see Alagadi beach other see me the way I want them to see me	.701
Knowledge (kno)			.928	.812	.027	.901
Kno2	Knowledge	.922
Kno1	Exploratory	.886
Kno3	New culture	.894
Novelty (nov)			.824	.541	.105	.735
Nov3	Different from previous experiences	.77
Nov4	Experienced something new	.708
Nov2	Unique	.801
Nov1	Once-in-a lifetime experience	.654
Refreshment (ref)			.815	.529	.176	.728
Ref2	Enjoyed sense of freedom	.832
Ref4	Revitalized	.555
Ref3	Refreshing	.733
Ref1	Liberating	.761
Involvement (inv)			.892	.735	.023	.857
Inv2	I enjoyed activities which I really wanted to do	.938
Inv3	I was interested in the main activities of this tourism experience	.875
Inv1	I visited a place where I really wanted to go	.747
Local culture (loc)			.885	.72	.094	.849
Loc3	Local people in a destination were friendly	.94
Loc1	Good impressions about the local people	.828
Loc2	Closely experienced the local culture	.770
Place Affect (aff)			.84	.637	.319	.798
Aff2	I feel happiest when I am in Alagadi beach	.815
Aff1	I have strong positive feelings for Alagadi beach	.774
Aff3	I am fond of Alagadi beach	.804
Place Dependence (dep)			.775	.539	.176	.734
Dep1	Alagadi beach is the best place for what I like to do	.859
Dep2	I would not substitute any other area for the types of things I do at Alagadi beach	.704
Dep3	I get more satisfaction out of visiting Alagadi beach than any other place	.619
Meaningfulness (mng)			.767	.526	.402	.725
Mng2	I did something important	.774
Mng3	Learned about myself	.619
Mng1	I did something meaningful	.771

Note: β: standardized factor loading; β is significant at the .001 level; AVE: average variance extracted; MSV: maximum shared squared variance; CR: composite reliability; α: Chronbach’s Alpha.

Table 2 also indicates that all items were loaded sufficiently under their factors; therefore, no items were dropped during this stage.

Results of cross-tabulation analyses

Table 3 supports the existence of contrarian cases in the dataset between statistically correlated variables (i.e. between meaningfulness and pro-environmental behaviour). These cases oppose the main net effect and our intuition. Conventional methods ignore the existence of such cases in the dataset (Olya & Akhshik, 2019). Ignoring the asymmetry of the measurement items creates an illusion during data interpretation, thereby compromising decision-making processes based on these data (Armstrong, 2011). Table 3 illustrates that 24 individuals (6% of the sample) did not derive any sense of meaningfulness from the area they visited but still intended to behave in a pro-environmental manner.

Table 3. Cross-Tabulation Analysis of Pro-environmental Behaviour Intentions (PEBI) * Meaningfulness.

			Pro-environmental Behaviour Intentions (PEBI)					Total
			Strongly disagree	Disagree	Neutral	Agree	Strongly agree
Meaningfulness	Strongly disagree	Count	2	9	9	2	0	22
		% within Meaningfulness	9.1%	40.9%	40.9%	9.1%	0.0%	100.0%
		% within pebi	6.9%	9.1%	4.7%	2.7%	0.0%	5.4%
		% of Total	0.5%	2.2%	2.2%	0.5%	0.0%	5.4%
	Disagree	Count	11	22	36	19	3	91
		% within Meaningfulness	12.1%	24.2%	39.6%	20.9%	3.3%	100.0%
		% within pebi	37.9%	22.2%	18.7%	25.7%	27.3%	22.4%
		% of Total	2.7%	5.4%	8.9%	4.7%	0.7%	22.4%
	Neutral	Count	13	41	89	30	8	181
		% within Meaningfulness	7.2%	22.7%	49.2%	16.6%	4.4%	100.0%
		% within pebi	44.8%	41.4%	46.1%	40.5%	72.7%	44.6%
		% of Total	3.2%	10.1%	21.9%	7.4%	2.0%	44.6%
	Agree	Count	3	21	51	23	0	98
		% within Meaningfulness	3.1%	21.4%	52.0%	23.5%	0.0%	100.0%
		% within pebi	10.3%	21.2%	26.4%	31.1%	0.0%	24.1%
		% of Total	0.7%	5.2%	12.6%	5.7%	0.0%	24.1%
	Strongly agree	Count	0	6	8	0	0	14
		% within Meaningfulness	0.0%	42.9%	57.1%	0.0%	0.0%	100.0%
		% within pebi	0.0%	6.1%	4.1%	0.0%	0.0%	3.4%
		% of Total	0.0%	1.5%	2.0%	0.0%	0.0%	3.4%
Total		Count	29	99	193	74	11	406
		% within Meaningfulness	7.1%	24.4%	47.5%	18.2%	2.7%	100.0%
		% within pebi	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%
		% of Total	7.1%	24.4%	47.5%	18.2%	2.7%	100.0%

Note: Cramer’s V = .336, p < 0.000; The marked numbers illustrate contrarian cases ∼ Meaningfulness → Pro-environmental behavior intention (24 cases) and Meaningfulness → ∼Pro-environmental behavior intention (30 cases).

Moreover, Table 3 reveals negative contrarian cases (30 cases, 7.3%) in which visitors who perceived meaningfulness did not intend to behave in a pro-environmental manner. The results reveal the heterogeneity of indicating PEBI and the asymmetric relationships between the outcome and its antecedents. This affirms that contrarian cases exist even when the associations between predictors and outcomes are positive and their effect size is significant.

Table 3 demonstrates how linear modelling may lead to disregarding important information in our dataset. Therefore, to comprehend complex relationships, configural models are recommended to attain a better understanding of the studied phenomena (Woodside, 2015).

Results of model testing

The results of this study’s configural model, depicted in Figure 1 as arrows A to C, are presented in Tables 4 and 5. Coverage and consistency are two criteria for selecting the consistent and sufficient recipes that are generated in fuzzy truth tables (Olya & Akhshik, 2019; Olya & Han, 2020). Occurrences of more than 1 and a consistency of at least .8 are considered acceptable levels for each recipe (Ragin, 2014).

Table 4. Configural Models of high and low scores of Outcome (model A, B and their negations).

Models for predicting high score of Outcome	C	RC	UC	Models for predicting low score of Outcome	C	RC	UC
A. Outcome =  f(hed, kno, inv, loc, mng, nov, ref, gen,mar,age,inc,edu)				∼A. ∼Outcome =  f(hed, kno, inv, loc, mng, nov, ref, gen,mar,age,inc,edu)
				M1. hed * mng* loc* nov* inv * kno * ref * age * ∼inc * edu* ∼gen* mar	.99	.15	**
M1. hed * mng* loc* ∼nov* ∼inv * kno* ref * age * ∼inc* ∼edu* gen * mar	.82	.23	**
Solution coverage: .23 Solution consistency: .82				Solution coverage: .15 Solution consistency: .99
B. Outcome = f(dep, ide, aff, gen,mar,age,inc,edu )				∼B. ∼Outcome = f(dep, ide, aff, gen,mar,age,inc,edu)
M1. aff * ∼dep * ide * age* ∼inc* ∼edu* ∼gen	.75	.32	.10	M1. aff * dep * ide * age* ∼inc	.94	.51	**
M2. aff * dep * age* ∼inc* ∼edu* ∼gen* ∼mar	.78	.25	.03	M2. aff * ide * age* ∼inc* gen* mar	.92	.27	**
				M3. aff * ide * age* ∼inc* ∼edu* ∼gen	.96	.23	**
Solution coverage: .35 Solution consistency: .72				M4. aff * dep * ide * age* ∼edu* gen* mar	.96	.23	**
				M5. aff * dep * age* ∼inc* ∼edu* ∼gen*	.96	.13	**
				M6. aff * dep * ide * ∼inc* ∼edu* ∼gen* ∼mar	.96	.14	**
				Solution coverage: .60 Solution consistency: .92

Note: RC = raw coverage; C = consistency; UC: unique coverage; * =  and; ∼ = negation; Outcome =  pebi $\cap$ revi $\cap$ reci; pebi =  Pro-environmental behaviour intention; revi = Revisit intention; reci = Intention to recommend; age = respondents’ age; gen =  gender; mar = marital status; inc = income; edu = education; hed = Hedonism; kno = Knowledge; inv = Involvement; loc = Local culture, mng = Meaningfulness; nov = Novelty; ref = Refreshment; dep = Place dependence; ide = Place identity; aff = Place affect; marital status and gender are dummy variables: 0 indicates: single and men, while 1 indicates: married, women respectively.

Table 5. Configural Models of high and low scores of REVI, PEBI, RECI and outcome with all the antecedents (model C and their negations).

Models for predicting high/low scores of REVI, PEBI, RECI, and outcome	C	RC
C1. REVI = f(hed, kno, inv, loc, mng, nov, ref, dep, ide, aff, gen, mar, age, inc,edu)
M1. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age* inc*edu* gen*∼mar	0.95	0.10
M2. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* gen*∼mar	0.96	0.10
M3. ∼hed*∼mng*∼loc*nov*inv* kno* ref* ∼aff*dep* ide* age*∼inc* ∼edu* ∼gen* mar	0.96	0.11
M4. hed*mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age*∼inc*edu* gen*∼mar	0.96	0.12
M5. hed*∼mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* ∼gen*∼mar	0.97	0.11
M6. ∼hed*∼mng*loc*∼nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc*edu* gen*∼mar	0.97	0.11
M7. hed*∼mng*loc*nov* ∼inv* kno* ∼ref* ∼aff* ∼dep∼ide* ∼age*∼inc* ∼edu* ∼gen* mar	0.98	0.11
M8. hed*mng*loc*nov*inv* kno* ∼ref* aff*dep* ide* age*∼inc*edu* ∼gen*∼mar	0.95	0.12
M9. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep∼ide* age* inc* ∼edu* gen* mar	0.95	0.12
M10. hed*mng*loc*nov*inv* kno* ref* aff* ∼dep* ide* age*∼inc* ∼edu* ∼gen*∼mar	0.96	0.12
Solution coverage: .23 Solution consistency: .88
∼C1. ∼REVI = f(hed, kno, inv, loc, mng, nov, ref, dep, ide, aff, gen, mar, age, inc,edu)
M1. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc* ∼edu* gen*∼mar	0.97	0.13
M2. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age* inc*edu* gen*∼mar	0.98	0.12
M3. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* ∼gen* mar	0.95	0.12
M4. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* gen*∼mar	0.99	0.11
M5. hed*∼mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* ∼gen*∼mar	0.95	0.12
M6. ∼hed*∼mng*loc*∼nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc*edu* gen*∼mar	0.95	0.12
M7. hed*∼mng*loc*nov*inv* kno* ref* aff*dep* ide* ∼age*∼inc* ∼edu* ∼gen*∼mar	0.95	0.13
M8. hed*∼mng*loc*nov* ∼inv* kno* ∼ref* ∼aff* ∼dep∼ide* ∼age*∼inc* ∼edu* ∼gen* mar	0.95	0.11
M9. hed*∼mng*loc*nov* ∼inv* ∼kno* ∼ref* aff*dep* ide* age*∼inc* ∼edu* gen* mar	0.97	0.13
M10. hed*mng*loc*nov* ∼inv* kno* ref* aff* ∼dep* ide* age*∼inc* ∼edu* ∼gen*∼mar	0.96	0.13
M11. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep∼ide* age* inc* ∼edu* gen* mar	0.96	0.13
M12. hed*mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* gen* mar	0.95	0.15
Solution coverage: .28 Solution consistency: .88
C2. PEBI = f(hed, kno, inv, loc, mng, nov, ref,dep, ide, aff, gen, mar, age, inc,edu)
M1. hed*mng*loc*∼nov* ∼inv* kno* ref* aff* ide* age*∼inc* ∼edu* gen* mar	0.96	0.14
M2. hed*mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age*∼inc*edu* "cgen"	0.96	0.18
M3. hed*mng*loc*nov* kno* ref* aff* ∼dep* ide* age*∼inc* ∼edu* ∼gen*∼mar	0.97	0.12
M4. hed*mng*loc*nov*inv* kno* ∼ref* aff*dep* ide* age* inc* ∼edu* mar	0.97	0.17
M5. hed*mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age* ∼edu* gen* mar	0.97	0.17
M6. hed*mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age*∼inc* ∼edu* mar	0.96	0.21
M7. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc* ∼edu* gen*∼mar	0.98	0.11
M8. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age* inc*edu* gen*∼mar	0.99	0.10
M9. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* ∼gen* mar	0.98	0.10
M10. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* gen*∼mar	0.99	0.09
M11. hed*mng*loc*nov*inv* ∼kno* ref* aff*dep* ide* age*∼inc*edu* gen* mar	0.98	0.13
M12. ∼hed*∼mng*∼loc*nov*inv* kno* ref* ∼aff*dep* ide* age*∼inc* ∼edu* ∼gen* mar	0.98	0.10
M13. hed*mng*loc*nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc*edu* gen* mar	0.98	0.14
M14. hed*∼mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age*∼inc*edu* ∼gen* mar	0.97	0.13
M15. hed*∼mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* ∼gen*∼mar	0.99	0.10
M16. hed*mng*loc*nov*inv* kno* ref* aff* ∼dep* ide* age* inc* ∼edu* gen* mar	0.95	0.14
M17. ∼hed*∼mng*loc*∼nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc*edu* gen*∼mar	0.97	0.10
M18. hed*∼mng*loc*nov*inv* kno* ref* aff*dep* ide* ∼age*∼inc* ∼edu* ∼gen*∼mar	0.98	0.11
M19. hed*∼mng*loc*nov* ∼inv* kno* ∼ref* ∼aff* ∼dep∼ide* ∼age*∼inc* ∼edu* ∼gen* mar	1.00	0.10
M20. hed*∼mng*loc*nov* ∼inv* ∼kno* ∼ref* aff*dep* ide* age*∼inc* ∼edu* gen* mar	0.96	0.11
M21. hed*mng*loc*nov*inv* kno* ∼ref* aff*dep* ide* age*∼inc*edu* ∼gen*∼mar	0.98	0.11
M22. hed*mng*loc*nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc* ∼edu* ∼gen* mar	0.97	0.12
M23. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep∼ide* age* inc* ∼edu* gen* mar	0.98	0.11
M24. hed*mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* gen* mar	0.98	0.13
Solution coverage: .44 Solution consistency: .91
∼C2. ∼PEBI = f(hed, kno, inv, loc, mng, nov, ref,dep, ide, aff, gen, mar, age, inc,edu)
M1. hed*mng*loc*nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc* ∼edu* mar	0.93	0.25
M2. hed*mng*loc*nov* kno* ref* aff* ∼dep* ide* age*∼inc* ∼edu* ∼gen*∼mar	0.95	0.16
M3. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc* ∼edu* gen*∼mar	0.95	0.14
M4. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age* inc*edu* gen*∼mar	0.98	0.13
M5. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* ∼gen* mar	0.97	0.14
M6. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* gen*∼mar	0.99	0.13
M7. hed*mng*loc*∼nov* ∼inv* kno* ref* aff* ∼dep* ide* age*∼inc* ∼edu* gen* mar	0.95	0.17
M8. ∼hed*∼mng*∼loc*nov*inv* kno* ref* ∼aff*dep* ide* age*∼inc* ∼edu* ∼gen* mar	0.97	0.13
M9. hed*mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age*∼inc*edu* gen*∼mar	0.95	0.15
M10. hed*∼mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* ∼gen*∼mar	0.98	0.14
M11. ∼hed*∼mng*loc*∼nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc*edu* gen*∼mar	0.98	0.14
M12. hed*∼mng*loc*nov*inv* kno* ref* aff*dep* ide* ∼age*∼inc* ∼edu* ∼gen*∼mar	0.96	0.15
M13. hed*∼mng*loc*nov* ∼inv* kno* ∼ref* ∼aff* ∼dep∼ide* ∼age*∼inc* ∼edu* ∼gen* mar	0.96	0.13
M14. hed*∼mng*loc*nov* ∼inv* ∼kno* ∼ref* aff*dep* ide* age*∼inc* ∼edu* gen* mar	0.99	0.15
M15. hed*mng*loc*nov*inv* kno* ∼ref* aff*dep* ide* age* inc* ∼edu* gen* mar	0.95	0.18
M16. hed*mng*loc*nov*inv* kno* ∼ref* aff*dep* ide* age*∼inc*edu* ∼gen*∼mar	0.95	0.15
M17. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep∼ide* age* inc* ∼edu* gen* mar	0.95	0.15
M18. hed*mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* gen* mar	0.96	0.17
Solution coverage: .44 Solution consistency: .84
C3. RECI = f(hed, kno, inv, loc, mng, nov, ref,dep, ide, aff, gen, mar, age, inc,edu)
M1. hed*mng*loc*nov* kno* ref* aff* ∼dep* ide* age*∼inc* ∼edu* ∼gen*∼mar	0.95	0.11
M2. hed*mng*loc*nov*inv* kno* ∼ref* aff*dep* ide* age* inc* ∼edu* mar	0.95	0.16
M3. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc* ∼edu* gen*∼mar	0.95	0.10
M4. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age* inc*edu* gen*∼mar	0.95	0.09
M5. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* ∼gen* mar	0.95	0.10
M6. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* gen*∼mar	0.95	0.09
M7. hed*mng*loc*nov*inv* ∼kno* ref* aff*dep* ide* age*∼inc*edu* gen* mar	0.96	0.12
M8. hed*mng*loc*nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc*edu* gen* mar	0.96	0.13
M9. hed*mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age*∼inc*edu* gen*∼mar	0.95	0.10
M10. hed*∼mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* ∼gen*∼mar	0.98	0.09
M11. ∼hed*∼mng*loc*∼nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc*edu* gen*∼mar	0.97	0.09
M12. hed*∼mng*loc*nov* ∼inv* kno* ∼ref* ∼aff* ∼dep∼ide* ∼age*∼inc* ∼edu* ∼gen* mar	0.98	0.09
M13. hed*∼mng*loc*nov* ∼inv* ∼kno* ∼ref* aff*dep* ide* age*∼inc* ∼edu* gen* mar	0.98	0.10
M14. hed*mng*loc*nov*inv* kno* ∼ref* aff*dep* ide* age*∼inc*edu* ∼gen*∼mar	0.96	0.11
M15. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep∼ide* age* inc* ∼edu* gen* mar	0.96	0.10
M16. hed*mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* gen* mar	0.95	0.12
Solution coverage: .31 Solution consistency: .88
∼C3. ∼RECI = f(hed, kno, inv, loc, mng, nov, ref,dep, ide, aff, gen, mar, age, inc,edu)
M1. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age* inc*edu* gen*∼mar	0.97	0.13
M2. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* gen*∼mar	0.97	0.13
M3. ∼hed*∼mng*∼loc*nov*inv* kno* ref* ∼aff*dep* ide* age*∼inc* ∼edu* ∼gen* mar	0.95	0.14
M4. hed*∼mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* ∼gen*∼mar	0.96	0.14
M5. hed*∼mng*loc*nov* ∼inv* kno* ∼ref* ∼aff* ∼dep∼ide* ∼age*∼inc* ∼edu* ∼gen* mar	0.95	0.13
M6. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep∼ide* age* inc* ∼edu* gen* mar	0.95	0.15
M7. hed*mng*loc*nov*inv* kno* ref* aff* ∼dep* ide* age*∼inc* ∼edu* ∼gen*∼mar	0.95	0.15
Solution coverage: .25 Solution consistency: .88
C4. Outcome = f(hed, kno, inv, loc, mng, nov, ref,dep, ide, aff, gen, mar, age, inc,edu)
M1. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* gen*∼mar	0.95	0.17
M2. hed*∼mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* ∼gen*∼mar	0.95	0.18
M3. hed*∼mng*loc*nov* ∼inv* kno* ∼ref* ∼aff* ∼dep∼ide* ∼age*∼inc* ∼edu* ∼gen* mar	0.97	0.18
Solution coverage: .22 Solution consistency: .90
∼C4. ∼Outcome = f(hed, kno, inv, loc, mng, nov, ref,dep, ide, aff, gen, mar, age, inc,edu)
M1. hed*mng*loc*∼nov* ∼inv* kno* ref* aff* ide* age*∼inc* ∼edu* gen* mar	0.99	0.12
M2. hed*mng*loc*nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc* gen* mar	0.96	0.13
M3. hed*mng*loc*nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc* ∼edu* mar	0.97	0.16
M4. hed*mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age*∼inc*edu* "cgen"	0.98	0.15
M5. hed*mng*loc*nov* kno* ref* aff* ∼dep* ide* age*∼inc* ∼edu* ∼gen*∼mar	0.98	0.10
M6. hed*mng*loc*nov*inv* kno* ∼ref* aff*dep* ide* age* inc* ∼edu* mar	0.98	0.14
M7. hed*mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age* ∼edu* gen* mar	0.98	0.14
M8. hed*mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age*∼inc* ∼edu* mar	0.98	0.18
M9. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc* ∼edu* gen*∼mar	1.00	0.09
M10. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep* ide* age* inc*edu* gen*∼mar	1.00	0.08
M11. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* ∼gen* mar	0.99	0.08
M12. hed*mng*∼loc*∼nov* ∼inv* kno* ref* aff*dep* ide* age*∼inc*edu* gen*∼mar	1.00	0.08
M13. hed*mng*loc*nov*inv* ∼kno* ref* aff*dep* ide* age*∼inc*edu* gen* mar	0.98	0.10
M14. ∼hed*∼mng*∼loc*nov*inv* kno* ref* ∼aff*dep* ide* age*∼inc* ∼edu* ∼gen* mar	0.99	0.08
M15. hed*∼mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age*∼inc*edu* ∼gen* mar	0.99	0.11
M16. hed*∼mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* ∼gen*∼mar	0.99	0.08
M17. hed*mng*loc*nov*inv* kno* ref* aff* ∼dep* ide* age* inc* ∼edu* gen* mar	0.99	0.12
M18. ∼hed*∼mng*loc*∼nov*inv* kno* ∼ref* aff* ∼dep* ide* age*∼inc*edu* gen*∼mar	0.99	0.08
M19. hed*∼mng*loc*nov*inv* kno* ref* aff*dep* ide* ∼age*∼inc* ∼edu* ∼gen*∼mar	0.99	0.09
M20. hed*∼mng*loc*nov* ∼inv* kno* ∼ref* ∼aff* ∼dep∼ide* ∼age*∼inc* ∼edu* ∼gen* mar	0.98	0.08
M21. hed*∼mng*loc*nov* ∼inv* ∼kno* ∼ref* aff*dep* ide* age*∼inc* ∼edu* gen* mar	1.00	0.09
M22. hed*mng*loc*nov*inv* kno* ∼ref* aff*dep* ide* age*∼inc*edu* ∼gen*∼mar	0.98	0.09
M23. hed*mng*loc*nov* ∼inv* kno* ref* aff*dep∼ide* age* inc* ∼edu* gen* mar	1.00	0.09
M24. hed*mng*loc*nov*inv* kno* ref* aff*dep∼ide* age*∼inc* ∼edu* gen* mar	1.00	0.10
Solution coverage: .38 Solution consistency: .95

Note: RC = raw coverage; C = consistency; ∼ = negation; Outcome =  PEBI $\cap$ REVI $\cap$ RECI; PEBI =  Pro-environmental behaviour intention; REVI = Revisit intention; RECI = Intention to recommend; age = respondents’ age; gen =  gender; mar = marital status; inc = income; edu  = education; hed = Hedonism; kno = Knowledge; inv = Involvement; loc = Local culture, mng = Meaningfulness; nov = Novelty; ref = Refreshment; dep = Place dependence; ide = Place identity; aff = Place affect; marital status and gender are dummy variables: 0 indicates: single and men, while 1 indicates: married, women respectively.

Additionally, fsQCA enables prioritizing more than one outcome (Olya & Al-ansi, 2018). This allows us to target the desired visitors to protected areas with a combination of all intentions to achieve financial goals and promote pro-environmental behaviours.

The identified intersections of the antecedents of MTEs and demographics on the combination of all intentions (i.e. Figure 1: Arrow A. Outcome = f(hed, kno, inv, loc, mng, nov, ref, gen, mar, age, inc, edu)) are presented in Table 4.

One causal recipe (A: M1) led to a high outcome (pebi ∩ revi ∩ reci) score (solution coverage = 0.23; solution consistency = 0.82). A: M1 (hed * mng * loc * ∼nov * ∼inv * kno * ref * age * ∼inc * ∼edu * gen * mar) shows that older, married, poorly educated women with lower incomes and a high level of hedonism, meaningfulness, respect for local culture, and knowledge who demonstrated refreshment but not involvement or novelty displayed a high score in the combination of all intentions. This result is also confirmed by previous studies (Olya & Akhshik, 2019), where older married women showed stronger intentions to support sustainable development and exhibited high intentions of pro-environmental behaviour.

Table 4 ∼A presents one recipe for a low outcome score: ∼A: M1 (hed * mng * loc * nov * inv * kno * ref * age * ∼inc * edu * ∼gen * mar). This suggests that older, educated, lower-income, married men who demonstrated high scores in hedonism, meaningfulness, respect for local culture, involvement, novelty, refreshment, and knowledge of the environment and surrounding areas felt, upon reflection, that they had behaved poorly while participating in nature-based tourism. These results contradict mainstream research, which suggests that visitors with more knowledge or education will be more concerned about the environmental issues of locations they visit (Kim & Stepchenkova, 2020).

However, this non-compliance is fully explained by the contrarian cases and non-linearity in this study. Thus, unlike conventional methods that use one model to predict outcomes, this state-of-the-art method uses causal recipes to simulate visitors’ intentions. In symmetric models, a low level of intention is presented in opposition to a high level, whereas fsQCA proposes varying solutions for the negation of outcomes (Table 4; ∼A: M1 and M2).

More results (B: M1, M2; ∼B: M1, M6) from the configural model highlight two causal recipes for the combination of place attachment and demographics (solution coverage: 0.35; solution consistency: 0.72). B: M1 (aff * ∼dep * ide * age * ∼inc * ∼edu * ∼gen) implies that uneducated older men with lower incomes who scored high in place identity and place affect but lacked place dependency achieved high scores in predicting the outcome (the combination of intentions). This outcome is a counterintuitive result, which may have been misinterpreted in studies that employed regression analysis or structural equation modelling. Because of this, several studies have reported that place dependency is correlated with intentions (Scannell & Gifford, 2010).

This does not, however, necessarily indicate that visitors who reported high dependency for a place intended to act pro-environmentally or revisit the area. As an example, a negation of the same outcome (∼B: M1 (aff * dep * ide * age * ∼inc)) indicates that older visitors with lower incomes and high scores in the place attachment dimensions (place affect, dependence. and identity) do not intend to act pro-environmentally, recommend the service, or revisit the area. Again, conventional methods would misinterpret relevant data hidden in the dataset (Armstrong, 2011)

The results of arrows C1, 2, 3, and 4 in Figure 1, which are the intersections of all possible antecedents for each of the outcomes (C1, C2, and C3) and the intersection of the outcomes (C4), are outlined in Table 5. Ten recipes were generated in fsQCA (C1: M1–M10) for the confluence of all antecedents that result in REVI. Another 12 recipes were extracted for the negation of outcomes using the same combination (C1: M1–M12). An example recipe is C1: M1 (hed * mng * loc * nov * ∼inv * kno * ref * aff * dep * ide * age * inc * edu * gen * ∼mar), which explains that older, single, educated women with higher incomes who scored high in hedonism, meaningfulness, local culture, knowledge, refreshment, and place affect, dependence, and identity, but who lacked involvement, are expected to revisit the area.

Moreover, Table 5 suggests recipes for attaining high and low PEBI scores among visitors (C2, ∼C2). It also provides recipes for achieving high and low scores in visitors’ intentions to recommend (C3, ∼C3), the details of which are also presented.

The extension of these recipes illustrates the inherent complexity of the studied intentions and the complex relationships of the antecedents. In considering all potential configurations (i.e. demographics, MTEs, and place attachment) to predict the intersections of intentions (REVI, RECI, and PEBI), three causal recipes explain the conditions (solution consistency: 0.90; solution coverage: 0.22) under which marine turtle visitors have the desired combination of intentions (Table 5; C4: M1, M2, and M3). For example, C4: M1 (hed * mng * ∼loc * ∼nov * ∼inv * kno * ref * aff * dep * ide * age * ∼inc * edu * gen * ∼mar) reveals that single, older, more-educated women with low incomes who scored high in hedonism, meaningfulness, knowledge, refreshment, and place affect, dependence, and identity while scoring low on local culture, novelty, and involvement had strong intentions to behave pro-environmentally, recommend the service to others, and revisit the area.

Based on the results shown in Table 5 for the negation of the combination of intentions and antecedents, 24 causal recipes explain the low scores associated with intentions (solution coverage: 0.38; solution consistency: 0.95), the details of which are presented in Table 5 (∼F: M1–M24). An in-depth assessment of the complexity of these relationships is presented in the following section.

Necessary condition analysis

Unlike sufficient conditions, individual necessary conditions are critically important components in recipes; without these conditions, the outcomes will not occur (Dul, 2016). Table 6 demonstrates the results of NCA to identify the conditions that are necessary to achieve the desired intentions. A cut-off level for consistency of .90 was used to select the necessary conditions that resulted in each intention, as well as combinations of intentions (Olya & Al-ansi, 2018; Tóth et al., 2015). Hedonism and place affect were found to be necessary for PEBI.

Table 6. Necessary condition analysis.

Conditions	Outcome
	PEBI	REVI	RECI	outcome
hed	0.9065	0.8754	0.8584	0.9451
∼hed	0.3725	0.3803	0.3497	0.5241
mng	0.7828	0.7350	0.7164	0.8641
∼mng	0.5773	0.5739	0.5492	0.7224
loc	0.8471	0.8306	0.8308	0.9046
∼loc	0.4284	0.4123	0.3722	0.5266
nov	0.7795	0.7366	0.7351	0.8685
∼nov	0.6211	0.6155	0.5596	0.7588
inv	0.7694	0.7947	0.7567	0.8687
∼inv	0.4760	0.4288	0.4357	0.5442
kno	0.8295	0.7996	0.8050	0.8376
∼kno	0.3793	0.4011	0.3725	0.5082
ref	0.8507	0.7684	0.7520	0.8781
∼ref	0.5085	0.5275	0.5085	0.6877
aff	0.9035	0.8710	0.8831	0.9439
∼aff	0.3678	0.3849	0.3266	0.5110
dep	0.8072	0.7548	0.7396	0.8471
∼dep	0.5549	0.5483	0.5221	0.7136
ide	0.8919	0.8591	0.8646	0.9276
∼ide	0.4010	0.3919	0.3579	0.5447

Note: PEBI =  Pro-environmental behaviour intention; REVI = Revisit intention; RECI = Intention to recommend; hed = Hedonism; kno = Knowledge; inv = Involvement; loc = Local culture, mng = Meaningfulness; nov = Novelty; ref = Refreshment; dep = Place dependence; ide = Place identity; aff = Place affect; The numbers indicate consistency; ∼ indicates negation condition; Bold numbers indicate necessary condition to achieve desired outcome (>0.9).

Moreover, hedonism, local culture, place affect, and place identity emerged as necessary conditions to simultaneously achieve all three desired intentions.

These results provide guidelines for destination management organizations (DMOs) to emphasize the necessary conditions to focus on financial and environmental goals while managing visitors to sensitive sites.

Predictive validity

To predict future outcomes and test the forecasting power of the extracted recipes, this study tested for predictive validity, as many scholars have suggested (Ferguson et al., 2017). Accordingly, in the first step, the sample data were divided into a subsample and a holdout sample. Next, a fuzzy XY plot was generated for one of the recipes (C2: M6 (pebi = hed * mng * loc * ∼nov * inv * kno * ref * aff * dep * ide * age * ∼inc * ∼edu * mar)), both in the subsample and the holdout sample (Table 5). A comparison of the consistency and coverage of the original sample, the subsample, and the holdout sample affirms the predictive validity of the extracted model.

Summary of findings

In this empirical study, a configural model to predict the favourable intentions of visitors to protected areas was developed and assessed. The findings indicate that a single antecedent is rarely sufficient to predict the desired outcome (Tenet 1). This means that recipes containing different antecedents are needed to fully explain high or low outcome scores (Tenet 2: The recipe principle). In our study, the combination of the studied variables in Table 5, C4: M1 (hed * mng * ∼loc * ∼nov * ∼inv * kno * ref * aff * dep * ide * age * ∼inc * edu * gen * ∼mar) resulted in high levels of intention scores among visitors. Although hedonism may be a necessary condition for this recipe to achieve strong intentions, hedonism alone cannot describe or predict this outcome.

Furthermore, a single recipe that sufficiently predicts the outcome is insufficient to describe the phenomenon by itself, as other recipes can generate the same results (Tenet 3: Equifinality principle). As presented in Tables 4 and 5, multiple recipes exist that will generate high or low scores for the desired outcomes.

The models’ causality requires the negated outcome to be unique (Tenet 4: The causal asymmetry principle). As presented in Tables 4 and 5, the causal recipes for low outcome scores (∼A, ∼B, and ∼C) are unique and not the mirror opposites of those for high outcome scores (A, B, and C).

Additionally, we argue that an antecedent (e.g. hedonism) can contribute positively or negatively to the prediction of an outcome, depending on the presence or absence of other ingredients in the recipe (Tenet 5). In this study, the presence or absence of hedonism contributed to high and low scores, respectively, in predicting the outcome (Table 5; C3: M10 and M11 or ∼C3: M2 and M3). As the sixth tenet, which suggests a high Y score, argues, a recipe is relevant for some, but not all, cases and the coverage of any single combination should be less than 1.00. The evidence is provided in an XY plot (Table 7). A given recipe for a high PEBI score is relevant for some, but not all, cases and the coverage are less than 1.00.

Table 7. Evidence of predictive validity on two subsamples.

Configural model	Raw coverage	Consistency
C2:M6. hed*mng*loc*∼nov*inv* kno* ref* aff*dep* ide* age*∼inc* ∼edu* mar	0.21	0.96

Note: The fuzzy XY plot unveils the asymmetric relationship of the causal model and provides the predictive validity.

The findings assert that visitors’ behavioural intentions must be modelled as complex phenomena using configural causal modelling, such as fsQCA, if solutions are to be considered outside of a single framework.

Conclusion and implications

The theory of planned behaviour (TPB) and theory of reasoned action (Fishbein & Ajzen, 1975) are typically used when scholars consider self-interest and rational choices as the sources of environmentally friendly intentions and behaviours. This is usually opposed to the beliefs of those who relate such intentions and actions to pro-social motives and utilize other theories, such as value–belief–norm (VBN) theory (Stern et al., 1999). However, strong controversies have arisen in the literature, questioning the adequacy of these theories in efforts to comprehensively explain individuals’ PEB or intentions (Bamberg & Möser; Onwezen et al., 2013).

Hence, different theories have recently been integrated to explain individuals’ intentions and behaviours (Chuang et al., 2018; Han, 2015; Rezapouraghdam et al., 2019). Using several theories, this study found that a complex interaction of variables can best predict nature-based tourists’ intentional behaviours, and using fsQCA as a set-theoretic approach facilitated the measurement of combinations of the studied outcomes. We successfully applied a model that predicted the combination of PEBIs in conjunction with REVI and RECI. This empirical research offers theoretical and practical contributions to the context of nature-based tourism in protected areas.

Travellers’ behavioural intentions are shaped by visitors’ past experiences (MTEs), the resultant emotions (attachment), and their demographic profiles. Lawler’s (2001) affective theory of social exchange asserts that the essence of the exchanges that visitors experience with their surroundings and other people during travel can result in positive emotions, such as attachment. This is also supported by social judgement theory and the concept of involvement (Havitz & Dimanche, 1997; Zaichkowsky, 1985) and can attach them to a destination (Xu & Zhang, 2016). This attachment is an innate human need that occurs when a person finds meaning in a place.

Other theories, such as social identity theory (Brown, 2000) and attachment theory, support the notion that an attached individual would endeavour to react positively (PEBIs, REVI, and RECI) in favour of the place to which they are attached. Tourism consumption system theory (Woodside & Dubelaar, 2002) and SOR theory (Mehrabian & Russell, 1974) can explain the whole picture as well by describing how MTE, place attachment, and peoples’ positive intentions are interrelated.

Although visitors’ pro-environmental intentions at the time of (re)visiting are vital to the ecological well-being of destinations, the existing literature has mainly focused on the symmetrical antecedents of visitors’ intentions to revisit and recommend (Gupta et al., 2018; Hasan et al., 2019; Jeong et al., 2020; Zhang et al., 2018). Our study contributed to the literature by providing a detailed explanation of how various combinations of tourists’ experiences, with dimensions of place attachment and visitors’ demographic details, can lead to positive intentions.

MTEs’ role in shaping travellers’ behaviours in nature-based environments has received little attention in the literature (Rezapouraghdam et al., 2021), and simultaneous observations of visitors’ PEBIs, REVI, and RECI are scarce. In addition, although the literature supports the role of place attachment in fostering individuals’ PEBIs, the evidence is inconsistent across individual studies (see Daryanto & Song, 2021). This has made it difficult to ascertain the overall impact of place attachment on PEBIs.

Our findings provide key insights for destination managers and marketers. First, the importance of the interaction between demographics and MTEs in provoking desirable visitor intentions as travel outcomes should be highlighted (see A: M1). Focusing on tours courting older women as participants is a good starting point for destination managers. Moreover, based on this high-scoring causal recipe (A: M1), all MTE components except for involvement and novelty had a considerable effect on the predicted outcome variables. This gives destination managers in sensitive areas a better grasp of how to make travellers’ experiences more memorable to achieve the desired tourism outcomes.

In other words, as participants become more aware of the impact of their travel, the meaningfulness of their visit to the destination increases. Therefore, destination managers should promote this dimension of travel by emphasizing how visitors’ presence can financially support the preservation of marine turtles, for example.

Destination service providers can enhance the knowledge dimension of MTEs by delivering information about marine turtles, thus increasing providers’ contributions to the ecology of the region and the turtles’ life cycle. They can do so through various strategies, such as media campaigns, posted signage, and employing knowledgeable tour guides.

These targeted campaigns can focus on the visitors who bring the most value to both the destination service provider and the protected area. Alternatively, these findings can be used to adjust the experience design to encourage pro-environmental behaviours based on other factors (e.g. demographics) to realize the greatest benefit from an MTE.

In sum, the proposal for various recipes to drive PEBIs, REVI, and RECI in this study indicates that any single ingredient (variable) is insufficient to predict the desired intention.

Various examples of management failures have been observed at tourist destinations and businesses due to an intense concentration on a single dimension rather than a holistic, comprehensive vision (Dos Santos et al., 2017; Sirakaya et al., 1996). The complex nature of tourism products and tourist behaviour demands a recipe that contains the necessary components unveiled by this study. However, a single necessary condition can bring about the desired outcome only when combined with additional necessary conditions. This study revealed that hedonism and place affect are necessary single antecedents and further signalled the importance of engaging visitors emotionally while gamifying the experience to generate the most benefit from educational tours.

Place affect can enhance the visitor experience (Chanchaichujit et al., 2020), lead to further support for tourism and generate well-being outcomes that align with sustainable development goals (Ramkissoon, 2020a). Therefore, managers at tourist destinations should advise their staff to engage with tourists and discuss the emotional moments they experienced during their visits to increase tourists’ sentiments and positive affect towards the destination. For example, visitors may describe their feelings as they witnessed the turtles laying eggs or the eggs hatching.

Those who manage these protected zones should also consider advertising and promoting the area as a green destination based on its contribution to nature preservation and the local community’s well-being. This will increase visitors’ place identity and can increase their PEBI.

Limitations and suggestions for future studies

This study’s participants were limited to visitors to Alagadi Beach in Cyprus. Future studies could include participants at similar destinations. The study also did not include other potential ingredients in the recipes, such as visitors’ nationalities. The inclusion of these other factors, such as background or culture, could generate more exact estimations from the prediction models.

Moreover, testing the influences of other variables, such as destination image, destination familiarity, and destination social responsibility motives may result in new insights that could lead to different paths engendering desirable outcomes. Comparing symmetrical (using first- or second-order variables) and asymmetrical models to gain further insights into how asymmetrical approaches compare to conventional models would also be beneficial. Finally, the results of this study were limited to visitors’ intentions; however, intentions may not always convert into behaviours. Therefore, future research should also measure visitors’ actual behaviours.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Appendix

Table A1. Results of EFA, and descriptive statistics of scale items (n = 406).

Factor		MEAN	SD	Eigenvalues (% of Variance)	Skewness		Kurtosis
					Statistic	Std. Error	Statistic	Std. Error
Place Affect (Kulczycki & Halpenny, 2014)				1.403 (3.509)
Aff1	.826	3.9754	.93723		−.946	.121	.641	.242
Aff2	.859	4.0837	.82724		−.946	.121	.984	.242
Aff3	.753	3.7414	.91604		−.492	.121	−.334	.242
Place Dependence (Kulczycki & Halpenny, 2014)				1.332 (3.330)
Dep1	.828	3.4754	.97278		−.569	.121	.034	.242
Dep2	.808	3.3966	.90151		−.216	.121	−.178	.242
Dep3	.725	3.2537	1.10763		−.295	.121	−.522	.242
Hedonism (Kim & Ritchie, 2014)				4.283 (10.708)
Hed1	.760	4.1010	.81780		−1.032	.121	1.562	.242
Hed2	.804	3.7980	.95767		−.553	.121	−.143	.242
Hed3	.810	3.8916	.96766		−.718	.121	.040	.242
Hed4	.808	3.8547	.99185		−.895	.121	.540	.242
Place Identity(Kulczycki & Halpenny, 2014)				3.337 (8.441)
Ide1	.849	3.9310	.96362		−.810	.121	.217	.242
Ide2	.865	3.7808	.94997		−.661	.121	.025	.242
Ide3	.868	3.8054	.95533		−.746	.121	.264	.242
Ide4	.797	3.8744	.96940		−.792	.121	.283	.242
Involvement (Kim & Ritchie, 2014)				2.226 (5.565)
Inv1	.850	3.5591	1.42818		−.618	.121	−1.011	.242
Inv2	.928	3.5567	1.40211		−.630	.121	−.960	.242
Inv3	.906	3.6010	1.33089		−.635	.121	−.838	.242
Knowledge (Kim & Ritchie, 2014)				2.964 (7.411)
Kno1	.919	3.7512	1.40007		−.950	.121	−.436	.242
Kno2	.932	3.8227	1.29313		−1.023	.121	−.085	.242
Kno3	.918	3.8941	1.30189		−1.165	.121	.214	.242
Local culture (Kim & Ritchie, 2014)				2.015 (5.038)
Loc1	.865	3.9532	1.01848		−.977	.121	.559	.242
Loc2	.850	3.5665	1.22091		−.482	.121	−.727	.242
Loc3	.918	3.8916	1.08552		−.866	.121	.058	.242
Meaningfulness (Kim & Ritchie, 2014)				1.172 (2.929)
Mng1	.721	3.5616	1.07262		−.408	.121	−.622	.242
Mng2	.769	3.4064	1.05870		−.342	.121	−.600	.242
Mng3	.758	2.9286	1.02189		.074	.121	−.428	.242
Novelty				2.832 (7.081)
Nov1	.684	3.1921	.84183		−.001	.121	.262	.242
Nov2	.773	3.2217	.79822		.166	.121	.106	.242
Nov3	.867	3.2906	.88848		.075	.121	−.045	.242
Nov4	.826	3.2734	.93842		−.066	.121	−.358	.242
Pro−environmental behavior intention (Kiatkawsin & Han, 2017)				5.391 (13.478)
Pebi1	.729	3.4015	1.12608		−.316	.121	−.772	.242
Pebi2	.788	2.8793	1.08763		.045	.121	−.726	.242
Pebi3	.882	3.4384	1.05405		−.492	.121	−.405	.242
Pebi4	.740	2.6626	1.07356		.223	.121	−.708	.242
Pebi5	.697	3.6724	.87691		−.525	.121	.228	.242
Pebi6	.872	3.4236	1.01972		−.459	.121	−.325	.242
Refreshment (Kim & Ritchie, 2014)				2.439 (6.098)
Ref1	.759	3.3596	.92362		−1.378	.121	.892	.242
Ref2	.815	3.4163	.92249		−.900	.121	.142	.242
Ref3	.767	3.5443	1.04551		−.724	.121	.045	.242
Ref4	.771	3.3227	.92260		−.380	.121	−.080	.242
Revisit Intention (Toyama & Yamada, 2012)	–	4.13	1.861		−.078	.121	−1.037	.242
Recommendation intention (Toyama & Yamada, 2012)	–	4.52	1.855		−.440	.121	−.793	.242

Note: λ is factor loading coefficient; SD is standard deviation. The sources of the scale items are presented in parenthesis. PEBI =  Pro-environmental behaviour intention; REVI = Revisit intention; RECI = Intention to recommend; hed = Hedonism; kno = Knowledge; inv = Involvement; loc = Local culture, mng = Meaningfulness; nov = Novelty; ref = Refreshment; dep = Place dependence; ide = Place identity; aff = Place affect; All items gauged by 5-point Likert scale ranging from strongly disagree (1) to strongly agree (5) except REVI and RECI that are ranging from Strongly disagree (1) to Strongly agree (7). Kaiser-Meyer-Olkin (KMO): .752; Bartlett's Test of Sphericity: Approx. Chi-Square: 9638.24, df: 780; Sig.: 0.000; Eigen-values > 1