Abstract
This study proposes and tests a structural model regarding the effects of cuisine experience and psychological well-being on hot springs tourists’ revisit intentions and examines the moderating effect that self-health perception has on the relationships between these constructs. A questionnaire was administered to 524 Taiwanese tourists at the Bei-Tou hot springs area. The results indicated that cuisine experience and psychological well-being influence hot springs tourists’ revisit intentions and only cuisine experience affects psychological well-being; however, the significance of these factors varied based on the self-health perception levels (high or low) of tourists in the sample. Implications for theory and practice are discussed.
The health and wellness benefits of travel and tourism experiences are well documented (Douglas, 2001; Goodrich, 1994; Gilbert Abdullah, 2004). In Asian countries, and particularly in Taiwan and Japan, hot springs areas have become popular tourist destinations because of the diversity and concentration of their hot springs (Taiwan Tourism Bureau, 2011a). The region’s reputation as a hot spring tourist destination has spread in recent years, stimulating the growth of the hot springs tourism market (C.-F. Lee, Ou, Huang, 2009; T.-H. Lee, 2010; McMorran, 2008). An understanding of tourists’ revisit intentions is crucial for the development of well-targeted and effective hot springs tourism marketing strategies (Lin Morais, 2009). This is because repeat tourists are more profitable than one-time visitors and more likely to recommend the destination positively through word-of-mouth (Baker Crompton, 2000; Um, Chon, Ro, 2006). Identifying the factors that determine whether tourists will return to a destination is a critical way for tourism authorities to assess their economic sustainability (Petrick, Morais, Norman, 2001).
Most studies focused on predicting tourists’ revisit intentions have applied the general model of tourists’ destination choices (Lin Morais, 2009), the concept of loyalty (Petrick et al., 2001), a combination of the models for both destination choices and loyalty (Um et al., 2006), or the theory of planned behavior (Ryu Han, 2010). Um et al. (2006) found that revisit intentions could be better determined through the relationships among their possible determinants than by the desire to replicate the previous experience. Another study modeled a decision-making process to revisit after the process associated with making general destination choices (Lin Morais, 2009). In their model of tourist destination choice, Woodside and Lysonski (1989) define a behavioral intention as “the traveller’s perceived likelihood of visiting a specific destination within a specific time period” (p. 8). The tenet of this model is that an individual’s ultimate decision (i.e., destination choice) is influenced by the relationships between internal and external variables, including marketing variables, destination awareness, affective associations, traveler destination preferences, and intention to visit. Ajzen (1991) presented the theory of planned behavior, which explains that cognitive and affective components of a person’s attitude predict intention; subsequently, intention influences behavior. Thus, the prediction of tourists’ intentions to (re)visit a destination, derived from attitudinal and behavioral studies of tourists, is based on tourists’ cognitive and affective constructs concerning a destination.
Indeed, recent developments in revisit intention research have found that assessing appropriate models of particular tourist behavior can help us understand why tourists revisit a specific destination (Petrick et al., 2001; Sparks, 2007; Um et al., 2006). Similarly, Lin and Morais (2009) suggested that researchers examining behavioral intentions associated with destination choices specify the types of their tourist samples and use appropriate predictor variables for understanding their behaviors. Studies on tourist repeat behavior have identified several important antecedents of revisit intention, including satisfaction (e.g., Baker Crompton, 2000), perceived value (e.g., Petrick et al., 2001), destination image (e.g., Bigné, Sánchez, Sánchez, 2001), motivation (e.g., Li, Cai, Lehto, Huang, 2010), and service quality (e.g., Baker Crompton, 2000). However, there is still insufficient understanding of how tourists’ revisit intentions develop. Our first consideration in the investigation of tourist behavior was the observation that very little information is available in the existing literature regarding behavioral models for hot springs tourists, who make up an increasingly important market segment. There is, however, evidence that behavioral intention might be an initiator of revisit behavior and could be usefully extended to apply to other destinations and contexts, including tourists, to further advance the knowledge (Petrick et al., 2001; Um et al., 2006).
Second, some independent variables that are related to perception of self-health, cuisine experiences at the destination, and the sense of psychological well-being that tourism activity triggers, may also contribute to the likelihood that hot springs tourists will revisit. Hallab (2006) and Kaspar (1990), for example, argued that an individual’s healthy lifestyle is displayed in his or her travel behavior. Hallab (2006) further proposed that components associated with vacationers’ healthy lifestyles include overall health, eating habits, stress and coping skills, and fitness level. Lin and Liu (2007) found that hot springs tourists’ healthy lifestyles comprise three distinct dimensions: medical care, stress and coping strategies, and eating habits. Thus, cuisine experience and psychological well-being are anchored in travel behavior with respect to hot springs tourists’ healthy living and/or self-health perceptions. Recent studies on the attributes of a hot springs–based destination identified food/local cuisine and experience of hot springs resources as two important appealing qualities that motivate tourists to (re)visit a hot springs destination (C.-F. Lee et al., 2009; T.-H. Lee, 2010; Lin Morais, 2009). The tourist cuisine experience is related to a tourist’s feelings about local food arts and cultures (Cohen Avieli, 2004). Although previous studies have suggested that local cuisine is an important factor in tourism destination marketing (Okumus, Okumus, McKercher, 2007; Quan Wang, 2004) and may be a main reason for travelling (Ryu Han, 2010), few studies have explored empirically the relationship between cuisine experience and revisit intention. Moreover, literature has shown that tourists enjoying hot springs tend to express emotional happiness (Kaspar, 1990; Mueller Kaufmann, 2001) and exhibit a distinct sense of psychological well-being (Goodrich, 1994; Hobson Dietrich, 1994). Lin and Liu (2007), for example, mentioned that various hot springs resources found mostly in natural environments allow tourists to simultaneously promote self-health and enjoy nature. Researchers found that springs/spa resources (e.g., sulfur, carbonide, and sodium bicarbonate springs; Douglas, 2001; Kaspar, 1990; Lin Liu, 2007) and destination resources (e.g., atmosphere and environment; Gilbert Abdullah, 2004) affect the sense of psychological well-being. Thus, both the use and enjoyment of hot springs resources and the destination itself are associated with psychological well-being that is related to the general tourist experience. Mannell and Kleiber (1997) also found that, having completed a specific tourism experience (e.g., a hot springs vacation), an individual’s psychological well-being has an impact on his or her future behavior. Nevertheless, the relationship between psychological well-being and revisit intention has yet to be sufficiently empirically confirmed.
Although previous leisure literature has regarded the effect of food experience as a stimulus that elicits psychological well-being that may affect behavioral intention (Cooper, Fletcher, Noble, Westlake, 1995; Godbey, 2003), this study expects that individual perception of self-health moderates the effect of tourists’ cuisine experience and psychological well-being on revisit intention; this is because self-health perception is defined as the subjective perception of overall health (Davies Ware, 1981; Hobson Dietrich, 1994; Jefferson, 1995). For tourists who have a high perceived level of self-health, the impact of psychological well-being on revisit intention may be stronger than cuisine experience, because tourists may actively pursuit emotional happiness as part of their tourism experience (Orsega-Smith, Mowen, Payne, Godbey, 2004) and may appreciate healthy, natural, cooked, or more traditional (responsible in a broader sense) food in their cuisine experiences to avoid obesity, which carries the risk of stroke and heart disease (Nesse Williams, 1994). In contrast, for those with low levels of perceived self-health, cuisine experience may influence revisit intention more than psychological well-being, because cuisine experiences may provide more pleasure for them than other activities (Cohen Avieli, 2004). Therefore, self-health perception may play a moderating role in the relationship among tourist’s cuisine experience, psychological well-being, and revisit intention.
Previous research has helped explain the determinants of revisit intention; however, few studies have examined the relationships between revisit intention, cuisine experience, psychological well-being, and self-health perception. These relationships must be tested in a specific destination setting to broaden understanding of revisit intention. Therefore, we subscribe to the emerging belief that tourists’ revisit intentions to a hot spring destination are influenced by their affections (i.e., the direct effects of cuisine experience and psychological well-being) and perceptions/cognitions (i.e., the moderating effect of self-health perception) regarding the destination; this supposition is derived from the theory of planned behavior and the general model of tourists’ destination choices (see Figure 1). Hence, the contribution of this study is twofold: first, to extend the revisit intention literature by providing empirical evidence of these relationships in a hot springs destination and, second, to explicitly consider the moderating effect of self-health perception on determining repeat visitation behavior. The findings should help hot springs destination marketing organizations (DMOs) better understand tourists’ revisit intentions so that they can promote hot springs tourism.

Proposed Model for the Study
Literature Review
Revisit Intention
Over the past decade, behavioral intention (Baker Crompton, 2000; Sparks, 2007) and revisit intention (Li et al., 2010; Um et al., 2006) have been widely studied in the tourism literature. Generally, behavioral intention has been characterized as the possibility for individuals to engage in specific behavior (Fishbein Ajzen, 1975). Following Fishbein and Ajzen (1975), Baker and Crompton (2000) defined revisit intention as the possibility for the tourist to be willing to repeat an activity or to revisit a facility/destination. They all argued that behavioral intention can be a good proxy of future behavior. Thus, if behavioral intention can be adequately measured, tourist future behavior can be effectively predicted.
Several researchers have identified the multidimensional/multi-item nature of revisit intention that is dependent on the specific study context (Bigné et al., 2001; Petrick, 2004; Petrick et al., 2001); this has meant that there has yet to be a definitive answer on how best to measure revisit intention. The work done by Lin and Morais (2009), for example, suggested that intention to revisit, intention to recommend, and resistance to change were three dimensions of revisit intention. Moreover, Bigné et al. (2001) found intention to revisit and intention to recommend were two dimensions through which one could measure revisit intention. These two dimensions of revisit intention were adopted in subsequent work by Petrick (2004).
Direct Effect on Psychological Well-Being on Revisit Intention
Psychological well-being has been an important subject of empirical inquiry for researchers in psychology, leisure, and tourism (Gilbert Abdullah, 2004). In general, psychological well-being may be defined as a feeling of happiness and satisfaction (Mannell Klieber, 1997). Gilbert and Abdullah (2004) proposed that the dimensions of psychological well-being consisted of happiness, quality of life, and life satisfaction. The happiness dimension represented a brief mental state of gaiety and elation, reflecting the temporary emotional condition of the person toward the current state of affairs. In contrast, quality of life and life satisfaction refer to a person’s judgment about how enjoyable they believe their life is overall. However, few empirical studies have confirmed that these three dimensions constitute psychological well-being. In the context of hot springs tourism, C.-F. Lee et al. (2009) suggested that hot springs were regarded as an attractive destination to domestic Taiwanese tourists because they were relaxing, enjoyable, and pleasurable. Thus, hot springs tourism may well boost psychological well-being in the long run.
The relationship between psychological well-being and activity participation has been studied more extensively in leisure research than it has in tourism research (Mannell Klieber, 1997). Leisure studies have tended to find that psychological well-being has a positive effect on leisure participation. Mannell (1993), for example, found that, among older people, those with a higher level of life satisfaction are more willing to freely choose and continuously engage in recreational activities than those with lower level of life satisfaction. In addition, Gilbert and Abdullah (2004) concluded that taking holidays can enhance happiness. As researchers have customarily used behavioral intention as a proxy indicator for actual behavior, the following hypothesis was proposed:
Hypothesis 1: Psychological well-being positively affects revisit intention.
Direct Effects of Cuisine Experience on Revisit Intention and Psychological Well-Being
In addition to research on the effect of psychological well-being on revisit intention, there has been extensive work exploring the influence of cuisine experience on tourist behaviors. Cuisine experience is defined as an individual’s feelings toward local cuisine and dining habits acquired through food consumption (Quan Wang, 2004). The study of cuisine experience has focused on the general setting in which individuals taste, experience, experiment, research, discover, and understand food or wine (Kivela Crotts, 2006). For example, following Cohen’s (1979) classification of tourist experiences, Hjalager (2003), discussed four types of tourist cuisine experiences: existential—experiencing local cuisines in order to understand local culture; experimental—trying out new restaurants and new types of foods, in order to gather up to date dining information; recreational—searching for familiar kinds of foods at the tourist destination; and diversionary—attempting to escape from daily routines though a culinary cuisine experience. In another study, Quan and Wang (2004) derived a conceptual model in which food experience was composed of peak experience—where food was the major draw for tourists, and support experience—where food was merely part of the package necessary to satisfy tourists’ basic needs. Thus, the existing literature suggests that cuisine is a major factor influencing tourist behavioral intention to visit a destination (Ryu Jang, 2006), and the desired experiences of excitement, education, escape, pleasure, relaxation, and lifestyle can be achieved in gastronomic tourism (Frochot, 2003). However, neither the conceptual study nor the model was empirically verified.
Hall and Sharples (2003) proposed that cuisine is an essential element of tourists’ experiences. In the context of spa tourism, Cooper et al. (1995) found that food service is one of the major attractions for tourists. Empirical research has shown that, for some tourists, the experience of food may very well be the only purpose of their travel (Au Law, 2002). Ryu and Han (2010) also found that the quality of local cuisine/gastronomy can positively predict tourist revisit intention. In addition, Sparks, Bowen, and Klag (2003) argued that the quality of tourists’ culinary experience influenced the likelihood that they would recommend the location to relatives and friends. Based on these empirical research findings, the following hypothesis was proposed:
Hypothesis 2: Cuisine experience positively affects revisit intention.
In addition, cuisine consumption is an integral part of tourism experiences (Okumus et al., 2007; Quan Wang, 2004). Tourism experiences have always been described in terms of pleasure, enjoyment, and relaxation (Mannell Kleiber, 1997), but a lack of tourism experiences has been found to decrease the quality of life and life satisfaction (Gilbert Abdullah, 2004). Therefore, it may be inferred that tourism cuisine experiences are one of the most important resources of human well-being. Moreover, Godbey (2003) claimed that food experience influences individuals’ sense of well-being. This leads to the following hypothesis:
Hypothesis 3: Cuisine experience positively affects psychological well-being.
Moderating Effect of Self-Health Perception
Self-health perception may be a moderating construct of the relationship among cuisine experience, psychological well-being, and revisit intention. Health is defined as “a state of complete physical, mental, and social well-being and not merely the absence of disease or infirmity” (World Health Organization [WHO], 1947). Following the WHO’s definition, Davies and Ware (1981) define self-health perception as an individual’s subjective beliefs and ideas about their overall health (including past, present, and future health and resistance to disease). In addition, several researchers have indicated that both physical and mental aspects should be taken into account when measuring self-health perception (Davies Ware, 1981; Hoyman, 1975). In the context of hot springs tourism, C.-F. Lee et al. (2009) suggested that hot springs proprietors should extend the appeal of their tourism experience by capitalizing on the associated health and medical associations.
In addition to the effect that cuisine experience has on leisure behaviors (Godbey, 2003), and its role as a stimulus for psychological well-being that affects tourist behaviors (Cooper et al., 1995; Goodrich, 1994), several authors have identified self-health as a motivation for tourism experiences (Hobson Dietrich, 1994; Jefferson, 1995). Hjalager (2003) and Nesse and Williams (1994), for example, argue that, among individuals who follow a healthy lifestyle, the relative impact of psychological well-being on behavioral intention may be stronger than the impact of food experience. In contrast, Kivela and Crotts (2006) found that the possibility of savoring a destination’s unique gastronomy played a more important role than other attractions on revisit likelihood. Therefore, self-health perception may play a moderating role in the relationship between tourists’ cuisine experience, psychological well-being, and revisit intention. Following these studies, the following hypothesis was derived:
Hypothesis 4: The relationship between psychological well-being, cuisine experience, and revisit intention is moderated by self-health perception.
Method
The data set comprised Taiwanese visitors to the Bei-Tou hot springs area aged 18 years and older. The study site was selected because it is one of the five oldest and most famous hot springs areas in Taiwan (Taiwan Tourism Bureau, 2011a). Recently, Taiwan’s Tourism Bureau and hospitality industry designated 2006 as the “year of hot springs tourism” in an effort to capitalize from the worldwide growth of health tourism (Taiwan Tourism Bureau, 2011c). The Bei-Tou hot springs area is being actively promoted and developed by the Taiwanese government as a domestic hot springs tourist destination (Taiwan Tourism Bureau, 2011b) through the development of extensive infrastructure and private investment. No official data on visitor numbers and characteristics, hot springs packages, or cuisine availability and variety were available for Bei-Tou at the time of this study. Observations during the fieldwork and comments by local tourism business people suggested, however, that most tourists came during weekends and extended holidays and stayed for 1 or 2 days. There are approximately 40 hot springs hotels/resorts in the area, and of these, between two and five have a high reputation for cuisine and hot springs services from tourists. There are also many ethnic restaurants in the area. Of these, the most popular is Japanese, although there are also Chinese-style and Western-style restaurants. Most Japanese- and Chinese-style restaurants featured local ingredients, and some of them provided special plans, which put more emphasis on an organic and healthy diet. As the resort has become more popular, new, fashionable, and innovate menus have been designed to attract tourists; as a result, many of these restaurants have been recommended by travel agency, brochures, and travel guides.
Data collection was conducted on 10 weekdays and 10 weekend days selected randomly between November 2009 and February 2010. This 4-month period was chosen because this is the most popular time for domestic tourists to visit the area in search of a warm and relaxing break from Taiwan’s cold winter. To gather the data, trained interviewers wearing ID badges that displayed their association with a local university, mostly as graduate students studying tourism and travel, approached tourists and asked them to complete the on-site interview. Following a method successfully used in previous studies in Bei-Tou (Fang, 2002; Yang, 2004), the data were collected outside selected hot springs establishments (e.g., hotels and restaurants) and in a public parking lot. If an individual refused to participate, the interviewer moved on to another person. The sample in this study was restricted to respondents who were asked to evaluate their health perceptions before and after tasting local cuisine and soaking in the hot springs and then complete a questionnaire. If respondents did not visit the hot springs and sample local food and beverages, their surveys were invalidated. The reason for this exclusion was that the model to be tested in this article incorporated two measures of cuisine experience and psychological well-being so that we could assess the extent to which tourists’ self-health perceptions were influenced by the experience of soaking in the destination’s hot springs. Interviewers were uniformly trained in the interview protocol and process used to collect the data.
Because the conceptual model included four constructs (i.e., self-health perception, cuisine experience, psychological well-being, and revisit intention) beyond demographics, the study instrument was a self-administered questionnaire with five sections. Before the formal survey was conducted, a pretest was administered to a convenience sample of 280 on-site hot springs tourists to determine whether or not our devised measures of the constructs were meaningful and internally consistent. The first section included a 10-item self-health perception scale. Subjects were asked to rate their level of health for themselves before visiting Bei-Tou hot springs area. Self-health perception (pretest αs = .87) comprised items assessing perceived physical health (four items with a 5-point Likert-type scale: 1 = strongly disagree, 5 = strongly agree; pretest αs = .88) and mental health (six items with a 6-point scale: 1 = never, 6 = always; pretest αs = .86). These items were developed based on work done by McDowell and Newel (1996). This assessment is considered appropriate for hot springs tourists because the 10-item validated scale has been used in a number of population samples; in the field of leisure studies, Orsega-Smith et al. (2004) successfully used these items to measure park-based users’ self-health perception. The second section included a 12-item cuisine experience scale (pretest αs = .87). Cuisine experience was assessed by having each participant indicate the extent to which they agreed or disagreed with scale items after tasting food and beverage in the hot springs area. Individuals were asked to respond to each item using a 5-point Likert-type scale (1 = strongly disagree, 5 = strongly agree). These items were developed based on the works of Hjalager (2003); Lin, Liu, Wang, and Wang (2011) effectively used this scale in recent studies to examine hot springs tourists’ cuisine experiences. Before examining the dimensional structure of the multi-item latent construct, an exploratory factor analysis was conducted with a pretest sample to decrease the error variance of indicator correlations before a confirmatory factor analysis was conducted to obtain a measurement model (Bollen, 1989). Using a varimax rotation, factors with eigenvalues greater than 1.0 were extracted. Items were kept in their respective factors when they had factor loadings greater than .40. The mean of each factor was calculated, and then each factor was treated as a composite indicator to measure a latent construct. This procedure was used to simplify the structural model while maintaining its theoretical integrity (West, Finch, Curran, 1995). The internal consistency of indicators measuring destination personality and cuisine experience was evaluated using Cronbach’s alpha coefficients. Generally, Cronbach’s alpha coefficients of .70 and higher are considered acceptable values (Nunnally, 1978). If, however, a factor has a reduced number of items (e.g., six or fewer), .60 and higher may be acceptable (Cortina, 1993). All factors had Cronbach’s α values of more than .60, suggesting acceptable levels of internal consistency for every scale. The final results of the exploratory factor analysis indicated that two cuisine experience factors—health exploration and life diversion—accounted for 61% of the variance.
The third section of the questionnaire included 15 items, focusing on individuals’ psychological well-being (pretest αs = .87). To assess individual psychological well-being, each participant indicated the extent to which they agreed or disagreed with certain items, after soaking in the hot springs. They comprised items assessing happiness (eight items assessed on a 9-point scale: 1 = never, 9 = always; pretest αs = .90), quality of life (one item assessed on a 9-point scale: 1 = very unhappy, 9 = very happy), and life satisfaction (six items assessed with a 5-point scale: 1 = strongly disagree, 5 = strongly agree; pretest αs = .85). These items were developed based on work done by Gilbert and Abdullah (2004) and Hill and Argyle (2002); they were successfully tested in the context of hot springs tourism by Lin and Liu (2007). In the fourth section, a 4-item, 5-point Likert-type agreement scale was used to measure tourists’ revisit intentions (pretest αs = .82). These items were developed based on the works of Bigné et al. (2001) and Petrick (2004). Similar to other hot springs tourism studies (Chao, Wang, Chang, Liu, 2011), the scale included two dimensions—intention to revisit (two items; pretest αs = .75) and intention to recommend (two items; pretest αs = .89). The fifth and final section assessed sociodemographic information (i.e., gender, age, marital status, education level, and monthly household income). The instrument was written in traditional Mandarin Chinese, the official language of Taiwan. The scales central to this study were adapted from literature published in English; therefore, two colleagues fluent in Mandarin and English and familiar with the relevant literature were asked to comment on the face validity of the measurements. The items reported in this article reflect the back-translation of the survey items from Mandarin to English. The original wording of the items in Mandarin is available from the lead author.
Results
A total of 558 subjects were asked to participate in this study. Of these, 539 agreed to complete the survey on-site—a participation rate of 97%. However, 15 of the 539 surveys conducted were unusable because participants did not answer the questions that were critical for this study (i.e., cuisine tasting, hot springs soaking, and self-health perception). A total of 524 usable surveys were kept for further analysis (94%). In the usable sample, approximately 57% were female. Their ages varied from 22 to 81 years and 65.3% were between 31 and 65 years old. The majority (56.2%) were married, educated (65.4% had at least a college degree), and had a relatively high income (67% reported monthly household incomes between NT$30,001 and NT$100,000 [US$1.00 = NT$30.1 as of January 2010]. The sociodemographic profile of respondents was similar to the profile of tourists visiting the Bei-Tou hot springs area surveyed by Fang (2002) and Yang (2004).
A confirmatory factor analysis was used to test the measurement structure of each latent construct (i.e., cuisine experience, psychological well-being, self-health perception, and revisit intention; see Table 1). The assessment of the adequacy of measurement model considered the goodness-of-fit indices, the reliability of indicator and construct, and the validity of construct (i.e., convergent validity and discriminant validity). First, the goodness-of-fit indices indicated that of the latent constructs—self-health perception (2 dimensions, 10 indicators), cuisine experience (2 dimensions, 12 indicators), psychological well-being (3 dimensions, 15 indicators), and revisit intention (2 dimensions, 4 indicators)—the measurement model of all of these latent constructs had an accepted fit with the data. Second, with respect to indicator and latent construct reliability, the t values of all indicators (measured items) exceeded the critical level of 1.96 (p < .05; Hatcher, 1994), and the reliability of all latent constructs exceeded the recommended level of .60 (Cortina, 1993). Finally, convergent validity is assured if each indicator’s loading on its latent construct was greater than twice its standard error (Anderson Gerbing, 1988). In this study, the indicator’s loadings had significant t values, ranging from 8.97 to 21.38, demonstrating convergent validity. In addition, discriminant validity is justified if the variance extracted estimated for any pair of factors is greater that the square of the correlation between these two factors. The results confirmed this criterion, providing evidence discriminant validity (Hatcher, 1994; Jöreskog Sörbom, 1993).
Means, Factor Loadings, and Reliabilities for Cuisine Experience, Psychological Well-Being, Self-Health Perception, and Revisit Intention
Scale: 1 = strongly disagree, 5 = strongly agree.
These items were reverse-coded but are presented the same way they were presented in the questionnaire.
Scale: 1 = never, 6 = always.
Scale: 1 = never, 9 = always.
Scale: 1 = very unhappy, 9 = very happy.
p < .05. **p < .01. ***p < .001.
Measurement Model Testing
Following the two-step procedure first proposed by Anderson and Gerbing (1988), we used LISREL 8.52 software to examine the hypothesized relationships among antecedents of revisit intention. The first step required the confirmed measurement models for each latent construct (cuisine experience, psychological well-being, and revisit intention). Two indicators (health exploration and life diversion) for cuisine experience, three indicators (happiness, quality of life, and satisfaction of life) for psychological well-being, and two indicators (intention to revisit and intention to recommend) for revisit intention were included in the measurement model. One composite indicator of psychological well-being (quality of life) was dropped because the standardized factor loading was less than .4 (Jöreskog Sörbom, 1993). Thus, two indicators for cuisine experience, two indicators for psychological well-being, and two indicators for revisit intention were included in the modified measurement model.
The assessment of the adequacy of the modified measurement model revealed that it reasonably fit the data (χ2 = 24.04, df = 6, p < .001, nonnormed fit index [NNFI] = .97, comparative fit index [CFI] = .99, root mean square error of approximation [RMSEA] = .07, goodness of fit index [GFI] = .97, standardized root mean square residual [SRMR] = .04). It should be noted, however, that the significant chi-square might be due to the fact that large samples tend to inflate the sensitiveness of chi-square statistics in structural equation modeling (SEM) analysis (Hair, Black, Babin, Anderson, Tatham, 2006; Hatcher, 1994). With respect to indictor and latent construct reliability, the t value of each indicator exceeded the critical level of 1.96 (p < .05; Hatcher, 1994), and the reliability for latent constructs exceeded the recommended level of .60 (.66, .72, and .87, respectively; Cortina, 1993). Furthermore, each of the loadings of the indicators’ had a significant t value, ranging from 9.20 to 15.44 (greater than twice its error), providing evidence of convergent validity (Anderson Gerbing, 1988). A variance extracted test was also used to examine the discriminant validity of all four latent constructs. According to Hatcher (1994), discriminant validity is justified if the estimate of variance extracted for any pair of factors is greater than the square of the correlation between these two factors. The results demonstrated that the discriminant validity of the four latent constructs was supported. Overall, the measurement model with three latent constructs and six indicators was confirmed and accepted for the analyses of the structural model.
Structural Model Testing
In the second step, the fit of the structural model was examined using the maximum likelihood method. The results of the goodness-of-fit indices suggested that the structural model had adequate fit (χ2 = 24.09, df = 7, p < .001, NNFI = .97, CFI = .99, RMSEA = .06, GFI = .97, SRMR = .04). The chi-square difference test between the modified measurement model and the structural model was also used to examine the adequacy of the structural model (Hatcher, 1994). The results indicated that the structural model did not provide a fit to the data; in fact, the fit was found to be not significantly worse than that provided by the measurement model (Δχ2(1) = 0.05, p > .05). Cumulatively, the data suggested that the structural model should be accepted as a parsimonious model for the sample of hot springs tourists (Hair et al., 2006).
The results of the structural model provided support for the hypothesized relationships between cuisine experience, psychological well-being, and revisit intention (see Figure 2). Consequently, the following the hypothesized relationships were supported: (Hypothesis 1) psychological well-being is positively related to revisit intention (β = .43, p < .001), (Hypothesis 2) cuisine experience is positively related to revisit intention (β = .34, p < .001), and (Hypothesis 3) cuisine experience is positively related to psychological well-being (β = .57, p < .001). Overall, the analyses revealed that cuisine experience and psychological well-being were important determinants of revisit intention.

Structural Model With Estimated Path Coefficient
Empirical Testing for the Moderating Effects of Self-Health Perception
According to Baron and Kenny’s (1986) definition of the moderating effect, the magnitude of each of the hypothesized relationships among psychological well-being, cuisine experience, and revisit intention might differ based on hot springs tourists’ self-health perceptions. To examine this, a multiple-group SEM analysis (with invariance tests) was performed on the data (e.g., Qu Lee, 2011; Tang, Jang, Morrison, 2012); the mean of the level of self-health perception was set as the moderating variable to split the entire sample into two groups. Fifty-three percent of the sample had a self-health perception below the mean (Group 1 [N = 280]), and 47% had a level of self-health perception above the mean (Group 2 [N = 244])—they were defined as low and high levels of self-health perception groups, respectively. To understand the sociodemographic profile of each group, the chi-square statistic in the cross-tabulation analysis was applied to evaluate whether there were any statistical differences between them. The results showed that the two groups were statistically different from each other based on gender, age, and monthly household income variables. Group 1 had a higher proportion of men (66%), whereas Group 2 had a higher proportion of women (57%). Members of Group 1 were younger (aged 22-30 years) than their counterparts in Group 2 (aged 41-65 years). Group 1 had a higher proportion of people with a monthly household income of less than NT$30,000 per month (36.4%), whereas the second group was characterized by a proportionally higher number of people (35.2%) with monthly household income levels ranging from NT$70,001 to NT$100,000. These groups were used to test the casual relationships of the model. The mean of the samples was used to ensure that the two subgroups were bimodal and segmented by the moderator (Hair et al., 2006). Prior to examining the invariance tests, a single-group analysis using confirmatory factor analysis had generated a suggested baseline model for each sample separately (Bollen, 1989). As suggested by Bollen (1989), hierarchical tests were used to examine the invariance hypothesis, after the baseline model was established. In this study, the invariance tests included a factor structure and factor loadings; moreover, the path coefficients were constrained to be equal in both groups. The invariance tests are supported if the difference of chi-square is statistically significant.
Two separate SEM models, where the model structure was defined with the entire sample, were performed and are shown in Figure 3. The findings indicated that the theoretical model fit subsamples of both tourists with a high level of self-health perception (χ2 = 12.59, df = 6, p < .001, NNFI = .97, CFI = .99, RMSEA = .05) and those with a low level of self-health perception (χ2 = 12.20, df = 6, p < .001, NNFI = .97, CFI = .99, RMSEA = .06). Moreover, successive invariance tests showed that an invariance of factor structure (χ2 = 24.79, df = 12, p < .001, NNFI = .97, CFI = .99, RMSEA = .06) and an invariance of factor loadings (Δχ2 = 12.38, Δdf = 6) are not rejected (p > .05), and the invariance of path coefficients (Δχ2 = 21.29, Δdf = 6) was rejected (p < .05). Hence, Hypothesis 4 was supported.

Models of the Antecedents of Revisit Intention for Both Tourists With High and Low Levels of Self-Health Perception
Discussion and Implications
Revisit intention is an important concept in comprehending tourists’ choices of products and future repeat behavior. Few studies have identified the salient predictor of revisit intention in the context of hot springs tourism (Lin Morais, 2009). This study represented the first attempt to integrate the relationships among cuisine experience, psychological well-being, and revisit intention. In this regard, this study’s findings offer a significant contribution to the literature about hot springs tourists’ repeat behaviors and choices and behaviors of tourists in general, particularly in light of the fact that several scholars have noted that the development of hot springs/health tourism requires a deep understanding of the concept and formation of revisit intention (Goodrich, 1994; Mueller Kaufmann, 2001) and general tourist behavior/behavioral intention (Decrop, 2000; Sirakaya Woodside, 2005).
In the case of hot springs tourists in Taiwan, our results indicated that revisit intention is influenced by psychological well-being and cuisine experience and that psychological well-being is influenced only by cuisine experience. This finding lent support to studies conducted by Douglas (2001), Kaspar (1990), and Godbey (2003). For instance, Godbey (2003) found that the eating habits of tourists during their trip had an impact on their sense of psychological well-being. Douglas (2001) argued that most European spa-goers visit spa destination predominantly for health reasons. In addition, Au and Law (2002) and Quan and Wang (2004) proposed that some tourists travel solely in pursuit of cuisine experiences. In the study context, psychological well-being may well be the health-promotion/care outcome of the hot springs experience, and cuisine experience might be regarded as compensation for consuming the unique cuisine of a hot springs destination. Therefore, an important theoretical contribution of this study is the identification of cuisine experience and psychological well-being as the salient predictors of revisit intention by hot springs tourists. However, these results may change when tourists’ health perceptions are considered.
Tourism researchers have suggested that studies predicting revisit intention should focus on the interpretation of behavioral intentions demonstrated by a variety of tourists (Decrop, 2000; Sirakaya Woodside, 2005); they believe that it is necessary to identify the relative important predictors of revisit intention for different types of tourists in order to implement the most effect marketing strategies (Petrick et al., 2001; Um et al., 2006). Accordingly, although it is important to note that psychological well-being and cuisine experience go some way to explaining hot springs tourists’ revisit intentions, a more in depth understanding of this relationship is necessary. In particular, it is important to assess whether psychological well-being and cuisine experience are particularly important for one specific type of tourist. To this end, this study finds that the relationships between psychological well-being, cuisine experience, and revisit intention vary among hot springs tourists, depending on whether they have high or low levels of health perception. Another significant theoretical contribution of our study is the acknowledgment that tourists’ behavioral intentions may be attributed to their cognitive and affective constructs concerning a destination. Additionally, we reinforce the importance of tourists’ self-health perceptions in health/wellness tourism. In the case of Taiwan, for hot springs tourists with a low level of health perception, the effect of psychological well-being on revisit intention was not found, indicating that the hot springs soaking experience was not instrumental in forming revisit intention. Conversely, for hot springs tourists with a high level of health perception, both psychological well-being and cuisine experience were found to have a synergistic effect on revisit intention. The findings revealed that although cuisine experience is universally important, psychological well-being is only important for a specific type of hot springs tourist. As documented by Quan and Wang (2004), food experiences not only offer tourists a source of pleasure and enjoyment but they also have a special meaning for some tourists. In addition, the fact that psychological well-being was not found to affect all tourists’ revisit intentions was surprising. Mannell and Klieber (1997) found that the psychological well-being of tourists after visiting hot springs affected future behaviors. However, this study did not find this. This may be because the hot spring soaking experience was not strong enough to generate the expected relationship. It is also plausible that the hot springs service is not sufficiently distinctive. Cooper et al. (1995) supported the notion that distinctiveness is an issue in spa tourists’ happiness. The results of the present study provide further evidence that the relationship between psychological well-being and revisit intention is not well understood.
Looking beyond the structural relationships between endogenous and exogenous variables, it is useful to discuss the observed differences in the strength of path coefficients across the two types of hot springs tourists. For example, psychological well-being was found to be the most important (i.e., strongest) predictor of revisit intention for people in this market segment with high levels of self-health perception; however, the most important predictor of revisit intention for tourists with low levels of self-health perception was cuisine experience. Earlier work on revisit intention indicated that the importance of predictors of tourists’ revisit intentions can vary significantly with the context of the vacation experience sought (Baker Crompton, 2000; Petrick et al., 2001; Um et al., 2006) and with the type of tourist (Lin Morais, 2009). Our findings support this assertion. Hjalager (2003) and Nesse and Williams (1994), for example, argued that, among individuals with healthy lifestyle, the relative impact of psychological well-being on behavioral intention may be stronger than the impact of food experience. However, Kivela and Crotts (2006) found that some pleasure travelers would return to the same destination because of its unique gastronomy rather than other attractions. However, owing to the limited number of previous studies on the significance of these factors and varied results that were dependent on the hot springs tourists’ self-health perception levels, these findings could not be compared with or validated by findings from other studies.
The influence of the indicators of cuisine experience (health exploration and life diversion) and psychological well-being (happiness and life satisfaction) and their level of influence on cuisine experience and psychological well-being is noteworthy. The indicator “health exploration” of cuisine experience, for example, was the strongest predictors across the two types of hot springs tourists. This finding supports previous research about gastronomy tourism (Hjalager, 2003). According to Hjalager’s ideas, gastronomy symbolizes tourists’ lifestyle and explains and predicts human healthy behaviors. Additionally, the indicator “happiness” of psychological well-being was the strongest predictor for both two types of tourists. This finding is consistent with Kaspar (1990) and Mueller and Kaufmann’s (2001) assertion that tourists enjoying hot springs tourism tend to express emotional happiness. Therefore, the salient predictors for hot springs tourists’ revisitation intentions reflect their enthusiasm for the pursuit of optimum health.
Overall, the findings suggest that when tourists visit a hot springs destination, they develop a behavioral intention to revisit that destination. More specifically, they intend to revisit if they experience psychological well-being and enjoy the cuisine experience. We noted that the significance of psychological well-being and cuisine experience varied according to hot springs tourists’ high or low levels of self-health perception. The results of the current study, therefore, offer some practical implications for DMOs. First, research suggests that the characteristics associated with psychological well-being are happiness and life satisfaction. According to Strack, Schwarz, and Gschneidinger (1985), experience of memorable pleasant events provides evidence of having a good time in the past. Thus, when DMOs promote hot springs activities, a slogan such as “Hot Springs, Happy Lives, Well-being Forever” conveys the message that a hot springs vacation will bring happiness and pleasure to families, friends, or groups. Slogans can also trigger tourists’ memories from activities at a hot springs destination and motivate them to revisit. Another finding of this study is that the cuisine experience of tourists from the Bei-Tou hot springs area involves health exploration and life diversion. To encourage spa tourists to depart from daily routines and try healthy and nutritious food alternatives, hot springs DMOs should promote comfortable and amiable environments that offer a la carte dining experiences. The assumption is that continuously changing food options allow tourists to experience extraordinary pleasure in dining and create anticipation for future visits to the same destination. Finally, we believe that the most important management-based contribution of this study is the observation that hot springs tourists with low levels of self-health perception were only affected by the destinations’ cuisine experience. Contrastingly, hot springs tourists with high levels of self-health perception were influenced by both psychological well-being and cuisine experience. As a result, hot springs DMOs must develop marketing strategies specific to their clientele, with their varied levels of perceived self-health. For example, promotions for hot springs tourists with low levels of self-health perception should include innovative and ever-changing menus in comfortable and friendly environments so that tourists can experience dining pleasure. Conversely, promotions for tourists in this market segment with high levels of self-health perception should highlight both psychological well-being (e.g., happiness) and healthy cuisine experience (e.g., health exploration and life diversion). Thus, the provision of a comfortable and happy hot springs activity environment and organic cuisine options could improve the revisit likelihood of tourists with a high level of self-health perception.
Conclusions, Limitations, and Future Research
This study developed and empirically tested a structural model explaining hot springs tourists’ revisit intentions and how this related to cuisine experience and psychological well-being; it also examined the moderating effect of perceived self-health on the relationships between these constructs. The proposed model provided a comprehensive understanding of cuisine experience and psychological well-being; it identified the salient factors that influence revisit intention in the context of hot springs tourism in Taiwan and provided reasonable explanations of the moderating effect of perceived self-health on revisit intention. These findings reveal that psychological well-being and cuisine experience were important predictors of revisit intention of hot springs tourists; their effect varied, however, based on perceived levels of self-health. Moreover, these findings are supported by the effect of cognitive and affective constructs on behavioral intentions of tourists in general. Future efforts to understand hot springs tourists’ revisit intentions and behavioral intentions of tourists in general should take the variables described above into consideration.
Nevertheless, there were several limitations to this study. Because of the limited geographical scope of this study, the findings and conclusions may not apply more broadly. Past studies also indicated that, in most cases, past visits/experiences (Mazursky, 1989; Ryu Jang, 2006) and accommodations (C.-F. Lee et al., 2009; T.-H. Lee, 2010) positively influenced visitors’ revisit intentions; limitations of this study prohibit validation of the above findings. Factors influencing revisit intention should be included in the model(s) for further studies. In addition, Gilbert and Abdullah (2004) reported that the relationship between psychological well-being and behavioral intention was mediated by tourist expectation. Hence, future research should examine the relationships between motivation, psychological well-being, and tourist destination preference. Recent empirical studies show that tourism literature on revisit intention recognizes that demand generated by recommendations is more important than demand generated by revisits to a destination (Namkung Jang, 2007; Park Njite, 2010; Prayag, 2009). The Internet represents a lucrative venue for tourism service recommendations (Chi, 2012); therefore, it might be interesting to include “Internet recommendation” as an item to measure the construct of revisit intention in future studies. Another interesting issue for future research is the comparison of behaviors associated with international demand for a specific destination with domestic demand behaviors. Although this study provides credible evidence for including perceived health as a variable in hot springs tourist destination behavior models, investigations should be extended to a variety of societies, tourist samples, or destinations to determine if the same model may be applied more widely. Finally, as Fleischer and Pizam (2002) make clear, older people are likely to be an important group in future health tourism. Further research should examine their role, and the role of those with specific health problems to see how their revisit intention is determined.
Footnotes
Authors’ Note:
This research was supported by the National Science Council of the Republic of China, Grant No. NSC 97-2410-H-035-003-MY2.
