Introduction Virtual communities have changed lifestyles and interpersonal relations, creating an internet community culture. They are a primary medium of emotional contact with others. American psychologist estimated that more than 350 million people have Facebook Addiction Disorder (FAD) (Hong et al., 2014; SickFacebook.com, 2010), some kind of virtual community addiction (VCA). An average of 728 million people use Facebook every day, and users spend much time on Facebook (Hong et al., 2014). 80.24% of university students think that Facebook is an important element of social culture (Thompson & Lougheed, 2012). VCA has been explored in psychology in the past. This study adds three psychological dimensions and sense of virtual community when investigating the antecedents of VCA. The results of this study may direct public attention to VCA and help develop guidelines for online marketing and operation of virtual communities. Literature review and hypotheses development Virtual community addiction (VCA) Many people are not aware of addictive in virtual community. While some addictive behaviors are considered mental disorders, an increasing number of studies has found potentially addictive behavior involving computers and the internet (Andreassen et al., 2012), including video game addiction (Fisher, 1994), Internet addiction (Beard, 2005), mobile phone addiction (Choliz, 2010), and online addiction (Griffiths, 2012). The definition of Internet addiction remains controversial. This study defines VCA as a tendency that users forced to use and even can not control not to use for the virtual communities. Virtual community codependency (VCC) Although codependency lacks a clear definition, there are only two perspectives on codependency that are relevant to this study. First, codependency is a learned helplessness, resulting in individuals who cannot create or participate in interpersonal relationships in the future (O’Gorman, 1993). Second, codependency is a pattern of dysfunction in interpersonal relationships. According to the social compensation theory, if people feel insecurity and negative social identity in real life interpersonal networks, they may spend more time using virtual communities as compensation. Therefore, we hypothesize: H1. Virtual community codependency (VCC) positively influences virtual community addiction (VAC). Sense of virtual community (SVC) Blanchard (2007) defined the SVC as members feel membership, a sense of belonging and attachment to the group through electronic communication. Abfalter et al. (2012) defined the SVC as members having a sense of ‘spirit of belonging together’ in a virtual community. Members are interacting more, making them develop close relationships that lead to a stronger bond (Abfalter et al., 2012; Chavis et al., 1986; Obst et al., 2002). Facebook is widely used in the virtual community, regardless of the number and use of the degree has been increasing. Thus, we conclude that high VCC will need individuals to have a desire to derive compensation from the virtual community that cannot be achieved in the real world. If people in this community have similar needs, priorities, and goals, increasing the use of Facebook will lead to an increase in VCA. We propose the following hypothesis: H2. The relationship between the Virtual Community Codependency (VCC) and Virtual Community Addiction (VAC) is mediated by the Sense of Virtual Community (SVC). Behavioral inhibition system (BIS) and behavioral approach system (BAS) Gray (1982) proposed the most widely applied pair of systems controlling behavioral activity, the behavioral inhibition system (BIS) and the behavioral approach system (BAS). These are intrinsic motivation systems in charge of desire and disgust (Carver & White, 1994). They are used to measure the sensitivity of individuals to punishment avoidance and to reward acquisition. When the sensitivity of an individual toward penalty/reward is higher, their response to a stimulus is greater (Linden et al., 2007). Studies also show that Internet addiction and BIS are related (Yen et al, 2009). Thus, we conclude that high VCC will be stimulated by BIS when there is an accident or a novel experience, and negative emotional reactions will seek to be resolved in the virtual world. Hence, Facebook users will to escape from the real environment to the virtual world, so as to addict on the virtual community. This leads to the following hypothesis: H3. The direct relationship between the Virtual Community Codependency (VCC) and Virtual Community Addiction (VAC) is mediated by the Behavioral Inhibition System (BIS). BAS is based on motivations for an individual to pursue and achieve goals. BAS is more active and more sensitive to external reward clues, leading to stimulated sensory seeking behavior (Beauchaine et al., 2001). If the incentive clues appear, individuals who have high BAS are more prone to short positive emotions and approaching behavior individuals with low BAS. Thus, we speculate that high VCC will be stimulated by BAS when there is reward responsiveness or when driven in pursuit of goals in the virtual community. It will also make it more likely to engage in fun seeking in virtual communities. Thus, Facebook users deeply trapped in the virtual environment by BAS stimulation and fail to pull themselves out. Therefore, we hypothesize: H4. The direct relationship between Virtual Community Codependency (VCC) and Virtual Community Addiction (VAC) is mediated by the Behavioral Approach System (BAS). Methodology Survey development and measurement We adapted a number of scales (Abfalter et al., 2012; Andreassen et al., 2012; Carver & White, 1994; Fischer et al., 1991) and compiled a 55 item self-report scale questionnaire. Following the procedure recommended by Churchill (1979), we pre-tested the scale on 30 college students to determine the reliability and validity of the constructs. Based on the feedback, we refined the measures and ensured that all of the questionnaire items were applicable to our research. Finally, 53 valid items were retained in the formal questionnaire. Data collection and sample profile We used a web-based survey to collected data from the Youthwant survey platform in Taiwan. This was a more effective way to search for respondents for this study. The formal questionnaires were free to answer by members of the network platform. The participants were told of the survey’s importance to trigger participant’s interest, and give a 10 bonus points as reward. A total of 224 questionnaires were returned, 3 invalid and 221 effective. Respondents were evenly distributed across various ages between 18 and 50 years. The majority of respondents were female (60.2%) and only about 23.1% were students. Most respondents had completed university education (60.6%). 93.6% had been used Facebook for more than one year, with 63.3% having over four years of membership. 70.6% use Facebook more than once a day and for 82.8%, daily usage is at least a half hour, while 71.0% exceed a half an hour each time they use Facebook. This shows that of all respondents were long-term users. Results The total effects are determined by testing the direct impact of VCC on VCA without the mediator constructs. The results (see Table 1) show a significant relationship between VCC and VCA (β11 = 0.572, p < 0.001), supporting H1 (VCC positively affects VAC). The mediating effect is slight but statistically significant. 0.096 of the indirect effect is the result of the effect of VCC on SVC, which in turn influences VCA. The direct relationship between the VCC and VCA is significant as mediated by the SVC. Thus, H2 was supported. The mediating effect involves testing the direct relationship between the VCC and VCA mediated by the BIS. The results show a significant relationship between VCC and VCA (β13 = 0.488, p<0.001). VCC positively and significantly affects BIS (β12 = 0.493, p < 0.001). The effect of BIS on VCA is also positively significant (β23 = 0.169, p <0.05). This was confirmed by the evaluation of the structural model results. The direct effect is 0.488 (t = 5.661, p < 0.001) and the indirect effect is 0.083 (0.493  0.169= 0.083) which is significant (z = 2.039, p < 0.05) by Sobel test. Thus, the direct relationship between the VCC and VCA is slightly mediated by BIS. H3 was supported. The VAF of the model is 33.70% (0.193 / (0.380 + 0.193)  100%). In this situation, a VAF larger than 20% and less than 80% can be characterized as partial mediation (Hair et al., 2014). The ƒ2 effect size is 0.117, showing that BAS has medium effects on VCA. Thus, H4 was supported. Discussion This study is a step by step quantification of how changes in virtual community codependency (VCC) are related to changes in virtual community addiction (VCA) directly and indirectly through one or more mediators. First, the results show a strong relationship between virtual community codependency (VCC) and virtual community addiction (VCA). Furthermore, we cannot deny that the sense of virtual community (SVC) has a small mediating effect on the direct relationship between virtual community codependency (VCC) and virtual community addiction (VCA). The increased use of Facebook when there is a sense of the ‘spirit of belonging together’ can lead to increased tendency to virtual community addiction (VCA). The results are supported by Fischer et al. (1991) and Hong et al., (2014). Limitations and future research directions There are limitations in this study. First, this study does not have a large sample. Second, with technology advances and use behavior changes, the popularity of virtual communities and dependency is rising. The virtual community addiction scale can only evaluate the current situation. This study investigated the latent variables that determine partial or complete mediation using the AVF of Hair (2014). The judgment of AVF uses rules of thumb and awaits meta-analysis of partial and/or complete mediation in the future. Finally, future research could examine virtual communities in other cultures.