52 / 372
發言或緘默：心理安全與自我效能在社會資本影響社群網站使用者知識分享行為上所扮演的中介角色
52
Board System (BBS) so that interested people could link to the questionnaire and answer
page. Only registered and relatively active Facebook users were applicable for this study. To
ensure that this requirement was met, five sleeper questions and contingency questions were
asked. For example, all respondents were first asked to specify which Facebook functions
they use more frequently, followed by questions about whether they had ever posted their
own (or their relatives’) photos on Facebook before. If answers to the first question
contradicted those to the second, the person was prevented from participating further.
Furthermore, responses that showed signs of logical inconsistency were filtered out to reduce
sampling errors. In total, 439 Facebook users completed the survey, which constituted our
valid sample.
3.2 Measurements of Constructs
Measurements for each of the six selected constructs were based mainly on items or
scales reported in previous studies. First, measurements for structural and cognitive social
capital came from Chiu et al. (2006) and Obst and White (2005), and measurements of
relational social capital were derived from Nahapiet and Ghoshal (1998). As Zhang et al.
(2010) note, most social capital studies were conducted in the organizational context (in
which members are physically bound), so we modified previous measurement items to suit
the SNS context (in which members are virtually bound). Second, we modified
measurements for psychological safety from Siemsen et al. (2009), who examine the
influence of psychological safety on employees’ knowledge sharing behavior. Third, we
measured knowledge sharing self-efficacy by combining Constant et al. (1996) and Kalman
(1999) studies, which centered on the organizational context. Finally, we measured
knowledge sharing behavior using items from Lin, Hung, and Chen (2009). We undertook
any necessary modifications and adaptations in accordance with the research purpose of this
study. All items were measured on 7-point Likert scales, ranging from “strongly disagree”
(1) to “strongly agree” (7).
4. Results
To examine whether the hypotheses were tenable, we undertook several statistical
analysis procedures. We report the descriptive statistics first, followed by the confirmatory
factor analysis (CFA) results to check fitness of the measures. Last, we test the hypothesized
structural equation model using LISREL 8.8.