中间选项在李克特量表中的应用给教育研究的启示.pdf

Hong Kong Teachers’ Centre Journal 《香港教師中心學報》, Vol. 11© The Hong Kong Teachers’ Centre 2012The use of midpoint on Likert Scale: The implications for educational research中間選項在李克特量表中的應用：給教育研究的啟示Kwok Kuen TSANGPhD Student, Faculty of Education, The University of Hong KongAbstractThis article is going to review the debate of the use of midpoints on Likert scale in order to understand the implications for scale construction. In this review, the major issues around this debate are identified. The first is methodological. The debate around this issue is the concern about whether the midpoints affect the reliability and validity of measurements. The second issue is epistemological. It concerns about how researchers exactly know the meaning of the responses into midpoints that they intend to measure. To some extent, the debate based on the methodological issue illustrates that both using and not using of midpoints are acceptable, because the midpoints may not really affect the reliability and validity. Therefore, the epistemological issue while designing the rating scale of a measurement is the focus of this article. The implications of the debate for educational research are also discussed in this article.Keywordsmidpoint, Likert scale, scale construction摘要本文回顧有關在李克特量表應用中間選項的爭論，希望著此了解中間選項對建構量表的作121用。

本文將指出，關於中間選項的爭論主要圍繞兩個主題一是方法論的，這一主題所關心的是中間選項對量表的信度和效度的影響；二是認識論的，這主題所關心的是研究者如何得知受試者對中間選項的理解跟他們所想的一樣在某程度來說，關於方法論上的爭論普遍認同中間選項不一定對量表的信度和效度有影響，所以從應用和不應用中間選項均可接受因此，在建構量表時，我們應更多地考慮認識論上的問題另外，本文還討論了中間選項的爭論對教育研究的啟示關鍵詞中間選項，李克特量表，量表建構In educational research, Likert scale is commonly used to measure different kinds of variables, such as teacher stress and burnout (e.g. Chan, Chen, Dworkin, 2002), self-efficacy (e.g. Brouwers, Tomic, Cheung, 2006), school and teacher effectiveness (e.g. Bangert, 2006; Kyriakides, Campbell, Kyriakides, Demetriou, Reezigt Reynolds, 2001), school organization (e.g. Bowen, Ware, Rose, Firestone, 1984; Firestone Herriott Cavanagh Moos, 1987; Pang, 1998; Wagner, 2006), and the like. The reason is that Likert scale empowers educational researchers to effectively operationalize the variables and then identify their relationships in order to improve our educational system.However, some researchers claim that the use of midpoints on Likert scale may affect research reliability and validity, but some other researchers disagree. It is necessary for education researchers to learn this debate about whether midpoint opinions are included in a scale, because the debate not only directly points to the problem of research quality but also the validity of research implications and recommendations to educational system. In this article, therefore, we will first review the debate in social research in general. On the basis of this review, then, we will suggest the implications of the debate for educational research. 122The use of midpoint on Likert Scale: The implications for educational researchDebate of the use of midpoints on Likert scaleMethodological issueOne concern among researchers about having midpoints on a Likert scale is the effects of the midpoints to the reliability and validity of measurements. Generally, the supporters of midpoint opinions claim that the midpoints can increase the reliability of measurement. For example, Courtenay and Weidemann (1985) assess the effects of midpoint answers (“don’t know”) to the Palmore’s Facts on Aging quizzes (FAQ) and conclude that the midpoint answers tend to enhance the reliability of FAQ. Another study conducted by Adelson and McCoach (2010) present similar findings. In that study, Adelson and McCoach compared the response pattern of elementary students who responded a mathematics attitudes instrument with a 4-point Likert scale with another group of elementary students who responded the same instrument but the scaling had an additional neutral point. The study shows that the scale including a neutral midpoint might be more appropriate for elementary students than the 4-point scale, because the reliability of the 5-point scale was statistically and significantly higher than the reliability of the 4-point scale. On the other hand, the adversaries argue that the high reliability may be resulted from response set (Cronbach, 1950), especially the tendency to choose the midpoint options. Weems and Onwuegbuzie (2001) conduct three studies to show that there was a high rate of midpoint choices among their samples. This to some extent implies response set to the midpoints exist. Different from the findings found by the supporters of midpoints, the response set in Weems and Onwuegbuzie’s studies seems to attenuate the reliability rather than enhance it (Weems Kulas, Stachowski, Matell Lee, Jones, Mineyama, Wong, Tam, then, they were asked to indicate the answers again on a continuous non-verbal scale (literally straight, blank line). After that, Worcester and Burns compared the answers between the verbal and 。