This column is devoted to highlighting the findings of current research into science education.

The notion of replicability may not apply in educational research in the strict sense that it does in the physical sciences, as we cannot reproduce the conditions of previous research, and contextual differences between research samples complicate the generalisability of findings (Taber, 2014). In addition to acquiring a core of factual knowledge about the geosciences, he would also like his students to gain an accurate understanding of important geologic processes, to be able to think critically about spatial data, and to learn to integrate processes and data into a holistic model of the Earth and Earth processes. Mobile/eReaders – Download the Bookshelf mobile app at or from the iTunes or Android store to access your eBooks from your mobile device or eReader. International Journal of Science Education, 37(14), 2284–2305. So, published studies in science education may adopt the questionable assumption that conceptual knowledge across one (or more) school topics should be considered a single dimension and inappropriately offer a high value of alpha as evidence of such unidimensionality. Science Education, 98(4), 705–737.

Introduction - José teaches a capstone course for geoscience majors, and has recently become head of his department's curriculum committee. The responses were examined and various tests were applied to the instrument, including the use of Cronbach’s alpha, where “Cronbach α scores greater than .70 were considered as indicative of acceptable reliability… After checking the reliability of each factor using the Cronbach α coefficient, items that decreased the reliability of each factor were removed” (pp.2099–2100). C�r��o4Ϟ+X�i�ga��(�+>N��zK�t<7��=�,�Q@� ��Y�/��'��G�M-�K C����ao��v�]|�mUx��� r��� Many, but not all, of the papers found to cite alpha values in the 2015 volumes of four science education journals (IJSE, JRST, RISE, SE) offered qualitative interpretations of the significance of the values calculated in relation to what was being measured (which, as discussed above, was usually considered as a form of reliability or internal consistency). International Journal of Science Education, 37(3), 446–468. An initial informal search suggested these concerns did not simply reflect the treatment of alpha in a few isolated papers but were more widespread and led to the identification of examples that illustrated points worthy of raising within the research community. Mun and colleagues’ value of alpha of 0.79 across their instrument was obtained despite items not all measuring the same thing and was in part a result of the composite instrument having more items than its (apparently less reliable) component individual factor subscales.

The Role of Gestures in Geoscience Teaching and Learning (Acrobat (PDF) 3.9MB Oct28 08) In everyday terms, a high value of alpha offers a guard against specific items being unique in the particular sense of eliciting response patterns unlike any of the other items (Cortina, 1993).

Principled improvement in science: forces and proportional relations in early secondary-school teaching. Herman, B. C. (2015). That is, they calculated the mean correlation for each of the scales with the five other scales and obtained (in ascending value) results of 0.09, 0.30, 0.31, 0.32, 0.39 and 0.51. Other authors similarly suggest or imply that a high value of Cronbach’s alpha is desirable when a test of knowledge includes items testing across a range of different science concepts. A label as a hidden persuader: chemists’ neutralization concept. ��*҄�N@��7{�j�_�N-Od[�����^�5"�d$���U�V%4׻�M�yA���zRhb����&��i���������:�Z��a����i��T�!�O�A��hQP�9���4�����qN_K. Research Report 253. (This might be indicated where the scale or instrument contains large numbers of items but may be better judged by inspection of the actual items where these are provided.). Moreover, “The internal consistency reliabilities using the Cronbach α coefficient was .79 for all items” (p.2100) and this overall measure of consistency was greater than that of any of the three components or their individual factors.

So in this study, items that were considered to be “loosely related to each other,” and not to measure understanding of a single concept, collectively gave rise to an alpha value “considered acceptable”, providing a warrant for the authors to claim “The AREPDiT proved to be a valid and reliable diagnostic instrument to explore the knowledge and misconceptions of pre-service teachers” (p.1683). Of particular interest here is that the alpha value obtained reflects the specific sample tested. A high value of alpha obtained from administering an instrument to a sample of students could be understood as suggesting that the items are measuring some common factor(s) rather than unique features associated with individual test items. Meredith Beilfuss, Science Education Department, Indiana University

Sandra K. Abell (deceased) was MU Curators Professor of Science Education, University of Missouri, USA. International Journal of Science Education, 13(4), 459–471. Mun et al., provide their final list of items (pp.2111–2112), and four of these are selected here to illustrate how a high overall alpha does not necessarily imply that all items are strongly related: E2. Alison Stokes, University of Plymouth