A prominent aspect of a subject specific language is the vocabulary, and this study is investigating the vocabulary used in different subjects in the Swedish national tests in science for year 6. This study is an expansion of a previous study of the vocabulary used in science tests in different subjects, i.e. biology, chemistry, and physics. Therefore comparisons are made regarding a Swedish national test in science for year 9 and science items from the Swedish version of TIMSS 2011 year 8.

When assessing students’ knowledge through written tests, it is important to know whether a student’s incorrect response is caused by a lack of knowledge or a misunderstanding of the question itself. Student related issues such as misinterpretations, linguistic inadequacy (e.g. poor reading skills or lack of vocabulary), unconscious adding or omissions of words can cause incorrect responses. Item related issues can be that the question is too linguistically complex for the students level, is ambiguously formulated, or typographically, grammatically or semantically erroneous allowing for other interpretations than the intended (Clerk & Rutherford, 2000; Harlow & Jones, 2004). When using international tests, inadequate translations also open for other interpretations. Translated items can be unintentionally altered, due to linguistic differences between languages and cultural influences on interpretations of words (Oakland & Lane, 2004). Therefore it is interesting to see if there are any differences regarding vocabulary when comparing national tests developed in Swedish and translated international tests such as TIMSS. If visible differences are found, it may point to that how the translation is performed and how the translation process is organized, can be of importance for all countries using translated versions of international tests and therefore should be taken into consideration when evaluating results from such tests.

Claims have been made that science could not function without its specialized terms (Wellington & Osborne, 2001), but scientific texts with a high ratios of technical terms can be difficult to understand for students (Fang & Schleppegrell, 2008). Swedish science textbooks have a lot of content specific technical vocabulary, thus distinguishing them from the language used in social science textbooks, which have a vocabulary similar to the colloquial language found in newspapers and novels (Ribeck, 2015).

A characteristic of scientific vocabulary is abstraction, mainly achieved by nominalizations; the changing of verbs or adjectives into nouns (Fang & Schleppegrell, 2008). However, scientific texts in Scandinavian languages more often use compound words instead of nominalizations than corresponding English texts (Ekvall, 2011). Although the main words are easily recognized in Swedish compound words (Holmegaard, 2007), such low frequent compound words might be an additional complication for students in countries with languages frequently using compound words. Investigating students understanding of Swedish extra-long words (>13 letters) from the social sciences, it was found that such words -often requiring a sociocultural understanding- were difficult to understand for students (Holmegaard, 2007).

The aim of this study is to investigate the vocabulary used in the Swedish national science tests for year 6, together and separated into subjects (biology, chemistry, physics), and comparing them with the vocabulary used in a Swedish year 9 national science test and the science items from the Swedish version of TIMSS 2011 year 8. The research questions are:

• How is the vocabulary use in the three year 6 Swedish national tests in science distributed when compared with the 10.000 most commonly used words in Swedish school textbooks, general written Swedish and a more limited vocabulary?

• Is there any differences between the coverage when comparing Swedish national tests in science for year 6 with corresponding tests for year 9 and the translated TIMSS 2011 year 8 questions?

The study analyses the vocabulary used in the Swedish National tests in science (2013-2015) for year 6 in the three subjects Biology, Chemistry & Physics. Results are compared with corresponding results from analysis of the 2009 Swedish national science tests for year 9 and TIMSS 2011 science items for year 8. A total of 13 science tests have been analysed, nine from year 6, three from year 9 and one from TIMSS year 8. Following the method of a previous analysis of the TIMSS 2011 science items (Persson, 2016) the words in each test are compared with three Swedish corpora representing different kinds of vocabularies in written texts, thus reflecting the vocabularies different students might have. The corpus LäSBarT (Heimann Mühlenbock, 2013) is reflecting a more limited Swedish vocabulary, i.e. that of a second language learner or young student. The corpus SUC (Stockholm University, 2014) reflects a more general vocabulary in written texts, and finally the corpus OrdIL (Johansson Kokkinakis, 2007) reflects the vocabulary used in school textbooks year 7-9. The words in each corpora are separated into frequency bands in steps of 1000, where inflected forms of words are regarded as different words. Each word initially indicated as not being present in each of the 10000 first words of each corpus has been manually lemmatized (converted into its base form) and rechecked, as a student familiar with a words base form and the Swedish language´s rules for inflection can be assumed to be able to deduce other regular inflected forms (cf. Kanebrant et al., 2015). The text coverage of the three National Tests for year 6 are then compared with each other, both separated into subjects and all subjects together. As only minor differences between the different years were found, the test items of each subject were added together. This resulted in a total of 83 biology items, 86 in chemistry and 78 in physics. The same procedure were applied to the 39 items from the 2009 Swedish National Test in science for year 9, whereas the coverage of the 197 TIMSS items had previously been calculated in the same way (Persson, 2016).

All tests follow the same expected pattern: the coverage is highest for the school textbook corpus, and lowest for the easy-to-read corpus, but with some differences. Comparing the year 6 items with the Swedish school textbook corpus, the vocabulary seems to be at an appropriate level. In fact, the coverage for all 10000 words is even slightly higher (91.3%) than for TIMSS questions (90.3%), and also higher than for year 9 questions (89.5%). When comparing year 6 items with a corpora mirroring a more general Swedish written language the text coverage drops. Looking at results for all 10000 words the average drop is approximately 4%. The year 9 questions shows an average drop of 1.5% but from a lower level. The coverage for both national tests is therefore close to the textbook coverage. This can be compared with TIMSS that shows a drop of 7%. However, interesting differences can be seen when comparing school textbook coverage with a more limited vocabulary. The average difference in coverage for the 10000 words is only 5% for year 6-tests, but more than 9% for year 9 and more than 12% for TIMSS. As a limited vocabulary can be interpreted as a hindrance for a student’s meaning making in TIMSS items, the year 6 national test has a coverage only slightly less than the coverage of a more general Swedish language. Thus, the Swedish national tests seems to be able to make use of the 10000 most commonly used words in Swedish school textbooks similar to how TIMSS does, but at the same time also manages –especially in in the case of year 6 questions – to use a language that does not make a limited vocabulary an obstacle for interpreting the questions, as seems to be the case for questions posed in TIMSS.

Clerk, D. & Rutherford, M. (2000). Language as a Confounding Variable in the Diagnosis of Misconceptions. International Journal of Science Education, 22, 703-717. Ekvall, U. (2011). Enhetligt på den finska sidan men varierat på den svenska. Om kemiböcker i svenska och finlandssvenska klassrum. In I. Eriksson (Ed.). Kemiundervisning och textbruk i finlandssvenska och svenska skolor – en komparativ tvärvetenskaplig studie. Stockholm: Stockholms Universitets förlag. Fang, Zhihui & Schleppegrell Mary J. (2008). Reading in secondary content areas: A language-based pedagogy. Ann Arbor: University of Michigan Press. Harlow, A. & Jones, A. (2004). Why Students Answer TIMSS Science Test Items the Way They Do. Research in Science Education, 34, pp.221-238. Heimann Mühlenbock, K. (2013). I see what you mean: assessing readability for specific target groups. [Diss.] Data linguistica. No. 24, Institutionen för svenska språket, Göteborgs universitet. Holmegaard, Margareta (2007). Långa ord - en svårighet för flerspråkiga studerande? In Lindberg, I. & Johansson Kokkinakis, S. (Eds). OrdiL: en korpusbaserad kartläggning av ordförrådet i läromedel för grundskolans senare år. ROSA 8. Institutet för svenska som andraspråk, Göteborgs universitet. Johansson Kokkinakis, Sofie (2007). Språkteknologiskt arbete i OrdiL-projektet. In Lindberg, I. & Johansson Kokkinakis, S. (Eds). OrdiL: en korpusbaserad kartläggning av ordförrådet i läromedel för grundskolans senare år. ROSA 8. Institutet för svenska som andraspråk, Göteborgs universitet. Kanebrant, E., Heimann Mühlenbock, K., Johansson Kokkinakis, S., Jönsson, A., Liberg, C., af Geijerstam, Å., Folkeryd, J. & Falkenjack, J. (2015). T-MASTER A tool for assessing students' reading abilities. In Proceedings of the 7th International Conference on Computer Supported Education (CSEDU 2015). http://dx.doi.org/10.13140/RG.2.1.1135.0241 (Retrieved 31-01-2018) Oakland, T. & Lane, H. B. (2004). Language, Reading, and Readability Formulas: Implications for Developing and Adapting Tests. International Journal of Testing, 4, pp.239-252. Persson, Tomas (2016). De naturvetenskapliga ämnesspråken. De naturvetenskapliga uppgifterna i och elevers resultat från TIMSS 2011 år 8. [Diss.] Uppsala: Acta Universitatis Upsaliensis. Ribeck, Judy (2015). Steg för steg. Naturvetenskapligt ämnesspråk som räknas. [Diss.] Data linguistica. No. 28, Institutionen för svenska språket, Göteborgs universitet. Stockholm University (2014). SUC, Stockholm-Umeå corpus. http://www.ling.su.se/english/nlp/corpora-and-resources/suc/stockholm-ume%C3%A5-corpus-suc-1.14045 (Retrieved 31-01-2018) Wellington, J. & Osborne, J. (2001). Language and literacy in science education, Buckingham, Open University Press.