The Common European Framework in its political and educational context What is the Common European Framework?

Appendix B: The illustrative scales of descriptors
This appendix contains a description of the Swiss project which developed the
illustrative descriptors for the CEF. Categories scaled are also listed, with references to
the pages where they can be found in the main document. The descriptors in this
project were scaled and used to create the CEF levels with Method No 12c (Rasch
modelling) outlined at the end of Appendix A.
The Swiss research project
Origin and Context
The scales of descriptors included in Chapters 3, 4 and 5 have been drawn up on the
basis of the results of a Swiss National Science Research Council project which took
place between 1993 and 1996. This project was undertaken as a follow-up to the 1991
Rüschlikon Symposium. The aim was to develop transparent statements of proficiency
of different aspects of the CEF descriptive scheme, which might also contribute to the
development of a European Language Portfolio. 
A 1994 survey concentrated on Interaction and Production and was confined to
English as a Foreign Language and to teacher assessment. A 1995 survey was a partial
replication of the 1994 study, with the addition of Reception, but French and German
proficiency were surveyed as well as English. Self-assessment and some examination
information (Cambridge; Goethe; DELF/DALF) were also added to the teacher
assessment. 
Altogether almost 300 teachers and some 2,800 learners representing approximately
500 classes were involved in the two surveys. Learners from lower secondary, upper
secondary, vocational and adult education, were represented in the following
proportions:
Lower secondary
Upper secondary
Vocational 
Adult
1994
35%
19%
15%
31%
1995
24%
31%
17%
28%
Teachers from the German- French- Italian- and Romansch-speaking language regions
of Switzerland were involved, though the numbers involved from the Italian- and
217

---

Romansch-speaking regions was very limited. In each year about a quarter of the
teachers were teaching their mother tongue. Teachers completed questionnaires in the
target language. Thus in 1994 the descriptors were used just in English, whilst in 1995
they were completed in English, French and German.
Methodology
Briefly, the methodology of the project was as follows:
Intuitive phase:
1.
Detailed analysis of those scales of language proficiency in the public domain or
obtainable through Council of Europe contacts in 1993; a list is given at the end of
this summary.
2.
Deconstruction of those scales into descriptive categories related those outlined in
Chapters 4 and 5 to create an initial pool of classified, edited descriptors.
Qualitative phase:
3.
Category analysis of recordings of teachers discussing and comparing the language
proficiency demonstrated in video performances in order to check that the
metalanguage used by practitioners was adequately represented.
4.
32 workshops with teachers (a) sorting descriptors into categories they purported
to describe; (b) making qualitative judgements about clarity, accuracy and
relevance of the description; (c) sorting descriptors into bands of proficiency.
Quantitative phase:
5.
Teacher assessment of representative learners at the end of a school year using an
overlapping series of questionnaires made up of the descriptors found by teachers
in the workshops to be the clearest, most focused and most relevant. In the first
year a series of 7 questionnaires each made up of 50 descriptors was used to cover
the range of proficiency from learners with 80 hours English to advanced speakers.
6.
In the second year a different series of five questionnaires was used. The two
surveys were linked by the fact that descriptors for spoken interaction were reused
in the second year. Learners were assessed for each descriptor on a 0–4 scale
describing the relation to performance conditions under which they could be
expected to perform as described in the descriptor. The way the descriptors were
interpreted by teachers was analysed using the Rasch rating scale model. This
analysis had two aims:
(a)
to mathematically scale a ‘difficulty value’ for each descriptor. 
(b)
to identify statistically significant variation in the interpretation of the