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

No 5.
Key concepts: performances: Descriptors are matched to typical performances at
those band levels to ensure a coherence between what was described and what
occurred. Some of the Cambridge examination guides take teachers through
this process, comparing wordings on scales to grades awarded to particular
scripts. The IELTS (International English Language Testing System) descriptors
were developed by asking groups of experienced raters to identify ‘key sample
scripts’ for each level, and then agree the ‘key features’ of each script. Features
felt to be characteristic of different levels are then identified in discussion and
incorporated in the descriptors (Alderson 1991; Shohamy et al. 1992).
No 6.
Primary trait: Performances (usually written) are sorted by individual
informants into rank order. A common rank order is then negotiated. The
principle on which the scripts have actually been sorted is then identified and
described at each level – taking care to highlight features salient at a
particular level. What has been described is the trait (feature, construct) which
determines the rank order (Mullis 1980). A common variant is to sort into a
certain number of piles, rather than into rank order. There is also an
interesting multi-dimensional variant on the classic approach. In this version,
one first determines through the identification of key features (No 5 above)
what the most significant traits are. Then one sorts the samples into order for
each trait separately. Thus at the end one has an analytic, multiple trait scale
rather than a holistic, primary trait one. 
No 7.
Binary decisions: Another variant of the primary trait method is to first sort
representative samples into piles by levels. Then in a discussion focusing on
the boundaries between levels, one identifies key features (as in No 5 above).
Appendix A: developing proficiency descriptors
209

---

However, the feature concerned is then formulated as a short criterion
question with a Yes/No answer. A tree of binary choices is thus built up. This
offers the assessor an algorithm of decisions to follow (Upshur and Turner
1995).