Table 7 Operators for articulating scores for holistic text image for each dimension

Strategies aimed within one dimension

1. Setting baselines

Assigning a particular and personal meaning to a score, used as a benchmark

I set 4 as the standard representing pass level, the compositions better than 4 can be given 5 or 6.

18

2. Classifying samples into level groups

Placing the scores into a broader scale before considering the more granular TEM8 rating scale

I firstly classified the compositions into three broad bands as ‘high quality group’, ‘medium quality group’, and ‘low quality group’, respectively, then decided on concrete score within each broad bands.

12

3. Avoiding extreme scores

Avoiding controversy and complaints by not scoring too high or too low

I tended not to give too high score since it was risky, especially on Language use dimension, and it is impossible for EFL students to obtain full score in this dimension.

8

Strategies aimed between dimensions

4. Differentiating between dimensions

Disentangling the unevenness of writing quality in different dimensions

There is very big problem on the structure of this composition, but its language has no serious problem.

29

5. Simplifying

Reducing the burden of assessing a particular dimension by considering its relationship to other dimensions

Since there are only 3 points for organization dimension, for this dimension, I just look at some macro-level feature. Some problems in micro-organization will be put into content organization for subtracting scores.

13

6. Balancing

Reallocating scores between dimensions, taking into account their distinctiveness and the discrepancy in scores

Since I subtract too many scores on language use dimension for it, I would take off the factor of language when scoring the content dimension.

11