Features of a levels-based mark scheme and their effect on marking reliability
Levels-based mark schemes are commonly used in the marking of extended response items but, between specifications, there is little commonality in their design, formulation and application.
This study firstly establishes a list of the variable design characteristics between levels-based mark schemes and then, secondly, analyses marking reliability with reference to these characteristics.
It finds that most of the variation in marking reliability is due to the vagaries of individual responses, which a holistic approach to item design might mitigate. It also recommends a number of small adjustments to mark scheme design which might improve marking reliability and increase the transferability of skills between the marking of different items, units and specifications.