We are pleased to publish a report on the performance of the Checkpoint listening and reading assessments. In summary:
Analyses conducted in 2015:
Estimated Checkpoint's reliability to be between 90.2 and 100% with a standard error of measurement of between 1.7 and 3.2%;
Showed that it is impossible to make accurate judgements about listening ability or reading ability on the basis of a spoken performance; separate measurement of listening, reading and speaking skills is essential; and
Showed Checkpoint's listening and reading items to be discriminating well between candidates of low, medium and high ability.
Rasch analyses conducted in 2024:
Showed that Checkpoint's listening and reading items:
Cover a range of difficulty which is well-suited to the ability of the candidate population; and
Measure effectively with many items close to or at values ideal for measurement;
Resulted in revision of four of Checkpoint's 96 listening and reading items to further enhance their measurement properties.
You can download the report here.