Assessment also needs to test students in a consistent way, so that the first student to be assessed is treated in the same way as the last. When marking 60 pieces of writing in the FL or conducting 60 oral exams, we need to be sure we are applying the same standards across all students, that fatigue or boredom are not distorting our perceptions, that we do not become stricter or more lenient as we work our way through the group of students, and that any marking scheme is adhered to closely throughout. That is, we must ensure our assessment is reliable.

The ways to achieve test reliability are:

• Sample widely: the more items included in a test, the more reliable it is likely to be.
• Restrict student choice: the more candidates can choose from a range of questions (eg in essay writing), the more varied the responses will be across the cohort (thus making marking reliability more difficult to achieve) and the greater the likelihood of a candidate's performance varying depending on when he/she takes the test.
• Write unambiguous items: this is best achieved by getting colleagues to check them or by pretesting them in some way.
• Provide clear, explicit instructions: again, scrutiny by colleagues at draft stage is the best way to avoid problems.
• Ensure students are familiar with the test format: students will not perform as well if the testing techniques are new to them. Past papers and practice questions are thus important for reliable testing.
• Provide uniform test conditions: test outcomes will differ unless conditions (timings, acoustics, support materials) are the same on each iteration of the test.
• Use as much objective marking as possible (see Module 14, section 14.2.3, on marking written work).
• For tasks requiring subjective judgement (eg orals, essays), use clear and explicit criteria.
• For more objective tasks (eg most listening and reading comprehensions), use a detailed marking scheme and ensure all markers adhere strictly to it.
• Train markers: markers should meet to discuss acceptable answers and sample essays, etc., at the start of the marking process; this is especially important where several markers are involved.
• Mark anonymously to avoid any unwitting bias. Studies suggest even having the name of an unknown student on a script can affect the marks awarded.
• Employ double marking, ideally of the blind variety (see 13.2.6).

(Adapted from Hughes, 1989: 36-43.)