In norm referenced testing, the percentage of the test cohort (ie the people taking the exam) who will pass or get certain grades is decided in advance. Eg if I have a class of 20 learners I might decide in advance that, when the test is marked, the top 25% (ie the 5 learners with the highest marks) will get Grade A, the second 25% Grade B, the third 25% grade C, and the final 25% will fail. This means that if in June I give the test to a very good group, someone with 60/100 marks may well fail - because 15 out of the 20 learners got higher marks than this. However, a learner belonging to a much weaker cohort, which takes the test in December, might gain Grade A with 60/100 - because s/he was in the top 5%.
This unfair discrepancy is avoided by criterion referenced testing, which measures learner performance against a set of predetermined factors - ie descriptions of what learners are expected to know and/or be able to do at a specific stage of their learning.
For example, some simple criteria for a writing test at B2 level might include eg :
- Divides the essay into logically ordered paragraphs, each focusing on a single topic (or two shorter but logically connected topics)
- Uses a topic sentence to alert the reader to the topic of the paragraph.
- Uses connective expressions accurately and appropriately to show the relationship between ideas
Criterion referenced testing means that all learners who meet these standards will pass - regardless of whether this is 100% of the cohort, 0% or anywhere in between. My learners in June and December who score 60/100 will get the same grade, regardless of whether the test is taken in June or December. Learners are judged by what they can do - not by comparison with others in the cohort.