Warrants. The “warrants” logically connect the observed data to
the claims. The need for elaborate warrants varies with the strength
of the link between the data and the claims to be made. Sometimes,
the link between the data and the claim is so clear that the warrant
becomes self-evident and requires little explanation. For example, it
is very straightforward to explain the logical link between a road test
that requires a test taker to demonstrate the ability to perform
typical driving tasks safely, and the claim that a test taker can operate
a motor vehicle without endangering the public.
At the opposite extreme, consider an IQ test. The highest level
claim is about the test taker’s level of “intelligence”. The tested
behaviors, however, include recalling strings of numbers, completing
puzzles, defining words, etc. A strong warrant is needed to explain
why the ability to recall strings of random numbers allows inferences
about a test taker’s intelligence. When there is a great difference
between the observed behavior and the claim that is to be made, the
warrant must be comprehensive and convincing.
Accompanying many warrants are potential “alternative
explanations” that must be examined and excluded to help ensure
that the logical link between the behavior and the claim described in
the warrant is correct. For example, excessively hard language in a
task may cause English language learners who possess the tested
KSAs to respond incorrectly to the task. An important part of the test
developer’s job is to reduce the likelihood that alternative
explanations for the warrants are correct.