In his excellent book A Question of Intelligence: The IQ Debate in America, former Fortune magazine editor Daniel Seligman describes what it’s like to take the WAIS-R IQ test. One page 8, he describes taking the Similarities subtest. According to the school board I attended as a kid, this test measures:
Verbal abstract reasoning
Ability to generalize, to make inferences and associations.
Reveals level of thinking: descriptive, functional, conceptual
Seligman writes:
The final subtest, Similarities, is a measure of abstract reasoning ability. It all seemed very easy. The examiner mentioned two nouns–like say, chicken and pigeon–and the testee responds by saying what they have in common. (In this made-up example, the answer would be that they’re both birds.) Nothing to it.
Unfortunately there’s no way to infer Seligman’s score on this subtest from the above paragraph so this test must be excluded when I calculate his IQ. It’s likely he did very well, but one can find the test very easy without getting full credit on all or even any of the items and partial credits make a big difference
Pumpkin Person 
Hey Pumpkin, I have a question about this subtest. Say for instance you get asked the similarity between [reference to specific test item redacted by pp, aug 22, 2018]
I can’t discuss specific items because it could compromise the integrity of the test
Pumpkin, do negative practice effects mean that the first attempt at the test was an overestimate? If not, then what do negative practice effects show?
Pumpkin, on the similarities subtest, is your abstract reasoning technically underestimated if you got a question wrong because of poor verbal expression?
Also, can the average person on the similarities test, if they understand the relationship between the words, explain it well enough to receive a point?