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 5-6, he describes taking the Vocabulary subtest:

…If you have spent your professional life as an editor of a first-rate magazine, you ought to have a pretty good vocabulary, and the WAIS confirms that I do.  The drill in this subtest is as follows.  Your examiner reads off a list of words–thirty-five in my case–and asks you to define them.  If you give a superior answer, going to the heart of the meaning, you get two points; if your answer seems to show only some general understanding of the context in which the word might be used, you get one point; if the answer totally misses the point, you get a goose egg.  I got two points on thirty-three of the definitions and one point twice.  When [examiner] Stern and I sat down after the test and went over the results,  I found myself sadly agreeing that the two one-pointers had in fact reflected somewhat wobbly answers.

Based on the above we can infer that Seligman got a raw score of 68/70 on the Vocabulary subtest.  According to the WAIS-R manual, this equates to a scaled score of 17 in the peak age group (20-34) and also 17 in Seligman’s age group (55-64), equivalent to an IQ of 135 (U.S. norms;  134 U.S. white norms) on this one subtest.

[Update Nov 20/2015: because WAIS-R norms were a decade old when Seligman was tested, and the Flynn effect increased WAIS Vocabulary scaled scores by 0.35 points a decade from 1978 to 1995 (Flynn, 2012), his scaled scores must be reduced to 16.65 (IQ 133, U.S. norms, IQ 132 U.S. white norms).]

On page 6 Seligman writes:

Many people do not quite see why vocabulary should be tested in an exercise that is supposed to be measuring mental ability.  Their objection:  that vocabulary mainly reflects acquired knowledge rather than the ability to learn.  In fact, vocabulary is a pretty good proxy for overall IQ:  if a professional tester had to make do with just one of the subtests, he would probably land on Vocabulary.  The reason it correlates so powerfully with IQ is that you build a vocabulary in a process that requires a lot of reasoning.  In your reading and listening, you are endlessly making inferences about different shades of meaning and the different contexts in which words are used.  A somewhat similar point might be made about the Information subtest (which many people also view as unrelated to intelligence).  You acquire a fund of information not by absorbing data in isolation but by noting the connections between different data.  Both Vocabulary and Information correlate about 0.80 with overall IQ…