Suppose a person was a lower level genius say about 140 IQ with a normal 4 year college education. Now you took that same person and he dropped out of school in the eigth grade and did virtually no studying while he was in school. How would he expect to score on an IQ test?
IQ test result really shouldn’t go up with more schooling. Here’s an article from Wired about IQ tests and researching the genetic components related to IQ:
So theoretically that person should get the same result from their IQ test whether they go to school for 4 years, or if they drop out of school and work a manual labor job and do no more studying. I might guess that the version of this person where they do have a college education will do better on the IQ test, because their mind has probably stayed more active than the manual labor version, but I wouldn’t think there would be a huge difference in score.
If you repeatedly take IQ tests you start to score a bit better, so I’d say there would be an effect similar to that for the person who went to school and had regular (non-IQ) tests.
I’m a Masters level psychologist working on my PhD; while neurocognitive assessment isn’t my area of expertise, I do have a fair amount of background in this area.
Like Sam Lowry said, differences in IQ would likely be quite minimal, but there are a few areas where some subtle changes might be noted. To describe the Wechsler Adult Intelligence Scale in a little more detail: there are four indices, each of which taps a unique facet of intellectual functioning, which are ‘added up’ to yield one’s full scale IQ score. The four indices consist of verbal comprehension, perceptual reasoning, working memory, and processing speed. In turn, each index has 2 to 3 sub-tests which assess for separate but related skill sets within the broader domain.
I would suspect that barring very extreme circumstances, working memory and processing speed scores would remain relatively unchanged. These get at how much info can you store in working memory at a time, and how quickly you can scan an array and make selections accurately, respectively. High school and college isn’t going to give you a special edge here.
Perceptual reasoning might increase a very small amount with exposure to more advanced mathematics courses. Specifically, one of the subtests involves solving increasingly complex word problems in your head (it’s pretty terrifying). At the extreme end, some of the computations might be completely unfamiliar to someone with only an eighth-grade-level exposure to math. However, the other two subtests which comprise this index would be virtually unaffected as they assess for visual abstract processing and problem solving ability and do not require familiarity with any formal concepts. All told, this might result in getting another 2 or 3 questions right on one subscale, which would translate to maybe an extra point on the full-scale IQ.
The one area where I would expect to see a mild but potentially notable difference would be on the verbal comprehension index. This one is comprised of three subtests: the first is essentially a sophisticated vocabulary test; the second pulls for factual knowledge acquired from one’s culture (it’s kind of a glorified trivia test); the third requires the test taker to articulate how two increasingly complex concepts are similar to one another (e.g, how are ‘acceptance’ and ‘rejection’ alike?) Having more exposure to history and literature would give someone a slight edge on these, particularly the first two. However, just because the hypothetical person dropped out of school at 8th grade automatically means he or she never reads a book or watches a documentary again…so that same info could be obtained in different ways. If I had to throw out a guess, though, I’d say this person might get another 2-3 questions right on each of these subtests as a result of the higher educational attainment…which would probably result in an extra point or two on each subscale’s standard score. In terms of overall IQ, you’d be talking about a 134-140.
All told, you’re talking about a range of 7 points lower to exactly the same score in my estimation. Might keep you out of the ‘genius’ designation, but nothing very dramatic.
What would be much more susceptible to change as a result of the situation described in the OP would be Achievement (as opposed to Intelligence Testing) scores. A measure such as the Wide Range Achievement Test (WRAT-4) or the Wechsler Individual Achievement Test, would vary dramatically based on education, as that’s what it’s supposed to measure. The WRAT, for example, has four subscales which assess for mastery of mathematics, spelling, reading, and sentence comprehension. Without those additional years of formal education, there are several areas that one simply wouldn’t know how to approach at all (especially the math portion) or would be much less likely to do well on (spelling). Unfortunately, the WRAT (and most of the other tests that I’m familiar with) “max out” at a certain point and aren’t much good at detecting achievement past the 12th grade mark. You’d definitely see a difference with an 8th grade drop-out compared to a college grad, but probably wouldn’t with a high school grad versus a 4-year college degree grad (or if you did, it would be extremely small) due to this ceiling effect.
As far as ‘practice effects,’ the IQ test is not much like the kind of exams given in school at all, and additional years of formal education wouldn’t give you an edge. Now, if you’d never held a pencil before or sat behind a desk, sure, but barring extreme cultural differences, not so much. That said, IQ scores can become inflated if you use the same measure repeatedly in too short a time span (I believe it’s a minimum of six months’ wait between re-administrations of the same test). This has little to do with getting “better” at taking IQ tests per se, but is more about being able to recall the specific items on some of the subscales as the test itself doesn’t change—it’s the same exact questions each and every time—until a new version is published (often several years or even a decade or more apart). In situations where you would need to do a re-test, (like, the first assessment was poorly done) you’d simply use another IQ test…like, instead of the WAIS you could whip out the ol’ Woodcock-Johnson.*
*Yes, that’s really what the “other” leading IQ test is called.
Very interesting answer, Orange Skinner!
The only reason I know what turpentine is made from – tree (usually pine) resin – is because I was asked this as part of an IQ test as a kid and didn’t know the answer. I looked it up afterward and have never forgotten. This bit of knowledge has yet to do me any good, but if I ever retake that same IQ test for some reason, and if the question hasn’t been replaced in the past 20+ years, then I’m ready for it. 
You probably know this, but in theory, it shouldn’t matter at all, since IQ is supposed to be (by definition) immutable and unrelated to academic achievement. In reality, the effect you see is an artifact of the fact that our testing systems are not perfect and that one can actually “study” for an IQ test to some extent if one knows the kinds of material that are likely to be on it.
This is one of the reasons. In an ideal testing world, the question on turpentine would be rotated with dozens, perhaps hundreds, of alternate questions that tested for the same underlying IQ element. For example, instead of being asked about turpentine, the person could be asked about gasoline, gum, or linoleum.
My info s 40 years out of date, but:
I remember getting two IQ tests in school: one at age 8 and one at 13.
The thinking at that time was that, since they relied on the ability to pick up meanings from the culture, they really could not be used with adults, as even the slowest will eventually pick up the points - one question I specifically recall from age 8: 4 or 5 drawings of a small row boat and a striped pole. The boat was pointing toward the pole, away from it, along side, etc. The Q was "Which boat is moored (or maybe ‘tied’) to the pole?
For an 8 yr old, knowing that was meaningful; for an adult, somewhere along the line, you pick it up.
I vaguely remember the Draft Board’s IQ test (we weren’t supposed to know it was an IQ test). One series of questions showed a cube flattened (various patterns on each of the 6 faces) and the assembled cube. We were to select the flattened drawing which corresponded to the 3-dimensional cube.
(I was a draftsman and could not get this right to save my life - heh)
So - how confident are we that these new tests can assess adult intelligence?
I was given the Lorsch-Thorndike at age 13. That gives you my starting point.
My second year psychology professor always maintained that IQ is an excellent measure of how well you can do IQ tests. It’s very difficult to design an IQ test that isn’t biased in some way.