After all, while psychometricians definitely have above-average IQs, they probably don't have IQs over, say, 160.

If psychometricians already know what the right answers are (and how to obtain the right answer), then wouldn't they score very high on the tests, even if they're not THAT smart? (making it possible for other people to score high on the tests if they studied psychometrics as well?) The format of the test isn't going to vary that much from administration to administration. Of course, the tests are timed, but it's entirely possible that the practice the psychometricians get from writing the tests could easily make them better at each of the subsets of an IQ test.

  • 2
    $\begingroup$ It's important to note that these tests have evolved generally over decades with through research and dozens of people constructing the test. A single person couldn't easily construct such a test but that's not the situation. $\endgroup$
    – Ben Brocka
    Jan 21 '12 at 15:42

Short answer: construct a usable chair out of toothpicks. Now, I certainly can't do this, but I can verify your answer easily enough

Long answer: Verification of answers is often a lot easier than producing answers. And thats not just anecdotal, there are computational and psychological reasons behind this fact.

Computationally - complexity is different for solving a problem than for verifying it. For example the clad of NP problem cant (probably) be solved in poly time, but they can be verified in poly time.

Psychologically - think recall versus recognition. Recognition just involves your brain lighting up when it sees the right answer, whilst recall or production requires an extensive search at the very least

So the answer is that the test makers don't have produce the answers themselves, and certainly not under test conditions. They simply have to be able to see that the right answer is right (either produced via a group, a computer, or a genius).

Also, remember that no one question separates high IQs from low. Perhaps a 140IQ could answer one of the hardest questions, but not all of them.

Lastly, remember than IQ isn't a single dimension, perhaps the test was written by one expert in each dimension, which could then identify a genius in all dimensions

  • $\begingroup$ +1 I have a feeling that there's something still missing here. The question is framed in a far more obvious way, I wonder if we cannot offer a more straightforward single point answer. There must be one, I still feel. $\endgroup$
    – Kris
    Jan 19 '12 at 10:47
  • $\begingroup$ Nice answer. But actually, IQ tests are a poor in measuring extremely high intelligence. Even if you assign a number to a person I doubt if gives any worthy information. Perhaps it is about lack of comparison, increasing number of dimensions or psychometricians being not able to make a task conceptually way beyond them. $\endgroup$ Jan 21 '12 at 10:32
  • 2
    $\begingroup$ Piotr's answer is not complete either. Intelligence tests are poor at measuring any range of intelligences they are not intended to measure. Most intelligence tests have items that discriminate among people from -1 SD to +2 SD (maybe +3). If someone falls outside that range, it won't measure them accurately. But that doesn't mean they can't. An intelligence test could certainly be developed to discriminate among people from +2 SD to +5 SD. It's just that there are relatively few people high enough to justify extensive effort at doing so (at least, for most test designers). $\endgroup$
    – richard
    Jan 21 '12 at 17:29

In a comment above, Kris hopes for a more obvious answer to the question than the [very good] answers zergylord gave. So here's one more explanation that strikes me as almost too obvious to mention: test making and test taking are temporally asymmetric. One can spend fifteen minutes creating a question that must be answered in less time. This kind of time differential is extremely diagnostic -- a smart person and a dumb person might both be able to solve the missionaries and cannibals problem, but the smart person does so much more quickly.

  • 2
    $\begingroup$ This is an excellent point. Also keep in mind that many of these tests are developed by lots of smart people working in teams. 20 intelligent people with unlimited time working to develop items can develop items that would make a single extremely intelligent person struggle with limited time. $\endgroup$
    – richard
    Jan 21 '12 at 17:32
  • $\begingroup$ but it's worth remembering that human effort (unlike, say, computer processor power) rarely scales according to any predictable algorithm. so just because 20 people are working on a problem doesn't mean they'll come up with a solution 20x better than a single person working on that problem. $\endgroup$
    – Krysta
    Aug 29 '13 at 20:34

Above answers are great, I just wanted to add that you do not need a 'smart' person to create IQ test... you or me could do it... Any set of questions would measure one or more dimensions of intelligence...

You would give your set of questions to many people and their results would show up as Gaussian. You would then assign 100 IQ point at the peak and subdivide the Gaussian into parts by standard deviation where each segment would be your 10 IQ point increment...

enter image description here

P.S Don't get me wrong chances are it wouldn't be the best IQ test but test nevertheless...


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.