I think some of the confusion might be alleviated with the understanding of how a
good IQ test (if there is such a thing) is developed. As Evo and others have pointed out, those Internet IQ tests are meaningless for reasons I hope will become clear momentarily.
In general, an IQ test
can contain a list of any questions:
any type of questions. But not all types of questions are created equal. So when designing a
good IQ test the creator should make the attempt to use the following guidelines:
- Questions should be unbiased as possible, not giving the advantage to anybody of a particular background, race, ethnicity, gender, etc.
- Questions should test problem solving skills, not knowledge, since knowledge is inherently biased.
Although not absolutely necessary to be considered an IQ test, the following guidelines are also usually followed under the premise (which is debatable) that a person with higher intelligence can do simple problems faster than someone with lower intelligence:
- Questions must be simple.
- The test must be administered using a strict time limit (this reason alone invalidates the usefulness of those online IQ tests).
- The test must contain more questions than anybody could possibly answer within the time limit. The test must have so many questions that it is impossible to "ace" the test (anything short of that means something is wrong with the IQ test).
- The test must be designed such that there is no advantage to guessing. Usually this is accomplished by making the test multiple choice. Testees are penalized an appropriate amount for each wrong answer and awarded for each correct answer. For example, if each question contains four choices, 3 points are awarded to the pre-score for each correct answer and 1 point is subtracted for each incorrect answer. That way if the testee randomly guesses on every single question, the resulting pre-score should be near zero.
The next step involves the control group. The control group should be comprised of as many people as possible (the more the better) and should be a statistical representation of the population at large. That's where things get tricky. How do you assemble a large group of people that maintains a statistical representation of the general population? I don't know, but that's the goal anyway. Then each person in the control group is administered the test in a controlled environment (with the specified, strict, time limit). The pre-scores of all the testees are then tallied and recorded.
The final step involves creating a mapping function. The mean pre-score in the control group data maps to an IQ score of 100. Simple as that. The standard deviation of the pre-scores data of the control group is calculated. This pre-score standard deviation
maps to an IQ standard deviation of 15. Using this mean and standard deviation mapping, a Gaussian normal curve is created to map pre-scores to IQ scores. The end result is a look-up-table. The index of the look up table is the pre-score, and the corresponding table element value is the IQ score. That completes the process of creating an IQ test.
When the test is then administered to someone, it must be administered in same controlled environment setting as the control group (and with the same time limit). This is another reason that invalidates the usefulness of those online IQ tests. When completed, the testee's pre-score is tallied and then mapped to to the IQ score using the look-up-table.
By the way, on a separate but related note, it is a fallacy to say "my IQ is...," or "so and so has an IQ of..." People do not
have and IQ. An IQ is not something that actually exists that people can have. It is better to say, "I received this score on an IQ test," or "so and so scored this when taking an IQ test."
