The general mental skills which IQ tests cover are the skills involved in problem solving which rely on as little learned knowledge/behaviour as possible, the so-called culture free tests. These skills are those needed to deal with the unfamiliar, unfamiliar in both the sense of being pedantically novel and in the sense of being absolutely novel.
Pedantic novelty is where a problem is truly novel, that is, it has never been encountered before in this exact form but similar problems have been encountered. An example would be using a key you have never seen before to undo a lock you have never seen.
Absolute novelty is where the problem is something completely beyond the experience of the person. The individual has not encountered the exact problem in the pedantic sense of it being an identical problem and they have no similar experience from which they can extrapolate a rational solution to the problem they are encountering. An example would be a Briton going on safari in Africa who has no experience of Africa and suddenly meeting a lion six feet away.
Modern IQ tests are designed as far as is possible to be absolutely novel. It is true that a person will encounter similar general types of problems if they take more than one supervised test or rehearse IQ questions. But a general type of IQ problem does not provide a similarity sufficient to allow extrapolation to solve a particular problem of the type. For example, in tackling IQ problems which involve spotting the odd shape or pattern in a sequence the solver knows what they have to do in terms of what result is required, namely, identify the odd man out, but that knowledge gives no hint as how they are to find the odd man out.
An IQ test is different from any other test because every other type of test either permits subjective judgments or is deliberately knowledge dependent to a greater or lesser degree. Even other psychometric tests seek the person’s opinion of their feelings or perception of some physical event, there being no right or wrong answers. Exams dealing with specific subjects allow candidates to score reasonably well without necessarily engaging in any high level intellectual activity – even someone taking a maths course can get a fair way simply by mastering functions which can be applied mechanically (the more advanced the course the less this tactic can be employed). An IQ test is different because ideally there is no knowledge of inert facts, such as historical dates, speculation, matters of opinions or mechanical functions such as the rules of arithmetic which can be applied blindly.
Another way of looking at the demands of IQ tests is to consider the long term effects of knowing both what a particular set of test questions were and the correct answers to those questions. Suppose a group is first given an IQ test and afterwards provided with the correct answers, ostensibly to mark their own tests. They are allowed ample time to study the answers but are not able to retain them in any written form. Afterwards they are told that they will be called back for further tests in one year’s time. When they are re-tested the subjects are asked to re-sit the IQ questions they first tackled a year before but without any indication that it was the same test. Even if the subjects realised that they were taking the same test again, which not all would do, it is improbable that many would be able to recall the answers from twelve months previously. The only way to answer correctly with certainty would be re-solve the problems.
Those who claim that IQ tests only test how good people are at taking IQ tests are saying nothing useful, for any test of whatever nature on any subject by definition only tests how good people are at taking the test. The important question to answer is whether what is tested is of value. IQ tests would appear to have value because there is a strong correlation between IQ scores and life outcomes. This persistent correlation can be explained in two ways: either it is merely the most colossal and continuing set of coincidences or IQ tests are measuring abilities which are either applicable to life in general or are abilities which correlate strongly with other abilities which are generally applicable to human life .