To some degree, the IQ tests are valid as indicators of a person's ability to handle new problems, new situations, obscure activities. However, I have seen complaints that many of the IQ tests are biased, either culturally or ethnically or even with respect to gender. The tests depend heavily on the ability of the test subject to read and understand the questions and problems. To the degree that such difficulty is linguistic, the application of the test becomes flawed.
Here is my analogy. If I shoot a shotgun at a large, distant sheet of graph paper and from that, determine coordinates on an X-Y plane, I can then apply linear regression techniques to determine a straight-line best fit for a line represented by the scatter pattern. I can do this - but the line I get from doing it has no meaning. The error is in the assumption that this mathematical treatment applies to the particular experiment. The more correct approach would be a radial frequency determination to see how quickly the shot pattern spreads.
If the IQ test you want to give me implicitly includes the assumption that I can understand it, and if that assumption is wrong for some reason, then it is going to give crap results. And therein lies the difficulty in IQ testing. What it is meant to test can be quite real - but can the test actually find that which it seeks?