For certain cognitive ability items, particularly older tests, there has been evidence of discriminatory impacts for particular groups. The items that lead to this effect generally are those that at least partially rely on learned or prior process knowledge or skill. A good example are verbal ability items, or numerical ability items, that rely on the test taker interpreting large chunks of text or complex graphs and tables of numerical data in unison, which generally lean on educational experiences rather than raw processing power. Traditional ICI tests are particularly guilty of this.
The items that produce this effect the least are abstract (or inductive) reasoning type items. They rely only on one's ability to take in information presented to them, process it, and solve the problem in front of them, without the benefit of prior knowledge (think pictures, patterns type problems). These items show the least difference across different groups, and are the best at tapping the overall fluid intelligence and general mental ability that we would be looking to capture, so we primarily utilise this type of item in our measuring.