A puzzle, however, usually sets up the question already for you. The work of saying what a word means, whether a sentence is even a statement, whether “all” and “some” and “only” and “if” have been used as they should be, those questions have all been answered by the puzzle setup.
The same is true of many “fallacy” lists. A name is given to an error in reasoning, and the student learns to put the name on a paragraph. But knowing a fallacy and how to avoid it and how to rectify it when we find it involves knowing how language works, how words get their meanings, and what sentences imply in their grammar. If you cannot say what the sentence was doing, the label of the fallacy is just a name for something we do not really understand.
Without attending properly to words and the art of grammar, all our reasoning that is trained over puzzles and logic games will rely on terms already set up, statements already taken for granted, and a purpose for the argument or statement. Then someone trained primarily in puzzles and logic games or detecting biases discovers that he does not know what was being said, how things are being defined, and so on; he only knows a few strategies for helping himself better find the truth and has sharpened his skills somewhat but not on real arguments and reasoning; he may also confuse contradiction with disagreement, and end up treating every set of opposing views as if the two people had said opposite things about the same subject in the same respect.
Inference without the grammar of statements is an incomplete tool. And this is part of why, even in modern logic books and courses, when a formal language is being developed by logicians for rigorous thinking or a programming language for precise instructions to computers, the grammar of the formal language or the programming language is defined first.