The problem with Zipf’s law in language is there are dozens and dozens of different explanations for it, each plausible and well-motivated. For any particular explanation to be convincing, it will have to make novel testable predictions, something which is rarely done.