Becker is saying here that if you see people change their behavior, it can be because either (1) they change their preferences (i.e., their utility function), or (2) they face changed prices or income. He argues that we get nowhere when we assume (1), especially because we can't see people's preferences. Instead, with (2), economists assume that people have stable preferences. He has a famous paper about this titled "De Gustibus Non Est Disputandum," which means "There is no disputing about tastes" in Latin. The economic approach assuming (2) is very powerful, and is able to generate many testable predictions by modeling behavior as a response to changes in prices or income. What do you think about this