According to Felix Geisendorfer's newest post on the Debuggable blog, he thinks that "private and protected methods and properties are one of the most stupid concepts of OOP."

This is a thought I first shared at CakeFest Orlando this year, but could not explain properly at the time.

He illustrates with an example of a protected "balance" variable in a BankAccount class. Sure, it's marked as private but less skilled programmers might not use it that way. He recommends a method without the getters/setters to help make the usage of the variable a bit simpler. He also suggests that using protected/private scoping helps to promote "crappy code" - using them to provide a sort of protection for code that you either don't want getting used or hiding it away so the API can't get at it.