I think this article is _closer_ to right than a lot of writing on programming paradigms is, but still misses a central idea.

Programming paradigms can't be defined intensionally, by listing their features, nor extensionally, by listing programming languages that fit each paradigm. Either way results in unsatisfactory definitions from which reasonable people will identify errors of excluded languages.

This is because programming paradigms aren't sets, they're cognitive categories. Programming paradigms are subjective mappings of conceptual framings of program structure onto specific languages. A language fits a paradigm if that mapping is subjectively natural.

Of course, people tend not to like subjective definitions, but this is more accurate.