But languages such as Nadeshiko (なでしこ) http://nadesi.com/, a Japanese programming language, may be a challenge for non-Japanese speakers. Its source, especially one- or two-liners, just look like natural Japanese sentences. That means there's no explicit word boundary, and the parsing is based on the particles instead of word positions. (Longer program shows indentations and structures, makes it resembles to other proramming languages).
It's not a joke or experimental language. It aims at scripting for end users, and has pretty decent GUI/database/network bindings. Even a book was published.
It's not a a problem for 99% of programmers, so that's the reason most of languages stayed in English without the need of being translated
This isn't true for the API though - the API contains a much greater lexicon.