Give a very simple programmatic task, a paper (or a blackboard) and a pencil. As you watch the interviewee elaborate his solution, and as you discuss it with them, you'll immediately see if programs are a deeply foreign object to their mind, or something they're fluent with.
The point is not syntax, forgotten semicolons, or even knowledge of libraries. Don't bother with design patterns either, the kind of tasks we're talking about is too simplistic for them to be meaningful: just watch whether translating a task into program snippets comes naturally to them. If they can do that, they can learn, and probably have learnt a great bit already.
What works well is some form of "tell me about a time when ...". The candidate gets to tell a story of their choosing, but the discussion is long enough that they can't keep track of their lies...
One thing with science on the side of truth is that the simplest questions (such as "what's the difference between a boolean and an integer?") can have extremely complicated and deep answers that lead to crazy discussions about formal definitions or implementations on different platforms and etc. And these are the kinds of things you can only learn by actually learning them, and trying to bullshit their way through an answer will quickly paint the faker in a bad place.