Lucid dreams can sometimes produce the illusion of reality. A Nightmare on Elm Street (1984), for example, was based on a series of articles about young men who experienced terrifying dreams, refused to go to sleep, and later died in their sleep. For them, the terror of their dreams became so real it killed them.
More recently, Inception (2010) used a similar setup (entering the dreams of others) in order to explore the possibility of an idea being implanted in one's subconsciousness while one was asleep and dreaming (the "inception" of the title). In this more focused investigation, it is argued that dreams within dreams ("deeper levels" of dreaming) provide access to deeper levels of the subconscious, and thus facilitate the planting of a new idea.
I've experienced dreams within dreams on several occasions, and there are two features of the phenomenon which Inception certainly got right: 1. the dream within a dream tends to be more surreal / bizarre / "unrealistic"; 2. after waking from it, there is a feeling of reality, one has just woken up, which may then be subverted by further dream weirdness. This is a feature which has been exploited to great effect in a variety of places in popular culture—not just the aforementioned movies, but also genre pieces such as Neil Gaiman's Sandman (e.g. the character trapped in an eternal sequence of nightmare wakings at the end of issue #1).
But there is a weakness with the dream within dream approach to inception—if one does succeed in convincing oneself that one is awake, then whatever is remembered of the levels of dreaming that had occurred is judged a part of dream land.
Compare this with another phenomenon: dreaming about dreams. Suppose for example, one experiences a rather vivid and lucid dream, which closely adheres to an actual event, changing it in relatively small respects. Then suppose, as sometime happens, after brief awakening, one dreams a more surreal and bizarre dream, a subconscious response to the incidents of the earlier one. When one awakes from the second dream, the certainty that one is in fact awake, and that the preceding events were in fact dreamed, is not evidence against the reality of the previous dream (in the way it is when dreams are embedded). In fact, if the second dream was sufficiently vivid, one can become uncertain about whether the incidents which inspired it were produced by reality or a previous dream . . .
Of course, the effect depends crucially on the plausibility of the first dream. The essential point here is just that arbitrarily plausible embedded dreams receive an evidential refutation which arbitrarily plausible dream-inspiring dreams do not. Suggestion for a future inception technique?