1) The inter-aural time difference. Sound takes time to travel, so any sound coming from left of the head will arrive at the left ear sooner than the right ear.
2) The head-related transfer function. The shape and auditory characteristics of the head and ear canal change the sound slightly depending on the frequency and the direction of the incoming sound. Since this direction is different for the two ears, your auditory system can compare the 'shape' of the sound it gets from each ear and use that to position the sound in space.
The inter-aural time difference is, by far, the more effective of the two, but it only gives horizontal bearing. It also can't give 'in front' or 'behind'. The head-related transfer function can give both horizontal and vertical bearing, but it's not that accurate. Together, these explain why humans are far more accurate at localizing sounds horizontally than vertically.
Binaural recordings try to capture both of these mechanisms through stereo recording with a mock head. That's why listening to a binaural recording gives you the feeling of sounds coming from a point in space, because it's replicating the same experience you'd get from actually being in that space.