Recognizing faces is distinct from recognizing other objects. Whether a face is seen immediately in total (holistic model) or only after processing spatial and configural informations is still subject of discussion. However, it is well documented that it works very quick.
There are several face recognition models trying to dissect this complex procedure. Obviously, the face is the most informative part for recognition. This does not exclude the use of extraneous information to assist in facial recognition such as gait, body shape, clothing jewellery, clothing, scars or voice. Exactly these markers are preferentially used by prosopagnosics. This compensational behaviour is certainly a reason that inborn prosopagnosia is quite unknown.