They deal with pretty noisy mediums though. How much of the input is actually transformed into what we perceive as relevant knowledge and stored long-term?