If (and it's a big if) this thesis (pay-walled study) is correct, then proponents of cognitive load theory have a lot of rethinking to do. The suggestion is that while brains do indeed store short-term and long-term memory, they store these using two separate processes. So a memory doesn't have to be squeezed through short-term memory before it becomes a long-term memory. This makes a lot of sense to me - people like Romeo Dallaire talk about detailed complex traumatic memories of wartime where the entire experience stored and plays back over and over, brushing by the limits of cognitive overload as if they didn't even exist. "Post-traumatic stress disorder hard-wires events in your brain to the extent they will come back in digitally clear detail to your brain. You don't actually remember them. You relive them."