Memory records need to be structured for rapid access, be malleable and non brittle. They need to conform more to the metaphor of imagery than numbers. A single digit error can easily destroy a typical data object, whereas with upwards of 90% image loss or error; effective object recognition can still remain.This is unprecedented in traditional database structures In fact, the human brain can actually repair faulty data sets to re-create the original.

Consider the recognition of poor handwriting. Or a low resolution, black and white photocopy of a face, which can be recreated at high resolution by a skilled artist This is done by instantiating the degraded 2D bitmap image to a high resolution 3D generic human head simulation - constructed in the artists own mind. The black dots on the photocopy face are instantiated as crystal clear human eyes from simple reality exposure precedent from the artists past - i.e. seeing actual human beings, and thus training his 3D model representations from reality.

Unique photocopy-hinted features are then morphed onto the generic architypal form, which is subsequently used as a reference for the artists hands to lay down a new render plane, that is, to paint a perfect reconstructed image There is an important question regarding the structure of memory traces stored by the human brain. Is there any chance the records exist primarily as 2D renders planes direct from the modalities - time linked to describe a behavior chain - similar to the way film frames can hold an animation scene.

One proposed method to test these theories is to generate simple, but unfamiliar 3D models. Present them to volunteer subjects as manipulatable 3D forms within a web interface. One set of subjects will be restricted from seeing a particular perspective, but which can be inferred from the remaining information. The other group has access to the complete model Both groups are then tested to see the speed and ability to recognise the missing perspective as part of the model.