Neuropsychology · Simplified
Where Recognition Breaks
"I know the face but I cannot place them" and "I have no idea who that is" are different failures at different steps.
The model
Five steps, five questions
Recognising a person runs through several steps in order, with reading their expression branching off early. Break one step and everything that depends on it fails, while everything before it and everything off the branch carries on.
Key terms
- Face perception
- Working out the structure of the face in front of you, before any question of who it is.
- Familiarity
- A sense that this face is one you have seen before, with nothing attached to it yet.
- Identity
- What you know about the person: their job, where you met them, what they do.
- Prosopagnosia
- An inability to recognise faces that is not explained by poor vision or by general memory problems.
The steps run left to right, with expression branching off after the first one. A step drawn with a broken outline and a cross is the impaired one. Everything downstream of it fails; the branch does not.
What this person can still do
What this shows
The profile locates the break
Five different impairments, five different profiles, and none of them is simply "worse at faces".
Key idea: Because the steps run in order, an impairment produces a signature: everything downstream fails and everything upstream survives. Someone who cannot tell two photographs of a stranger apart has a problem at the first step, and it will affect everything. Someone who finds every face familiar but can say nothing about anybody has an intact first two steps and a break after them. Someone who knows exactly who the person is and cannot produce the name has only the last step gone, which is the mildest version and also the one everybody experiences occasionally. And because reading expression branches off early, it can survive an impairment that destroys identification entirely, which is the observation the branching was proposed to explain.
A box-and-arrow model of this kind is a summary of which abilities come apart, not a map of anything in the brain, and drawing it as a neat sequence overstates how orderly the process is: there is feedback between the stages, familiarity and identity are not cleanly separable in practice, and real cases are usually partial rather than all or nothing. Developmental prosopagnosia in particular does not fit a single-step story well. The model earns its place by predicting which profiles should exist and which should not, and it is those predictions rather than the boxes that are the science. Nothing here describes a real person.
The longer version adds a matched general-visual control and a parallel voice-recognition route, then runs the inference backwards from fictional profiles to every component compatible with each. It is at Face Recognition Detective in the main collection.