Neuropsychology
Lesion-Symptom Inference Trap
One fictional case, one tempting inference, and six complications that arrive after you have already made it.
Simulated — an invented case, an invented scan, a schematic map rather than an atlas
Learning objective
By the end you should be able to say why a lesion is evidence about what a task needs, not a label for what a region does.
About 30 minutes. Nothing you do here is saved or sent anywhere.
Localisation is not the villain
This is not an argument that lesion evidence is worthless. Lesion studies are among the few methods that speak to necessity rather than mere involvement. The argument is narrower: the step from "damage here, difficulty there" to "this region does that" needs converging evidence.
- Make an initial localisation inference and commit to a confidence.
- Reveal the complications one at a time. After each, re-rate how confident you are in the general claim.
- Watch the gap between your line and the ceiling that the evidence can actually carry.
The referral
Person M is fictional. She is 68 and was admitted after a sudden onset of difficulty producing speech. Two weeks later her speech is effortful, short and produced with obvious effort; she understands conversation on the ward without apparent difficulty. A scan report describes damage centred on the left inferior frontal region.
The evidence room
The claim you are rating throughout is the general one: fluent speech output is localised to the left inferior frontal region.
Key terms
- Localisation
- Assigning a function to a place in the brain.
- Diaschisis
- Reduced functioning in undamaged tissue that is connected to the damaged area.
- Reorganisation
- Change elsewhere in the brain in the period after an injury, so that what you test months later is not fixed by the lesion alone.
Step 0 - the referral
Set your confidence in the general claim, then reveal what the notes did not say. Each complication changes what the evidence can carry, whether or not it changes your mind.
Your claim against the evidence
The referral, as written
How your confidence has moved
| Step | What was revealed | Your confidence | Evidence supports at most |
|---|
What is on the map, in words
| Mark | Feature | What it means |
|---|
The outline is a schematic drawn for teaching. It is not an atlas, the shapes are not to scale, and no real scan, image or template is reproduced.
Challenge - which claims survived?
All six complications are on the table. Sort these four claims by how well this one case supports them.
What this demonstrates
A lesion answers a necessity question
What damage can tell you is that a task could not be done without this tissue, in this person, on this occasion. That is a genuine and hard-won kind of evidence, and functional imaging on its own cannot supply it: activation shows involvement, not necessity. What a lesion cannot tell you is what the tissue was doing. Necessary is not the same as responsible, and "region X does Y" is a different claim from "Y needs X".
A one-word label describes a volume
Lesion reports name a region because a name is useful, but the damage is a volume that crosses whatever boundaries were drawn on it, and two cases with the same label can have very different lesions. The white matter underneath is damaged in almost every cortical stroke, and it is often what matters. Worse, the volume is not random: strokes take the shape of the vessel that failed, and one vascular territory supplies several functionally distinct regions together. That co-occurrence is systematic, so it does not average away as the sample grows. Voxel-based lesion-symptom mapping has to model it explicitly, and even then the resulting maps are biased towards the middle of common territories.
The deficit you measure is not the lesion alone
Early after an injury, swelling compresses structurally intact tissue, and regions connected to the damaged area show reduced activity without being damaged at all. Deficits measured in the first weeks therefore overstate what the lesion accounts for, and some of the recovery that follows is those effects resolving rather than anything reorganising. The other end of the measurement is just as loose. Premorbid level does not appear on any scan, so a score at the population average can be a substantial loss for one person and unremarkable for another. Education, language background, sensory function, mood, fatigue and effort all sit between a low score and any statement about tissue - and language organisation itself varies between people, more so in left-handers.
The dissociation can be a property of the questions
"Comprehension seems good" is a claim about ward conversation, where context, gesture and expectation do a great deal of the work. Tested with sentences whose meaning depends on word order, the same person can be well below the control range. Two tasks with the same name are not interchangeable evidence, and the cleanest dissociation on a case summary is sometimes an artefact of how one side of it was measured.
For teaching elsewhere: take this activity as one self-contained block of HTML, on the clipboard or as a file. Either way it is styled so that it will not disturb the page you put it into.