Neuropsychology · Simplified
How Sure Can One Case Make You?
A single case with a clear lesion and a clear deficit. The inference feels much stronger than it is.
The case
One patient, one lesion, one deficit
Mrs D, 61. Two weeks after a stroke she speaks in short effortful phrases, leaves out small grammatical words, and takes a long time to find nouns. She understands most of what is said to her. Her scan is reported as damage to the left inferior frontal region.
Here is the claim on the table:
"The left inferior frontal region is where fluent speech production is carried out."
The solid line is what you said. The shaded band above the dashed line is confidence the evidence does not support. Each complication lowers the dashed line; nothing you learn raises it.
What it can carry
Necessary, not responsible
Key idea: What a lesion case can establish is that some of the damaged tissue was necessary for the task, in this person, at this moment. It cannot establish what that tissue does, because the label names the centre of a lesion rather than its extent, because strokes take the shape of the vessel that failed rather than of a function, and because some of the affected tissue is not damaged at all but merely swollen or temporarily quiet. Every one of those is a reason the deficit might belong to tissue other than the tissue named. A claim about one person is also not a claim about a region: to say anything about the region you need many people, and the vascular problem does not average away as that sample grows, because the same regions keep being damaged together.
The ceiling drawn on the chart is a teaching device, not a calculation. There is no formula that turns three complications into a number, and a different neuropsychologist would draw the line somewhere else. What is not arbitrary is the direction: every complication here removes support and none adds any, so the most the evidence can carry only ever falls. The case is fictional, the scan findings are invented, and nothing here is a diagnosis or a description of any real person.
The longer version adds three further complications, a claim-sorting exercise separating single-case from population claims, and a full trace of every rating. It is at Lesion-Symptom Inference Trap in the main collection.