If someone close to you has had a stroke, you may have seen this: he knows exactly what he wants to say, his eyes are urgent, his hands are working — and only one or two words get out, a sentence broken into pieces that arrive one at a time.
Or you may have seen the opposite: fast, fluent speech, the intonation rising and falling in all the right places, and half a minute in you realize he is not saying anything at all.
From the outside both are "trouble talking," but what has broken is clearly not the same thing. Before the 1860s nobody could open a living skull to see where language lived. The only method was to wait until someone whose function had broken died, then look at what was damaged inside. That method is called the lesion method, and it gave the field its first hard evidence — along with its deepest pitfall.
Three things here: how the map got drawn, what made it remarkable in its time, and why it is no longer used as an address.
In April 1861, on a surgical ward in Paris, Broca met a fifty-one-year-old patient named Leborgne. The man had been ill for twenty-one years. He understood questions, answered with gestures, could handle numbers, showed no obvious collapse of intelligence — and the only thing that came out of his mouth was one syllable, tan. On the ward they simply called him Tan. He died six days later; Broca performed the autopsy and found a patch of destroyed tissue in the lower, rear part of the left frontal lobe. That same year he obtained a second case: Lelong, eighty-four, with about five words left, and a lesion in a matching location.
What is worth looking at is not the part he lost but the part he kept. Patients like this can often still swear, still count, still finish a phrase worn smooth by use — and yet cannot assemble one new sentence for the moment in front of them. (Whether Leborgne himself also retained a curse is disputed in the historical record; but such "speech automatisms" are documented in many later cases.) From this the neurologist Hughlings Jackson drew a distinction still in use: automatic speech — greetings, oaths, memorized material — and propositional speech — building a new sentence for this instant — can come apart, one collapsing while the other stays. So "being able to talk" was never one channel. It is at least two.
In 1865 Broca put it more bluntly: we speak with the left hemisphere. Asymmetry was now on the table. One piece of history deserves to be recorded straight: back in 1836 the French physician Marc Dax had already reported orally, in Montpellier, a link between left-hemisphere damage and loss of speech; the manuscript went unpublished until his son Gustave brought it out in 1865.
State the strength of the evidence plainly: a single case plus a naked-eye autopsy, tiny sample, no re-examination, no controls. It was the only route available then, and the source of every difficulty that followed.
In 1874 a twenty-six-year-old named Wernicke described patients who were almost the mirror image. Their speech was fluent, at normal speed, the sentences sounded like sentences, the grammatical parts were present — but the words were wrong, or invented on the spot. Ask one how he got here today and what comes back has roughly this texture: "Well, we sort of went, the two of us together, and the... the thing you put up, and then it was fine." You can hear that it is English. You cannot hear what he is saying. More importantly, such patients often cannot follow what you say either, and do not notice that their own output is wrong. The lesion sat in the upper, rear part of the left temporal lobe.
Either patient alone is just a curiosity; together they have force. One is effortful but on target; the other is effortless and off target. That mirror pattern — what is broken in you is intact in me, and the reverse — is called a double dissociation, and it is the strongest kind of evidence for inferring a division of labor from patients. With one patient you cannot tell whether the damage hit the function itself or some incidental thing it happens to lean on. The second, mirrored patient is what removes that confusion.
In 1885 Lichtheim drew the whole thing as a diagram: a rear region for understanding, a front region for producing, a line connecting them, plus a direction leading off toward concepts. Its strongest achievement was not explaining the known but predicting the unknown — cut only that connecting line, and a patient should understand and speak while being unable to repeat back what was just said. No such syndrome had been systematically documented. Later it was found, and named conduction aphasia.
A model that predicts an unseen syndrome is a good model in any field. So for nearly a century that diagram barely moved.
Broca preserved the brains of Leborgne and Lelong, and for over a hundred and forty years nobody was willing to cut them open. In 2007 Dronkers and colleagues took a different route: don't cut, scan. High-resolution magnetic resonance imaging. The result was unkind — in both men the damage extended far beyond the surface patch Broca had seen, running deep; Leborgne's also involved the insula and a set of deep structures, and in both men the fiber bundle connecting the front and rear language regions was cut. The founding cases of this field were never a case of "only that small patch was damaged."
The converse evidence is just as hard. Mohr in 1978 collected cases where the infarct was confined to that small region and its immediate surroundings: these patients go through a period of being unable to speak, then the clumsiness of articulation clears up quickly, and no lasting aphasia is left behind. To produce the textbook syndrome — persistent, telegraphic — the infarct has to be much larger, taking in the insula and a wide surrounding area. The region's name was used to name a syndrome, but damaging only that region does not produce it.
The third cut lands on the impression that these patients merely cannot produce, while comprehension is fine. Caramazza and Zurif in 1976 tested with a particular class of sentence: "The girl that the boy is chasing is tall" — who chases whom is settled purely by structure, and swapping the two nouns reverses it. The control was "The apple that the boy is eating is red," where an apple cannot eat a boy and the meaning locks the answer down. Patients were fine on the second class and clearly fell behind on the first: the telegraphic quality on the output side and a collapse on the input side arrive together. (Later work found this pattern varies a lot between individuals, and not everyone diagnosed this way looks like this; it is still argued about.)
The fourth cut lands on the names themselves. Tremblay and Dick in 2016 asked 159 researchers and clinicians where exactly the two regions are. The answers scattered without converging, with boundaries differing several-fold in size. A name in use for a century and a half, with no agreement even on its extent, is hard to keep using as a precise address.
Even the elegant prediction only half survived. Conduction aphasia was predicted as "that connecting line is cut," but when Buchsbaum and colleagues in 2011 overlaid the lesions of fourteen such patients, the densest overlap circumscribed a small patch of cortex at the temporal-parietal junction rather than the fiber bundle; the current view is that damage to that bundle is not a prerequisite for the syndrome. The syndrome was called correctly, the location was not.
So what is used now? The most common replacement framework stops asking "which block does what" and asks "which route connects what to what": a dorsal route carrying heard sound to articulatory movement, strongly left-lateralized; and a ventral route carrying sound to meaning, laid down on both sides. It incidentally explains something the old model could not — why one-sided temporal damage rarely wipes out comprehension: the route that handles meaning has a second copy on the other side.
The last cut comes from across languages. The signature of this syndrome is telegraphic speech — function words and endings dropped, content words left. But that summary was made from English-speaking patients. Miceli and colleagues in 1983 looked at Italian-speaking patients: they don't drop, they substitute. The reason is plain — an Italian verb reduced to a bare stem is not a word at all, so you cannot get it out of your mouth without an ending, and the damage can only show up as filling in the wrong one. The cross-language aphasia materials assembled by Menn and Obler show that in richly inflected languages (Italian, Hebrew and the like) substitution is more common than omission. The same damage develops into a different shape inside a different language system — and taking one language's patients as the definition of the syndrome means filing a property of the language as a property of the brain.
At minimum it shows the region is one necessary condition for the task. But there can be many necessary conditions, and they can be about routing, or blood supply, or attention. What further class of evidence would separate "participates in" from "is responsible for"? Does electrical stimulation, which switches off a small patch temporarily, count as a cleaner experiment?
If the description of symptoms has to be stated over a particular language, what is left of a cross-linguistic diagnostic standard? Is there a way to describe the damage rather than describing how much morphology this language happens to have? And where would a signer with aphasia fit into that description?
A term used for a century and a half is simultaneously a teaching scaffold, a clinical shorthand, and a search keyword in the literature. Retiring it is expensive; keeping it transmits the impression that there is a definite place there, generation after generation. How would you handle a name that is known to be imprecise and enormously convenient?