By What struck me about this case is how much of our learning happens below awareness, the habits and instincts we never remember picking up. A mirror-tracing test, given to a man who could no longer form new conscious memories, showed that as sharply as anything in the record. He got better at it across ten trials, still had the skill three days later, and at the end had no recollection of ever having done the task.
The psychologist Brenda Milner put a five-pointed star in front of the man known as H.M. The star was drawn with a double outline, leaving a narrow channel between the two lines, and his job was to trace that channel with a pencil without straying over either edge. He could see his own hand and the star only as reflected in a mirror. The mirror reverses every movement, which makes the task awkward for anyone on a first attempt.
Milner reported the result in 1962, in a chapter written in French for a conference volume called Physiologie de l'hippocampe. Most people today meet the finding through Larry Squire, author of a 2009 review of H.M.'s legacy published in Neuron shortly after the patient died. Squire puts Milner's data this way in that review: "H.M. acquired this mirror-drawing skill during ten trials and exhibited excellent retention across 3 days. Yet at the end of testing, he had no recollection of having done the task before."
What the star test could not show is where in his brain that learning was happening.
H.M. was Henry Molaison, whose name became public only after his death in December 2008 at the age of 82. He had minor seizures from the age of 10 and major ones after 16, Squire's review recounts, and by 1953, when he was 27, they were so incapacitating despite high doses of anticonvulsant drugs that he could not work or lead a normal life.
The surgeon William Scoville offered him an operation that the 1957 write-up, co-authored by Scoville and Milner, would call "frankly experimental".
It removed the medial structures of both temporal lobes, meaning the inner parts, including much of the hippocampus, a structure lying on the inner side of each temporal lobe that this very case would help tie to memory, and the cortex next to it. The surgery did control the epilepsy. It also left him with a severe and permanent memory impairment.
Milner, who had done her doctoral work at McGill University, found a man whose intellect and perception were intact but who forgot the events of his day almost as fast as they happened.
He described his own state as "like waking from a dream". He was studied for five decades, mainly by Milner, her former student Suzanne Corkin and their colleagues.
A man who could not hold on to what had just happened to him was getting measurably better at a difficult motor task, and holding on to the improvement across days, while denying he had ever tried it.
Squire calls it "Perhaps the most unexpected discovery about H.M." In his words, the demonstration "provided the first hint that there was more than one kind of memory in the brain", and it suggested that motor skills must sit outside the medial temporal lobe.
The mirror star is often credited to that 1957 paper, which established the amnesia and became one of the most cited papers in neuroscience. The paper does not mention mirror drawing.
Its abstract says the memory loss "left early memories and technical skills intact", which refers to things H.M. could already do before the operation rather than to anything new he picked up afterwards. The new learning is in the 1962 chapter.
For a while, researchers treated motor skills as a special exception and assumed everything else about memory was wrecked in patients like H.M. That reading did not hold.
Later work showed amnesic patients could fully learn to read mirror-reversed words, a perceptual skill rather than a hand movement. In 1980, according to Squire's review, he and his colleague Neal Cohen used results like these to split knowledge into two broad classes. Declarative memory is conscious knowledge of facts and events, which is what most people mean by memory. Procedural memory is skill-based knowledge that builds up gradually, with little ability to say what is being learned.
The framework then grew past two. Declarative memory depends on the medial temporal lobe, the review says, and "nondeclarative" became an umbrella term for skill learning, habit learning, simple conditioning, emotional learning, perceptual learning and priming, which is the way a recent exposure speeds up later processing of the same material.
The structures Squire lists for the nondeclarative side sit outside the medial temporal lobe, in places such as the basal ganglia, the cerebellum and the amygdala. Pinning that down took other patients and other tasks, along with animal work, not the star test.
As Squire puts it, the finding "revealed that memory is not a single faculty of the mind and led ultimately to the identification of the multiple memory systems of the mammalian brain."
Daniel Schacter, a memory researcher at Harvard, said in an email that the mirror drawing result set off a great deal of later work trying to pin down exactly which tasks amnesic patients could improve on, sometimes to normal levels, "while lacking conscious memory" of the practice. "These findings paved the way for distinctions between different kinds of memory," he wrote, naming the explicit or declarative side that is impaired in amnesia and the implicit or nondeclarative side that survives.
Squire and his co-authors Knowlton and Mangels, in a 1996 Science paper, gave a probabilistic classification task to amnesic patients and to patients with Parkinson's disease who did not have dementia. Participants had to work out, trial by trial, which of two outcomes tended to follow a given combination of cues.
The amnesic patients learned the task normally while being severely impaired at remembering the training sessions themselves. The Parkinson's patients remembered the training sessions and failed to learn the task.
The authors took that double dissociation as evidence for parallel systems, with the neostriatum, meaning the caudate nucleus and putamen inside the basal ganglia, doing the slow incremental learning that habits are made of.
A 2015 paper in Brain by Jacobsen and colleagues starts from the clinical observation that musical memory in Alzheimer's disease and other dementias is "surprisingly robust". The team scanned 32 healthy young adults at 7 tesla. Tesla is the unit of a scanner's magnetic field strength, and 1.5 and 3 tesla remain the clinical standard, so 7 tesla is a much stronger research scanner.
In those young adults the authors identified regions that encoded long-known music, in the caudal anterior cingulate and the ventral pre-supplementary motor area.
In 20 Alzheimer's patients and 34 healthy older controls, the paper reports, those same regions showed relatively little shrinkage and relatively little disruption of glucose metabolism compared with the rest of the brain, although amyloid build-up there was not much lower. The authors say this may explain the pattern.
The 1962 chapter is in French and rarely read first-hand, so nearly everyone knows the finding through later summaries, including this one.
H.M.'s damage also went beyond the hippocampus into the parahippocampal gyrus in front of it, which Squire says is why his impairment was so severe, since damage confined to the hippocampus produces a significant but considerably milder deficit. Because the star result rests on a single patient, it opened a question rather than settling one, and the memory systems framework was assembled over decades from other patients and from animal work.
The music study maps regions in healthy young brains and then compares tissue measures across groups of patients and controls. That explains a tendency. It does not predict what any one person with dementia will keep.
"Some researchers interpreted the early findings on mirror drawing as showing that motor learning is uniquely preserved in amnesic patients, but it turned out that preserved implicit/nondeclarative memory in amnesic patients extends beyond motor learning to a variety of tasks that do not require conscious memory of past experiences," Schacter said.
Sources
- B. Milner, "Les troubles de la mémoire accompagnant des lésions hippocampiques bilatérales", in P. Passouant (ed.), Physiologie de l'hippocampe, CNRS, Paris, 1962, pp. 257 to 272.
- L. R. Squire, The legacy of patient H.M. for neuroscience, Neuron, 2009.
- W. B. Scoville and B. Milner, Loss of recent memory after bilateral hippocampal lesions, Journal of Neurology, Neurosurgery and Psychiatry, 1957.
- L. R. Squire, Memory systems of the brain: a brief history and current perspective, Neurobiology of Learning and Memory, 2004.
- B. J. Knowlton, J. A. Mangels and L. R. Squire, A neostriatal habit learning system in humans, Science, 1996.
- J. H. Jacobsen and colleagues, Why musical memory can be preserved in advanced Alzheimer's disease, Brain, 2015.
- E. de Vries and colleagues, Clinical 7 Tesla magnetic resonance imaging: Impact and patient value in neurological disorders, Journal of Internal Medicine, 2025.
- Email from Daniel Schacter, October 2026.
