Neuron

NeuroView

The Legacy of Patient H.M. for Neuroscience

Larry R. Squire

1,2,

Veterans Affairs Healthcare System, San Diego, CA 92161, USA

Departments of Psychiatry, Neurosciences, and Psychology, University of California, San Diego, La Jolla, CA 92093, USA *Correspondence:lsquire@ucsd.eduDOI 10.1016/j.neuron.2008.12.023

H.M.isprobablythebestknownsinglepatientinthehistoryofneuroscience.Hisseverememoryimpairment,whichresultedfromexperimentalneurosurgerytocontrolseizures,wasthesubjectofstudyforﬁvedecadesuntil his death in December 2008. Work with H.M. established fundamental principles about how memoryfunctions are organized in the brain.

In 1952, Brenda Milner was completingher doctoral research at McGill Universityunder the direction of Donald Hebb. Atabout this time, she encountered twopatients (P.B. and F.C.) who had becomeseverely amnesic following unilateralremoval of the medial structures of theleft temporal lobe for the treatment of epileptic seizures ( Penfield and Milner,1958 ). This unfortunate outcome wasentirely unexpected, and it was proposedthat in each case there had been a preex-istent,butunsuspected, atrophic lesioninthe medial temporal lobe of the oppositehemisphere. In that way, the unilateralsurgery would have resulted in a bilaterallesion, an idea that was confirmed atautopsy some years later for patient P.B. After the two cases were presented atthe 1955 meeting of the American Neuro-logical Association, Wilder Penfield (theneurosurgeon in both cases) receiveda call from William Scoville, a neurosur-geon in Hartford, Connecticut. Scovilletold Penfield that he had seen a similarmemory impairment in one of his ownpatients (H.M.) in whom he had carriedoutabilateralmedialtemporalloberesec-tion in an attempt to control epilepticseizures. As a result of this conversation,Brenda Milner was invited to travel toHartford to study H.M.H.M. had been knocked down bya bicycle at the age of 7, began to haveminor seizures at age 10, and had majorseizures after age 16. (The age of thebicycle accident is given as 9 in somereports; for clarification seeCorkin,1984.) He worked for a time on anassembly line but, finally, in 1953 at theage of 27 he had become so incapaci-tated by his seizures, despite high dosesof anticonvulsant medication, that hecould not work or lead a normal life. Sco-ville offered H.M. an experimental proce-dure that he had carried out previously inpsychotic patients, and the surgery wasthen performed with the approval of thepatient and his family.When Milner first visited H.M., she sawthat the epilepsy was now controlled butthat his memory impairment was evenmore severe than in Penfield’s twopatients, P.B. and F.C. What sheobserved was someone who forgot dailyevents nearly as fast as they occurred,apparently in the absence of any generalintellectual loss or perceptual disorder.He underestimated his own age, apolo-gized for forgetting the names of personstowhomhehadjustbeenintroduced,anddescribed his state as ‘‘like waking froma dream

every day is alone in itself

’’( Milner et al., 1968, p. 217).The first observations of H.M., and theresults of formal testing, were reporteda few years later ( Scoville and Milner,1957 ). This publication became one of the most cited papers in neuroscience(nearly 2500 citations) and is still citedwith high frequency. H.M. continued tobe studied for five decades, principallyby Brenda Milner, her former studentSuzanne Corkin, and their colleagues( Corkin, 1984, 2002; Milner et al., 1968 ).He died on December 2, 2008, at theage of 82. It can be said that the earlydescriptions of H.M. inaugurated themodern era of memory research. BeforeH.M., due particularly to the influence of Karl Lashley, memory functions werethought to be widely distributed in thecortex and to be integrated withintellectual and perceptual functions.The findings from H.M. established thefundamental principle that memory isa distinct cerebral function, separablefrom other perceptual and cognitive abili-ties, and identified the medial aspect of the temporal lobe as important formemory. The implication was that thebrain has to some extent separated itsperceptual and intellectual functionsfrom its capacity to lay down in memorythe records that ordinarily result fromengaging in perceptual and intellectualwork.

The Medial Temporal Lobe MemorySystem

The early paper is sometimes cited incor-rectly as evidence that the hippocampusis important for memory, but this partic-ular point could not of course be estab-lished from a lesion that, by the surgeon’sdescription, included the hippocampus,amygdala, and the adjacent parahippo-campal gyrus. As Milner subsequentlywrote, ‘‘Despite the use of the word‘hippocampal’ in the titles of my paperswith Scoville and Penﬁeld, I have neverclaimed that the memory loss was solelyattributable to the hippocampal lesions’’( Milner, 1998 ). Indeed, the original paperends, quite appropriately, with the state-ment:It is concluded that the anteriorhippocampus and hippocampalgyrus,eitherseparatelyortogether,arecritically concerned in theretention of current experience. Itisnotknownwhethertheamygdalaplays any part in this mechanism,since the hippocampal complexhas not been removed alone, but

Neuron

, January 15, 2009

2009 Elsevier Inc.

always together with uncus andamygdala. ( Scoville and Milner,1957, p. 21).The findings from H.M. were initiallymet with some resistance, especiallybecause of the difficulty for many yearsof demonstrating anything resemblinghisimpairmentintheexperimentalanimal.Effortstoestablishananimalmodelinfactbegan almost immediately when Scovillehimself came to Montreal and did thesame surgery in monkeys that he haddone with H.M. But these monkeys andothers with medial temporal lesionsseemed able to learn tasks that H.M.could not learn. Only much later did itbecome understood that apparentlysimilar tasks can be learned in differentways by humans and monkeys. Forexample, the visual discrimination task,which is learned gradually by the monkeyover hundreds of trials, proved to involvewhat one would now call habit learning.In the monkey, this kind of learningdepends on the basal ganglia, not themedial temporal lobe. Eventually, taskswere developed for the monkey thatwere exquisitely sensitive to medialtemporal lobe lesions (for example, theone-trial, delayed nonmatching to sampletask), and an animal model of humanmemory impairment thereby becameavailable ( Mishkin, 1978 ).Cumulative work with the animal modelover the next decade, together withneuroanatomical studies, succeeded inidentifying the anatomical componentsof what is now termed the medialtemporal lobe memory system ( Squireand Zola-Morgan, 1991 ): the hippo-campusand theadjacentperirhinal, ento-rhinal,andparahippocampal corticesthatmake up much of the parahippocampalgyrus. This information showed whichstructures within H.M.’s large lesionwere important for understanding hisimpairment and, more broadly, whatstructures are important for memory. Afewyearslater,animproveddescriptionof H.M.’s lesion was obtained withmagnetic resonance imaging (MRI) ( Cor-kin et al., 1997 ). MRI had been delayedbecause of concerns that clips placedon the dura during surgery made H.M.ineligible for imaging. However, thoroughinquiry revealed that the dural clipsconstituted no risk. At this juncture, several points becameclear. First, H.M.’s lesion was lessextensive than described originally by thesurgeon in that it extended a little morethan 5 cm caudally from the temporalpole (not 8 cm). As a result the posteriorparahippocampal gyrus was largelyspared(specifically,theparahippocampalcortex or what in the monkey is termedarea TH TF). Second, the reason thatH.M.’smemoryimpairmentwassoseverewasthatthebilateraldamageincludedtheparahippocampal gyrus (anteriorly) andwas not restricted to the hippocampus.Damage limited to the hippocampuscauses significant memory impairmentbut considerably less impairment than inH.M. Third, memory impairment moresevere than H.M.’s could now be under-stood, as when the damage includes thestructures damaged in H.M. but alsoextends far enough posteriorly to involvethe parahippocampal cortex (patientsE.P. and G.P.;Kirwan et al., 2008 ).In the early years, the anatomy of themedial temporal lobe was poorly under-stood, and terms like hippocampal zoneand hippocampal complex were oftenused to identify the area of damage. Withthe elucidation of the boundaries andconnectivity of the structures adjacent tothe hippocampus and the discovery thatthese structures are important formemory, vague terms like hippocampalcomplex became unnecessary (thoughone can still find them in contemporarywriting).Itisnowpossibletoachievecare-ful descriptions based on anatomicalmeasurement and modern terminology.H.M. not only motivated the develop-ment of an animal model of humanmemory impairment and the subsequentdelineation of the medial temporal lobememory system. As described next, thestudy of H.M. also led to fundamentalinsights into the function of the medialtemporal lobe and the larger matter of how memory is organized in the brain.

Immediate Memory and Long-TermMemory

H.M.’s intact intellectual and perceptualfunctions, and similar findings in otherpatients with large medial temporallesions, have been well documented. A key additional finding was that H.M.had a remarkable capacity for sustainedattention, including the ability to retaininformation for a period of time after itwas presented. Thus, he could carry ona conversation, and heexhibited an intactdigit span (i.e., the ability to repeat backa string of six or seven digits). Indeed,information remained available so longas it could be actively maintained byrehearsal. For example, H.M. could retaina three-digit number for as long as 15 minby continuous rehearsal, organizing thedigits according to an elaboratemnemonic scheme. Yet when his atten-tionwasdivertedto anewtopic,heforgotthe whole event. In contrast, when thematerial was not easy to rehearse (in thecase of nonverbal stimuli like faces ordesigns), information slipped away inless than a minute. These findings sup-ported a fundamental distinctionbetweenimmediate memory and long-termmemory (what William James termedprimarymemoryandsecondarymemory).Primary memory [immediate memory]

comes to us as belonging to therearward portion of the presentspace of time, and not to thegenuinepast( James,1890,p.647).Secondarymemory[long-termmemory]isquite different. An object which has been recol-lected

is one which has beenabsent from consciousness alto-gether, and now revives anew. It isbrought back, recalled, ﬁshed up,so to speak, from a reservoir inwhich, with countless otherobjects, it lay buried and lost fromview. ( James, 1890, p. 648).Notably, time is not the key factor thatdetermines how long patients like H.M.canretaininformationinmemory.Therele-vant factors arethe capacity of immediatememory and attention, i.e., the amount of material that can be held in mind andhow successfully it can be rehearsed.The work with H.M. demonstrated thatthe psychological distinction betweenimmediate memory and long-termmemory is a prominent feature of how thebrain has organized its memory functions.

Multiple Memory Systems

Perhaps the most unexpected discoveryaboutH.M.,givenhisprofoundandglobal

Neuron

NeuroView

Neuron

, January 15, 2009

2009 Elsevier Inc.

memory impairment, came when BrendaMilnertestedhisabilitytoacquireavisuo-motorskill( Milner,1962 ).H.M.wasshowna five-pointed star, with a double contour,andaskedtotraceitsoutlinewithapencil,but in a condition when he could onlysee his hand and the star as reflectedin a mirror. H.M. acquired this mirror-drawing skill during ten trials andexhibited excellent retention across 3days. Yet at the end of testing, he hadno recollection of having done the taskbefore. This demonstration provided thefirst hint that there was more than onekind of memory in the brain and sug-gested thatsomekindsofmemory (motorskills) must lie outside the province of themedial temporal lobe.For a time, it was rather thought thatmotor skills were a special case and thatall the rest of memory is impaired inpatients like H.M. Later it became appre-ciated that motor skills are but a subsetof a larger domain of skill-like abilities, allof which are preserved in amnesia. Thedemonstration of a fully preserved abilityto learn the perceptual skill of mirrorreading suggested a distinction betweentwobroadclassesofknowledge:declara-tive and procedural ( Cohen and Squire,1980 ). Declarative memory is what ismeant when the term ‘‘memory’’ is usedin everyday language, i.e., consciousknowledge of facts and events. Proce-dural memory refers to skill-based knowl-edgethatdevelopsgraduallybutwithlittleability to report what is being learned.In the years that followed, otherpreserved learning abilities began to bereported for amnesic patients, and theperspective shifted to a framework thataccommodated multiple (i.e., more thantwo) memory systems. As Endel Tulvingwrote:But even if we accept the broaddivision of memory into proceduraland propositional forms

thereare phenomena that do not seemto fit readily into such a taxonomy( Tulving et al., 1982, p.336).Subsequently, the terms declarativeand nondeclarative were introduced withthe idea that declarative memory refersto the kind of memory that is impaired inH.M. and is dependent on the medialtemporal lobe. Nondeclarative memoryis an umbrella term referring to additionalmemory systems. These include systemsthat support skill learning, habit learning,simple conditioning, emotional learning,as well as priming and perceptuallearning. The structures with specialimportance for these kinds of memoryinclude the basal ganglia, the cerebellum,the amygdala, and the neocortex. Thestarting point for these developmentswas the early discovery that motor skilllearning was preserved in H.M. Thisfinding revealed that memory is nota single faculty of the mind and led ulti-mately to the identification of the multiplememorysystemsofthemammalianbrain.

Remote Memory

H.M.’s memory impairment has generallybeentakenasreflectingafailuretoconverttransient, immediate memory into stablelong-term memory. A key insight aboutthe organization of memory, and medialtemporal lobe function, came witha consideration of his capacity toremember information that he hadacquired before his surgery. The firstexploration of this issue with formal testsasked H.M. to recognize faces of personswho had become famous in differentdecades, 1920–1970 ( Marslen-Wilsonand Teuber, 1975 ). As expected, H.M.wasseverelyimpairedatrecognizingfacesfromhispostmorbidperiod(the1950sand1960s), but he performed as well as orbetter than age-matched controls atrecognizing faces of persons who were inthenewsbeforehissurgery.Thisimportantfinding implied that the medial temporallobe is not the ultimate storage site forpreviously acquired knowledge. The earlydescriptions of H.M. conform to this view.Thus, H.M. was described as havinga partial loss of memory (retrogradeamnesia) for the 3 years leading up to hissurgery, with early memories ‘‘seeminglynormal’’ ( Scoville and Milner, 1957, p. 17).Similarly, about 10 years later it wasremarked that there did not appearto have been any change in H.M.’scapacity to recall remote eventsantedating his operation, such asincidents from his early schoolyears, a high-school attachment,or jobs he had held in his late teensand early twenties ( Milner et al.,1968, p. 216).Subsequently, a particular interestdeveloped in the status of autobiograph-ical memories for unique events, whichare specific to time and place, andmethods were developed to assess thespecificity and the detail with which suchrecollections could be reproduced. Inthe earliest efforts along these lines, assummarized by Suzanne Corkin ( Corkin,1984 ), H.M. produced well-formed auto-biographicalmemories,fromage16yearsor younger. It was concluded that H.M’sremote memory impairment nowextended back to 11 years before hissurgery. The situation seemed to changefurther as H.M. aged. In an updateprepared nearly 20 years later ( Corkin,2002 ), H.M. (now 76 years old) wasdescribed as having memories of child-hood, but his memories appeared morelikerememberedfactsthanlikememoriesof specific episodes. It was also said thathe could not narrate a single event thatoccurred at a specific time and place.Essentially the same conclusion wasreached a few years later when newmethods, intended to be particularlysensitive, were used to assess H.M.’sremote memory for autobiographicalevents ( Steinvorth et al., 2005 ). Theselater findings led to the proposal that,whatever might be the case for factmemory, autobiographical memories,i.e., memories that are specific to timeandplace,dependonthemedialtemporallobe so long as the memories persist.There are reasons to be cautious aboutthisidea.In2002–2003,newMRIscansof H.M. were obtained ( Salat et al., 2006 ).These scans documented a number of changes since his first MRI scans from1992–1993 ( Corkin et al., 1997 ), includingcortical thinning, subcortical atrophy,large amounts of abnormal white matter,and subcortical infarcts. These findingswere thought to have appeared duringthe past decade, and they complicatethe interpretation of neuropsychologicaldata collected during the same timeperiod. Another consideration is thatremote memories could have been intactin the early years after surgery but thenhave faded with time because they couldnot be strengthened through rehearsaland relearning. In any case, the optimaltime to assess the status of past memoryis soon after the onset of memoryimpairment.

Neuron

, January 15, 2009

2009 Elsevier Inc.

Neuron

NeuroView