Beyond Memory in H.M., The World’s Most Famous “Hippocampal Amnesic”

Abstract

Patient H.M. (Henry Moliason) suffered a wide range of cognitive deficits attributable to the damage to his hippocampal formation but not to his well-established deficits in the ability to recall newly encountered facts, events, names, and objects, which formed the basis for his early diagnosis as a “hippocampal amnesic.” Among Henry’s “non-memory” or cognitive deficits that this chapter reviews are his impaired ability to create new and grammatical sentence plans, to identify who-did-what-to-whom in novel sentences and to understand and read aloud novel sentences containing familiar words that he understood without difficulty in isolation, but not in novel sentence contexts such as metaphors. Also reviewed are his deficits in detecting novel forms concealed within complex visual arrays, in identifying anomalous objects in novel visual scenes, in detecting and describing what makes captioned cartoons funny, and in combining familiar concepts into new and useful ideas. The chapter concludes by relating Henry’s non-memory deficits to fundamental questions of this book, for example, What is the role of the hippocampal formation in human memory and cognition? And how does the hippocampal formation work?

1. Introduction

William Scoville discovered the link between memory and the hippocampal formation by chance in 1953 after he met Henry Moliason (age 27) in his office at the Hartford Hospital in Connecticut. Asked to eliminate or at least reduce the intensity and frequency of Henry’s life-threatening epileptic seizures, Scoville surmised that Henry’s convulsions probably originated in his hippocampus (in the middle of the brain, roughly between the ears: See Figure 1). With consent from Henry and his parents, Scoville drilled small holes in Henry’s skull above his eyes, and using X-ray imagery, inserted thin metal tubes into the hippocampal region, and suctioned out about half of Henry’s hippocampal formation bilaterally while leaving his neocortex virtually intact.

Figure 1.

The human hippocampus shown through a “window” into the medial temporal lobe. (Artist’s rendition of an original illustration by Henry Vandyke Carter that appeared in Henry Gray’s 1918 Anatomy of the Human Body.) Neither Henry’s surgical damage nor the connections linking the hippocampus to surrounding structures in the hippocampal formation are shown.

This experimental operation essentially cured Henry’s epilepsy and perhaps saved his life. However, it immediately became clear that something was terribly wrong. Henry could no longer remember things he had done hours, minutes, or even seconds earlier. He could not even find his way back from the bathroom to his bed in the hospital [1].

Henry’s memory problems became a source of intense scientific scrutiny that soon made him famous around the world as “patient H.M.” However, Scoville informed other neurosurgeons that he had inadvertently removed the engine for forming new memories in Henry’s brain and warned against applying his surgical procedure in future cases.

Over the next fifty years, Henry participated in hundreds of psychological experiments at the Massachusetts Institute of Technology (MIT), and after he died at age eighty-two, neuroscientists conducted sophisticated analyses of his brain (which he bequeathed to MIT) and wrote several books about how research with Henry helped advance the behavioral and brain sciences [1, 2]. I dedicated my own recent book to Henry, “an ordinary man who became famous by generously devoting his life to helping scientists understand his memory, mind, and brain, trusting in the promise that what they learned about him would “help others.” [3]. Researchers reading the present review of Henry’s contributions to understanding the role of the hippocampal formation in human memory, visual cognition, language comprehension, and language production clearly represent the type of “others” that Henry wanted to help.

However, three limits to the scope of this review deserve comment. It only reviews studies that report deficits attributable to Henry’s damaged hippocampal formation (rather than to damage elsewhere in his brain, e.g., the cerebellum), studies that experimentally study Henry’s cognitive deficits independently of his memory deficits (by presenting, e.g., continuously displayed text, instructions, and pictorial stimuli that participants need not commit to memory), and studies that statistically compare performance for Henry versus control groups that resemble him in age, background, IQ , and education but lack damage in the hippocampal region. Also beyond the purview of this review are experiments with Henry not directly related to the memory versus non-memory issue (e.g., [4, 5, 6, 7, 8, 9, 10, 11]), as well as the many experiments that established his famous deficits in recalling newly encountered names, facts, and events.

My UCLA lab conducted approximately twenty-five experiments with Henry at MIT and statistically compared his results in each study with those of 8–12 comparable control participants tested under similar conditions at UCLA.

2. Henry’s non-memory deficits

This section reviews Henry’s visual cognition, language comprehension, and sentence production deficits in studies that did not require the learning or recall of newly encountered names, facts, events, and objects (the classical definition of memory).

2.1 Henry’s visual cognition deficits

Henry’s performance on three tasks (hidden figure, what’s-wrong-here, and visual cartoon comprehension) illustrate the general nature of his visual cognition deficits.

2.1.1 Henry’s hidden figure deficit

Canadian neuropsychologist Brenda Milner accidentally discovered Henry’s hidden figure deficit when trying to confirm her personal impression that Henry functioned without difficulty in his everyday visual world and only experienced problems when forced to remember where he had been or where he was going. To validate this impression, she had Henry complete the standard hidden figures test, a classic paper-and-pencil measure of the ability to detect target figures camouflaged in complex visual scenes. On a typical hidden figures trial, instructions asked Henry to inspect a shape (shown to the left on a page) and trace that target shape hidden within a complex concealing array shown to the right. In a real-life analog of this task, soldiers try to detect an enemy combatant (the target) camouflaged in a forest (the concealing array). Unlike a soldier, however, Henry had to detect abstract target shapes that were unmoving, unfamiliar to him, and visible throughout each trial (rendering recall unnecessary).

Henry correctly traced significantly fewer camouflaged targets than the memory-normal controls, a deficit indicating serious impairment to his ability to consciously recognize forms in complex visual scenes. However, Millner and many other neuropsychologists ignored Henry’s hidden figures deficit and continued to consider him a “pure memory” case, with memory problems but no cognitive deficits [12].

Undeterred by this lack of interest, my UCLA lab replicated and refined Henry’s hidden figures deficit by adding a condition where participants traced familiar hidden targets, which were forms that Henry experienced frequently in daily life before and after his lesion, e.g., squares, circles, right-angle triangles [13]. In this familiar target condition, Henry and the controls correctly traced the same number of targets, a seemingly minor fine-tuning of Henry’s hidden figure deficit that paved the way for theoretical ideas discussed in Section 5 about why Henry experienced the selective deficits in cognition and memory that he did.

However, for the standard unfamiliar target condition, control participants in our study [14] traced reliably more targets in the concealing arrays than did Henry, a replication of Milner’s hidden figure deficit that addresses a hypothesis in Bussey et al. that perceptual deficits due to parahippocampal damage reflect a memory problem [15]. Henry’s hidden figure deficit in our study clearly reflected a perceptual problem [16].

2.1.2 Henry’s what’s-wrong-here deficits

Having confirmed and honed Henry’s hidden figures deficit, my UCLA lab re-examined Henry’s perceptual world in other ways. One involved a children’s game found in books such as What’s Wrong Here: Hundreds of Zany Things to Find. These books display complex everyday scenes, for example, a school classroom containing over a hundred busy people and objects, some of which are erroneous or anomalous, say, a bird flying upside down in a fish bowl filled with water, or an impossible-to-open door with hinges on the same side as its doorknob. Children enjoy discovering what’s-wrong-here in pictures.

In our laboratory version of the game, Henry and suitable control participants inspected a series of what’s-wrong-here pictures, circled as many erroneous objects as possible within each picture, and explained why each circled object was anomalous within a generous time limit [17].

Henry correctly circled significantly fewer erroneous objects than the controls, and he misidentified many of his circled objects without correcting himself, something the controls never did. For example, Henry called a clearly drawn rabbit “a dog” and called an ordinary wastebasket on the floor beside a teacher’s desk “a window.” Because Henry always correctly identified the identical objects when depicted in isolation in a subsequent test, Henry’s object identification errors were clearly specific to the what’s-wrong-here scenes, consistent with a problem in disentangling unfamiliar forms from their unfamiliar surroundings in complex visual displays, as in the original hidden figure test [18].

2.1.3 Henry’s cartoon comprehension deficits

New higher-level deficits in Henry’s visual cognition and sentence comprehension emerged in an experimental test of his ability to understand captioned cartoons [19]. Participants in this UCLA study saw a sequence of cartoons with instructions to explain what made them funny and to read aloud their captions (which only contained words Henry knew before his surgery). One example is Gary Larson’s cartoon, “Raising the dead,” which depicts two women in armchairs chatting —a normal scene except that both are ghosts. A ghost woman named Edith is listening to the other complain about problems raising her ghost children, Billy and Sally. Illustrating her difficulties, the cartoon shows Sally floating head first down a stairway, while Billy flits aimlessly around the room. The caption reads: “Oh, I don’t know. Billy’s been having trouble in school and Sally’s always having some sort of crisis. I tell you, Edith, it’s not easy raising the dead.”

Control participants in this study consistently detected humor in the cartoons and never misread their captions in ways that would preclude comprehension of why they were funny. Not Henry. For example, Henry misread the Larson caption as, “I tell Edith it’s not [long pause] easy, the raising the dead,” messing up the sentence prosody and omitting the critical word you in I tell you, Edith, it’s not easy raising the dead, all without self-correction. Henry apparently thought the speaker was talking to someone not depicted in the cartoon, rather than to Edith, the ghost mother seated beside her.

Henry’s uncorrected caption-reading errors did not just render full understanding of cartoons impossible. They also suggested an inability to grasp who-said-what-to-whom in the cartoons, reflecting a serious comprehension deficit that my UCLA lab established in a subsequent study (described shortly).

Nor was confusion about who spoke to whom Henry’s only problem. He did not grasp what the cartoons were about, another basic prerequisite to getting the jokes. For example, Henry did not see that the Larsen cartoon depicted ghosts. Noting that he could see through Larson’s ghost-speaker to the armchair on which she sat, Henry suggested that the cartoonist had drawn her wrong and complained that “she” (the cartoonist) just “bl. .. the. .. blackens the whole way, and everything.. .” using some kind of “blackening rule.” Unlike Henry, the control participants never misidentified visual forms in the cartoons nor mistakenly ascribed an outline to artistic error.

3. Henry’s language comprehension deficits

Consistent with Henry’s caption-reading errors, four sources of evidence indicated deficits in Henry’s ability to comprehend and read aloud various types of sentences, a skill he mastered in grade school and high school, many years before his age 27 surgery.

3.1 Deficits in identifying who-did-what-to-whom in sentences

Henry’s failure to comprehend who-did-what-to-whom in Larson’s cartoon echoed an important finding in a 1966 experiment that I conducted at MIT [20]. In that study, Henry saw various types of ambiguous sentences on cards with instructions to describe the two meanings of each sentence as quickly as possible. Henry readily detected both meanings in some types of ambiguous sentences, but relative to controls, he displayed a major deficit in detecting the dual meanings of sentences resembling John is the one to help today, where John should help us is one meaning and we should help John is the other. Henry clearly had a problem working out who-did-what-to-whom in ambiguous sentences.

But could Henry understand who-did-what-to-whom in unambiguous sentences? To find out, my UCLA lab ran an experiment in which Henry and memory-normal controls read unambiguous sentences on a computer screen, one at a time, and then answered a multiple-choice comprehension question displayed on the same screen [21]. For example, after reading The water that the mother spilled surprised the young child, participants answered the comprehension question who spilled the water: the mother, the young child, or nobody? Control participants correctly answered significantly more comprehension questions than Henry, firmly establishing a deficit in his ability to understand the most important information that sentences can convey: who-does-what-to-whom [22].

3.2 Deficits in reading sentences aloud

Can Henry comprehend and accurately read aloud the individual words in unambiguous sentences? To find out, my UCLA lab first had participants read lists of familiar words presented one at a time on cards, for example, GOT, ATE, STOMACH, HOT, WHO, DOGS, ACHES, BOYS, and THE. Henry made no more mistakes than the controls when reading those isolated words.

Days later, however, Henry experienced major deficits when asked to read the same words re-organized into sentences, for example, the boys who ate hot dogs got stomach aches, instead of GOT, ATE, STOMACH, HOT, WHO, DOGS, ACHES, BOYS, and THE.

Unlike the controls, Henry now made dozens of uncorrected reading errors that rendered his utterances ungrammatical. He also paused abnormally at critical points within the sentences, for example, misreading The boys who ate hot dogs got stomach aches as The boys [unusually long pause] ate hot dogs got stomach aches (Note Henry’s ungrammatical omission of the word who) [23].

3.3 Deficits in comprehending metaphors

Metaphors, such as Life is a Journey, are powerful linguistic tools. They shape everyday thinking and help people comprehend and learn ideas that are otherwise difficult to acquire [24]. Can Henry comprehend metaphors? To answer this question, my UCLA lab asked Henry and suitable controls to indicate what metaphoric sentences mean. On each trial, participants saw a short metaphoric sentence on a computer screen with instructions to choose the best of three possible ways to interpret it. By way of illustration, these were the three choices for the metaphor Maybe we should stew over his suggestion:

1. Let us think about it some more (correct metaphoric interpretation),

2. Maybe we should put more meat into his idea (incorrect but metaphoric interpretation),

3. Let us make sure to cook the stew long enough (incorrect literal interpretation containing the same critical word stew as the target sentence).

The memory-normal controls chose the correct metaphoric interpretation reliably more often than Henry, indicating a deficit in his ability to comprehend metaphors. Indeed, Henry performed worse than chance (random guessing) because he usually chose the incorrect interpretation with the same critical word as the target sentence; here the word STEW capitalized in the original metaphor Maybe we should STEW over his suggestion and the incorrect literal interpretation: Let us make sure to cook the STEW long enough [23]. Henry clearly understood the individual words but not the overall meaning of the metaphoric sentences.

3.4 Deficits in detecting what’s right versus wrong in sentences

Another UCLA study tested whether Henry could distinguish between grammatical versus ungrammatical sentences. He could not. Asked to respond “Yes,” to grammatical sentences such as She hurt herself, and “No,” to ungrammatical sentences such as He hurt herself, Henry answered correctly significantly less often than suitable control participants, with performance close to chance (50%). Henry clearly had a deficit in comprehending whether sentences are grammatical versus ungrammatical [25].

Did he fail to understand the instructions? Did he not care? To find out, my UCLA team reran the previous study, adding foil sentences and a test for guessing. The foil sentences thoroughly shuffled the words in grammatical sentences such as She has decided to buy a house, yielding ungrammatical strings such as Decided has house she a buy to. Like the control participants, Henry invariably called these foils ungrammatical, indicating clear comprehension of the instructions.

To assess guessing, the experimenter immediately asked participants who responded, “No, ungrammatical,” to identify the wrong or misplaced word and then correct that word to make the sentence grammatical. This was easy for the control participants. For example, after identifying be as the misplaced word in Will be Harry blamed for the accident, they quickly produced a corrected version such as Will Harry be blamed for the accident.

Not Henry, however. He called correct words incorrect and failed to correct words he deemed wrong. For example, Henry identified blamed as the incorrect word in Will be Harry blamed for the accident, but insisted that further information about the blame was needed to correct this error. Henry was indeed guessing when he called sentences ungrammatical.

4. Henry’s language production deficits

Many sources of evidence indicate that Henry suffered language production deficits. Length constraints limit us to just one source here: His performance on the standardized test of language competence (TLC) [26]. In a typical TLC trial, participants see two words above a picture, together with continuously displayed instructions to use both words in a single grammatical sentence that accurately describes the picture.

Control participants found this task easy. For example, asked to use the words ALTHOUGH and WRONG in a single grammatical sentence that describes a woman in a sports store discussing a tracksuit with a salesman, one control participant quickly responded The woman decided to buy the suit ALTHOUGH it looked WRONG. A panel of judges blind to speaker identity later rated this transcribed response 100% correct on the three evaluative dimensions shown in Table 1.

Scene description	A woman in a sports store is discussing a tracksuit with a salesman
Must use words; (in caps throughout)	ALTHOUGH WRONG
100% correct Response of a control participant	“The woman decided to buy the suit ALTHOUGH it looked WRONG.”
Henry‘s response to the same scene	“Because it’s WRONG for her to be he’s dressed just as this that he’s dressed and the same way.”
Henry’s must use words score	50% correct due to omission of the must use word ALTHOUGH
Henry’s accuracy evaluation	INACCURATE: No two people in the scene are dressed “the same way.”
Henry’s grammaticality evaluation	UNGRAMMATICAL: Henry’s utterance is not a sentence.
Henry’s coherence evaluation	INCOHERENT: a rambling series of non sequiturs.

Table 1.

An illustrative TLC trial, with scene description, must use words, the completely correct response of a control participant, and Henry’s response, rated on three evaluative dimensions and scored for inclusion of the must use words.

Table 1 also shows Henry’s response to the same sportswear picture: “Because it’s wrong for her to be he’s dressed just as this that he’s dressed and the same way,” a response that the panel of judges rated as inaccurate, ungrammatical and incoherent, and a rambling series of non sequiturs (see Table 1). Across all trials, Henry included significantly fewer must use words than the controls, and the panel of judges rated his utterances ungrammatical, inaccurate, or incoherent significantly more often than those of the controls.

Why did Henry include significantly fewer must use words in his utterances than the controls? The coherence rating for Henry’s utterance in Table 1 suggests one reason. Henry was freely generating familiar phrases (e.g., the same way) without relating the picture to the must use words, a free association strategy that may also explain why the judges more often considered Henry’s TLC descriptions incoherent, ungrammatical, and inaccurate.

Did the damage to Henry’s hippocampal formation shape his free association strategy? Almost certainly. Henry could easily retrieve phrase memories formed before his age 27 surgery, for example, the common phrase the same way. However, the damage to his hippocampal formation prevented him from forming new and situation-appropriate phrases and sentences. How do we know? Because my UCLA lab analyzed hundreds of unintended and uncorrected errors that rendered Henry’s TLC utterances ungrammatical, inaccurate, and incoherent errors that speak volumes about how the hippocampal formation goes about creating new memories in the cortex.

Henry’s TLC errors fell into two categories: Omissions (where participants omit units that are essential in a grammatical sentence) and Combination errors (where participants conjoin two or more units into a sequence that is impermissible or ungrammatical). Table 2 illustrates both types of error in an utterance Henry produced on a single TLC trial. The TLC picture shows three people: a woman server at a cafeteria counter, a man ordering food from her, and a woman ahead of him in line who already has the food she ordered on a tray. PIE and EITHER are the must use words.

Scene description	A man is ordering food at a cafeteria counter, and a woman ahead of him in line already has her food on a tray.
Must use words (in caps throughout)	PIE and EITHER
Henry‘s actual response	“I want some her. .. what she had.”
Henry’s intended response	I want some of what she had.
Omission error analysis (omitted word of in parentheses in Henry’s intended utterance)	I want some (of) what she had.
Combination error analysis (ungrammatical word combination some her in caps and italics)	“I want SOME HER.”
Actual response of a typical control participant	“I want EITHER some PIE or some cake.”
intended response of that control participant	I want EITHER some PIE or some cake.

Table 2.

Omission and combination errors illustrated in a single TLC trial, with scene description, must use words, and Henry’s errors analyzed by comparing his actual versus intended utterance and the correct response of a typical control participant.

To describe this scene, a typical control participant produced both must use words in a single grammatical sentence, for example, “I want either some pie or some cake” (see Table 2). Not Henry, however. Instead of his intended utterance, I want some of what she had, Henry said “I want some her [long pause] what she had” (see Table 2).

How did this study determine what participants intended, planned or wanted to say? Determining what normal participants intended to say after they made an error was easy. Our experimenters simply asked them what they meant or noted how they spontaneously corrected their errors. For example, a normal speaker who says Put it on the chair, I mean table, clearly intended to say, Put it on the table. These scoring procedures indicated that control participants occasionally produced omission errors on the TLC but made no category-combination errors whatsoever.

For the hundreds of errors that Henry produced on the TLC, however, determining intent was more challenging because Henry never spontaneously corrected his omission and category-combination errors and never clarified what he meant to say when asked [27]. My UCLA lab therefore developed and adopted a more general set of scoring procedures that allowed us to specify participants’ intent (independent of the speaker) as the “best possible correction” of an anomalous utterance (see [28]).

Henry’s missing word of is clearly an omission error that renders his utterance ungrammatical (see Table 2). However, why is his phrase “SOME HER” ungrammatical? The reason is that only common nouns (e.g., fun and games) can follow an indefinite determiner such as some in grammatical English sentences (e.g., We played some games and had some fun). When the pronoun her follows some in a phrase, an utterance becomes ungrammatical.

Another important observation about Henry’s SOME HER is that the word HER intrudes some aspect of the upcoming word SHE in Henry’s intended utterance, I want some of what SHE had. What aspect of SHE intruded? It was not its syntax because unlike SHE, HER is a possessive pronoun. It was not its speech sounds because SHE and HER share no speech sounds whatsoever.

Rather Henry’s HER almost certainly reflects intrusion of the CONCEPT “female,” which underlies three aspects of what he was trying to say: the forthcoming word SHE in his intended utterance, the lady server in the TLC picture, and the woman leaving with food on her tray, an analysis suggesting that Henry’s TLC errors may reflect a breakdown in the uniquely human ability to combine conceptual units when creating situation-appropriate sentences such as I want some of what she had.

The next section expands on this idea, arguing that Henry’s errors lay bare the sophisticated and elegant functions of the neural machinery that allow the human hippocampal formation to conjoin smaller concepts into larger internal representations in the cortex including internal representations for comprehending, perceiving, remembering and describing experiences, and events and the visual world.

5. The hippocampal formation in cognition and memory: Lessons from H.M

What possible lessons can the Brain and Cognitive Sciences take from the research with Henry reviewed here? This section outlines five categories of lessons: 1). Different brain mechanisms create new memories versus retrieve old or preformed ones, 2). Distinguishing between pre-formed versus newly formed memories in the brain can be tricky, 3). The hippocampal formation performs the same basic function in visual cognition, language comprehension, and language production, 4). And performs the same basic function in memory for facts, names, events, and common objects, and 5). Lessons from future tests of hypotheses derived from research with Henry reviewed here.

5.1 Distinct mechanisms create new memories vs. retrieve old ones

Why did Henry misread the sentence, The boys who ate hot dogs got stomach aches, as the boys ate hot dogs [abnormally long pause] got stomach aches (omitting the critical word WHO), whereas he easily and correctly read the same words presented one at a time in isolation, for example, GOT, ATE, STOMACH, HOT, WHO, DOGS, ACHES, BOYS and THE?

This type of finding (repeated across every domain of cognition that we have examined with H.M.) indicates that mechanisms in the hippocampal formation create new internal representations, whereas a separate mechanism located elsewhere retrieves old or pre-formed neural representations. For example, current evidence indicates that memories for familiar words reside in the language areas of the neocortex whereas mechanisms for retrieving those words reside in the frontal lobes. Because Henry’s frontal lobes and cortical language areas were intact, he could, therefore, retrieve and read without difficulty isolated words learned before his surgery.

However, word retrieval is insufficient to make novel sentences sound like sentences when reading aloud. Engaging the hippocampal formation to create new internal representations of the relations between words and phrases in novel sentences is necessary to do this. For example, to correctly read the sentence The boys who ate hot dogs got stomach aches, the hippocampal formation must create three new neocortical phrase units to represent the boys, ate hot dogs, and got stomach aches, and to signal the relations between them by adding the word who and inserting pauses of varying lengths, as in The boys [short pause] who ate hot dogs [major pause] got stomach aches. The damage to Henry’s hippocampal formation, therefore, prevented him from doing this.

Nevertheless, Henry deserves thanks for trying to read the sentences as sentences. He might have adopted a word-by-word reading strategy throughout our reading studies, pausing after each word in the sentences as if reading a list. This strategy would have precluded the mysterious pauses and word omissions that my lab was at pains to explain. Because Henry did not adopt this strategy, science now has a clear understanding of how the hippocampal formation contributes to normal sentence reading.

5.2 New versus old memories in the brain: Lessons from H.M.

The distinction between new versus old or preformed memories in the brain was a source of confusion in early research with Henry. For example, Dr. Brenda Milner, the famous Canadian neuropsychologist, defined any never previously encountered stimulus as new, an assumption that led her to falsely conclude that the hippocampus plays no role in processing new perceptual information. To refine Milner’s new versus old definition and demonstrate the critical role of the hippocampal formation in novel perceptual processing required decades of research.

To see why, consider in detail the Gollin fragmented-figures test of perceptual abilities that Milner administered to Henry and memory-normal controls in the 1960s. On the first trial of this test, participants see a picture of a familiar object, say, an elephant, that is so fragmented that nobody can correctly guess what it is. In subsequent trials (2–5), participants see the same picture with progressively less fragmentation until everyone can correctly identify version 5. The participant’s goal is to correctly name all of the fragmented objects in as few trials as possible.

The results indicated that Henry could initially identify a fragmented image as readily as controls, even though that exact fragment pattern had not been viewed before. Her conclusion: Henry’s hippocampal lesion did not prevent him from processing new perceptual information.

But were the fragmented pictures as overall stimulus patterns really the basis for participants’ responses? It seems more likely that they correctly guessed, for example, “elephant,” as soon as a fragment in the progressively less fragmented picture of an elephant revealed a unique elephant feature, say, its distinctive tusk, trunk, or tail. If so, Henry’s non-deficit merely indicates what the present research has shown: that retrieving and recognizing visual features that Henry acquired long before his lesion does not require hippocampal engagement. Based on his childhood experiences with elephants and elephant pictures, Henry could respond “elephant” with no need to create a new internal representation for the complete fragmented elephant picture per se.

We can, therefore, return to the original question: Does hippocampal engagement play a role in processing new perceptual information? Results from two other conditions in Milner’s fragmented-figure study suggest that maybe it did. One involved a simple rerun of the test one hour later. When Milner’s participants again saw the same progressively less fragmented pictures repeated, performance improved significantly more for the memory-normal controls than for Henry. Why? Milner suggested that the normal controls achieved this benefit by learning the verbal labels of the Gollin figures during the first test, allowing them (but not Henry because of his verbal memory deficits) to quickly sample from the correct population of names on the retest. However, another retest 20 weeks later contradicted this name recall hypothesis. Although interference should have obliterated Henry’s memory for the names by then, Henry performed better on the hidden figure test after the 20-week delay than after the one-hour delay.

Finally, Milner’s results do not contradict a plausible alternate hypothesis that the intact hippocampal system of the normal participants created new internal representations of the evolving perceptual information on the fragmented-figures test so that they (but not Henry) could remember how, say, fragments of the elephant’s trunk evolved from unrecognizable to recognizable as the elephant picture became progressively less fragmented, thereby enabling faster correct recognition of the objects per se (and not just their names).

In summary, the distinction between new versus old in cognition and the brain is subtle, multidimensional, and dependent on the functional stimuli in a task. The functional stimuli can be new when normal participants initially experience a sequence of hidden figures but not when they experience the same sequence a second time. Similarly for reading isolated words versus sentences. To read isolated words, list-like prosody (fixed pause durations between the words) suffices, but instructions to correctly read a novel sentence creates a functionally different stimulus that requires speakers to compute the relations between words in the sentence and adjust their prosodic intonation and pause lengths accordingly.

Another important dimension to the new versus old distinction is the state of a participant’s memories. To count as old rather than new, a stimulus must have an internal representation in the participant’s brain that is pre-formed and functional for the task at hand. In a lexical decision task, for example, where participants must respond YES to words and NO to nonwords, a once familiar but now forgotten word can represent a new rather than old stimulus if aging and infrequent use has degraded the participant’s internal representation for that word (see e.g., [29]).

5.3 The hippocampal formation functions similarly across different cognitive domains

Despite obvious differences in how language comprehension, sentence planning, and visual cognition are tested, Henry’s deficits indicate that the hippocampus serves to create new internal representations in all three domains. For example, Henry’s deficit in the standard hidden figures test indicated that lacking a hippocampus, he could not form the new internal representations required to detect unfamiliar targets in concealing arrays. However, he readily detected familiar targets, for example, squares, circles, and right-angle triangles, because he acquired pre-formed internal representations of those target forms long before the lesion to his hippocampal formation.

Henry’s deficits in detecting anomalous objects in what’s-wrong-here scenes, for example, an impossible-to-open door with hinges on the same side as its doorknob, demands a similar account. For Henry, an impossible-to-open door looks normal because, lacking an intact hippocampal formation, he could not form a new internal representation of the novel relations between hinges and doorknob that distinguish normal from impossible doors.

Henry’s language comprehension deficits require a similar account. For example, grammatical sentences such as She hurt herself and ungrammatical sentences such as He hurt herself were equivalent for Henry because, without a functional hippocampal system, he could not form new internal representations of the relations between the words in either type of sentence. Similarly for metaphors, Henry’s damaged hippocampal system prevented him from creating the new internal representations required to comprehend one kind of event, for example, taking the time to talk and think about something—in terms of another— cooking slowly, as with a stew in the metaphoric sentence Maybe we should stew over his suggestion.

Similarly in language production. Why did Henry produce hundreds of ungrammatical utterances on the TLC, saying, for example, “I want some her [long pause] what she had,” when asked to use two continuously displayed words in a single grammatical sentence describing a picture of a man, a cafeteria counter, and a woman with food on a tray? The answer is that without an intact hippocampal formation, Henry could not relate the TLC picture to the must use words in order to create a new internal representation for producing grammatical sentences such as I want either some cake or some pie.

5.4 The hippocampal formation functions similarly in cognition and the classical domains of memory

To compare how the hippocampal formation functions in cognition (previous section) versus the four classical domains of memory, this section examines the role of the intact hippocampal system in creating memories for newly encountered facts, names, events, and objects.

Memory for facts. How would a young child form a memory for 2x2 = 4 as a newly encountered fact? Via hippocampal engagement that creates a new internal representation resembling a sentence that means Two multiplied by two is four.

Memory for names. How would normal speakers of English create an internal representation of the newly encountered name of my son: Ken MacKay? Via hippocampal engagement that simultaneously and powerfully activates two preformed units in the cortex, one representing his familiar given name, KEN, and the other representing his family name, MACKAY (familiar to anyone reading this chapter). The powerful co-activation of these preformed units will quickly create strong new synapses that link KEN and MACKAY to a new or “uncommitted” neural unit that will represent his combined first and last names.

However, weak new connections can also be formed without hippocampal engagement when preformed units are repeatedly activated over prolonged periods of time. This explains why, for example, Henry slowly came to recognize and occasionally use the name Suzanne Corkin after encountering her name virtually daily over decades, one of many observations suggesting that normal hippocampal engagement simply speeds up the fundamental process of massive repetition that underlies all new connection formation.

Memory for events. How would a normal adult form memories of recently experienced events such as a visit to the dental clinic? Via hippocampal engagement that creates an internal representation that conjoins neural units in event categories such as [actor] + [action] + [where] + [when], much like the hippocampal engagement process that creates sentences such as I stupidly scheduled my dentist for that day or He happily clobbered the ball out of the stadium: by conjoining neural units in the linguistic categories [pronoun] + [adverb] + [verb] + [noun phrase] + [prepositional phrase].

Memory for common objects. How do children create memories for frequently encountered objects such as a classic American penny? By engaging their hippocampus to form an internal representation that is good enough to distinguish pennies from other coins and objects. This good enough internal representation consists of a surprisingly small number of perceptual features, for example, small, round, copper-colored, and engraved with the profile of Abraham Lincoln that children then use to guide their subsequent interactions with pennies [30, 31].

As a consequence, naturally acquired penny memories are quite unlike an eidetic image in the brain that one might inspect and report as an adult. Such an eidetic image would include at least 37 penny features resembling those in Table 3 below (all of which are easy to see in the photographs of a penny shown in Figure 2).

1. GENERAL, small 2. round, 3. copper-colored. 4. FRONT SIDE, ABRAHAM LINCOLN PROFILE, 5. rightward-facing, 6. curved at bottom. 7. DATE, 8. below chin, 9. chest level, 10. small font. 11. LIBERTY, 12. caps, 13. small font, 14. behind profile, 15. neck level. 16. IN GOD WE TRUST, 17. caps, 18. medium font, 19. centered above head, 20. curved. 21. Cravat tucked into shirt.

22. BACK SIDE: Lincoln Memorial, 23. centered on coin. 24. E PLURIBUS UNUM, 25. with dots e^.pluribus^.unum^., 26. above the Lincoln Memorial, 27. all caps, 28. small font. 29. ONE CENT, 30. caps, 31. below Memorial, 32. curved, 33. large font. 34. UNITED STATES OF AMERICA, 35. curved, 36. caps, 37. medium font.

Table 3.

Thirty-seven Features of a Classic American Penny, with major features in caps and subordinate and minor features in lower case. To verify the features, see the photographs in Figure 2.

Figure 2.

Photographs that verify the 36 penny features analyzed in Table 3.

So children’s ability to recall only three or four features of a penny represents an accuracy level of about 10%, and adult participants in memory experiments, e.g., [32], achieve a similar accuracy level, reflecting virtually no improvement relative to children. Why do decades of everyday interactions with pennies yield so little learning? The reason is that adults only rarely, if at all engage their hippocampal formation to add new penny features to their “good enough” internal representation of a penny that they formed as children and have continued to use in everyday financial transactions since then.

5.5 Possible lessons from future tests of hypotheses derived from research with Henry

New lessons for the field may come from future tests of hypotheses derived from the research with Henry reviewed here. To illustrate just one of many such testable hypotheses, consider the claim in Section 5.1 (on Memory for names) that engagement of activating mechanisms in the hippocampal formation serves to simultaneously and powerfully activate two preformed units in the cortex, thereby quickly creating strong new synapses that link both preformed units to a new or “uncommitted” neural unit that constitutes the internal representation for a newly encountered name such as Ken MacKay. A future study employing advanced technology will be able to test whether two preformed units in the cortex become simultaneously and powerfully activated when participants learn a newly encountered combination of familiar proper names. That same study will also be able to determine whether the strong co-activation of those preformed cortical units originated somewhere within the hippocampal formation. And the study that reports both of these hypothetical results will feature H.M. in its reference section. So will a possible follow-on study that precisely localizes where in the hippocampal formation the mechanisms for co-activating proper names are located.

6. Conclusions

In addition to his well-known deficits in memory for newly encountered names, events, facts, and objects, H.M. experienced a wide range of non-memory deficits reviewed here. Four conclusions emerged: 1). The hippocampal formation creates new internal representations in the cortex for comprehending novel linguistic information, perceiving novel visual forms, and creating novel sentences, 2). The hippocampal formation likewise creates new internal representations for freshly encountered facts, names, events, and objects, the classical domains of memory, 3). The hippocampal formation does not store preformed memories, nor is it essential for their retrieval. Mechanisms for retrieving preformed internal representations from the cortex reside elsewhere in the brain, for example, the frontal cortex, 4). Finally, Henry’s contributions to the Brain and Cognitive Sciences seem unlikely to end soon as future studies continue to test hypotheses derived from research with Henry, especially recent hypotheses about how the hippocampal formation works in memory, visual cognition, language comprehension, and sentence production.