But in spite of this impoverished test performance, Henry’s semantic storehouse did retain tiny traces of his experiences during the postoperative years. We found tantalizing evidence that his amnesia was not absolute. His recognition memory was not totally wiped out. When we asked him to select the best definition for words and phrases among four choices printed in a test booklet, he recognized fifty-six percent of the definitions for pre-1950s words and phrases and thirty-seven percent for the post-1950s words. The latter score, though impaired, was better than chance (twenty-five percent). At the same time, his ability to recall these words and phrases was clearly compromised. He successfully fished up sixty-one percent for pre-1950s words and phrases but only fourteen percent for post-1950s words and phrases.17
Likewise, Henry’s performance on the word vs. nonword task (“Is the following a real word?” thweige) and the name-categorization task (“Is or was the following a famous person?” Lyndon Baines Johnson) also demonstrated some, albeit scant, knowledge of post-1950s words and names. When the examiner asked Henry to define words that had entered the dictionary after he became amnesic, he usually did not know the answer. But rather than just say, “I don’t know,” he would harness his intellect and make an educated guess. Many of his responses construed literal meanings from bits of words and phrases. He defined angel dust as “dust made by angels; we call it rain,” closet queen as “moths,” cut-offs as “amputations,” and fat farm as “a dairy.” The brain areas in Henry’s cortex that housed his semantic store for old words and names could not engage in learning new words and names because the critical interactions between these key cortical areas and intact hippocampal circuits were missing. For the most part, he could not consolidate new semantic memories. Returning to the question of whether an amnesic person placed in a foreign country would learn the language but forget having been there, it appeared from Henry’s performance that even in a natural learning environment in which the individual heard, spoke, read, and wrote the language in meaningful contexts, he would be unable to expand his word knowledge with new entries. This deficit was part and parcel of Henry’s profoundly impaired declarative memory.18
Extending this line of thinking, we conducted several experiments in the mid 1990s following up on Henry’s lack of semantic learning for new words. The question that intrigued us was whether this deficit in acquiring novel information through declarative memory would carry over to nondeclarative memory, which we knew was intact. Although he could not consciously recollect new words, would he show priming with these words? Would he be able to process them normally through preserved nondeclarative-memory circuits? Specifically, we wondered whether Henry’s priming performance would be different with words that came into common use after the onset of his amnesia and thus were novel to him—post-1965 words like granola, crockpot, hacker, and preppy—compared with old words like blizzard, harpoon, pharmacy, and thimble.
A graduate student in my lab created four repetition-priming tasks. Two of them assessed Henry’s word-completion priming—one using words that were in the dictionary before his 1953 operation and the other using words that entered the dictionary after 1953. The two other tasks measured Henry’s perceptual identification priming—one with pre-1953 words and the other with post-1965 words. All four tasks engaged Henry’s nondeclarative memory circuits. In this experiment, each priming task had a study condition and a test condition.19
In the study condition for word-completion priming, Henry read words aloud as they were presented one at a time on a computer screen. The test condition occurred one minute later when he saw three-letter stems one at a time on the computer screen (GRA- for granola, THI- for thimble). Half of the word stems corresponded to words in the study list, and the other half to unstudied words. The examiner asked Henry to complete each stem with the first word that jumped into his mind. If he completed THI- to thimble, rather than to more common words that begin with THI- (think, thin, thief, thick), then he was showing the effect of prior exposure to thimble, the essence of word-completion priming. The word-completion priming score was the number of stems completed to studied words, like thimble, minus the baseline score of words completed by chance to similar unstudied words. Henry showed priming whenever he completed significantly more word stems to studied than to unstudied words.20
Perceptual identification priming also included a study condition and a test condition. During the study condition, we flashed words on a screen, one by one, for less than half a second and asked Henry to simply read each word aloud. In the test condition, Henry again saw words presented very briefly on the screen and again read each one aloud. He had seen half of the test words in the study session, and the other half were new words he had not studied. His perceptual-identification priming score was the number of correctly identified studied words minus the baseline score of correctly identified unstudied words. Henry again showed priming. When the words flashed briefly on the screen during the test, he was able to read more of the studied than unstudied words. During the study condition, each time Henry read a word, the experience left behind a memory trace that beefed up the representation of that word. Then, during the test, when he saw studied and unstudied words for less than an instant, he was more likely to read the studied words because of their enhanced mental representation.21
Henry gave us clear-cut results on these four priming tests. For the pre-1950 words—blizzard, harpoon, pharmacy, thimble—he showed normal performance on both kinds of priming. On stem-completion priming, he completed more stems to studied than unstudied words, and on perceptual identification priming, he correctly read more briefly presented studied than unstudied words. We found a different breakdown of results with the post-1950 words—granola, crockpot, hacker, preppy. Henry’s perceptual-identification priming was still normal, but his word-completion priming score was zero. Why? Henry lacked the necessary preexisting semantic representations of the novel words in his mental dictionary for word-completion priming to occur. He did not need this representation for perceptual-identification priming because the task engaged low-level visual processes independent of language. This finding of impaired word-completion priming with novel words, but robust perceptual identification priming with the same words, told us that different mechanisms sustain each kind of priming—one disrupted in Henry and the other still functioning.22
Henry’s differing results for these two nondeclarative-memory tasks—conceptual priming and perceptual priming—were important because they highlighted the fact that the two procedures activated brain circuits at different levels of information processing. Word-completion priming operated at the level of word knowledge stored in the temporal and parietal lobes. The word granola was unfamiliar to Henry and not represented in his mental dictionary—his semantic store. Consequently, when he read it in the study list, he did not have a representation available to benefit from the extra processing needed to complete the stem GRA- to granola in the test. He had no difficulty, however, completing GRA- to grandmother because it was already part of his preoperatively stored world knowledge. This experiment confirmed our previous work showing that Henry had normal word-completion priming when we tested him with words that were familiar to him.23
In contrast, perceptual-identification priming operated at the more elementary level of visual perception. Henry simply read granola aloud in the study list, and the related processing in his visual cortex was all he needed in the test to say the word aloud when it flashed on the screen. Henry showed comparable perceptual-identification priming for pre-1953 and post-1953 words because his visual cortex in the back of his brain, where these computations were performed, processed meaningful (blizzard) and non-meaningful (granola) letter strings in the same way.24
Amnesic patients, including Henry, characteristically show learning on nondeclarative memory tasks, such as priming. Our finding that Henry was impaired on word-completion priming with novel words was an exception. This deficit resulted directly from Henry’s inability
to consolidate and store post-1953 words. In this priming task, he read words like granola, crockpot, hacker, and preppy on a computer screen, but he had no entry in his mental dictionary for them to activate. He could not benefit from reading the words and was, therefore, unlikely to respond with granola when he saw GRA-. His performance showed that the circuits in his brain that supported this kind of priming were distinct from those that sustained his normal perceptual-identification priming with the same words. This division of labor occurs in our brains as well.25
Although Henry demonstrated over and over that he had lost his ability to make new semantic memories, to learn new facts and retain them, he occasionally astonished us by remembering things we never expected him to know. One day, in casual conversation, research associate Edith Sullivan asked Henry what he thought of when he heard the name Edith. She was flabbergasted when he said, “Edith Bunker.” Edith Bunker was a fictional character from All in the Family, a TV show that began in 1971. The next day, she brought up the subject again and asked him: “What’s the name of the star of the program?”
“Archie Bunker,” Henry said.
She asked him what Archie Bunker called his son-in-law, adding, “it’s not a very pleasant name.”
After a long pause, Henry said: “Meathead.”
These surprising and seemingly random new memories appeared from time to time like driftwood washing up from an empty sea, and they felt like small miracles to those of us accustomed to seeing Henry fail to remember. In the early years, Henry’s mother often believed he was improving—she told us, “He knows things that he has no business to know.” In hindsight, it is clear that Henry’s amnesia was permanent, and these fragmentary scraps of memories were the exception not the rule. Compared with any normal person, Henry’s ability to remember the experiences of his life was always abysmal. In 1973, he was not able to identify common names in the news at the time, “Watergate,” “John Dean,” or “San Clemente,” despite the fact that he had heard them repeatedly on television every night. He did not know who the president was, but when told his name began with an “N,” he said, “Nixon.”
In July 1973, I asked Henry whether he could tell me about Skylab. He replied, “I think, uh, of a docking place in space.” He also correctly said that there were three people in Skylab at the time, but immediately added: “But then I had an argument with myself, then was it three or five?” When I asked him, “What’s it like to move around up there?” he answered, “Well, they have weightlessness—I think of magnets to hold them on metal parts so they won’t float off away, and to hold them there so they can move around themselves and stay in one area, and they won’t move away unvoluntarily [sic].” Henry accepted new technologies, such as computerized tests, without blinking, but he could not keep up with changing times in other ways—he once incorrectly stated that a hippie was a dancer. The world changed, but for the most part Henry was left behind.
The little islands of remembering that Henry treated us to over the years spurred us to investigate his semantic memory more intensely. We decided to again examine his knowledge of celebrities because it allowed us to take advantage of his extensive exposure to magazines and television through which he constantly took in information about famous people and noteworthy events. A 2002 experiment when Henry was seventy-six further characterized the depth of his semantic learning by probing for details of individuals who became famous after his 1953 operation. Our previous studies had not explored the depth of the material he was able to acquire postoperatively. The fragments of knowledge he had demonstrated in our prior experiments were sufficiently sparse that either declarative or nondeclarative learning could have supported them. (Two other groups of memory researchers had each shown that a severely amnesic patient, exposed to many weeks of training, could gradually learn and retain new semantic facts. The patients demonstrated their scant factual knowledge through performance but could not consciously remember the learning episodes. The learning was nondeclarative.)26
In our study, two graduate students addressed an ongoing controversy. Some researchers predicted that regions beyond the hippocampus proper could support some semantic—consciously accessible—learning. But another lab argued that amnesic patients’ episodic and semantic memory circuits are equally impaired, so that semantic learning would be impossible in someone like Henry who had no episodic memory. We designed experiments to reveal the route through which we acquire semantic knowledge. The new theoretical issue that motivated our experiments was whether all new information enters your brain in the form of episodes and later becomes general knowledge. For instance, the first time you ate peach ice cream and discovered you loved it was at the beach when you were twelve years old. Over the years, you forgot that episode but continued to rate peach ice cream as your favorite. This fact began as an episodic memory and later became a semantic memory. So the question was, do all memories have to start like this, as episodes, or can they bypass your episodic circuits and enter your brain as semantic knowledge? Because the hippocampus is necessary for episodic learning, we could also phrase the question in this way—can semantic learning occur without a functioning hippocampus? Because Henry’s hippocampal damage was complete and because he had virtually no episodic memory, he was the perfect case for testing this hypothesis.27
In our initial experiment, Henry heard the first names of famous people and was asked to quickly complete the name with the last-name that immediately came to mind. Because this task did not ask for a famous name—any name would do—implicit (nondeclarative) memory could support performance, automatically without awareness. For example, Henry completed Ray with Charles, not because he knew that Ray Charles was a famous individual, but because he had formed an unconscious associative link between Ray and Charles when he had heard or read those two names together. Charles just popped into his head. Henry completed fifty-one percent of first names with the last names of individuals who were famous before his operation, and a surprising thirty-four percent with the last names of individuals who were famous after his operation. He completed Sophia with Loren, Billie Jean with King, and Martin Luther with King. Although Henry’s ability to acquire this kind of information was clearly impaired relative to healthy participants, the results suggested that he had at least thin knowledge of postoperatively famous individuals, sufficient to support a link between their first and last names.28
The next day we presented meaningful cues for each individual—“famous artist, born in Spain in 1881, formulated cubism, works include Guernica.” After this cue, Henry heard, “When I say Pablo, what is the first word that comes to mind?” This kind of cueing boosted Henry’s last name identification equally for individuals who came to prominence before and after his operation. The fact that he benefited as much from semantic cueing about post-1953 names as about pre-1953 names suggested that his new knowledge, like his preoperative knowledge, was incorporated into schemata. These organized semantic networks, clusters of related information, are capable of supporting conscious recall. This finding adds to the body of evidence that Henry was capable of limited declarative, semantic learning.29
In a critical companion experiment, we examined the scope of Henry’s new semantic knowledge by focusing on the amount of detail he could provide about celebrities. First, he saw two names side by side. One was famous and one was a name randomly taken from the Boston area telephone directory. When asked, “Which is the name of the famous person?” he was correct ninety-two percent of the time for names he had come across before his operation and an impressive eighty-eight percent for those he had encountered afterward. Then, for each individual Henry selected as famous, we asked him a key question: “Why was that person famous?” Even though the semantic information Henry provided for post-1953 celebrities was impoverished compared with the responses of control participants, and compared with the information he himself furnished for individuals who were celebrities before 1953, the result was astonishing. He was capable of providing accurate
, distinguishing information for twelve people who became famous after 1953. He knew that Julie Andrews was “famous for her singing, on Broadway,” that Lee Harvey Oswald “assassinated the president,” and that Mikhail Gorbachev was “famous for making speeches, head of the Russian parliament.”30
This investigation demonstrated that some semantic knowledge can be acquired in the absence of any discernible hippocampal function. Henry, despite a lesioned hippocampus, showed that he was capable of learning information about celebrities who came to fame after his operation, providing robust, unambiguous proof that at least some semantic learning can occur in the absence of episodic learning.31
While it is interesting to note the extent to which semantic learning can occur without hippocampal function, it is equally interesting to note the ways in which this learning was different from that of our control participants. Henry was capable of generating semantic knowledge about only a fraction of the individuals who were well known to controls. Moreover, the information he generated about the postoperative people familiar to him was relatively sparse compared with control participants and compared with the amount of information he was able to generate about preoperatively famous individuals. Henry, for example, could not provide even the sex of some of the names he selected as famous. For instance, he said that Yoko Ono was “an important man in Japan.” Further, while the control participants were better at generating knowledge about individuals who became famous recently compared with remotely, consistent with the general pattern of forgetting typically observed in healthy people, Henry showed the reverse pattern of performance. In addition, his ability to produce information about postoperatively famous individuals was sporadic. For example, in prior attempts to assess his learning about famous individuals, he had successfully identified Ronald Reagan as a president and Margaret Thatcher as a British politician, but was unable to generate their occupations during this investigation. And during the 2002 investigation, he indicated that John F. Kennedy had been assassinated, whereas on prior occasions, he had said that Kennedy was still alive.32
Permanent Present Tense Page 30