In 1960, Ella Fitzgerald was at the height of her career, widely regarded as among the best jazz singers in the world. An experienced performer by any measure, she would typically perform 45 weeks a year. On Saturday, February 13, 1960, Lady Ella took the stage at Deutschlandhalle in Berlin in front of 12,000 fans. After singing a repertoire of popular songs from memory, including “Misty,” “The Lady Is a Tramp,” and “Summertime,” she and the band began “Mack the Knife,” first made popular by her collaborator Louis Armstrong five years earlier. Although many might not have seen the show it came from, The Threepenny Opera, the audience knew the song well—­it had been a number 1 hit for Bobby Darin just a year before—and they cheered as soon as they heard the first line, “Oh the shark has pearly teeth, dear . . .”

Then something extraordinary happened.

After the third chorus, about a minute and a half in, Ella forgets the words. Without losing a beat, or her composure, she continues to sing perfectly in time: “Ah, what’s the next chorus, to this song now? This is the one now, I don’t know . . .” She continues improvising lyrics, occasionally inserting a remembered word or two from the song. When the next chorus comes around, still stymied, she riffs on the song’s history by referencing Darin’s and Armstrong’s versions, then self-deprecatingly sings that she’s “making a wreck of ‘Mack the Knife.’ ” The band keeps playing and after modulating up a half step, a new chorus starts. Having now made up words to two entire choruses on the spot while never losing the rhyme scheme, she begins to scat while imitating Louis Armstrong. The audience goes wild.

In the history of live performance, this stands as one of the most mesmerizing, thrilling demonstrations of mastery. And what is mastery if not the ability to deal with the unexpected as though it was expected? To take an error and turn it into something better than if there had been no error at all? This level of mastery requires deep memory for the tools of one’s craft, arrived at through thousands of hours of practice, memorization of procedures, facts, and conventions, until one’s art or craft reaches a state of automaticity, what some call the flow state.

We witnessed this type of mastery in a wholly different domain with “Sully” Sullenberger, the pilot who improvised a safe recovery of US Airways flight 1549 in 2009 after it lost power in both engines over New York. Sullenberger describes what went through his mind, and his words could equally apply to Ella’s extraordinary performance. “Because I had learned my craft so well . . . I knew my profession so intimately, I could set clear priorities, and so I chose the highest priority items. And then I had the discipline to ignore everything I did not have time to do as being only distractions and potential detriments to performance.” Landing the plane required that he not think deliberately so much as act out of training and instinct, on automatic pilot as it were—the flow state. Without memory, there can be no flow. But memory is not an all-or­-nothing entity. It flows in bits and pieces; it stops and starts and sputters and spurts. Our left cerebral hemisphere then stitches the pieces together with what it thinks are plausible inferences to fill in the gaps.

Although there was less at stake in Berlin that day than on the Hudson, Ella exhibited the same kind of presence of mind and mastery as Sully did—she prioritized and didn’t panic. Ella’s highest priority was that the show must go on, and she ignored everything nonessential. Getting the words exactly right was far less important than keeping it swingin’ with the familiar melody and rhythms.

When we listen to that magical recording, we hear that Ella has at her disposal a number of cues to the nature of the song—her memory for “Mack the Knife” has not been completely slashed. Her memory for the melody and rhythm are intact and precise. We hear her playing with the beat, as she’s done throughout the song so far, using the rhythmic technique of swing that propels each phrase into the next in a long­short, long-short rhythm; she creates groove, even when forgetting the lyrics. The next line is supposed to be “There’s a tugboat, down by the river, don’t you know?” Do you notice what Ella did? She remembered the last word of the line: know. She also remembered, or reconstructed, the internal rhyme of lyricist Marc Blitzstein, with the long o- sound of don’t when she sings “I don’t know.” Ella improvises her way into a self-deprecating last line that ends with the name of the song, just as her previous chorus did. She then skips to her memory of the final chorus, where the lyrics name-check characters from the opera: Jenny Diver, Sukey Tawdry, and Lucy Brown (Louis Armstrong added Lotte Lenya, a nod to one of the original leads in the show in both Germany and the United States, and wife of composer Kurt Weill).

Ella’s stunning ad hoc performance earned her a Grammy award and was so admired that when Frank Sinatra sang “Mack the Knife” with the Quincy Jones Big Band in 1984, he incorporated some of Ella’s Berlin lyrics into his version, adding her name to Bobby Darin’s and Louis Armstrong’s in the list of singers who had previously covered the song. (Ella herself reprised pieces of the improvisation during a return to Berlin in 1962.)

As the seldom-sung final stanza of “Mack the Knife” goes,

There are some who are in darkness, and the others are in light
And you see the ones in brightness. Those in darkness drop from sight.

Ella was the bright light, and her performance of forgetting stands as one of the great monuments to memory and forgetting in the history of American music. The truth is that memory and forgetting are forever entwined. Ella retained an exquisite memory for the compositional structure, rhyme scheme, melody, the underlying harmony, accent structure, phrasing, and rhythms of the song. She had intact memory for the song’s history, who had sung it before, and the particular timbre of Louis Armstrong’s voice.

Most of us, of course, don’t possess the kind of mind—or voice—that Ella Fitzgerald did. But we all possess neural networks that represent the music we’ve heard, and these form vital windows for neuroscientists into how we build and access memories in general. The musical circuits in the brain contain potent clues as to how and why music can be an effective tool in treating injury, disease, and a range of mental and physical ills ranging from PTSD and depression to Parkinson’s and Alzheimer’s.

*

Our memory is largely governed by a brain structure called the hippocampus, which is shaped like a seahorse. It’s found in both halves of the brain, deep down inside, and it arranges the storage of short- and long-­term memory, including the connection of certain sensations and emotions to those memories. I’ve come to believe that memories aren’t actually stored in the hippocampus the way you might store a book on a shelf. In the library of the brain, the hippocampus is the card catalog that tells you what shelf the memory is on, in which wing of the library.

The hippocampus evolved in humans and other mammals to help us find things: Where’s the water? Where’s the food? Where’s the shelter? (Who let the dogs out?) Memory for where things are is extremely important to secure food and shelter. The hippocampus evolved for that, not for remembering things like your PIN number or the lyrics to a song, but we use it for those tasks anyway. In just the last five years or so we’ve learned that the best way to improve your general memory is to exercise your geonavigation skills because they make the hippo­campus function better, regardless of what type of information you are wanting to store in it. Go for a walk in a place you’ve never been! Especially on an uneven surface, where you have to negotiate obstacles like tree roots or holes. Or drive to a place you haven’t been before. If you’re not in a hurry, turn off the GPS in your car. All of this strengthens the functioning of the hippocampus. As does music.

Memory serves many disparate functions. Although it feels as though it is a single, unified faculty, it is a basket full of separately evolved capabilities. It’s certainly helping us with much more than spatial geonavigation. Memory tells us what foods have poisoned us or others, which people and animals are allies, foes, predators, or prey. Without memory, we could not have language, for we would not remember what words mean or how grammatical structure conveys their meaning.

The different kinds of memory systems humans have are sensory memory, procedural memory, semantic memory, episodic memory, geographic/spatial memory, and autobiographical memory. Sensory memory includes the afterimage you may have beneath your eyelids after seeing something, or the echo in your mind’s ear after you’ve heard something. Procedural memory, what we often call motor memory, keeps track of things like how to tie our shoes, how to use tools, how to drive a car, or brush our teeth.

Memory for specific events, or episodes, is called episodic memory, and memory for facts and general knowledge is called semantic memory. A quick rule of thumb is that if you remember where and when you experienced something, that is episodic. If you only remember the thing itself, it is probably semantic. Our semantic memory system is the storehouse for remembering everyday objects, historical events, and cultural practices—­things like the capital of Canada is Ottawa; the square root of 9 is 3; and if there’s an emergency, call 9-1-1. Extending the rule of thumb, if you know something but you don’t have a specific recollection of learning it, it is likely a semantic memory. Of course there are cases in which they blend: I remember learning my times tables by practicing them on my walk to school; I remember when I learned the name of the album Magical Mystery Tour because I was over at my friend Glen McClish’s house and his older sister played the title cut on her JCPenney record player. The term “semantic” in this context comes from the Greek word “semantikos,” which means “significant.” The idea behind the term is that semantic memory stores the significant or meaningful information that we have acquired about the world around us, and it is thought to underlie our ability to reason, communicate, and understand language.

Autobiographical memory is the most intimate form of memory; it is necessary for our sense of self. I played little league softball and I was the centerfielder. I honeymooned in Hawaii. I know how to drive a manual transmission. These combine your personal, episodic history and other general self-knowledge: Am I someone who likes vanilla ice cream or chocolate? Am I an early bird or a night owl? What do I like to do in my spare time? Who were the great loves of my life? This autobiographical memory is separate from episodic memory. Like episodic memory, it indexes events that you were a part of, but it indexes much more: the memories of the things you did and thought, insights, places you’ve been, your political views, your values and morals—all the things that make you you, things that no one else knows.

These different memory systems (geographic/spatial, sensory, procedural, episodic, semantic, and autobiographical) are subject to loss individually as the result of injury, disease, or trauma. The nature of memory, however, is that the different systems often provide access points to one another, cross-referencing and cross-correlating. Even autobiographical memory, which draws on both episodic and semantic memory, can be separately impaired; that is, dissociated from the other forms of memory. President Reagan, before he was diagnosed with Alzheimer’s, but probably suffering from it already, famously confounded an episode (a movie he acted in) with his autobiographical memory—he thought that the things that happened to the fictional character happened to the nonfictional him.

Memory is the heart of who we are, and the very private sense of what it is like to be us. Inextricably linked to memory formation are emotion and attention. Not everything we experience makes it into memory. If you are really paying attention to something, it’s much more likely to be encoded. Often the unconscious is deciding: What does this experience mean to me? What might I learn from this? How can I use this later? Say I walk into a house that I’m thinking of buying: I’m going to notice different things than if I’m thinking of robbing it. As a buyer I don’t care about the expensive coin collection in the glass case; I care about the leaky roof. Context profoundly influences what gets stored in memory! Also, we tend to remember those things that carry the biggest emotional wallop. This makes evolutionary sense: an emotional event, such as finding water when you’re lost and thirsty, or seeing a family member eat a plant that poisons them—needs to be encoded for survival.

A popular metaphor holds that memory is like a digital recording of our lives—when we want to recall something, we find the file in our memory banks and hit play. And we can fast forward, rewind, jump ahead, play in slow motion—our memory is scan­nable, and its “playback” is under our control. But this metaphor is not very good, because memories are subject to distortion or change over time, making it difficult to rely on what we pull out of our mental attic.

In reality, memory is more like film that has been edited and doctored, with entirely new scenes inserted, old ones rewritten with new characters, backgrounds, and dialogue, and some distorting lenses applied. Our recollections can be influenced by a range of external factors such as context, mood, and expectations, causing them to become altered. This can result in different people having different memories of the same event (in the book I describe how three Beatles and George Martin all had entirely different—and incompatible recollections—of critical Beatles sessions). Over time, our own memories of an event might change, sometimes drastically, as we recontextualize the experiences—we do this without conscious awareness and can’t help ourselves. These intruding distortions can also affect retrieval as we search our memory banks. If you’re searching your memory for that time you had shrimp scampi with your old high school friend Jim Ferguson, you may not find it because you and Jim never had shrimp scampi—you only talked about it—and this detail is somehow lost, merged, or rewritten. The maddening thing about all of this is that we are so certain that the thing happened just like we remember it. This often is not true; psychologist Elizabeth Loftus has shown how easily and readily false memories can form.

Understanding the neuroanatomy of the different kinds of memory is crucial for working out how music therapy and music interventions can and do work. Memories of the emotional and autobiographical features of music—Barber’s Adagio for Strings always makes me sad; Van Morrison’s Days Like This is the album she and I listened to in bed—are prone to distortion like any memory. But the perceptual components of music hold a privileged position: they tend to be more accurate than other memories, even for nonmusicians. Memory for the absolute pitches of songs that we know well are preserved in memory with great precision; the same is true for tempo, timbre, melody, harmony, and spatial location (such as, the cellos are supposed to be on my right). Memory for the structure of the lyrics, if not each and every word, is also well preserved. We remember, with astonishing accuracy, the rhythms and notes that lyrics are attached to, the accent structure, the rhyme scheme, the pauses. Even if we only know a song fairly well, it’s likely we’ve encoded enough of this componential information that we can take a good guess at what the missing words are.

Why this privileged position for music? In general, it is easier to remember information when there is some sort of organization scheme inherent in the material. The most successful mnemonic devices are ones that contain a structure predictable enough that we can learn it (ideally, just through passive exposure, as we do with our native language). And music—even more so than language—is a highly structured medium, meaning that the structure itself can help to scaffold one’s memories. Because of this structure, we don’t need to remember every single musical detail in order to reconstruct our musical memory. Any one of the individual component attributes of rhythm, melody, timbre, lyrics, and phrase structure can help us to remember others, allowing us to make plausible inferences about what comes next.

Suppose I sing:

Somewhere over the rainbow, skies are blue
And the dreams that you dare to dream really do __________

You may have forgotten what comes next, or you may never have heard the song and so never knew. But the mutually reinforcing constraints of rhythm, accent structure, meter, melody, and rhyme (it probably rhymes with blue) constrain the possible completions. How would you finish this? Really do fit my shoe, really do go to the loo, really do make a stew? Those might work in a Weird Al Yankovic song, but here they seem out of place. Really the only thing that fits is what Yip Harburg wrote for that lyric: really do come true. The lyric is memorable because structure and semantics—logic—reinforce it.

*

Musical memory is unique in that it combines procedural memory (motor memory) with auditory memory and a special kind of semantic memory, memory for the structural rules that hold everything together. Because there are so many redundancies in musical memories, they are among the most robust in our lives, surviving even advanced Alzheimer’s disease.

But musicianship requires much more precision; a multifactorial network of brain regions must cooperate. Ella was lucky in that she only lost the words. In other cases, the sounds can remain in our brains as auditory memory, but we have difficulty getting them out. I learned about this first-hand, in recent years, from my friend Rosanne Cash.

In 2008, the New York Academy of Sciences invited Rosanne and me to perform music together and talk about what goes into creating music, both from the musician-composer side and from the scientific side. Rosanne had undergone brain surgery that affected her musicianship a year before. Rosanne is a rare jewel among marquee performers. She is more focused on the experience of those around her than on her own experience—she directs her attention to the crew, the sound engineer, the stagehands, the other musicians. She is the anti-diva. Knowing that I’d be flying down from Montreal in the winter—not a good time to bring an instrument on a plane—Rosanne offered to bring me one of her guitars to play. We had a wonderful time. We had played together informally at her house, but this was the first time we had performed in public. It was so completely natural and comfortable, it felt like we’d been doing it for years. The event led to a series of similar engagements during which we talk about music and the brain, play a few songs, talk some more, and play a few more songs. It’s a novel and intimate format; the audiences say that they feel like they’re eavesdropping on a pair of friends having dinner at the next table in a restaurant.

Prior to her surgery, Rosanne had been experiencing headaches for as long as she could remember. As she got older, they became debilitating. “Once, I even dropped to my knees, the pain was so intense,” she recalls. “But a lot of times, singing, playing music myself, I would move out of the headache. It would just dissolve. That’s an interesting thing about music. You know, people say it’s very healing. It is very healing, literally.” After ten years of misdiagnoses, she underwent an MRI that showed she had a “Chiari malformation,” a rare abnormality I had to look up, that caused a bit of cerebellar tissue to slide down into the spinal cavity in the neck, pushing against the brain stem, “the geographical equivalent of starting in Vancouver and wandering down to Houston,” Rosanne said. The only solution was neurosurgery. When Rosanne arrived in pre-op, the nurse asked her what she was there for. “I’m having a decompression craniectomy and laminectomy for Chiari 1 and syringomyelia,” she said. Rosanne knew far more about it than I did.

The cerebellum is responsible for the exquisite orchestration of movement, whether it’s lifting one’s leg up a stair, opening a jar of jam, or playing a musical instrument. Any disturbance there, no matter how slight, can have outsized consequences for movements from large to small to minuscule.

Following the surgery, Rosanne tried to play the piano again. Her cerebellum, unceremoniously poked and prodded, was having none of it. She could move her fingers just fine, but all of those precise movements and sequences that had been so carefully rehearsed over a lifetime were now lost. Undaunted, she set off on a path of cognitive and musical rehabilitation, getting out her second-grade piano books and teaching herself those pieces, one note at a time. At first it was frustrating. But gradually the dormant patterns came back to life. What had taken her decades to learn the first time around, she relearned within six months. It is an extraordinary property of memory that once you’ve learned something, even if you think you forgot it, you can retrain yourself and it comes back much more quickly than you think it will, given the fits and starts that characterize the early stages.

Photo: Walter Wlodarczyk

*

The intricate processing of music unfolds component by component, commencing from the very moment we perceive its resonating sounds. The waveform, whether it reaches our ears through live performances, speakers, or earbuds, undergoes a deconstruction process that gives rise to neural representations across three distinct dimensions: frequency, amplitude, and duration. These neural representations then traverse specialized pathways where they are transformed, gradually assuming the more familiar attributes of pitch, loudness, and rhythm. Through further specialized processing circuits, the rich world of music materializes into a mosaic of the nine dimensions that engage and shape our perception: melody, contour, harmony, timbre, meter, tempo, tactus, spatial location, and distance.

Accomplished musicians possess the capacity to construct higher-order representations encompassing chords (clusters of pitches), harmonic rhythm (progressions of chords), and phrase structure (such as the distinction between 8-bar and 12-bar blues). As these higher-order representations crystallize through repeated listenings or deliberate practice, they often coalesce into meaningful “chunks.” Chunking reduces memory load, by transforming a collection of disparate parts into a cohesive entity.

Say you’ve got a favorite song on your playlist, and you’ve heard it hundreds or even thousands of times. Each time you hear it, it lays down a trace in your memory, and repeated hearings strengthen that trace. Some of the features of this song are invariant across multiple listenings—the melody, tempo, and rhythms, for example. Other features are variable, such as how loud it is, where it is coming from in space (near or far, left or right), and whether or not it is tangled up with other noises, such as the sound of your windshield wipers swishing, birds singing in your backyard, or a crowd of people chattering in a café. Each of these listenings forms a distinct trace that registers the unique features, and at the same time reinforces the previous traces that encoded the common features. Contrast this with listening to a live music performance, where the band may play the song differently each time—they may change up the rhythms, the melody, and all sorts of things. From an information processing perspective, these are distortions of the original melody. Song recognition remains accurate in the face of very large distortions, or alterations to the original way we heard it. Indeed, jazz improvisation relies on the idea that part of your brain is keeping the melody and chord changes in your head while the soloists riff on it, sometimes into a completely different melody or tonal space. Each such listening stores whatever that particular performance had in common with your previous listenings, as well as whatever is new.

Every time you experience something, whether it’s going on in the world around you or it’s a thought inside your head, it lays down a trace in your memory. When you experience something many, many times, the traces overlap and the memory becomes exceptionally strong. If each experience is slightly different, the traces still exist in a kind of package, like sticks of dried sage bound into a single stalk of incense. The traces themselves are instantiated in a pattern of synaptic connections. All that synaptic activity is governed by characteristic neurochemical activity, a balancing act across up to 100 different neurotransmitters. That high-fidelity, accurate, and absolute memory we have for the pitches, tempos, and timbres of well-known songs essentially falls out of this model for free: once you’ve heard those same compositions hundreds of times, each time writing a memory trace into your brain, you have essentially rehearsed the song, as a musician might, without even knowing it. This sort of massive rehearsal leads to overlearning, a mastery of something that causes it to become automatic—like Ella singing the melody and rhythm of “Mack the Knife” or Sully landing an Airbus 320 without having to think much about it.

Each experience, thought, or piece of information creates a unique synaptic network. When we hear a particular piece of music, some of the 80 billion or so neurons in our brains become connected in a unique way, specific to that song, and specific to that particular episode. Those neurons become members of a special subset of neurons that represent that song. Brain scans from my laboratory at McGill show that when people imagine listening to music versus actually hearing it, the activation patterns are nearly identical. That is, the act of remembering music causes activation of the same neural circuits that were active when we heard the music in the first place. That special, dedicated subset of neurons that were tied together in listening comes online again when remembering. What are otherwise disparate and isolated neurons spread throughout the brain re-form to once again become members of that original experience group; the neurons are re-membered into their original membership group, into their original formation. Remembering is re-membering.

In this respect, Ella’s memory for music appears to be not a recording, but a series of recordings, each of a musical feature— like the multitrack recordings producers make that have drums on one track, bass on another, and so on. So here, imagine that one track got erased by mistake and you still have the others to work with. It is fairly easy to fake your way through. Or imagine that you’re used to playing with a five-piece jazz combo and the piano player doesn’t show up one night. Each musician modifies what they play to some extent, to account for the change of information the audience will hear. On-the­-fly adaptations like this are possible because music contains such rich, mutually reinforcing and redundant structure. And a similar dynamic colors how we access memories—all of which are cross-indexed in our brains, with multiple, almost limitless entry points.

*

Some entry points to memories are universal (“name a song normally sung at a birthday party”), some are well known within a culture (“name a popular Beyoncé song”), some are very specific to you or your cohort of friends (“name the song we danced to over and over again that night on the beach”). In cognitive psychology, we call these entry points cues. The cues can be almost anything: sensory, emotional, autobiographical, factual, geographic, associational, a lyric, a piece of melody . . . the list goes on.

Think of a song that makes you sad. Think of one that always makes you feel happy. Think of a song that is about driving; a song that reminds you of your first kiss; one that you like to sing out loud; a song with the name “Jane” in the title. Each of these serves as an entry point to memory, a cue. But the meaning of a song can’t be summed up by an appeal to the notes that constitute it. If an especially emotional memory becomes associated with a song at any point, you will probably never hear it the same way again: memories are not static or passive, but rather dynamic and evolving. That song you loved when you were dating so-and-so might be difficult to listen to after an acrimonious breakup. Hearing it and experiencing those negative feelings will cause these new negative feelings to become attached to the memory and then stored along with it. As the most recent feelings stored, they may dominate the older ones so that it will be hard to recall a time when you ever enjoyed that song.

The fascinating phenomenon of state-dependent memory retrieval emerges as a consequential companion to these observations. When our spirits are high, we retrieve joyful memories effortlessly. Conversely, when we find ourselves submerged in sorrow, the retrieval of any gleeful recollections can prove arduous, leading us to conclude that our existence has perpetually been tinged with melancholy. This sets in motion a discouraging cycle of escalating despair, rendering it increasingly challenging to emancipate ourselves from the clutches of despondency. Trauma operates on a similar principle—whenever we recollect a traumatic encounter, entwined with all the adverse sentiments it entails, liberating ourselves from this perpetual cycle becomes a daunting endeavor: each triggering stimulus evokes a sense of panic, imprinting itself as an additional memory trace atop the accumulated others. Even music associated with a distressing period of one’s life can summon forth fresh negative memories in individuals afflicted by depression, exacerbating their condition.

The multifeatured, multifaceted aspect of music, and the different ways it can be accessed, are what allows it to be so powerful with Alzheimer’s, depression, PTSD. When nothing else gets through, a little snippet can shoehorn its way into consciousness, mood, and memory itself.

Neuroscientist Amy Belfi found that music-evoked autobiographical memories are more vivid than autobiographical memories evoked by other familiar cues, such as photographs. In all our musical adventures, from childhood on, our brains are seeking similarity and patterns. Our brains are giant prediction machines, trying to figure out what will come next to us in the world. To do that, we need to extract patterns, commonalities, and create groupings of objects that are similar. We make sense of orchestras because we don’t hear the 100 or so instruments as separate—­our ears naturally group similar sounds together. All the violins automatically get merged in our head into a section (or two or three when they are playing different parts). Similarity can also apply to melodies as they are handed off from one instrument to another. Our brains apply certain rules that are inborn, to define similarity in several ways. An example is that notes that are close to one another in pitch space tend to get grouped together, and they separate—pop out—from other note clusters that are far apart from them. This is what allows us to hear a bass line as distinct from a guitar or clarinet line. There are over a half dozen rules for detecting similarity, most of them worked out by the Gestalt psychologists back in the late 1800s; these “configurationists” rejected the dominant notion that one could understand a complex phenomenon by studying the parts one at a time.

This penchant for similarity draws us toward music that has something in common with music we’ve heard before, the strongest example being different versions of the same song, followed by songs that are based on the same structure (from 12-bar blues up to sonatas and symphonies). When we’ve learned some of these fundamental laws that govern the music we listen to—that is, when we’ve built up dedicated brain circuits to do the work—our brains take over and handle the ongoing tracking of melody, rhythm, and other musical features without our conscious control. Our brains notice small and large deviations from what they expect, and those get encoded as separate memory traces, allowing us to build up a larger vocabulary of patterns, enriching our experience of music.

All of these musical elements are being stored along with extramusical and contextual memories—smells, tastes, feelings you’re experiencing, the place you are, the people you’re with. Each of these independently serves as a retrieval cue. This is why a smell can trigger a memory, and particularly if it is an odor that you don’t experience all the time, and so is uniquely associated to a very specific memory. That is the key, really—the more specific a memory cue, the easier it is to retrieve what you’re looking for. If I ask you, “Do you remember that shower you took three Wednesdays ago?” you’re unlikely to have a specific memory for it—it’s blended in with all the other showers you’ve taken recently. But if I ask, “Do you remember that shower you took recently when the water suddenly turned ice cold?” that, presumably, only happened once in the last few weeks, and so the uniqueness triggers the memory.

All these processes can be tied together: unique memory cues, multiple trace memories, similarity, and memory for individual features of music. When I simply imagine a song without hearing it, that hippocampal index tells me, These are the circuits that were involved, the neurons that fired, the firing rates of those different neurons—all of that info in the neural library card catalog I described. But I don’t get it back perfectly, which is how I know it’s a recollection and not the actual thing—unless I’m hallucinating to the point where I can’t tell the difference between reality and imagination. This is another point about memory—neurochemical tags tell us when something we are thinking about is a memory, versus something that we dreamt or hallucinated. That’s a good idea, evolutionarily. You don’t want to pick a fight with Caveman Og just because you dreamt that he stole your food. But this system of neurochemical tags sometimes goes awry. When that happens we can have experiences such as déjà vu, the feeling that we’ve been somewhere or done something before when we really haven’t; as you are actively encoding the experience and putting it into memory, something goes wrong and it gets spit back out as though it is coming from memory, rather like a kitchen sink backing up. Jamais vu occurs when you’re doing something you’ve done many times, but it feels like the first—as though you’ve never (jamais) done it. Jamais vu is a state that every musician and actor seeks to attain when they are performing. The late Burt Bacharach probably sang “Alfie” more than a thousand times. But as pianist and producer Shelly Berg said, “Every time he sings it, it sounds—and feels—like he is singing it for the first time ever. And that’s what audiences connect with.” Psychotherapy can often trigger déjà vu and jamais vu because it aims to recontextualize memories.

We also connect anything we’re hearing with what we’ve heard before, and that deeply enriches the listening experience. Our musical memories begin before birth and evolve as we grow. As long as we are listening, we keep adding to those memories, reinterpreting the new in light of the old, and in the process, recontextualizing the old. Your brain is changing all the time, and so are your emotional states, and so the “music medicine” you receive is essentially a brand-new medicine each time. The brain that hears that favorite song today is different from the brain that heard it last month. The journey through music is a never-ending one, guiding us through moments of joy and sorrow, discovery and nostalgia, a faithful friend that is always there to lift us up or help us through rough times. As long as we keep listening, we move forward, one note at a time. ♦

Excerpted from I Heard There Was a Secret Chord by Daniel J. Levitin. Published in the United States by W.W. Norton, August 2024. Copyright © 2024 by Daniel J. Levitin. All rights reserved.