A patient in the late stages of Alzheimer’s may not be able to identify the room she is sitting in, the year she is living in, the meal she ate twenty minutes ago, or the face of the daughter holding her hand. She may have lost the ability to follow a sentence longer than a few words, the ability to dress herself, the ability to recognise her own reflection in the mirror. The disease, by this point in its progression, has destroyed the hippocampus, has spread through the temporal lobes, has begun consuming the cortical regions that hold most of what made her recognisably herself. And yet, if someone in the room begins humming “Moon River” or “The Tennessee Waltz” or whatever song she danced to at her wedding sixty years earlier, the patient may sing along — word for word, melody intact, the emotional cadence of the song coming back to her face as clearly as if no time had passed at all. This is one of the better-documented and stranger features of advanced Alzheimer’s. It has been observed in care facilities for decades. Until comparatively recently, nobody knew exactly why.
According to an announcement from the Max Planck Society describing the 2015 study by Jörn-Henrik Jacobsen and colleagues at the Max Planck Institute for Human Cognitive and Brain Sciences in Leipzig, the first complete neuroanatomical explanation for this phenomenon was published in the journal Brain in August 2015. Jacobsen’s team conducted a two-part study designed to answer a very specific question. The first part, using 7-Tesla functional magnetic resonance imaging on 32 healthy young volunteers, mapped the precise brain regions that activate when a person listens to music they have long known — the songs of their youth, the standards of their cultural environment, the music that has been heard so many times across so many years that it has become part of their baseline mental furniture. The team identified two regions that showed strong, specific responses to long-known music: the caudal anterior cingulate cortex and the ventral pre-supplementary motor area. Both regions sit roughly in the frontal portion of the brain. Neither had previously been understood as a primary repository of musical memory.
What the second part of the study found
The second part of the study examined the same two regions in the brains of 20 patients diagnosed with Alzheimer’s disease, alongside 34 healthy age-matched controls, to determine the degree to which these specific musical-memory regions had been damaged by the disease. Three separate Alzheimer’s biomarkers were analysed: cortical atrophy (the actual loss of brain tissue, measured by MRI), glucose metabolism (a measure of how actively the tissue is still functioning, measured by FDG-PET), and amyloid-beta deposition (the presence of the protein plaques that are the earliest physical marker of the disease, measured by flobetapir-PET). As reported in the Jacobsen et al. paper itself in Brain, the musical-memory regions showed substantially less cortical atrophy than the rest of the brain in the Alzheimer’s patients, and substantially less disruption to glucose metabolism. The regions had not been spared by the disease in any absolute sense — amyloid-beta deposition was present in the musical-memory regions at roughly the same levels as the rest of the brain — but the standard progression by which Alzheimer’s typically destroys brain tissue (amyloid accumulation, followed by metabolic decline, followed by physical atrophy) had been substantially slower in these specific regions than in essentially every other part of the cortex.
The interpretation that follows is direct. The musical-memory regions of the brain are not immune to Alzheimer’s. They develop the same early-stage amyloid pathology that the rest of the brain develops. They are, however, unusually resistant to the later stages of the disease — the metabolic collapse and physical atrophy that destroy the function of the affected tissue. In a patient with advanced Alzheimer’s, whose hippocampus and temporal cortex and prefrontal regions have been largely destroyed, the caudal anterior cingulate cortex and the ventral pre-supplementary motor area may still be working — and may still contain, intact, the neural representations of the music that the patient learned over a lifetime of listening.
Why music memory is structured this way
The deeper question of why musical memory should be stored in these specific regions, and why these regions should be unusually resistant to Alzheimer’s progression, remains substantially open. Per a commentary by Clark and Warren published in the same August 2015 issue of Brain, the most plausible current explanation involves several converging factors. Musical memory is, neurologically speaking, not a single memory system but a hybrid: it draws on the declarative-memory system (which stores what you know about a song), the procedural-memory system (which stores how to produce the sounds with your voice or fingers), and the emotional-memory system (which stores how the music makes you feel). The hybrid nature means that musical memories are distributed across multiple brain regions and reinforced through multiple types of neural encoding, making them substantially more redundant — and therefore more resilient to localised damage — than the autobiographical memories that depend more narrowly on the hippocampus and surrounding temporal-lobe structures.
The redundancy hypothesis is supported by the broader clinical literature on what kinds of musical memory persist into late-stage Alzheimer’s. As detailed in the PubMed indexing of the Jacobsen paper and the related clinical literature on musical memory preservation, what survives most reliably is what neuroscientists call procedural musical memory (the motor patterns required to sing or play a song) and semantic musical memory (the recognition of familiar tunes and the ability to identify their lyrics). What is lost earlier is episodic musical memory — the autobiographical context surrounding a song, the memory of where and when the patient first heard it, the association of the music with specific people or events. A patient in advanced Alzheimer’s may sing every word of a song she has not heard in fifty years without being able to identify what the song is called, when she first learned it, or whose memory it is now embedded with. The song itself remains. The story around the song has gone.
What the finding has meant in practice
The clinical applications of the neuroanatomical finding have, in the decade since Jacobsen’s paper was published, become substantially more sophisticated than the broad practice of “playing music for dementia patients” that preceded it. Music therapy is now a recognised intervention in memory-care settings around the world, with personalised playlists assembled from the specific songs each patient remembers from their formative decades — typically the music they heard between roughly ages 15 and 25, when the brain’s musical encoding is most active. The Music & Memory organisation, founded by Dan Cohen in 2010, has now placed personalised playlists in more than 5,000 care facilities worldwide. The documented effects, in well-conducted studies, include reduced agitation, improved mood, brief but real windows of lucidity, and the kind of moments in which a patient who has not recognised her own daughter in months will turn toward the music, sing along, and look at her daughter with what appears, briefly, to be recognition. The disease has not been cured. The neurons damaged earlier in the disease’s progression have not been restored. What the music does is reach a part of the brain that the disease has not yet destroyed, and allow the patient — for the duration of the song — to be, in some recognisable sense, herself again. The two specific brain regions that make this possible were not identified until 2015. The phenomenon they explain has been observed in care facilities for as long as we have had care facilities. The mechanism, once you know where to look, is structural. The mercy, once you recognise it, is real.
