Why Music Sticks In The Mind

AAnyone who has watched a person with advanced dementia fail to name a spouse and then sing four verses of a hymn without hesitation will understand why musical memory is so often described as a separate faculty. The impression is strengthened by the ease with which songs learned in adolescence return decades later, apparently untouched by the decay that removes almost everything else from the same period. Popular accounts explain this by positing a dedicated store, a part of the brain reserved for music and spared by the processes that erode ordinary recollection. The proposal is attractive because it converts a puzzle into a plain anatomical fact, and because it fits a long tradition of treating music as something set apart from language. It is, however, the explanation that specialists have found hardest to support, since the evidence for a distinct musical store is largely an inference from the very phenomenon it is supposed to explain.

BYusra Behrend, who directs the Coleridge Institute for Music Cognition, argues that the comparison on which the whole impression rests has never been conducted fairly. A song heard on the radio in adolescence may be encountered several hundred times over a decade, whereas the news bulletin that followed it is heard once; if repetition alone accounted for the difference, no special store would be needed. To test this, her group assembled a set of spoken proverbs and a set of unfamiliar songs, and exposed 220 participants to each item exactly forty times over eleven weeks. When the two sets were tested a year later, the advantage held by the songs had shrunk from the fourfold gap reported in survey research to a difference of about 8 per cent. Behrend maintains that what remains is real but modest, and that most of what the public regards as the power of music is simply the arithmetic of exposure.

CRasmus Teigen, a neurologist at the Falkbridge Memory Clinic, contends that the exposure argument, however tidy, cannot reach the cases that provoked the debate in the first place. Patients whose episodic memory has been destroyed by disease can still perform songs they knew beforehand, and performance of that kind, he insists, draws on procedural systems that survive damage to the structures supporting recollection of events. Between 2016 and 2019 his unit followed 47 such patients and found that 34 could complete lyrics they had not sung for years, while none could report where or when they had learned them. That dissociation, Teigen argues, is not something a difference in repetition can produce, since repetition should strengthen both kinds of memory together. Music, on this account, is not kept in a private room but is unusual in being encoded as an action as much as an item, and actions are held in a system that disease reaches late.

DThe dissociation is striking, yet Teigen's evidence carries limitations that he acknowledges only in passing. Every patient in the sample had received formal musical training before the age of twenty, a selection that is hard to justify given the conclusion drawn, and the songs used were chosen by relatives who already believed the patients would know them. Behrend adds a further objection: completing a familiar lyric requires almost nothing beyond recognition, whereas reporting when a song was learned demands the reconstruction of an episode, so the two tasks differ in difficulty as well as in the systems they engage. A dissociation between an easy task and a hard one, she notes drily, is the least surprising finding in the discipline. Teigen concedes the point for lyric completion but replies that his patients also reproduced tempo and phrasing that no recognition test could have supplied.

EBoth camps agree that the effect, whatever its source, is not evenly distributed across a life. Songs first encountered between the ages of roughly ten and twenty-five are recalled far more readily than those met later, a bias that has appeared in every survey conducted since the pattern was named in 1986. It holds only for material that was salient at the time; songs merely present in the background during those years show no advantage at all, which suggests that the window matters less than what was done inside it. There is a second restriction that is easy to overlook. The advantage attaches to the sung form of a text and does not transfer to the same words when they are spoken, so that participants who learned a lyric to a tune and were later tested on the words alone performed no better than those who had learned the words as prose.

FThe most defensible position is that the two accounts describe different portions of the same effect rather than competing for the whole of it. Exposure, as Behrend has shown, explains why ordinary listeners overestimate the gap, and it does so without any appeal to specialised anatomy. Procedural encoding, as Teigen has shown, explains why a residue survives conditions that destroy almost everything else, though it has yet to be demonstrated in anyone who was not musically trained. What follows for practice is less dramatic than either side implies. Programmes that play familiar songs to patients rest on the least contested part of the evidence, namely that retrieval of a well-rehearsed sequence is easy; the further claim that such programmes restore access to the episodes surrounding the song remains, at present, an expectation rather than a finding.