Explorations in tunemaxxing

Musical earworms are inevitable and frequent for me. This can be pleasant. Some music is of poor quality, so earworms from it are less pleasant. Much music has lyrics in languages, so earworms from it evoke complex, task-unrelated thoughts.

Earworms become a net positive insofar as they replicate high-quality, wordless music. Repetition gets boring is unpleasant, so it'd be better to form earworms from a wide variety of music. If the countable form of music is "tunes", and consistently pursuing a value is described by the gerund suffix "-maxxing", then I should be able to casually describe methods to make the above conditions true as "tunemaxxing".

§ Exposure

To remember a tune, you must hear it. I listen to music, grouped as album or soundtrack, whenever alone enough, usually while doing something else. Proper enthusiasts might listen to music on its own, with full focus. As I listen just to tunemaxx, a lesser, peripheral attention to the music suffices, if perhaps more slowly.

I pick albums that I know to be wordless. If, by my subjective intuition, it's also high-quality, I listen to the album again for a few repeats. After a round, when I hear many fragments of the tracks, I anticipate what follows. After several rounds, music from that album will sometimes spontaneously appear as full earworms.

§ Reading

Memory is associative, so tunes come to mind in large part sith you observe and experience stimuli related to those tunes. If you form more associations from cheap stimuli to tunes, the tunes so associated should form earworms more often.

A tune's title is a sensible choice for such a cheap stimulus. Spaced repetition is a sensible choice for how to intentionally associate memory.

My Memoire couldn't play audio, so I used the more popular Anki to associate tunes to titles. For each track in several albums I liked, I made a card that showed a title on the front and played the tune on the back.

But most tunes are multiple minutes long. Anki cards ought to be maybe ten seconds long, or you miss much of their efficiency. The back of each card holds not an entire track, but a crucial six-second excerpt.

I tried to get a program to automatically extract "crucial" excerpts, with the heuristic that the loudest interval is the most important. This gave mediocre results. I ended up picking excerpts manually.

However I picked these excerpts, it was near-impossible for me to correctly recall tunes from just reading the titles. Perhaps it would be easier if I built up the reverse association first.

§ Labelling

Anki's note types make it easy to add reversed versions of cards. So I made another card for each track that played an excerpt and asked for the title. These cards were much more usable.

Memory associations are much stronger from cue to response than the reverse. When the tune is just the cue, studying the pair only helps me recall the tune as something that I think of more, not reliably prompted by anything else. This is the same result I get from just listening to music, only weaker, sith the excerpts are up to a hundred times shorter than full tracks.

The main value of tune-to-title cards was to familiarise me with track titles that I barely noticed when first listening.

§ Advancing

It turns out studying title-excerpt pairs is basically ineffective.

• Titles of wordless tracks have little logic. Learning is much more rote than semantic, i.e. hard. Play a video game, and track titles might make sense by association to stages in the game, but I never played the games.
• Separate tracks in the same soundtrack may share motifs closely enough to mix them up.
• Titles, the cue of the pair, scarcely appear in real life to trigger the response, at least until you deliberately look at song titles.
• Most tunes hold detail and content not covered in even the most interesting excerpt. Read the title and you'll recall the crux, but the crux summons the rest of the tune only if you've already heard the crux in context enough.

You can address the last issue by listening to a tune in full, which gives a sequence of segment-successor pairs. To reap the efficiency of spaced repetition, pick several excerpts from each tune, and split each excerpt to form a pair.

So for each track in another album, I made two cards that each play a four-second piece and ask for the next four seconds.

§ Reinforcement

Segment-successor pairs are, so far, also ineffective. It turns out most music is too complex to be summarised with just a few excerpts per track.

I could split each tune into a continuous chain of Anki cards, and thereby memorise full tracks. But most tunes have some relatively indistinct quiet parts — common questions to multiple answers are the bane of flashcards — and the overhead of an SRS would make the endeavour less efficient than normal listening.

These spaced repetition techniques failed, but their rigour showed what caused me to remember tunes well.

• Re-listening to an album long after my first few cycles brings it back to mind much more strongly than any one of those first listens.
• Trying to perform a tune helps me remember it more clearly for long afterward.
• Hearing even a fragment of a tune in a new, more vivid context solidifies the title-tune association.