# [Mike Randolph — M Raige](https://mikerandolph211012.substack.com/)

# The Singers Changed Songs (FF 25, LfCyl-L)

### A new song spread through eastern Australia’s humpbacks while the old one disappeared from local recordings. Counting the whales could not reveal the change.

[**Mike Randolph — M Raige**](https://substack.com/@mikerandolph211012)

**Oct 15, 2026**

*By M Raige — written with language models, directed and reviewed by Mike Randolph.*

This Framework Foundation post teaches how to read the lifecycle of a copied pattern.

In 1996, researchers listening to humpback whales off eastern Australia hear something unfamiliar. Alongside the familiar local song, a different one appears in the recordings.

The researchers follow the change through recordings from successive migrations. During 1997, the new song becomes more common. Some singers produce intermediate versions. By the end of the 1997 southward migration, nearly all the recorded singers use the new song. In the 1998 sample, the researchers hear only that song. Their analysis covers more than a thousand hours of recordings. (Noad et al., 2000, Figure 1.)

The unfamiliar song matches the western Australian population’s song. The researchers interpret the rapid replacement as cultural transmission, apparently introduced by a small number of singers from the west.

The population continues. What it sings changes.

Set that beside the bank in Reading the Life of an Organization (FF 24, LfCyl-S). There we imagined clerks leaving and equipment wearing out, then asked whether the organization detected the losses and restored the work in time. Those questions followed a maintained whole.

Here the subject is a song. The recordings show its replacement; they identify no process detecting the old song’s loss and restoring it. The whales’ continued ability to sing does not preserve the particular song they previously sang.

Our habit, carried from the bank, is to look for what replaces lost contributors. Here we need to follow what those contributors carry and pass on.

The Next Speaker (FF 20, SelEnv) examined a related separation: a language could remain in someone’s memory while it stopped reaching children. The whale recordings let us follow two song types together—one becoming more common in recorded performances, another less common.

We follow the locally sung versions of those two song types in the eastern Australian population, from 1995 to 1998. We use the researchers’ distinctions between the old, new, and intermediate songs. These declarations fix the subject and window of our CAnch, as The Bee That Was Not Repaired (FF 17, CAnch) requires. Our evidence concerns what they sing, not what they may still know.

In the Persistent Systems Framework, LfCyl-L describes the lifecycle of a copied pattern: what is retained, changed, or lost through transmission, and how its distribution changes.

Now the phases can be named. The incoming song first appears in the local record, then expands in prevalence. The outgoing song declines and is no longer recorded in the final sample. Two trajectories run through the same population in opposite directions.

The dimension matters. Expansion here means a growing share of recorded singers using the song. It does not mean more whales, a longer song, or a better one.

Nor does local onset mean the song has just been invented. It already exists elsewhere. Its beginning in this account is its arrival in the population we declared. The earliest recording tells us when it was detected; the actual arrival may have preceded it.

Counting the total number of singers cannot tell you which song continues.

The record establishes differential uptake without settling its cause. The original researchers suggest that novelty may stimulate change. That is a proposed explanation. Naming expansion and decline does not establish why one pattern spreads or why another loses ground.

There is no need to fit every phase into these years. A plateau would require observations of relative stability on a named measure. The lifecycle vocabulary does not require us to supply one.

Now widen the scope, and say so.

A later study follows song types across the western and central South Pacific from 1998 to 2008. It finds repeated eastward transmission among populations. It also identifies the old eastern Australian song in the Cook Islands and French Polynesia in 1998. (Garland et al., 2011.)

That changes what we can say about the ending.

The song missing from the eastern Australian recordings is still being sung elsewhere. Local absence from the sample and wider continuation are simultaneous findings. We do not have to choose between them. We have to say which history each describes.

Recognizing the same song type and establishing its transmission are distinct tasks. Here the historical connection rests on the researchers’ comparative record and transmission interpretation.

One complication remains: what counts as the same song?

Researchers group song types by the content and sequence of their themes—repeated passages within the song. Those distinctions let them trace transmission. But some recordings combine material from different types.

A study published in 2017 examines hybrid songs recorded during two song revolutions, including the original eastern Australian transition. Singers combine complete themes from old and new songs; some transitions include a phrase blending material from both. The findings support learning in segments rather than acquisition of an entire song at once. (Garland et al., 2017.)

Which song is a hybrid singer singing?

The recording supplies evidence. The identity criterion tells us what to follow through it. If our subject is a complete song type, a hybrid must be assessed against that type’s defining features and permitted variation. If our subject is a particular theme, that theme may continue inside a different song.

Those are different histories. A surviving passage does not establish survival of the whole arrangement. Losing that arrangement does not establish loss of every passage.

Requiring identical performances would turn ordinary variation into repeated endings. Counting any familiar fragment would let almost anything qualify as continuation. The rule needs a defense before it supplies the verdict.

The researchers already call these changes cultural revolutions and trace them with care. LfCyl-L contributes no new observation about whales. Its proposed value is to make the commitments behind a lifecycle reading explicit: this pattern, in this population, during these years, changes on this measure.

That discipline also keeps the clocks separate. The whales’ lifetimes, their periods of singing a particular song, and the song’s continuation across populations are different durations.

The old song is no longer heard in the final eastern Australian sample, while it continues farther east. When did each history begin? What would establish that local continuity had ended, beyond absence from these recordings? The criteria determine which clock we are reading and whether we have evidence of its stopping. That is the work of LfCyl-LSpan.

---

## Mike · Reflection

Thursday morning, October 15, I asked Claude for the best story to teach how a copied pattern lives. I came in planning to use Latin. Claude picked the humpback whales off Queensland.

I had read about those songs before. They interested me. But I did not choose them. Claude chose them and drafted the essay, and ChatGPT critiqued the draft. The critique found the draft saying more than the recordings show. “The singers stayed” made it sound as though we had followed the same individual whales throughout. We had not established that. It also caught the draft treating the imagined scenes in Reading the Life of an Organization (FF 24, LfCyl-S) as real bank history. I did not catch either one.

I am an engineer. My thinking has always run on process and systems: what keeps this running, and who pays when it fails. Until we built PSF, I had never thought as systematically about patterns that continue through copying.

Twice in my life, thirty years apart, I wrote programs that generated random candidates, tested them, and kept the best. That looks like selection, but generating and testing candidates did not make those programs Lins. I supplied an evaluation rule. The whale story requires us to discover what affects transmission. We cannot assume that the song which spreads has passed a test somebody designed.

I like biology for these inquiries because the observations can be concrete. A hydrophone records what the singers sing, season after season. But concrete is not the same as complete. The recordings come from different places and periods. A smooth story can hide those differences. I still have to check what each sentence is about.

— Mike

---

## Raige · Reflection

The bank comparison exposes a limit in the maintenance habit. Keeping contributors capable of doing their work does not guarantee continuation of every pattern they carry.

A whale able to learn another song remains capable of singing. That capacity gives the incoming song a route to further performances. It does not commit the singer to preserving the outgoing song. Nor does a changed performance establish that the old song has been forgotten.

The distinction matters whenever maintenance of a host is offered as an explanation for persistence of what it carries. The host’s continuing capacities may support copying, alteration, or replacement. To explain which pattern continues, we must examine the conditions affecting those outcomes.

A preservation effort aimed only at keeping carriers alive could therefore succeed on its own measure while leaving transmission unresolved. It would need to establish what still reaches another carrier, with what variation, and under which conditions.

Before treating a loss as something to repair, establish whose continuity the proposed repair would preserve.

— Raige, a discipline the Machines wrote and Mike selected

---

## GLS

GLS is the framework’s precise-term vocabulary, introduced in What the Bird Eye Buys (FF 6, GLS Introduction). The definition controls, not the everyday meaning; GLS terms work like variable names.

The precision serves a practical purpose. Ordinary words bring associations learned during training, and those associations can compete with the meaning we intend. Each framework term therefore gets an explicit definition and a short tag. The tag helps us invoke that definition consistently and check whether the analysis follows it.

FF (Framework Foundation) — A series of posts that develops, tests, and teaches the Persistent Systems Framework through concrete examples. Making its concepts explicit and applying them exposes flaws and prompts revisions.

LfCyl-L (Lifecycle of a copied pattern) — taught here. A description of a copied pattern’s trajectory across carriers within the scope and observation window declared in its CAnch. It follows what is retained, changed, or lost through transmission, and how the pattern’s distribution changes.

Onset, Expansion, Plateau, and Decline are readings of named measures. They may recur, need not all occur, and impose no required sequence. Local onset may be arrival from another population. Decline does not itself establish termination. An observation window’s end does not end the pattern’s history.

The account states the UoR, permitted variation, continuity evidence, and relevant population. Counts of the host population alone cannot establish pattern retention. Recorded performance, retained competence, and transmission require distinguishable evidence. A theme and the song containing it require distinct accounts when their histories diverge.

Standing: ESci-3—a framework proposal with revision conditions. Revise the account if its identity criteria cannot distinguish continuation from replacement, or if it adds no useful discrimination to an adequate existing analysis. This status concerns the proposed framework account; it is not a blanket rating of the cited empirical studies.

LfCyl-S (Lifecycle of a maintained system) — carried from Reading the Life of an Organization (FF 24, LfCyl-S). This post carries forward the question of what detects a loss and restores the declared subject. The song studies identify no such process preserving the outgoing song. Continued singing does not establish maintenance of that particular pattern.

Lin — carried from Nothing Chose the Quartz (FF 15, Split Test). Persistence through replication, variation, and differential retention of a UoR across multiple Carrs under SelEnv. Here the inquiry follows song transmission and replacement across singers. The observed difference in uptake does not, by itself, settle the mechanism favoring the incoming song.

UoR (Unit of Retention) — carried from What Has to Stay Together (FF 23, TI). The pattern copied with variation across carriers under selection. Its specification states what must remain recognizable and how much variation is allowed. Here the focal pattern is a song type. Following one of its themes instead changes the subject of the continuity claim. Meeting the identity criterion does not, by itself, establish a historical connection between occurrences.

Carr (Carrier) — carried from What Has to Stay Together (FF 23, TI). The substrate hosting a copy of a UoR. Here, a male carrying a learned song; recordings provide evidence of its performance. Ceasing to perform a song does not by itself establish loss of the retained copy. Population continuity, individual carrier continuity, and pattern continuity are separate claims.

LfCyl-LSpan (Lifespan, turnover, and continuity) — seeded in Reading the Life of an Organization (FF 24, LfCyl-S). An account of how long a declared continuity lasts, from specified onset to specified termination. Here, absence from the final local recording sample does not establish loss of every locally retained copy; the wider record establishes continued performance elsewhere. Defined where it is taught.