Storing and remembering information is not a one-time event, we also want to effectively transmit it to future generations.


April 14, 2021

Baskets for the mind.

If it is a human thing to do to put something you want, because it’s useful, edible, or beautiful, into a bag, or a basket, or a bit of rolled bark or leaf, or a net woven of your own hair, or what have you, and then take it home with you, home being another, larger kind of pouch or bag, a container for people, and then later on you take it out and bat it off, share it, or store it up for winter in a solider container or put it in the medicine bundle or the shrine or the museum, the holy place, the area that contains what is sacred, and then next day you probably do much the same again–if to do that is human, if that’s what it takes, then I am a human being after all. Fully, freely, gladly, for the first time.

This paragraph is from Ursula K. Le Guin’s The Carrier Bag Theory of Fiction, where she says novels are like carrier bags: words hold meanings, and books hold words. I think she’s right, and that her idea can be extended to how human culture works in practice. As a species we rely on bags and containers for ideas, too. Humans need places to store their knowledge, to remember. Our own memory is our first bag, though it comes with plenty of limitations. But the bigger and larger, bag, the bag that contains all other bags, is what Joe Henrich calls our collective brain: the knowledge we store across all members of our social group.

This collective brain has a huge impact on the level of complexity of our society–you could almost argue this massive bag is our production possibilities frontier. And in pre-writing times, the role of population size was so important that there is evidence of populations quite literally forgetting how to do things. The mechanism makes intuitive sense: human memory is not infinite, so the more functioning brains our society has, the more information we can store as a whole (and the more social learners there are). One of the most famous (and debated) cases is Tasmania. It’s possible that Tasmanian aboriginals lost the ability to make complex tools, fishing spears, and even cold-weather clothing due to a rapid decrease in population. As an aside, you could argue that Ward-Perkins makes a similar point with regards to the fall of Rome, but this is more linked to economic complexity than population size.

April showers and Australian aboriginals

Aside: We know little about the origins of physical containers, interestingly. It could be (via Richard McElreath) that our first bags were baby slings, large leaves, or shells, but the archeological evidence is sparse as many of these are perishable.

So population size matters; just like a backup drive, the fact that the same information is stored in multiple places drastically decreases the probability of failure. But storing and remembering information is not a one-time event, we also want to effectively transmit it to future generations. Consider a catastrophic event that could kill or even wipe out an entire society: perhaps a tsunami, or a massive earthquake. How can (pre-writing!) humans of the present generation warn future generations of potentially life-shattering events that they know will happen?

I can think of at least three ways to relay this kind of information intergenerationally. The first one is through our collective memory and oral tradition. This can be done through a simple rule that perhaps everyone knows about and follows, even if the causal mechanism remains unclear. There’s a bunch of these examples in The Secret of our Success, of which my favorite is the hypothesis that divination is a sophisticated form of randomizing human behavior. An alternative is to encode this information in foundational myths, songs, or other oral memes that have a higher probability of being remembered. Here’s an example of this:

In about 1943, a band of hunter-gatherers faced a severe and enduring drought in the Western Desert of Australia. With their normal water sources failing, an old man named Paralji led his band to increasingly distant water holes, only to find them dry or insufficient. After traveling far across their vast territory and checking more than two dozen water holes, Paralji faced having to lead his band to their last tribal water refuge, a place he had only been to once in his life, during his manhood initiation rite a half century earlier. When the band finally arrived, their last refuge was jammed with people from at least five other tribal groups.

Soon local food supplies at the refuge began to fail. Confronting disaster, Paralji recalled the ceremonial song cycles that his people periodically performed at rituals. The songs told of the wanderings of ancestral beings and included a sequence of places and names. Relying on these ancient lyrics to direct him, Paralji headed off into territories unknown to him, followed by several young men and their families. Combining the information in the songs with trail markings, Paralji led the group along a chain of fifty to sixty small waterholes and across 350 kilometers of desert, eventually arriving at Mandora Station, on Australia’s west coast. The group had been saved by their ritual songs and by the distant memories of an old man.

The bottom-line is that the crucial information encoded in ceremonial songs saved the hunter-gatherers. This is a fascinating story, but the fact that Paralji was the only one to remember this and correctly infer that there was useful information in ritual songs gives you an idea that catastrophe was only narrowly averted, and that this system is quite fragile. The weakness is that encoding information this way requires mapping the relevant piece of information to the problem at hand, which may not be intuitive at all. How many bands perished because no one could make the relevant connection?

Another weakness is that information can easily get distorted to the point where it may stop being useful. Folk songs, proverbs, and sayings are popular in societies across the globe, and it’s likely most readers can come up with at least a few of them. But would you be able to explain their specific meaning? A few days ago I had a conversation with a friend about the expression “En abril, aguas mil”, which is a Spanish proverb with an English equivalent (“April showers”). There’s also a French equivalent, though it’s about March. My original understanding was that in April it rained a lot, so if you had asked me about total rain volume I would not have guessed that April is actually rather dry in the UK, and definitely not the wettest month in central Spain (that happens to be November). The point about April showers is that they’re frequent and not very long–there’s a useful follow-up to the Spanish proverb that clarifies this: “En abril, aguas mil, y todas en un barril”, meaning there’s plenty of showers and they can all fit in a barrel.

The point of this digression is that we have hundreds of expressions that used to have a specific meaning that was well-known, and that one could map to a specific problem and context, but that can easily be lost over time if it’s not reinforced or required. And that’s precisely what happens with catastrophic events: it can be decades before they show up again.

A second way of transmitting information is through physical objects. This has the advantage of not relying (as much!) on human knowledge, except in terms of how to interpret it. Some time ago, it was common for coastal populations in Japan to build monoliths that tracked the highwater marks of previous tsunamis, and warned people to avoid building their homes and towns anywhere between the monolith and the coastline. Note that these particular monoliths have writing on them, which makes it drastically easier to convey a message (as long as the alphabet and language are alive and can be understood). But this text is not required as long as the knowledge about their meaning is widespread.

Monoliths everywhere. Source: Smithsonian Magazine

As an aside, are modern traffic signals be an example of this tradeoff? The most common traffic signals, such as stop signs, traffic lights, do not contain any words (the stop sign being an exception, and even if there were no word it would be recognizable), but no one would think of coming up with separate signs that show a different geometric shape for every speed limit sign, because we already have a different encoding that does the job more efficiently: numerals.

Information compression and grandmothers

These signposts are performing a crucial task: they are compressing information. Think about how you would convey the all information required to be able to tell where it is appropriate to build a town or a home. Perhaps you would need a map of the coastline that highlights danger zones, but you don’t have this because your society hasn’t invented writing yet. Or perhaps you could use some sort of natural landmarks to guide your thinking, but this would require memorizing every single mountain, rock, bay, et cetera. Think about encoding this information in bits and bytes, and it soon becomes obvious that it is unwieldy and far less efficient than a good old signpost. In the monolith scenario, all you need to do is remember a rule: don’t build a home between a monolith and the sea. Yes, it is incomplete and can lead us astray. If the monolith’s been moved, or the coastline has changed, it can easily under or overestimate real danger. Or perhaps you’re wading into a previously inhabited area, so you wonder whether this absence of a monolith implies that building is allowed, or just that there wasn’t anyone there to find out (in which case your data would be censored, or truncated). The article itself states that when the last tsunami hit, many of the stones (many of which were build at the end of the 19th century) were swept away by the sea. This shows they’re imperfect prediction machines, but they still contain useful information.

Maybe, deep down, monoliths are a generative statistical model (and so are most other things).

The third way to transmit information requires no technology or infrastructure. It relies on people who, by virtue of having been on this planet for a long period of time, have a vastly larger sample size of events to draw from. This is the elders, and especially grandmothers, whose longevity is (to some extent) an abnormality in the animal kingdom.

As Henrich notes:

…in species with cultural learning, while aging individuals may be physically declining, they still possess transmittable know-how that makes them increasingly valuable to younger generations.

In our particular case, it’s not just that grandmothers have more know-how, it’s also that they have been been exposed to many more catastrophic low probability events. A common example from the animal kingdom is the 1993 drought in Tanzania. Elephant calf survival was higher in groups where the elephant matriarch was older, presumably because they remembered an earlier drought that had happened 30 years before, and were therefore able to migrate to areas with forage and water.

But there’s plenty of evidence for this in human pre-literate societies. Diamond wrote about it back in 2001:

…the information encompasses wisdom about how to survive dangers — such as droughts, crop failures, cyclones and raids — that occur at long intervals but that could kill the whole tribe if it did not know how to react. For instance, on Rennell Island in 1976, I was quizzing local people about trees: for each tree species, what (if any) animal ate its fruit, and did people also eat it? I soon learnt that trees were divided into three categories: fruit inedible to humans, fruit normally eaten by humans, and fruit eaten only in famine times such as “after the hungi kengi”.

I had never heard of a hungi kengi, and my middle-aged Rennellese informants did not know what fruits to eat after it. Instead, they led me to a hut in which lay a frail old woman, who named a dozen fruits to eat during that time.

It turned out that the hungi kengi was the most destructive cyclone in modern Rennellese history, datable to around 1910. It had destroyed gardens, forcing people to rely on ‘famine foods’ that were normally not eaten but that old survivors of the previous cyclone remembered as safe and nutritious. In 1976 that old woman must have been in her early eighties, because she said that she had been a child not quite of marriageable age when the hungi kengi struck. Upon her knowledge would hang the survival of villagers bearing her genes, when another cyclone should strike.

How much of a difference does this make? That depends on the length of the intervals between catastrophic events, and what distribution they follow, but that’s a separate matter. It seems that individual experience would allow for a much more accurate mapping of problems and solutions than other oral traditions. It’s true that the information held in the older Rennellese woman’s memory could be encoded in a popular myth or song (and perhaps it was!). But the fact that this information is seldom used (because this sort of event is rare) means that eventually it would erode, or lose details crucial to its usefulness (and who knows, maybe even becoming something more akin to April showers).

This is all to say that keeping information alive is hard, and it’s even harder to keep it alive for extended periods of time, and over many generations. And yet, as humans, we’ve managed to keep facts in our carrier bag for millennia. We must be doing something right.