Hacker News new | comments | show | ask | jobs | submit login
Individual behavior in clonal fish despite near-identical rearing conditions (nature.com)
65 points by r721 on June 11, 2017 | hide | past | web | favorite | 35 comments

It seems possible to me that it would be advantageous for there to be randomness baked into personality, to avoid things like 'decision making deadlock'.

For instance, suppose a tribe of headstrong individuals who would never back down in an argument, and would resort to violence before losing face. It might be beneficial for random personality variation to exist to allow them to settle such a debate nonviolently. Maybe one individual randomly becomes able to laugh off conflict, or becomes more meek, or is more easy going.

It's kinda like how four cars that pull up to a four way stop at the same instant decide who goes: some randomness in the timing of their reactions leads to one pulling out a little before the others, and they're allowed to go, gracefully resolving the deadlock. Four naively designed AIs might get stuck in that situation (of course, simultaneity is implausible, but there are similar cases that might be more realistic).

> It seems possible to me that it would be advantageous for there to be randomness baked into personality, to avoid things like 'decision making deadlock'.

There is no need for randomness there. This scenario is handled by frequency-dependent selection like the 'hawk and dove' model or the male/female sex ratio. If aggressiveness becomes too common, then it gets selected against until it's driven back down to a frequency where it's neutral.

This sort of frequency-dependent selection is one of the behavioral genetics interpretation of why we have personality factors, actually: see "The Evolutionary Genetics of Personality" http://www.unm.edu/~gfmiller/newpapers_sept6/penke%202007%20... , Penke et al 2007; & "The Evolutionary Genetics of Personality Revisited" http://www.larspenke.eu/pdfs/Penke_&_Jokela_in_press_-_Evolu... , Penke & Jokela 2016

(I ran into a nifty mouse study on personality the other day demonstrating this sort of thing: https://academic.oup.com/beheco/article/26/5/1285/242100/Tem... )

I apologize in advance if I'm being overly pedantic, this weekend I took the time to do a cursory study of chaotic complex, systems emergence and fractals and i'm still trying on this new knowledge, but is it really considered random? (I dont know the precise definition)

Even if raised in identical conditions, a small perturbation in a social interaction (say one od the fish is looking directly at an event whilr another only catches a glance at it from the side) can bubble up to produce large changes in behavior?

It's not (probably) random in the sense of true, quantum mechanical randomness, but random can be thought of existing 'with respect to the certain knowledge'. So if you flip a coin and have as input the precise physics of the launched coin and the air, the result isn't random; if you only know 'a coin was flipped' then it's essentially random.

So I imagine chaotic systems' outcomes seem random if you have only approximate the initial conditions (since small perturbations will greatly modify the outcome) (that's probably a good, equivalent definition of a chaotic system, also).

QM effects are considered random and influence just about every real world system.

PS: At the 'other end' relativity actually influences a lot of chemistry because electrons are just that fast in orbit around atoms. But, that's generally ignored by just about everyone. https://en.wikipedia.org/wiki/Relativistic_quantum_chemistry

Two thoughts here.

If you look at fingerprints or the spongy interior of bones, clearly there's a lot of detail there. But the position of every skin whorl or bone void is not specified directly by a gene. Identical twins don't have identical fingerprints. This doesn't matter for finger or bone functionality, as long as the material works on a statistical, higher level. If you think of it as a blueprint for a bone, like for a building, the specification just says "x percentage of void with average cell size y and variance z", just like a building specification doesn't specify the location of every grain of sand.

And of course then it also grows in response to how it's stressed. Anyway, this is because it wouldn't make sense to store all the exact information, you just store some metadata and generate the actual stuff when building. Also you do testing and build more after that.

With the brain, it's the same, the amount of neurons and synapses is large compared to the amount of genes or "blueprints". But in a brain, as opposed to a bone or a finger, because the whole thing is designed to amplify weak signals, detail differences can have more effect on the function of the whole. There is this whole internal testing pattern going on while the brain is being built, but it seems it's not so perfect that clear differences creep in easily.

Yes, I was surprised that the article didn't mention neurogenesis. It's worth pointing out a result from game theory: Given related individuals (using the example of two shepard brothers who may signal each other for help) there are many signaling equilibria. The particular path is sensitive to random initial conditions. Interestingly, even for a simple game where a non-signaling altruistic solution dominates, a signaling solution reliably emerges and converges.

Game Theory Evolving, Herbert Gintis, 2000 (first edition), pp 321-323.

> It seems possible to me that it would be advantageous for there to be randomness baked into personality

There is also a question of possibility. Do we really understand how few hundred megabytes of information driving a soup of self-replicating peptides could encode the details of large, complex organisms, completely with all their behaviors, regardless of environment and random flukes? It's a miracle that this stuff works even as far as we can observe it working.

And on a somewhat tangential note:

People discovered genes and sorta jumped to conclusion that there must be genes for every adjective they have ever come up with to describe organisms. They used genetics to justify various eugenics movements. 100 years later, we still have trouble finding genes responsible for complex features.

I somewhat trollishly call it "intelligent design" or, more specifically, "anthropomorphically-intelligent design" because it really is an assumption that life is a sort of machinery designed in the same way the current generation of humans would go about engineering that kind of stuff, i.e. describe the "features" we think organisms have and figure out ways to "implement" them using biological processes we assume to be possible.

I'm not saying that I know a better way. I'm just not really surprised when science delivers less than people hope and hype.

It's not just decision making deadlock that receives a benefit.

The survival of my offspring is often dependent on them expressing strategic diversity from one another. The only feature that you share in common with your ancestors who survived snowball earth is the uncertainty of inheritance through mutation. If uncertainty is adaptive, it makes sense that it would also be encoded into the genome.

That's basically the thesis of the paper, there is some randomness completely independent of both genes and environment. I think it's super cool but I can see how some (white) people would find this upsetting.

You've repeatedly been violating the HN guidelines by posting inflammatory, unsubstantive comments and dropping gratuitous flamebait. If you continue to do this, we're going to have to ban your account, so please (re)-read the following and fix this:



This study's result should really not be surprising to anyone. The idea that somehow genetics + big picture environmental conditions are determinative of outcome seems so superficially unlikely to be true to me that I'm kind of baffled that anyone thought otherwise.

I'm sure many aspects of an individuals development are largely determined by genetics and big picture environment (anything else would be non-adaptive). But it kind of seems obvious that within the range of reasonably adaptive differences, the outcome is likely often chaotic. That should be the default presumption, and absent some clear reason to expect otherwise, is what they should have expected.

I think you're engaged in hindsight and rather glibly biting that bullet. Large swathes of sociology, medicine, nutrition, health research, governments (even where genetic influences are accepted) are premised on the assumption that one's environment and shared environment especially causally affect outcomes in ways that can be measured and quantified and used in research and provide paths towards understanding, control, and amelioration. Parenting, or air pollution, things of that ilk. It certainly is not obvious to most people that those are all minor effects of small importance compared to random developmental noise or other things.

Of course no one would be surprised to find non shared-environment variance components which were non-zero, but it's very surprising to find that they are so large: often larger than shared-environment effects, and sometimes as large or larger than heritability. This is one of the most interesting and counterintuitive findings of behavioral genetics, and experiments like OP put this in extremely stark relief. If clones in as exactly identical environments as researchers can create and perceive still wind up with such dramatic individual differences, how is science even possible on what causes the differences?


- "Three Laws of Behavior Genetics and What They Mean" http://people.virginia.edu/~ent3c/papers2/three_laws.pdf , Turkheimer 2000

- "Why Are Children in the Same Family So Different? Nonshared Environment a Decade Later" https://pdfs.semanticscholar.org/4091/dbfb736a4e793281e6b1aa... , Plomin et al 2001

- "Epidemiology, genetics and the 'Gloomy Prospect': embracing randomness in population health research and practice" http://ije.oxfordjournals.org/content/40/3/537.full , Smith 2011

- "Top 10 Replicated Findings From Behavioral Genetics" https://www.gwern.net/docs/genetics/2016-plomin.pdf , Plomin et al 2016

My guess is the evolutionary advantage of phenotypic heterogeneity. Phenotypic heterogeneity is linked to increased population-level fitness.

Phenotypic heterogeneity exists in the context of cellular noise without any known environmental triggers [1]. Gene expression can have switches that turn genetically identical population into phenotypically different subpopulations.

Evolution seems to push towards higher levels of expression noise in stress response genes, but genes that control basic functions are strongly shielded against expression noise[2]. Phenotypic heterogeneity can provide advantages when individuals are under stress. Predator can't predict behaviour as easily for example.

Most of these studies relate only to small micro-organism, but I hypothesize that they may be factor in larger organisms as well. I would like to see followup study where that fish are divided in strong stress and low stress environment. If fish in strong stress environment show more behavioral differences from each other, than fish that grow in low stress environment, my hypothesis might have some merit.


[1]: Nature, Nurture, or Chance: Stochastic Gene Expression and Its Consequences http://www.cell.com/abstract/S0092-8674(08)01243-9

[2]: Single-cell proteomic analysis of S. cerevisiae reveals the architecture of biological noise http://www.nature.com/nature/journal/v441/n7095/full/nature0...

I have a shaky understanding of "non-shared environment", I get the sense that it is based on the context of the study. e.g. If I do a study on genetically identical mice and control their sugar intake, "sugar intake" would be shared environment and something like "exposure to lead" would be non-shared. If I did the same study and controlled their exposure to lead, it would be "shared" in that case.

Is that an accurate understanding? Is "non-shared environment" basically the "everything else" category after genetics and shared environment have been controlled for?


> If clones in as exactly identical environments as researchers can create and perceive still wind up with such dramatic individual differences, how is science even possible on what causes the differences?

Another point in favor of simulation theory? :)

> Is "non-shared environment" basically the "everything else" category after genetics and shared environment have been controlled for?

Yes. For instance, in a study of childhood development, "shared environment" is the things that can plausibly be controlled since they are inflicted on all children in a family (e.g., household income, or frequency of family dinners), and "non-shared environment" is the "other" category that captures the infinite number of uncontrolled influences that different children within the family experience differently (e.g., having a scary fall down the stairs).

> If clones in as exactly identical environments as researchers can create and perceive still wind up with such dramatic individual differences, how is science even possible on what causes the differences?

Obviously science can't find the causes of differences that are uncaused.

But science doesn't have much trouble dealing with identical objects behaving differently under identical circumstances; quantum physics and statistics are both pretty successful.

That's science—even a failed experiment can have significance. So long as it was methodically done and well recorded, work like this is still valuable in that it improves our understanding of the world.

Ohh, absolutely. I'm glad they did the study. I'm just responding to the expectation stated in the paper. I really don't understand what sort of mindset would lead someone who has actually given it serious thought to expect what they did.

Phenotypic plasticity is a well documented phenomena in biology, this study seems to help confirm it exists in fish. In plants it gets pushed to interesting extremes due to their sessile nature. For instance etiolation occurs when plants are grown in the absence of light. They adapt by growing long thin brittle stems, to optimize the individual's morphology for longer reach, and minimal metabolic impact.

"Personality" in animals is just an extremely variable/adaptable manifestation of this.

>not surprisingly [...] the outcome is likely often chaotic.

Ok so if we have assumption that to make Identical[DNA + environment] equal to a chaotic outcome, then we need to add variable, random. Random would be determimed by environmental noise or possibly even (even more random) random quantum noise.

Why should we expect the effect of Random to be large enough to create easily identifiable personalities, or more accurately, behavior patterns?

So to be fair about this: I'm not aware of a specific reason it was so obvious to you (you don't provide any).

Your intuition in hindsight will always seem like common sense to you. That's why it's good we do these types of experiments which question our intuition.

> Why should we expect the effect of Random to be large enough to create easily identifiable personalities, or more accurately, behavior patterns?

Well, my main reason for thinking it is that complex systems tend to be chaotic unless there's some reason for them not to be. Two pendulums attached to each other produce a chaotic system. A living being is a lot more complex. When things aren't chaotic, I'm looking for an explanation.

I suspect that a little chaos is probably beneficial to the species/gene, so to the extent that there is a reason to expect anything, chaos seems likely (obviously only when the chaos isn't actively maladaptive).

My other source for the idea is a little introspection and observation. I can see the effects of small things in my own live and those of others.

>complexity tends towards chaos

Sure, that's a good point, but to play devil's advocate, I could also say something like: "life tends towards patterns of similarity" and point to anything from fractal patterns in life to similar behavioral outcomes in social strata or in culture.

> life tends towards patterns of similarity

I would argue that this is only true to the extent that there's a reason for the similarity. Similarities arise when they serve an adaptive purpose. And, to be fair, those clonal fish are a lot more similar than they are different. They're all still very similar looking fish!

My whole point was that the authors of the paper seemed to expect the similarity to continue beyond what there's any adaptive reason for, and that's just crazy!

I think we're a bit disconnected b/c I haven't done a good job of communicating my idea.

I'm not saying you don't have good ideas about what conclusion we should expect, I'm saying it's a good idea not to worry too much about such ideas in any place where we can conduct an experiment. Humans are notoriously bad at this and shouldn't do it if we have a better way (and we do, experimentation). See:

Orwell wrote about the problem of arguments in an essay On Politics, which can be found in the published book (not the essay) Why I Write. He discussed the aquatic screw, and how leading figures published ideas in leading print newspapers and journals arguing both for and against its effectiveness. It was only when an experiment was conducted that we, collectively, could know well-enough to consider it fact which one worked more efficiently.

Feynman said, pretty much: 'if it turns out nature is this way or that way so be it, it doesn't matter how good or how beautiful your ideas are, nature is going to come out to be the way she is, that's the beauty of science.'

Worth pointing out the movie The Boys From Brazil which explores the same concept in humans: https://en.m.wikipedia.org/wiki/The_Boys_from_Brazil_(film)

A bit corny but worth watching.

Also, (spoiler alert) -- Man in the High Castle.

I am not biologist (that is the following is pure speculation based on ignorance) but it feels to me that the method of their cloning leaves in opportunity of minute changes in the offspring DNA.

This then still leaves open possibility that different behaviour is cause by genetic mutations but this variation it much more amplified that presumed previously.

I would be happy if somebody with more knowledge finds time to comment on this.

Would a DNA sequence comparison on the clones not be sufficient to rule out such a basic mistake?

No, because DNA sequencing has lots of errors in it. This is why you might've seen numbers like '30x' or '50x coverage': to get a decent sequence, it gets done again and again and hopefully the random errors wash out. Identical twins or clones are not perfectly genetically identical, but the differences are so rare that any difference in your sequence is almost certainly a sequencing error unless you do an expensive number of passes like 70x costing several times as much. Interestingly, this comes up a lot in criminal contexts now, since you can, if you want to spend the money, figure out which of two identical twins left DNA at a crime scene.

Yes, it should, but how precise is this method and does it take the full DNA into account?

Edit: the fish reproduces gynogenetically, that is the offspring contains only maternal DNA. But this does not mean that mutations in DNA do not occur.

They also say to use genetic sampling which translates into my understanding as an approximate method.

Biological systems are non-linear. In other news Jurassic park will be opening today!

Free will has been proven!

Would this be called "personality"?

We need Pshycohsitorians a la Harry Seldon

Applications are open for YC Winter 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact