Side topic about the genome-language analogy

I don’t want to derail the conversation on the argument clinic thread so thought I would post this separately. I am not referring to any particular post, but there have been a few allusions to it and I have seen it in the wider world of the internet.
I am just wondering about arguments for the degeneration of the genome, giving an analogy to a well-formed sentence and the impact of random changes. I probably overlooked it, but I was curious about whether the concepts of pleiotropy or epistasis are relevant when thinking about this analogy.
On pleiotropy, the fifth international edition of Futuyma’s textbook “Evolution” pg 97 states “Virtually all mutations that have any phenotypic effects show pleiotropy. Pleitropy therefore plays a crucial role in evolution: genetic changes that alter one trait invariably have side effects on other traits”. doesn’t this effect cause problems for the analogy from language. If the change to a letter in one string of the text causes it to read nonsensically it could at the same time increase the meaning in another string. There is also the possibility of the opposite, with antagonistic pleiotropy.
Bergstrom and Dugatkin’s third edition of “Evolution” define this as “The antagonistic pleiotropy hypothesis proposes that the same gene (or genes) that codes for beneficial effects… also codes for deleterious effects in other contexts…”. I would also assume both effects being negative, and much less frequently both being positive are possible.
The thing I am getting at is whether this is a defeater for the analogy from language?

The other concept I was considering was epistasis. Futuyma on pg 91 defines this as “Epistasis is the situation in which the effect on a phenotype (or fitness) of an allele at one locus depends on the alleles at another locus.”
Similarly, Bergstrom and Dugatkin pg 957-960 (kindle)talking about the Mc1R and the Agouti loci in mice state “Not all genes interact to produce the straightforward additive genetic effects that Nilsson-Ehle observed in his wheat kernel color system. When the alleles at two or more loci interact in nonadditive ways to determine phenotype, we refer to this as epistasis… Because of epistasis, the phenotypic effects of these loci are context dependent; the phenotypic effects of alleles at one locus depend on the context that is set by the alleles at another locus. Natural selection then operates on allele combinations that determine particular phenotypes. Some allele combinations increase in frequency, whereas others may be eliminated from the population.”
With this in mind, for the analogy from language to work there would have to be instances where the effect of a removal or changing of a letter is mediated by another string of some length elsewhere in the text. This doesn’t seem possible, so seems like another defeater for the analogy.

Clearly I am not familiar with the wider literature on these topics, so this is just a “hey, I recently read about this, and could it apply to the analogy” sort of thing. Wondering what folk think

2 Likes

To paraphrase George Box, all analogies are wrong but some are useful. No analogy fully captures the properties and behaviors of whatever it is analogizing; if it did, it would be an identity. Over the domain where there is correspondence between the two parties to an analogy, there may be useful opportunities for illustration and possibly even inference.

I think it is fair to say that pleiotropy and epistasis are much more common in biology than language contexts. But you will always be able to find incongruities around any analogy, so their existence alone doesn’t necessarily undermine an argument from analogy. One needs to consider whether those features are relevant to the particular argument being made. Sometimes pleiotropy and/or epistasis may be important to consider, but not always.

Take the recent specific example of a sentence from which 7 letters of a sentence were changed and 4 characters dropped to render it, if not unintelligible, certainly difficult to read and with some ambiguity about the ancestral wording. A gene subject to a comparable number of random mutations (in absolute number or fraction of total positions) would likely experience a significant reduction in fitness as well. What’s missing in that particular example is not a consideration of pleiotropy or epistasis so much as a consideration of mutation rate and purifying selection.

And even then, that doesn’t mean that a language analogy is useless. We could incorporate considerations of mutation rate and purifying selection into our analogy. Pre-Gutenburg manuscript reproduction involved a copying process that introduced single letter substitutions and word elision, but at a rate much lower than 8 errors per sentence per copy. And errors could be identified and removed so that they were not subsequently proliferated. As a result, current editions of various texts from antiquity are considered reliable representations of the originals.

At the same time, we certainly need to be careful that if we are using any analogy to guide our reasoning that our inferences don’t rely on properties or behaviors of one half of the analogy that don’t apply to the other half (or a lack of properties or behaviors of one half that actually are present on the other side). There are plenty of ways that language analogies can lead one astray when thinking about biological systems. But the specifics of each situation matter. If we simply observe that there are significant differences between language and genes or proteins and thus toss out all analogies between the two, we miss out on the utility that analogies can offer when properly scoped to help illustrate ideas for folks who are more familiar with one than the other.

3 Likes

Whilst fully agreeing with @AndyWalsh’s statements, I would like to put my own response to arguments from analogy in the following explicitly absurdist terms:

  1. In Alice’s Adventures in Wonderland, the Mad Hatter asks the question: “Why Is A Raven Like A Writing Desk?”

  2. Many varied answers have been proposed to this question, e.g. [1], [2], [3].

  3. An analogy works through similarities between the things being compared.

  4. Therefore a raven and a writing desk are analogous.

  5. Ravens can fly.

  6. Therefore, by analogy, writing desks can fly.

So, we are left with the conclusion that either (i) we must believe that writings desks can actually fly, or (ii) we must accept that there’s a problem with arguments from analogy.

Now the caveat may be made that this was a bad analogy. But, as Andy pointed out, “no analogy fully captures the properties and behaviors of whatever it is analogizing”. Even a good analogy (in a generalised way) may not be sufficient, if it does not capture the specific point of comparison. Therefore to be valid, you first have to demonstrate that the comparitors are sufficiently similar on that point. But such a demonstration renders the analogy redundant – as to prove the analogy valid, you already need to know enough about it to analyse it without need of the analogy. Therefore it would seem that an argument from analogy is either redundant or fallacious – which is why they are always a bad argument.

1 Like

That is their only utility. Analogies are not arguments.

3 Likes

Thanks for the response all. I will try to respond sensibly this weekend. Been a bit of a tough week for thinking

Definitely agree that no analogy is perfect. What I was doubting though is whether this analogy is too imperfect to be a useful one.

Agreed that mutation rate and purifying selection are important to think of, but to my mind they are secondary to the analogy. The analogy as it stands could operate under any mutation rate so long as time is then factored in to acheive the same result. It doesn’t really matter too much whether the rate is high or low, the end result is purported to be the same - a lack of sense in the code (language in this instance).
Negative selection likewise, the analogy isn’t primarily concerned with whether mistakes can be removed and the original restored to some degree, it is claiming as far as I can see that changes render the code senseless at some point whether that persists or not. I think that the analogy does break down because of purifying selection to some degree, but even there such selection isn’t consistently robust that no errors accumulate. But that is not the main point of the comparison as far as I can see.

True, no need to jettison the baby with the bathwater. That said, I think that it is simply misleading unless sufficiently caveated that is can be problematic and I rarely see the caveats. That said, I am not massively familiar with the apologetics literature any more

“Useful” for what? For convincing somebody who is skeptical about “the degeneration of the genome” that they should accept this analogy as a valid argument for it? If that is the case, then I suspect any imperfection would be problematical, as it would imply that there are ways in which the genome does not behave in the same way as a linguistic sentence, meaning that the skeptic has no particular reason to believe that they degenerate in the same way.

Useful for informing even those that may agree that the genome degenerates, without misleading. When giving an analogy we ought to understand what the hearer will likely take from it and mediate our message in accordance with that. If the hearer is going to come away from it with a misunderstanding that will further cement their view, then it is not in my mind a “useful” analogy