I'm a bit confused here...
The initial dog pile on Nick seemed (to me) to have as one of its main points something
like "Look, old man, once you formalize something mathematically we don't need to
care what any of the words might mean or imply in any other context, it is just math,
stop thinking that the words matter!"
And now there have been several posts by EricS, at least one by Glen, and I
think Marcus and Frank are in there somewhere as well, claiming that the words
are crucially important and we need to take them much more seriously.
So.... where does that leave us? Is everyone now onboard with the metaphors
mattering quite a bit?
I'll also note that "function" can't do the work on its own to explain evolution. We still need to know why
some functions are favored by selection and others are not. EricS seemed to indicate that we assess "fit" by
determining if animals are "happy".... but the metaphor of "fit" is like a key in a lock. To
explain evolution you need the matching of form-and-function-to-a-particular-environment. That matching *sometimes*
increases reproductive success, and *sometimes* the traits in question are hereditary.
Population genetics combined with field research can be very powerful along
those lines, but the math of population genetics on its own, floating out in
the ether, can't do it at all.
Best,
Eric
<mailto:[email protected]>
On Tue, Mar 31, 2026 at 6:10 AM Santafe <[email protected]
<mailto:[email protected]>> wrote:
Hi Nick,
Two smaller replies to what have become two sub-threads:
> On Mar 30, 2026, at 15:42, Nicholas Thompson <[email protected]
<mailto:[email protected]>> wrote:
>
> DES, EPC, FW
>
> So far as I understand, the argument flowing from Fisher makes no claims
about the kind of trait that produces reproductive success other than that it is
the kind that produces reproductive success. FW, if that's not a tautology, it's a
pretty tight circle.
As usual, let’s decamp to more neutral ground in the hope of having an
ordinary negotiation.
Suppose that, in your overweening pursuit of the study of metaphor, you
never noticed that there is a once/4-year gathering called The Olympics. Also
never learned what any of its so-called “events” are, what they are about, how
they work, and how one differs from another. My hypothetical here is meant to
define a condition of having “very little prior information” about some
phenomenon that we can, nonetheless, still reasonably unambiguously
circumscribe.
But a quick inspection shows that a subset of the participants (who all
together seem to be called “athletes”) are given metal disks and stand on some
kind of 3-tiered podium, while other athletes do not. Being a statistician — a
skill so helpful in the study of metaphor that it was worth taking the time out
to learn — you immediately recognize that this is a kind of marking that can be
used to partition the athletes. Taking notice, for the first time, of some of
the conversation in the society around you, who seem not nearly so devoted to
metaphor and thus have time to do other things, you gather that these marked
people seem to be called “winners” (or better, “medalists”, this “winning”
thing is a finer sub-partition; I’ll mis-use “winner” to label the most salient
marking for this little parable). It’s handy to have such a term, for use in
later sentences, so they become less tedious than the ones I have been typing
so far.
You also note that while there is only one 3-tiered podium and metal-disk
set per one “event”, there seem to be many such distinct “events”, so some kind
of event name gives you a second kind of marking you can put on the athletes.
Moreover, interestingly, the “event” label is again a proper partition (or at
least seems to be; this one is less cut-and-dried than the observation that
everyone carrying a metal disk is not someone not-carrying a metal disk, so we
are wary; the event label seems to be a bit more abstract): every athlete is in
some “event” set, and it appears that no athlete is in more than one of them.
As with the “winners” label, you learn that there are conventionalized names
for the events, and you can find a look-up table if you need one or another of
them.
Now, I can make a list of statements that seem to be of two different kinds
(scare quotes here indicate my statisticians’ attribute labels; in my condition
of very little prior knowledge, I don’t claim I have any more semantics for
them than I listed above):
1. Every “winner" is someone marked as having won something.
2a. Every winner in the “gymnastics” event is shorter than the average
over all the participants;
2b. Every winner in the “high jump” event is taller than the average over
all the participants;
… (we could presumably look for other such summary statistics that seem to
be unusually regular and to carry different values in different “events”).
I would say sentence 1 is “a tautology”, or close enough to it for the
purpose of this negotiation. Maybe I should use EricC’s good, and slighly more
flexible term, “truism”.
Now you may write a protest email: But the sentences 2a, 2b, have not told
me what constitutes “competition” in these “events”: “gymnastics” and “high
jump”, and given me the rule book for scoring them. Okay. And they didn’t
cook your dinner and do the dishes afterward either. Life is hard. And more a
propos (breaking my little 4th wall here), the path to a fully-adequate
“causal” theory through statistical inference is like the Road to Heaven:
narrow, tortuous, and inadequate to many things one can rightly want to know.
That’s what other sciences are then for.
But if you claim: The sentences 2a and 2b didn’t give me _any information_
about these “events”, and couldn’t have, because they are tautologies, I would
say you made an error. Of course, the real Nick would not say that, so we are
all safe.
The above parable is, of course, about selection. I didn’t say anything
about heredity. But if I had happened to note that height is a fairly
heritable trait, I could have spun out a much longer story, and defined some
Bayesian-posterior conditional probabilities, which would be shown to have
properties such as: the posterior probability, under various ceteris paribus
conditions, for a child of a high-jump winner to turn out another high-jump
winner is higher than for that child to turn out a gymnastics winner, and so
forth. The amalgamation of both of those stories would go in the direction of
Fisher’s fundamental theorem. It would leave out all the stuff that Fisher
left out of emphasis in his mad pursuit of his covariance term as an analog to
the thermodynamic 2nd law (a non-valid analogy, as it turns out to be easy to
show), and that Price included didactically (and here, to EricC’s answer):
that I didn’t even mention that the tall people might get drafted into
wars and put into an infantry to fire rifles over tall dijks, while the
short people might be drafted into Special Forces and sent on missions to
attack through underground tunnels, and so the number of survivors could depend
on many factors about which war their country had started, in what theater, and
against what opposition, etc. These are the world of everything-else that
Fisher lumped together into “deterioration of the environment”, as Steve Frank
(and I think also Price) lays out. They are probably not well-analogized to
“mutation”, but in genetics, mutation also goes into the same bin in the Price
equation — _outside_ the term of Fisher’s fundamental theorem — as the
“deterioration” effects. The accounting identity is flexible enough that we
don’t need analogies to use it; we can formulate a version for whatever
statistics our phenomenon-of-interest supplies.
Anyway; at issue: Seriously, do we have a problem in scientific work, of
people being unable to gain partial knowledge about phenomena through sentences
of the kinds 2a, 2b, because they can’t tell the difference between those and
sentence 1? In the world where I live, I don’t see evidence for this mistake.
Eric
