The limits of undirected selection

General Discussions
1

Edited to add: I’m not suggesting anyone here is practicing totally undirected selection; this is more of a thought experiment to illustrate the different ways that these two kinds of traits behave under selection.

GtS emphasizes wide genetic diversity, wide crosses, and minimal care of plants to select the best ones. All this is to the good if properly used. However, one thing I was recently considering is that there are two kinds of relevant traits.

If one threw seeds for a wide selection of varieties for a particular crop into an unmaintained garden, and harvested whatever survived for a few generations, without any intentional selection, the plants would improve in many ways: they would become more disease and pest resistant, more adapted to the local soil and climate, and more generally vigorous (assuming that enough plants survived the initial seasons.)

A number of other traits, however, would automatically deteriorate; all those which make them desirable crops. Dry seed crops would become more prone to shattering. Many fruits would develop thinner flesh and larger seed cavities, particularly the squash. Average fruit size would probably go down. Leaves and stalks would probably become tougher and possibly bitterer, as might fruits; that’s one way to ward off pest pressure. Biennial crops might become more annual, bolting in the first year. Brassicas like cabbage or broccoli would revert to kale-like forms. This process would go faster if one included wild or wild-type plants in the original mix.

The “landrace” or “adaption” gardening method derives its strength from the first part of the equation; the automatic or near automatic increase in vigor and fitness. But if we aren’t careful, we could end up having problems with the second half. It can take a lot of careful selection to maintain the desirable agronomic and gastronomic qualities of a crop. In turn, that requires a garden layout with enough order and space so that we can tell which plants are producing what, and methods which allow us to remove undesirable plants before they irretrievably cross into the year’s seed production.

Examples: this year my GtS cucumbers produced some more-bitter fruits. It was impossible to tell which plants were producing what, and in any case it was too late in the year to remove all the bitter plants, remove all the fruit and flowers, and wait till a new batch of fruit was produced, so I’m not saving seed from my patch this year. My GtS sweet peppers produced some hot peppers, and the same situation applies, though in this case I’m saving seed, though not contributing it to GtS; since peppers cross less than cucumbers, I am hopeful I can clean the contamination out. My beans are all tangled on the same trellis; if some pods are tougher than others, how could I figure out which plant is which? They are all tough once mature . . .

My non-GtS but genetically mixed summer squash are more vigorous than the varieties I started with, and make decent eating, but none of them are as tasty and productive as the Costata Romanesco I added to the mix. (The C.R. dies early from powdery mildew, though, which the mixed plants do not.) And there is always the possibility that the squirrels will plant seeds from some of the many, many ornamental gourds and jack-o-lantern pumpkins the neighbors put out, and the cull squash that I dump in the compost, and that very inferior genetics will creep in. So many traits are only noticeable at the end of the season, after the crosses have already happened. So with the squash, I am going to start over, using known varieties and doing hand pollination, not to keep them pure, but to mix and cross them in a controlled way, allowing me to choose both parents and develop something superior in all respects to the starting varieties.

None of this is an attack of adaption gardening! But it is something I wanted to share as I consider how my gardening methods impact my ability to select excellent crops.

7 Likes
2

Yup, all good thoughts. It’s very important to know what your highest priority is.

I think continuing selection is essential. I think most things — and domesticated plant populations are certainly among them! — tend to gradually get worse unless they are continually maintained. You never know when a genetic sport that contains an undesirable trait that helps with survival might happen, for example.

I think there are often tradeoffs, and everyone has to figure out for themself what their top priorities are, and what things they are willing to let go. For instance, one gardener might want intense flavors and very few seeds in the fruits of their watermelon landrace, while another might highly value drought tolerance and be perfectly okay with meh flavors and small, seedy fruits.

In life and in gardening, you get what you prioritize, and you may or may not get your second, third, etc. priorities. It’s well worth trying for all of them, but you can never make everything you want your top priority. You have to choose what matters most to you, and be willing to give up some of your lower priorities sometimes, in order to optimize for that.

Just a random thought: I like to roast and eat the seeds of any squash that isn’t tasty enough to save seeds from. That way, I get to eat some protein (very valuable nutrition!), and I don’t have to worry about any not-so-great squashes volunteering. :wink:

Selecting based on the mother fruit only is a rather easygoing, slapdash means of selection, but I’d say it works pretty well. I don’t mind if the less-ideal plants managed to fertilize some of the seeds that I saved from the best plants — maybe they had something useful to contribute to the next generation, like high productivity or better drought tolerance or something. It’s okay if you feel differently. Selecting based on both parents is likely to yield awesome results much faster. It just takes a lot more time and attention. One of those examples of needing to choose what you want to prioritize! :smiley:

4 Likes
3

Hi Emily, I agree, there are always trade-offs. It is just that I think we should be more aware of this one.

As far as selection based only on the female plant, I agree that it can be a good compromise between getting the ideal genetics and the productive use of limited time. But selection based on the fruit is not necessarily the same as selection based on the female plant.

For an example, let’s take my tangle of cucumbers. Let’s say that I select some nice-looking and good-tasting fruits (though this is already complicated by the fact that by the time I harvest they are overly mature for ideal eating.) If the plants are too tangled to correlate fruits with plants and observe all the plant traits, I’m not actually selecting the female parent for yield or other qualities. All I know about a given parent is that it survived sufficiently to produce at least one fruit!

And things get worse with other situations; for example, the beans I mentioned above. If I allow a tangle of green bean vines to dry down and just save all the seeds, it is unclear if I’m selecting for anything at all beyond basic survival; I’m not selecting for the quality of the beans, which is impossible to correlate under such conditions. This condition to actually pretty close to my “throw a bunch of seeds out there an do no selection” scenario, only modified by the fact that beans have low outcrossing rates.

In other words, my general point is that we might be doing a lot less selection than we think we are doing, such that traits other than vigor unnecessarily decline.

1 Like
4

All those remarks seem to me based on your set up and your realizations that it may not be enough (relatively to your objectives) to throw a bunch of seeds and see who survives, and I agree that to build nicely on Joseph’s vision - which helps a lot to free the minds regarding seed saving and reveals untapped adaptation potentials - we then, depending on our context, the species we grow, have to do that extra step to get it in line with our objectives. And that step is ssssso interesting in itself! Personnally doing it in winter as I’m growing mostly summer crops: time to imagine, dream…. But it’s not contradictory to the initial message : it’s just the way forward : a mix between a crop, a land and the gardener’s personnality.

Some of my set ups may sound ultra complicated for a beginner, but yes I also had to create ways to do proper “to-the-point” assesment of individuals, and as much as possible all along the season, to make my set ups coherent with my not only survival, but also taste, yields, and overall plant health objectives… otherwise I was feeling like not doing a really clever move. And especially this year because my 2024 set up and (subjective!) taste selection of melons and watermelons has proven to be of zero efficiency in 2025! Was a shock: my “best of the best” of last year were not better than my “average” resown in 2025! In other words: the tens of hours of selection in 2024 were proven to be of zero efficiency! Ooooops…

One thing I have done in these years of selection post initial grex is giving big spacings between my squashes and doing some interplanting to be able to assess each plant individually. Otherwise, the intertwinning made me unable to do any yield assessment…

Anyway: I feel like the reflections on our set ups are a steps forward, and a true reflection of personnal understandings in our context, based notably on wins and failures, and personnality. It’s so creative!

3 Likes
5

I totally agree! :blush:

In case this is interesting, I tend to do my selection for favorable traits of the parent plants before they even flower. My number one criteria, drought tolerance, I don’t have to select for because my climate does it for me . . . :laughing: But as for the rest, I keep an eye out for any plants with traits that annoy me and pull those plants out before they start flowering.

In general, this mainly means “the presence of thorns” (or “worse thorns than average” if complete thornlessness is hard to get in that particular species — pepo squashes are rarely completely thornless, for instance). Occasionally it also means “you seem to attract far more bug pests than usual.” (This usually applies to squash bugs on squashes.) Or, if the leaves are edible and I consider the flavor important, it might mean “these leaves are untasty.” (This usually applies to brassicas and lettuce.)

I select positively for earliness and producitivity, in that I’ll plant a whole lot of extra seeds from any plant that tastes good that I know gave me a tasty harvest sooner than average (or gave me way more of a harvest total). But I don’t select against plants that are late or not very productive. Generally, my climate applies a harsh enough selection pressure that I’m happy with anything that makes tasty food for me.

3 Likes
6

I like this discussion and there are many small practical ways I can relate to it. For example, how to select for tender pods on pole beans when they get tangled up and all have different qualities at the time when you harvest them for seed. Some kind of identification system is usually necessary in that case.

Or the very relatable: I both want to plant pumpkin closely together to get good coverage, and I would also like to assess how many fruits each plant give. I personally like the vining trait too, because those plants seem more vigorous and can root as they go. It gets challenging to select for plants with multiple fruits that way.

I do not know anyone in this community that practices “undirected selection” and it is the first time I come across the expression. Perhaps you are thinking of the use of a first-year multiple cross of named varieties? A lot of members use that method to create an initial grex to then locally adapt from.

Most people have highly personal selection criteria for many things they grow , esp. the ones they care a lot about. Tasting always brings direction, consciously or not.

The more I think about it, and the more seasons I save seed, the more it becomes apparent that I am one outnof many selection pressures for the crops I care about. Whether my selection pressures are conscious or not, they usually have some kind of intention or direction to them. My habits, my personality and the community that I am part of (and that takes part in tasting or harvesting or storing the crop) is also what the crop adapts to. Conscious direction probably plays a similar role in breeding as it does in the rest of the human organism: a very powerful one, that can create entirely new patterns, but proportionally very small role.

5 Likes
7

Depending on your available space, nothing prevents you from adding in varieties that have the culinary traits you desire. I do, on a continual basis. If those produce seed, i save it and add it to the grex. If not, then they were not destined for glory.

4 Likes
8

Thanks for the thoughts, everyone! I appreciate hearing about everyone’s experiences.

To clarify, I’m not suggesting anyone here is practicing totally undirected selection (the only thing that come nearest are first-year trials to see if anything survives, with selection picking up in the second year.) The thought experiment of undirected selection was to highlight the radical difference in how the two groups of traits behave under different selection regimes. (For instance, if one doesn’t save a fixed number of seeds per fruit or work around the problem in some other way, cucurbit fruit cavities will “automatically” increase in size at the expense of the flesh, since more seeds increases a plant’s influence on the next year’s gene pool.)

My purpose in bringing this up is rather to emphasis that some traits are harder to select for than others, and that in many cases garden layouts and gardening habits will have to be modified to allow for successful selection. I think I’m feeling some surprise at just how hard it is to maintain quality in my plants, even as plant vigor seems to be increasing rapidly.

3 Likes
9

Thomas, I wonder if your selection for taste didn’t seem to work due to non-genetic factors that influence taste?

2 Likes
10

I don’t think so as it was in the same soil just nearby, with a comparable weather pattern and similar soil and plant practises. Though the overall qualiy was good, but no significative difference between the different levels of taste as sorted subjectively in 2024. I’m talking of a taste selection based on 300 watermelons approx and something like 200 melons. Did sort the watermelons and so the seeds in three categories : average, good, very good.

Because I view the subjectivity as the main cause of the relative “failure”, I’ve decided in 2025 to use more “objective” criterias, so I did:

  • clarify my harvest criteria (tendril opposite to the fruit had to be dry to assess maturity, instead of the “sound” that made the watermelon)
  • regarding taste I did a brix survey, discarding everything under 10 deg. Brix, so 10% of sugar.

Then I did a secondary evaluation, in surch of superior aromas and qualities.

So everything under 10 is simply discarded, above 10 something like 80% is dried qnd marked as “good”, the rest being my best of, the type of sensorial quality I want to see in later years: not only sugars but aromas. We’ll see next year as I’ll reproduce a set up that allows me to do those assessments, in other words: verify my progresses.

3 Likes
11

Maybe another issue is that taste could in theory be a very polygenic trait, and that in an unstable grex of an outcrossing species this makes any progress in the right direction very slow? Presumably many of your “best of the best” seeds actually represented a natural cross between “best of the best” plants and merely “good” plants, and similarly with the “good” ones?

3 Likes
12

This was my thoughts too. So patience could be a friend.

On a related note. I have selected pumpkin for taste three years now and this year I was surprised by how high the baseline was. Very few blands and mehs so far and some we really liked. They were significantly better than last year. I believe something is improving.

3 Likes
13

I don*t think the detoriation is that fast. It could be if there is some random hard selection that bottlenecks the population, but generally I would expect much more than just few years for detoriation. More likely is that “undesirable traits” are actually condition related. Like if you try to start selecting for drought tolerance in squash, many will produce smaller than normal fruits with possible thinner flesh just because the conditions didn’t allow for any better. Like my last years survivor plot produced at most 1kg maximas. This year in my normal plot those were 5kg at the most and most 3-5kg. Those were little smaller than where those came from in -23 (at least those that were from my seeds), but that might have been because this years dry period was right at the time when they started to produce.

Some of my melons are 5-6 year in practically unassisted selection in a way that I haven’t done much selection on the fruit size or taste. I only cull those that naturally are slower to give more space for the others. I started with some that had little sour fruits, but that seems to have disappeared. Only thing that I remember is one fruit that was more ripe cucumber tasting that I eliminated, but otherwise I have used pretty much all seeds that have been from the fastest fruits. Size also doesn’t seem like is affected, which I don’t think would change much in something like melons where total seed amount isn’t affected much by size. I have grown other varieties with them, but I don’t think that would have been that crucial as old has been always the main source and they have freely crossed. Also F3 of the half wild tomatoes seem to have little bigger fruits on average than F2s, as is F2s of the half wild chilies. Again only selection I did was cull those that were smaller than others, which seems to have favoured those with slightly bigger fruits. So it might be true that cherry sized is favoured by natural selection if you leave mix of cherry to beefsteak to fend for themselves, but that selection might also favour cherry over currant sized tomatoes.

But you are right that it should not only be left for nature to select, but the involment doesn’t have to be more than just little pushes toward right direction. Like if you see something makes way more/less seeds than other, and seeds aren’t the harvest, then just to sow them accordingly. Favour those plants that seemed to be more tolerant and add some of the others for diversity. Remove something that is going away your goals etc.

I don’t know if starting all over with the squash will help you much. Even if you can keep them clean, F2s are always a pandoras box waiting to opened. Some unwanted traits will appear anyway. I would use what you have and instead just self them in controlled way. That way you can then select the ones that at the end of the season have performed as you like.

2 Likes
14

I was thinking if the variance is bigger than the difference between different populations and so it’s hard distinguish if selection has moved the needle any way. Like if you can select average 5% improvement it would be very hard to notice without measuring brix from every fruit. I would just hold the course and keep favouring those with best taste. Add some diversity from others if they have some distinguishing trait.

2 Likes
15

Inlaws of my brother asked what type of pumpkin (probably with the hope finding it in the store) when my brother shared some of the one I gave him. To be fair, those had quite exceptional sweetness and, although I haven’t bought any squash from stores in ages, must have been quite a bit better than any that you can find in stores.

3 Likes
16

Hi Jesse,

As far as my squash, I probably wouldn’t be starting over if I had a huge space available, but I can only plant a few hills of squash each year, and those hills need to both produce food and move the project forward. I made the mistake of mixing in some genetics of lower eating quality (not terrible, but they don’t have the greatest taste, tend to go tough at smaller sizes) when the Costata Romanesco is already everything I could want in terms of taste, and even in general vigor and tolerance to cold soil in the spring. It just isn’t resistant to powdery mildew. So a directed cross with a PM resistant zucchini seems like it would achieve my objectives faster.

PM resistance is an interesting example of this “selection is more complicated than it looks” dynamic. Plants that aren’t resistant set fruit just fine, and even mature seeds. The problem is that by the time it shows up, it is too late to start over. If the handful of plants I allowed to mature fruit aren’t resistant, well, there goes that year of selection . . . and in any case, they will have crossed with the ones that aren’t. Whereas if I make hand crosses and record them, I can save from fruits where both parents are resistant. Plant survival doesn’t actually select for PM resistance in any way, since the plant doesn’t “care” about cranking out new young squash for me right till the frost. (Just like human survival might not put much evolutionary pressure on keeping functional joints into one’s 80s.)

1 Like
17

I still don’t this is the best solution to your problem. If you now make new crosses and later notice there was some unwanted resessive trait (which there is likely to be) it will take you 2 or more years to even notice this. If you have just small sample it might take couple more years for it to first show up. It might not be easier with the current population either, but that you would be able to work right away. I’m still a little doubtful that there can be a diverse population that doesn’t have any unwanted traits. Resessive traits can always linger on for years without detection. It might look like you have gotten rid of it and then when they cross pollinate it comes back. It’s difficult in any case.

1 Like
18

It’s where it’s both a reflexion of our understandings and personnalities, and we can offer new intuitions or understandings or questions to others :

here is the synthesis or interpretation I made in 2024 from Joseph’s way of selecting allogamous like squash or melon over the years + my strategy which was relative to super diverse grexes with a hundred varieties of watermelon and melons for example.

b2c803539b8e4f2425506ae9d2148466b0e83ada_2_1326x1000 (1)
b2c803539b8e4f2425506ae9d2148466b0e83ada_2_1326x1000 (1)1326×1000 265 KB

That is :

  • 2 years of crossings and zero selection (year 1 = AN 1 and year 2 = AN 2), so to say the “grex” stage
  • then three year of taste selection, for me starting in 2024 (AN 3),
  • and eventually the “maintenance” stage Joseph is talking about, with from then on a super gentle selection to maintain the qualities selected + marginal addition of nice varieties (1-2% of plants per year)

As I had big wide grexes in a climate where all varieties were fruiting (nothing compared to Utah!) I felt that the high potential of those grexes was all in the inner diversity which I didn’t want to loose too much in the selection process (that reflexion was mostly inspired by Cecarelli’s chapter on philosophies regarding assembling of evolutionary populations, a.k.a. modern landraces, and specifically by the 2nd philosophy outlined page 79 as I intend to bring my crops to new agroecological frontiers, that I won’t delve into here). So I decided to go through a rather slow selection process : trying to maintain some diversity at each step.

So that was my context and mindset. But as the outlined strategy (getting rid of the “so-so” at year 3 harvest , ofthe “average” at year 4 harvest, and then of the “good” at year 5 harvest, then maintaining excellence from year 6 on…) was purely based on my own synthesis and theorizing, I had to assess the result of that strategy. The 2025 melon and watermelon deployment of those different qualities in different plots was the occasion for that.

And so to build on my experience and trying to reply to your remarks, my understanding right now is that :

  • the harvest criterias are crucial to make any progress in selection (talking particularly for places where the growing season is long! Where frost isn’t a selection factor!). An hour long video of Agrobio Perigord details a decade long zero progress in participatory selection of corn… despite tens of thousands of Euros in it (employees helping) + thousands hours of voluntary work by farmers… due to a lack of focus, lack of understanding of heredity, too much criterias in selection, lack of methodology… But essentially : subjectivity in the field! They were looking for too many things at the same time… So now they are focusing on just one or 2 objective criterias and they measure their progresses… eventually!

Harvest criterias are also of particular importance if you want to share your seeds to others : for ex I’m a very excellent harvester of watermelons with the “sound method”, so my selection was consistent with that, but when it went to farmers places in 2025 (about 6 farms) they were incapable of reproducing my “feeling” with the sound… So I had to go back to a more objective criteria (= opposite tendril must be dry), and ALL my selection will be coherent with that trait from now on. Same story for full-slip / half-slip / no slip harvest indications in melons in my very (too?) wide melon grex which encompasses about everything in the huge cucumis melo genome - much bigger and more complex than watermelon : only one type of watermelon against 7 to 9 types in melon!

So as my 2024 harvest wasn’t strict on simple objective criterias, and adding to that because there was so much diversity in those mixes, there was a first reason for that mess in the fields in 2025… Despite tens of hour of - what I thought was - “selection” in 2024!!!

  • Complementary understanding from that 2025 years of selection is that, even if it was nice in theory to do a relatively slow selection, to stay coherent with new agroecological qualities I envisionned, in practise there could be too much of a dilution effect regarding fruit quality, that due to the fact that, as most of us doing selection with the fruit of an outcrossing plant, the outstanding fruit quality we keep seeds from is diluted proportionnaly to the “crap” of the surrounding plants. Meaning, in my case, and for melons especially as it appeared to be sssso inconsistent : I’ll be increasing the selection pressure drastically in 2026!!! And also to be fair I should add that, still for melon, the dilution of quality, due to an initial lack of understanding of what I was doing and consequentially of smart selection in first years, maybe too advanced! Next year will tell : either it’s too advanced (and so I’ll throw away all my melon seeds! bye bye melons…) or I measure progresses, and so I will keep on selecting with clearer and more objective harvest and taste criterias. And that shoud do.

For melons in 2026, I’ll strictly plant full-slip types (and so full-slip will also be THE harvesting criteria), and will plant only “excellent” melon seeds of the 2025 selection. I cannot do more with what I have : I won’t risk any “dilution effect” any longer. Then I’ll do a double taste evaluation with a brix minimum threshold + a secondary taste evaluation for complex aromas that no tool can tell, and only those superior types should be replanted in 2027.

In watermelons it’s less of a problem as - even if I didn’t make any progress in 2025 - overall quality is good to very good. And yields are there too. Did yields comparable to better than Sugar Baby in most cases (even in farms in complete different soils and climates).

In moschata and maxima taste selection in 2 months from now will tell.

Finally : I agree with @MalcolmS overall view on the topic : I’ve seen more and more vigor in my populations over the years, which is probably due to my intense oversowing and culling of 95% on early vigor over 3 years (anyone can check his/her supplementary vigor by simply growing "new varieties” on the side of his/her population), but in my experience there is still a fair amount of work on taste selection to get something consistent out of that first emerging trait. And that is more or less crucial depending on species and personnal objectives.

Objective harvest and taste criteria + selection pressure adapted to the consistency observed in the population, seems to me as the base formula of taste selection when dealing with our populations.

I hope that “dilution effect” is not a fiction - as I never heard of such thing ! - but it’s how I see the risk associated with wide grexes and/or too weak taste selection pressure in allogamous.

1 Like
19

I have also realised that maybe harder selection is necessary. Not only because making faster adaption, but also not to spend time saving too many seeds even for my own hard selection and sharing. For example sow 50-80% of the area from the very best (like 1-3 plants) from last year and, if there is, from the year before. Rest of the remaining survivors or interesting plants. With tomatoes that might be mostly those that I picked late in the season that didn’t show symptons of LB after harvesting in storage. Rest from the first fruits to favour earliness, although I don’t know if I have to go just LB some time. I do select for earliness with early culling and the remaining differences might be too small to care about. But many crops like melons or squash is quite easy because I can’t have too many selection criteria besides earliness. Unless summer is exceptionally hot. But I might need to rethink how much new blood I want to include in those populations at the moment and rather see if the selection is paying dividents.

1 Like
20

We humans aren’t the only animals selecting crops. These tasty fruits developed their flavour as an attempt to atract animals to eat the fruits and spread the seeds. The tastier and easier to eat for its usual predators, the more of them can be found. If we humans stop eating and selecting them, then these crops will become tastier for whatever other species is feeding on them.
But as long as we keep to our niche, eating the things we love to eat best, instead of eating whatever rubbish we are sold at the groceries just for the convenience, we should be fine.

1 Like