As one researcher put it, “It is theoretically and practically impossible to precisely specify a supposed common denominator for all these [GMO] products.”

This causes real problems. People argue about GMOs because they are worried about safety, or environmental integrity, or human rights. But because the category is so porous, any policy governing “GMOs” — whether encouraging or discouraging them — can work directly against those values.

By now you are probably protesting that you’ve seen definitions that work well enough. (I’ve doubtless taken a stab at providing one at some point.) But let’s zoom in to take a look at how well those definitions fare in practice.

Definition: Moving genes from one species to another

Moving genes between species is called transgenesis. This seems like a hard line, and mostly (we’ll get to this later, when we talk about nature), it really is. Scientists today can perform some gene imports that simply weren’t possible in the past, like splicing a bacterial gene into eggplant.

But defining GMOs as transgenics leaves out a lot: It doesn’t encompass gene silencing, where scientists use lab techniques to turn off a gene, and many forms of gene editing — which has just become radically easier with the CRISPR breakthrough, and is therefore booming. The biotech company Cibus is altering the genes of plants in hopes of serving the growing demand for non-GMO foods. Though Cibus is modifying genes, it is just making internal tweaks, not introducing any foreign DNA, so it’s not making GMOs under this definition — a definition that is enshrined in U.S. policy, by the way.

This definition also leaves out mutagenesis: exposing plants to radiation or mutagenic chemicals to scramble or “modify” genes. This, as you might imagine, is more likely to cause unintended outcomes than transgenesis. All the critical attention to GMOs, and all the regulation aimed at catching any potential risk from transgenesis, has led companies to focus more on mutagenesis. As Bloomberg reporter Jack Kaskey found:

Earnings at BASF’s agriculture unit rose 27 percent in 2012 from the previous year, partly because of higher demand for mutant seeds in Eastern Europe, according to the company’s latest annual report. “The flexibility is there to use this technology quite broadly,” says Jonathan Bryant, vice president of the global strategic marketing group for herbicides at BASF. “Because it’s a conventional breeding technique … it’s very amenable for a wide range of seed companies.”

In other words, our attempts to minimize risk by regulating “GMOs” has created a boom in a riskier form of genetic modification.

The solution seems obvious: Expand the definition! Aren’t you committing genetic modification anytime you are mucking about with genes? What if we define a GMO as …

Definition: Anything that couldn’t occur naturally

The problem with this is that it casts so broad a net that it catches all sorts of things that we don’t normally think of as GMOs. This definition would include grafting trees — the technique of splicing the branches of one species onto the root stock of another. Grafting is an old-timey way of breaking the species barrier and combining the best traits of two organisms that couldn’t breed. But grafting doesn’t actually alter genes, so we could exclude it with a little tweak: “an organism in which the genetic material has been altered in a way that does not occur naturally through fertilization and/or natural recombination.” This is how the European Union defines GMOs.

But even with the tweak, the definition catches a lot of cool varieties that farmers have come to depend on. For example, there’s a French variety of wheat called Renan which is especially useful to farmers who don’t use pesticides because it is very disease-resistant. It was bred in the 1940s and has passed on its traits to many other wheat varieties. Stephen Jones, at the celebrated Washington State University Bread Lab, is currently working with a descendant of Renan adapted for the West Coast of the U.S. Its genes have spread far and wide.

Renan was made by breeders who coaxed the genetic material from wheat, and two other distantly related species, to combine. To do this, they bathed the plants in colchicine, which keeps chromosomes from hooking back up after they split during cell division — doubling the number of chromosomes in the plants. Later, they exposed the plants to X-rays to scramble some of the DNA, but eventually they got the combination of traits that they wanted. As Ontario farmer and former crop-science professor Terry Daynard wrote:

“So what’s the difference between Renan and many other GM crop varieties? Not much it appears except for the fact that Renan contains much more transgenic material, has not undergone the large amount of testing for safety and environmental impact as other GM events, and little is known about the mechanisms of the transferred genes.

Think of that while you enjoy your baguette at an organic café sur la Rive Gauche, Paris.”

Renan hasn’t undergone testing because the European Union decided that GMOs created with these techniques were exempt. There were just too many good crops coming from these techniques — and already in circulation — to subject them all to the scrutiny that has thwarted many transgenic crops. The same was true of mutagenesis. There are just so many wonderful mutants: disease-resistant cocoa, my favorite organic brown rice, barley used for craft beer and expensive whiskey,the Star Ruby and Ruby Red grapefruits, peas, pears, peanuts, peppermint, and thousands more.

In the U.S., when the National Organic Standards Board defined GMOs in 1995, it started with something similar to the E.U., but sought to revise the rule in 2012. When the board members began doing research for the revisions, they found that, under their existing definition, lots of organic farmers were growing GMOs.

Exploring this issue has brought to the attention of the subcommittee that engineered genetic manipulation of plant breeding materials has already occurred in many of the crop varieties that are currently being used in organic farming. A partial list:

• Disease resistant tomatoes (embryo rescue to introduce resistance genes)
• wheat and barley (double haploid technology using wheat and corn crosses along with embryo rescue and colchicine gene doubling)
• hybrid corn parent lines (double haploid to get homozygosity in 1 generation)
• Seedless tangerines and mandarins (mutations through irradiation)
• Brassica hybrids (cell fusion from radish traits)

Many of these techniques that were used in initial crosses that have now passed down through many generations may not be traceable any longer.

This came up again in 2014, when Humboldt County, Calif., voted to ban GMOs under the 1995 organic definition. In an editorial, biology professor Mark Wilson noted that the measure could backfire:

[I]t seems likely that many local farmers are growing crops that somewhere in their history underwent gene doubling, or cell fusion. The farmers who planted these crops are almost certainly unaware of their breeding histories.

What would happen if one of the massive GMO corporations (which have been conspicuously silent), or even just a mischievous student, stepped in, not to challenge the Measure P definitions but to force the county to actually follow the letter of the law? Farmers would have 30 days to destroy crops that they had planted in good faith, and more importantly crops that there is no rational reason to think are somehow dangerous. A major disruption to our local farmers is possible.

No one has tried to push strict enforcement to force organic farmers to destroy such “GMOs,” fortunately. Still, a purist might insist that that’s the only sensible route: At least we can draw a hard line between basic, vanilla, plant reproduction, and genetic alteration that wouldn’t occur in nature, right? Right? Wrong.

Nature is a genetic engineer

The biggest problem with defining GMOs as the product of “unnatural sex” is that nature has tried it all. As genetic engineers, we have gotten a little kinky in dabbling with some corn bondage. That’s nothing: Nature is the Marquis de Sade. It’s been having massive interspecies orgies for millennia, and is raising the resulting children.

For a long time, people thought that genes didn’t jump around from one species of plant to another. But then some scientists found that a gene that had somehow jumped between rice and barley. Then, scientists started finding more. At this point, it looks like basically all plants have some cross-species hanky-panky in their family tree.

When I asked on Twitter for examples that defied the lay definition of GMO, I got lots of responses, including notes about sweet potatoes …