Why this planner will never give you a companion-planting chart
The grid of good and bad neighbours is the most reproduced object in garden writing and one of the least examined. Here is what happens when you check it.
The chart is the product everywhere else
Open any garden planner and you will find a companion-planting matrix: a grid of crops down one axis and across the other, cells marked "plant together" or "keep apart". Basil with tomato. Carrots love tomatoes. Never plant beans near onions.
It is the easiest feature in the category to build. The data is a list of pairs; nobody asks where the pairs came from; and a user who follows it and has a good season credits the chart.
This planner does not have one, and will not get one. That is a deliberate omission, not a gap in the roadmap.
Where the pairs actually come from
The bulk of the modern companion-planting canon descends from early-twentieth-century biodynamic and anthroposophical writing, propagated through mid-century popular gardening books. It was not assembled from trials. Individual pairings are rarely tested, and when they are tested, most of them do not survive.
The corpus records that finding as a claim in its own right, with the same grade and the same sources as anything else it says:
Companion planting charts encode real relationships.
Refuted
Most companion planting advice you'll find online is folklore. Here's what we can actually defend, and why.
what is actually known
The bulk of the modern companion-planting canon descends from early-20th-century biodynamic and anthroposophical writing, propagated through mid-century popular gardening literature. Individual pairings are rarely trialled. A minority have real named mechanisms — volatile masking, living mulch, insectary provisioning — and are recorded here as a good hunch. Most have nothing.
what this rests on - one pointer
history of the companion planting literature
A grid cannot state a mechanism. That is the whole problem with it. "Plant together" is not a claim you can check, because it does not say what is supposed to happen, to which plant, by what route, or how much. The cell is unfalsifiable by construction, which is exactly why it has survived a century without being examined.
What replaced it here
Every rule in this planner has to name a mechanism in one sentence, without the word "helps". If it cannot, it is downgraded or deleted. That single constraint is what makes the chart impossible: almost no cell in a companion-planting grid can pass it.
What can pass it is a much shorter list. Corn shades and supports; squash foliage intercepts light at the soil surface and slows evaporation; a legume fixes nitrogen through a bacterial symbiosis on a timescale you can state. Those are mechanisms. They are modelled here, and each one carries its evidence.
Some of the mechanisms are genuinely weak, and those are shown weak rather than dropped. The aromatic-herbs-with-brassicas pairing is the honest edge case: something is reported, the mechanism is unresolved, and the trials disagree with each other.
Aromatic species interplanted with brassicas are reported to reduce cabbage-worm damage. The mechanism is unresolved and trial results are mixed.
A good hunch · suboptimal if ignored → advise
Mechanism: Unresolved, and the rule says so rather than picking one. Volatile masking of host-finding cues and disrupted landing are both proposed. The trials that report an effect measured DAMAGE, which cannot separate fewer eggs laid from fewer larvae surviving or more predators arriving - so no mechanism here is established, only candidates.
Evidence: confirmed against 5 sources in the research-based literature.
the 5 sources
Univ. of Minnesota Extension, 'Companion planting in home gardens', extension.umn.edu/planting-and-growing-guides/companion-planting-home-gardens (read 2026-07-25): 'In an Iowa study, thyme, onion and nasturtium helped to reduce cabbage looper and imported cabbageworm damage in broccoli' - citing Riesselman, Leah B. (2006), 'Companion Planting: A Method for Sustainable Pest Control', Iowa State Research Farm Progress Reports. Thyme is this guild's canonical masking species, so the one trial found is on the plant the corpus actually prescribes.
Same page, on the wider claim base: 'While gardening charts online will provide you with long lists of plants that repel insects, these are not always accurate or backed by research', and of marigolds specifically, 'Sometimes yes and sometimes no.' That is the mixed picture this rule's effect_size asserts, now cited rather than claimed.
PRIMARY READ: the maintainer read Riesselman (2006) itself on 2026-07-25, at dr.lib.iastate.edu/farms_reports/295, and promoted this rule on it - not on Minnesota's summary alone. The trial's own measurements are not transcribed into this pointer; what the corpus asserts is still the DAMAGE claim the summary reports, and the read did not overturn it.
GAP that survives the read: the effect is measured as damage, and damage cannot separate fewer eggs laid from fewer larvae surviving or more predators arriving. That is why this claim does not mention egg-laying and why the mechanism is stated as unresolved. Verified means the source was read, not that the mechanism is known.
NEGATIVE RESULT, recorded so it is not cited by its title: Morkeliune et al. (2024), 'Effect of White Cabbage Intercropping with Aromatic Plant on Yield, Mineral and Biochemical Composition', Plants 13(13):1870, intercrops cabbage with calendula, French marigold, thyme and sage across 2017-2019 and measures NO pests, eggs or oviposition at all - yield and biochemistry only. It cannot support this rule. Nor can the '42% fewer egg clusters under sage' figure circulating in search results, which traces to commercial content sites and to no study.
See this rule in the planner
That is what a real companion claim looks like when you write it down properly. It is much less satisfying than a green cell in a grid, and it is the truth.
The specific beliefs, answered
A chart is also where a lot of individual folklore hides. The corpus keeps 19 of these as first-class records, each with the corpus's answer and, where one exists, a source that refutes it outright. Two of the most commonly charted:
Marigolds repel most garden pests and protect the whole garden.
Contested
Marigolds don't repel pests in general. Their one real effect is narrow: a dense planting of the right Tagetes, grown before a crop, suppresses root-knot nematodes in the soil (not a scent that wards off flying insects). Enjoy them for colour and pollinators; use the nematode trick on purpose if you need it.
what is actually known
Marigolds have one real, narrow effect: some Tagetes suppress root-knot nematodes in the soil when grown as a dense preceding stand, via the root compound alpha-terthienyl — a soil biofumigant effect on a specific soilborne pest, handled as a role in this corpus, not a scent that wards off insects. The broad folk claim — that a few marigolds dotted among the vegetables repel aphids, whiteflies, cabbage moths, or rabbits by smell — has no reliable field support.
what this rests on - one pointer
University of Florida IFAS Extension, ENY-056 'Marigolds (Tagetes spp.) for Nematode Management' (read 2026-07-14): the effect is a dense preceding cover crop, not a scent repellent — 'alpha-terthienyl inhibits the hatching of nematode eggs'; 'Planting should be dense to ensure the best nematode control'; 'Marigolds cannot eradicate nematodes.' Backs the narrow root-knot-nematode mechanism; the general pest-repellent claim is unsupported.
Basil improves the flavour of adjacent tomatoes.
Contested
There's no evidence for this and no mechanism anyone has proposed. Plant basil next to tomatoes anyway — you'll want them at the same time in the kitchen.
what is actually known
No proposed mechanism. No trials. Extremely widely believed.
Neither is a scandal. Marigolds do something real and narrow; basil is pleasant beside a tomato and the flavour claim is not supported. The point is that you can see which is which, instead of reading a cell that says "companion" and inferring something that was never claimed.
What you get instead
When this planner tells you to do something, it cites the rule that says so, and the rule names its mechanism, its evidence grade and its sources. When the evidence is thin, the page says thin. When nothing is known, it says that too, and treats the absence as the finding.
That is a slower product to build and a harder one to sell than a grid. It is also the only version of this that is worth trusting.
All explainers · Plan a bed