How to Judge Deer Research

Most deer-resistance advice online is anecdote repeated until it sounds settled. The published work is better, and it is not all the same kind of better. Knowing which kind you are reading tells you how far it travels.

The three kinds of evidence

Practitioner consensus. A committee of professionals agrees on ratings. Cheap, broad, and unverifiable.

Captive pen trials. Known deer, measured food, controlled comparison. Precise, and the animals are penned and often hungry.

Field studies. Real plants, wild deer, actual gardens or woodland. Realistic, and much noisier.

Each answers a different question. Confusing them is where most bad advice comes from.

What the Rutgers list is

The four-tier scale this site uses is the first kind, and it says so. It was compiled with input from nursery and landscape professionals, extension personnel and Master Gardeners in northern New Jersey. It is not a feeding trial. It names no deer species anywhere.

That is not a criticism. No trial covers 500 plants, and practitioner consensus across hundreds of gardens is genuinely informative about the middle of the range, which is where disagreement lives. Everyone knows hostas get eaten and lavender does not. The value is in the calls on hydrangea, rose and weigela.

But it means the list carries local assumptions. Rutgers files rosemary under annuals, because northern New Jersey cannot overwinter it. That is a hardiness judgement from one state travelling nationally inside a deer rating.

Our methodology page states which parts of our own data are verified and which are not, for the same reason.

Why pen trials and field trials disagree

Two Connecticut studies on predator urine reached opposite verdicts, and the reason is instructive.

The 1991 work sprayed urine directly onto foliage weekly and found it significantly reduced browsing. The 2010 work tested commercial products as sold and put coyote urine in the ineffective group. Same state, same animal, opposite conclusion.

The difference was concentration and delivery. One tested a substance; the other tested a product. Both were right about what they measured, and only one describes what you can buy.

This is the single most useful habit to develop: ask what was actually tested. The coyote urine that rates High in Colorado's table was 100 per cent undiluted. Registered products contain 3.5 to 5 per cent.

What captive trials cannot tell you

Penned deer cannot leave. A trial can tell you whether an animal prefers treated or untreated food when both are in front of it, and it cannot tell you whether a deer would walk past your garden.

The most-cited comparison in this field, twenty commercial repellents on captive black-tailed deer over eighteen weeks, is excellent within that frame. It found no product eliminated browsing, that contact repellents beat area repellents, and that eight of the nine best performers gave off sulphurous odours.

Read as ranking, that is strong. Read as a promise about your hedge, it is not, and the authors do not claim otherwise.

The fence-height data has the same shape. Deer were deterred 0 per cent of the time by fences under five feet and 100 per cent at just under eight, but those were motivated deer being tested on what they can clear in containment. It is a ceiling of ability, not a prediction of behaviour.

What good evidence looks like when you find it

Two examples worth knowing, because they show what is possible.

The Georgia Milorganite work counted 500 individual soybean plants per plot across five properties and reported browsing day by day: 23.9 per cent of treated plants browsed at day 3 against 54.4 untreated, and 80.2 against 100 by day 30. That is a field trial reporting decay honestly rather than quoting its best week.

A systematic review of the repellent literature found odour repellents based on meat and blood, and on predator scent, to be the strongest categories. Pooling is how you see past any single study's quirks.

The regulatory gap nobody expects

Worth knowing before treating an EPA registration number as a quality mark.

For predator urine, the agency's own registration document says: efficacy data were not submitted and are not required. More broadly, EPA generally does not require efficacy data for a non-public-health pesticide claim. A label reading "repels deer" has never had to be demonstrated to anyone.

Registration is a statement about safety. It is not evidence that a product works, and products exempt under the minimum-risk rule are not reviewed at all.

What the research agrees on

Across kinds and regions, a few findings hold.

Preference is relative, never absolute. Every serious system describes tiers, and Rutgers' own text says no plant is deer proof.

Pressure changes everything. The same plant rates differently at different deer densities. This is the largest single source of contradiction between published lists.

Season matters more than most lists admit. Late winter browsing bears little resemblance to summer browsing, because deer lose weight on winter browse even where it is unlimited.

Aromatic foliage is the most reliable predictor, and there is now a mechanism: deer judge by volatiles before tasting, and were reluctant to begin foraging at all near high-terpene plants.

Delivery beats ingredient. Applied to the plant beats distributed through an area, repeatedly, across studies that disagree about everything else.

Using it

Treat a published rating as a starting probability and calibrate against your own garden. If something rated Rarely Damaged is being eaten on your property, your pressure is higher than the rating assumed, and the right response is to move the whole plan one tier more conservative.

That single adjustment is worth more than any individual plant recommendation on this site.

Common questions this raises are answered in the plant FAQ, and everything we rely on is listed at sources.