Pruning Flowering Shrubs and Roses: Deducing Instead of Memorising

What wood does it flower on: the question that replaces the calendar

There are only two behaviours, and everything else follows from them. A shrub that blooms on old wood formed its flower buds during the summer after its previous flowering, and it carries them all winter on stems that are already woody. A shrub that blooms on new wood starts from nothing in spring: it grows first, and it is that fresh growth which will carry the flowers a few weeks later. The first is therefore pruned right after flowering, because that is the only moment when it has not yet rebuilt its buds and still has the whole season ahead to do so. The second is pruned at the end of dormancy, before bud break, because the cut destroys nothing that already exists and because it is precisely that cut which triggers the growth that will flower.

This key makes a species by species calendar unnecessary. Lilac, forsythia, weigela, spring-flowering spiraeas and spring-flowering viburnums do not bloom at the same time, but they all follow the first rule. Buddleia, rose of Sharon, Japanese spiraea and panicle hydrangea all follow the second. That leaves only two boxes to remember instead of thirty lines of a table, and above all the rule keeps working in front of an unfamiliar shrub, which no calendar can do.

Getting the timing wrong does not kill the shrub. It grows back, often even more vigorously than before; it simply lacks its flowers, for one season. What makes the mistake so frustrating is not its severity but how quietly it repeats itself. A gardener who cuts back their forsythia in late winter sees no damage at all: the shrub comes back green and dense, so they conclude that the pruning did it good, and they do it again the following year. Three years on, they have a beautiful shrub that no longer flowers, and nothing about the plant’s appearance points to the cause.

The truly irreversible pruning accidents almost never come from timing but from severity and species: cutting back into bare wood a shrub that cannot regrow from old wood. It is also worth knowing that the rule of flowering wood allows for three families of special cases, which do not invalidate it but sit at its edges: shrubs that follow the rule but that any cutting weakens, such as spring-flowering viburnums, which bloom on old wood from buds formed as early as the year before, so that they need almost no pruning at all and an annual cut thins their flowering even when timed correctly; repeat-flowering cultivars, which bloom on both old and new wood at once; and shrubs grown for their foliage, for which the question does not even arise.

Working it out yourself, on a shrub whose name you do not know

Four clues point the same way, and the first is enough in most cases. Flowering date, read against the midpoint of the growing season rather than a fixed month, sorts almost every garden shrub. A shrub that flowers in the first half of the season, between bud break and midsummer, simply has not had the physical time to build its flowers since bud break: they were already there beforehand. It therefore blooms on old wood. A shrub that flowers in the second half, from high summer into autumn, has had all the time it needs to form them on fresh growth. This criterion carries over unchanged to either hemisphere, because it is relative to the season rather than to a calendar month.

Second clue, to check during or just after flowering: what is the flower actually sitting on? If the flower cluster is perched at the tip of a green, soft, leafy shoot that appeared this year and connects further down to brown wood, the shrub blooms on new wood. If it emerges straight from a woody, rigid, bark covered stem all the way up to the flower, it blooms on wood at least a year old. Forsythia gives the clearest demonstration of this: its flowers emerge from bare wood, before the leaves even appear.

Third clue, the most direct and the most reassuring, because you can check it secateurs in hand right before cutting: on a shrub that flowers on old wood, the flower buds are already formed and visible throughout winter, and they look nothing like the leaf buds. They are bigger, rounder, plumper, often clustered, while a growth bud is pointed and pressed flat against the stem. On a lilac, the difference is obvious at arm’s length. On a shrub that will flower on new wood, there is nothing of the kind: every bud looks the same.

Fourth clue, useful when you want to date a stem rather than classify it: bud scale scars. Each year of growth leaves at its base a tight ring of fine circular scars, marking where the terminal bud’s scales fell away at bud break. Between two such rings lies exactly one year of growth, which lets you tell, on any given stem, where last year’s wood ends and this year’s begins, and so count its age by hand.

When nothing settles it, there remains the controlled test, which costs you one stem. At the end of dormancy, cut back a single stem, mark it with a tie, leave everything else untouched and wait for flowering. If the cut back stem flowers just like the others, the shrub blooms on new wood and you can safely prune it at that time of year. If it is the only one that fails to flower, it is an old wood shrub, and you have just learned that for the price of one branch instead of the price of an entire season.

Species by species, and the two genera that trip everyone up

The table below gives the flowering wood, the timing expressed as a phenological marker, and the expected severity of the cut. It is not meant to be learned by heart: it serves as a check after you have worked things out for yourself, and as a fallback when a shrub falls into one of the exceptions.

Hydrangea is the first trap, because the common name covers species that behave in exactly opposite ways. Hydrangea macrophylla, the mophead and lacecap hydrangea, along with Hydrangea serrata and the oakleaf hydrangea, Hydrangea quercifolia, all bloom on old wood: a hard prune in winter or early spring removes buds that are already formed. Hydrangea paniculata, with its pointed, cone shaped flower spikes, and Hydrangea arborescens, with large round heads carried on flexible stems, bloom on new wood and can be cut back hard at the end of dormancy. Without a label, the shape of the flower head alone is usually enough to decide. One category still blurs the picture: repeat-flowering macrophylla types, which appeared with 'Endless Summer', introduced in 2004, and which form buds on both old wood and new growth at once, so they flower again even after a winter that has killed off the old buds. For these old wood hydrangeas, two moves beat a single prune: leave the faded flowerheads in place through winter, where they protect the buds just beneath them from frost, then cut them off in early spring, working back to the first pair of clearly swollen buds. Quercifolia is the exception in the group: it needs only a minimal spring prune, limited to dead wood and stems that have grown too long.

Clematis is the second trap, and it has long been codified into three pruning groups. Pruning Group 1 gathers the early flowering clematis, such as Clematis montana, alpina, armandii and cirrhosa, which bloom on the previous year’s growth: no regular pruning at all, at most a light tidy up right after flowering. Pruning Group 2 gathers the large flowered, late spring hybrids, of the 'Nelly Moser' type, which bloom on short shoots arising from last year’s wood, and often a second time later on the current year’s growth: very light pruning at the end of dormancy, limited to dead wood and cutting back to a pair of healthy buds, followed by a second pass after the first flush. Pruning Group 3 gathers the late summer clematis, viticella, tangutica and the herbaceous species, which bloom on the current year’s growth: cut all the stems back to the lowest pair of healthy buds, between 15 and 30 cm from the ground, at the end of dormancy once the buds start to move. Without a label, the date of first flowering tells you the group.

Photinia falls outside the flowering logic altogether: it is grown for the red colour of its young foliage, and every cut triggers a flush of fresh red leaves. The trouble is that these young leaves are exactly what Entomosporium maculatum, the fungus responsible for the shrub’s leaf spot, colonises, and trials show the amount of disease tracking the amount of pruning. In a garden where the disease is already present, prune less often rather than more, schedule the cuts for the dormant season rather than summer, and pick up fallen leaves, since splashing rain spreads the infection straight back from them.

Where to place the cut: the bud decides the direction

A cut is never made into thin air. Pruning cuts fall into two basic types, heading cuts that shorten a stem back to a bud, and thinning cuts that remove an entire branch back to its point of origin, and both are made in relation to something that will take over afterwards: a bud, a side shoot, or the swelling at the base of a branch. Between the cut and whatever takes over, almost nothing should remain. On a heading cut, the right distance above a bud is around five millimetres. Any closer and you injure or dry out the bud itself. Any further and you leave a stub that the plant no longer feeds: it dies, and a dead stub is exactly the entry point for wound fungi. Coral spot (Nectria cinnabarina), which covers dead wood in small orange pustules, is a weak pathogen that gets in almost only through wounds and stubs, then works its way down the stem.

The direction the retained bud faces decides what comes next. The shoot that will extend the stem grows in whichever direction that bud points: it is the one real lever you have over the shrub’s future shape. Choose an outward facing bud to open up the centre of a bush rose or a shrub that is becoming congested, and, as an exception, an inward facing bud to fill a gap. You still need to look at how the buds are arranged, though. In species with opposite buds, lilac, weigela, viburnum, maple, buds come in pairs: you are not choosing a direction but an axis, and the cut is made straight, just above the pair. In species with alternate buds, rose, forsythia, buddleia, spiraea, you genuinely get to choose.

The angled cut exists so that water does not sit on the wound right against the bud: the slope tilts away from it. Its usefulness is real but modest, well below that of the distance to the bud and the sharpness of the blade. A blunt blade crushes and tears the fibres instead of slicing them, and a crushed wound stays open longer. On species with opposite buds, an angled cut makes no sense in any case: it would sacrifice one of the two buds.

As soon as you remove an entire branch, making a thinning cut, the rule changes. At the base of every branch sits a slight swelling, the branch collar, where the tissues of the branch and those of the trunk interlock. That is where the plant builds the chemical barrier that seals the wound off from the rest of the wood, and that is where the healing callus grows from. Cutting flush and removing that collar opens up the trunk itself and gives rot the clearest possible route in; leaving a stub prevents the callus from forming at all. So cut just in front of the collar, without cutting into it. And do not paint the wound: wound sealants slow down the natural compartmentalisation process instead of helping it. The one known exception concerns plums and cherries, where a sealant applied straight after the cut can help keep out the spores of silver leaf disease. On a heavy branch, work in three stages, an undercut first, a full cut further out, then the final cut in front of the collar, otherwise the weight tears away a strip of bark right down the trunk.

What the cut sets in motion: apical dominance and the sap-drawer

As long as a stem’s terminal bud is in place, the buds below it stay dormant: this is apical dominance. Remove it, and the ones below wake up within hours. The textbook explanation, a flow of auxin coming down from the tip that would be cut off by pruning and instantly release the buds, no longer holds up: those buds break dormancy up to twenty four hours before any measurable change in auxin. The initial trigger is a redistribution of sugars, previously captured by the growing tip and now flowing toward the buds as soon as it disappears; auxin steps in afterwards, to settle the new hierarchy among the resulting shoots. The observation itself does not change, and it is the whole mechanism behind pruning: cutting a plant does not make it grow, it lifts an inhibition.

This gives a rule of severity that common sense gets backwards: the harder you cut back, the fewer buds break, and the more vigorous each of them turns out to be. A buddleia cut back to around 30 cm sends up only a handful of stems, but they are enormous and carry flower spikes that easily reach 30 cm long; the same buddleia left uncut piles up old wood and covers itself in small, sparse spikes two or three metres up. Light pruning produces the opposite: many weak, short shoots, all clustered just below the cut. That is exactly the effect wanted on a hedge kept trimmed with shears, and exactly the effect suffered by a border shrub shortened at the same spot every year: a dense crust at the surface, and an interior that thins out for lack of light.

Since auxin travels down the stem, the angle of that stem matters: apical dominance is strongest when the stem stands upright. Bend a long climbing rose stem down to the horizontal and you have not cut it at all, yet you have removed its hierarchy: buds wake up along its whole length and the stem flowers from end to end instead of only at the tip. That is the real reason for training climbing roses and wisteria horizontally or in a fan, well before it is a matter of looks.

The sap-drawer belongs to the same family of reasoning, applied to larger cuts. When shortening an older branch, you do not simply stop wherever is convenient: you cut just above a side branch that you keep, and that side branch, the sap-drawer, keeps pulling sap into the portion you have retained. Its rule is a rule of diameter: it must measure at least a third of the diameter of the branch being removed. Too small, and it fails to keep circulation going in the retained wood, the tip dries out, and the cut ends up creeping back further down in the following seasons.

The term is very often misused, applied to a vertical water sprout arising from a main branch, or invoked to justify leaving some random length of branch in place. It is exactly the opposite: a sap-drawer is chosen deliberately, and a branch left in place without being chosen is nothing more than a stub with a flattering name. On the thin stems of a border shrub the question does not arise; it becomes central as soon as you reduce large branches on an old specimen, and that is the point where a rejuvenation prune is distinguished from a partial felling.

Roses, category by category

A word of vocabulary first, because getting it wrong costs more than any error of timing. On a climbing or shrub rose, two things need to be told apart. The main canes, or framework, are the large principal stems that give the plant its structure and that you train against a support; they last several years and are only ever cut on purpose, to renew one of them. The side shoots, or laterals, arise from these main canes and are what actually carries the flowers. Almost all of a climber’s annual pruning falls on the laterals, almost never on the main canes. Confusing the two is the fastest way to undo five years of framework building in one afternoon.

The costliest mistake is to prune a once-flowering climber in late winter, at the same time as the other roses. A repeat-flowering climber blooms on shoots produced during the current season from a permanent framework: you can therefore prune it throughout dormancy without touching a single flower. A once-flowering climber, most often a rambler, flowers only once, on old wood; cutting it back in late winter amounts to removing all the wood that was going to flower. What makes this mistake particularly costly is that it feeds on itself. The plant responds to the cut with long, extremely vigorous new stems that are completely flowerless for the year. The gardener sees a plant in glowing health without a single bloom, concludes it needs more pruning, repeats the operation the following winter, and can carry on like this for years without ever seeing a rose. The telltale sign of a rambler that has had a winter prune is exactly that: plenty of powerful new growth, no flowers at all.

Without a label, one season of watching is enough to decide. A single spectacular flush, then hips and nothing more: that is a once-flowering rose, pruned right after it blooms, removing at the base the stems that have just flowered and training the long new shoots of the year in their place. Successive flushes running right through to the end of the growing season: that is a repeat-flowering rose, pruned during dormancy, with laterals cut back to two or four buds and the main canes kept intact.

Among bush roses, the difference between categories comes down to what you are aiming for. On a large-flowered rose, a hybrid tea, you want one perfect bloom on a strong stem: this is the most severely pruned category. At the end of dormancy, once growth is starting again and hard frosts are no longer a threat, cut the vigorous stems back to four to six buds, at 10 to 15 cm from the base. Common practice, less firmly established, calls for cutting weak stems back harder still, to two to four buds at 5 to 10 cm, and for removing wood more than three years old, said to produce only mediocre flowers from then on. On a polyantha or a floribunda, you are after a mass of flowers in clusters rather than a single bloom: you therefore leave the stems noticeably longer, on the order of 25 to 30 cm, with more buds retained, and content yourself with cutting one old stem back close to the ground now and then to force fresh growth from the base.

Three categories remain that need very little pruning. Landscape and ground-cover roses get by almost without any: light trimming is enough to keep them in check, and after several years, cutting them back to around ten centimetres from the ground revives the clump. Old roses, the gallicas, damasks, albas and centifolias, are mostly once-flowering and bloom on old wood, exactly like a rambler: light pruning right after flowering, removal of the oldest stems at the base, never a late-winter cutback. Be careful, though: several old classes are repeat-flowering, the Portlands, the Bourbons and the Hybrid Perpetuals, and these are pruned during dormancy instead; it is the behaviour you actually observe that decides, not the class name. Modern shrub roses and English roses, almost all repeat-flowering, need light pruning during dormancy, a reduction of about a third, and the occasional removal of one old main cane.

On all repeat-flowering roses, deadheading spent blooms brings on the next flush. Cut above a well developed, five leaflet leaf that faces outward, rather than just below the flower, which would leave a length of stem with no bud capable of regrowing. Skip this step on roses grown for their decorative hips, and it serves no purpose at all on once-flowering roses. One last point puts everything else in perspective: trials that pruned roses with hedge shears, with no care at all for where the cut fell, produced flowering comparable in the short term to secateur pruning, though the plant’s structure and health were judged worse over the long term. In other words, timing and the choice of what to remove matter far more than the perfection of any single cut.

What you do not prune, or barely

The first category forgives nothing: shrubs with no dormant buds left in old wood. Lavender, rosemary, broom, rock rose, heathers and almost every conifer, including thuja and Leyland cypress, will not regrow from bare brown wood. In these plants, the inside of the clump or hedge is a permanently dead zone: you can prune within the green as much as you like, you will never reclaim anything beyond it. A lavender that has turned into a woody stump cannot be rejuvenated, only replaced; a thuja or Leyland hedge cut back into the brown stays brown. The exception among conifers is yew: it is the only one that regrows readily from old wood, which allows an overgrown yew hedge to be reclaimed by cutting back one side one year and the other the next. This is a distinction worth knowing before you plant a hedge, not fifteen years afterwards.

The second category covers shrubs that get damaged for nothing. Spring-flowering viburnums bloom on old wood, from buds already formed the year before: they need almost no pruning at all, an annual cut thins their flowering even when correctly timed, so you limit yourself to dead wood and any branch that is genuinely in the way. The same goes for magnolia, witch hazel and daphne, which heal slowly and whose natural shape is exactly what is wanted of them. This needs saying plainly, because the sheer volume of gardening literature suggests the opposite: a shrub that is never pruned does not die. Pruning is a tool, not a compulsory chore.

The third category does not rule out the cut itself but the season for it. Ornamental Prunus, cherry laurel included, are pruned during active growth rather than during dormancy. Silver leaf disease, Chondrostereum purpureum, releases its spores mainly in damp weather during the cold season, and they settle on freshly exposed wood: pruning during the active period removes most of that risk. The same seasonal reasoning applies to nesting: in many countries, destroying or disturbing an occupied nest is an offence, and a hedge tackled in the middle of nesting season almost always shelters one. Finally, never prune in frosty conditions, where wood snaps instead of slicing cleanly and the wound stays open for as long as the cold lasts, and avoid late pruning, which triggers soft new growth that cannot harden off before the first frosts.

One last point, which concerns not the plant but the person pruning it and what they leave on the ground afterwards. Several very common hedging shrubs are toxic, and it is their clippings that end up handled by the armful. In yew (Taxus baccata), every part contains cardiotoxic taxine alkaloids, with the sole exception of the red flesh of the aril, whose central seed remains toxic; the taxines persist in clippings even once they have dried out. Cherry laurel leaves (Prunus laurocerasus) contain cyanogenic glycosides, on the order of 1 to 2.5 percent of the plant material, which release hydrogen cyanide when the tissue is crushed, and poisonings of cattle, sheep and goats that had eaten pruning clippings are documented in the veterinary literature. Holly berries contain saponins that irritate the digestive tract, and both the leaves and berries of privet contain terpenoid glycosides: in either case, ingestion by a child can cause vomiting and diarrhoea. The practical upshot fits in one sentence: a heap of yew or cherry laurel clippings should never be left within reach of a grazing animal, within reach of a child, or fed to any animal at all.

The blade: what actually spreads, and when it matters

The question is almost always framed the wrong way round. People ask what to disinfect with, when the real question is what actually spreads by way of a blade, and under what conditions. Not every shrub disease travels the same way, and most of what damages a shrub after pruning does not arrive through the secateurs but through the wound itself: wound fungi, such as coral spot or silver leaf disease, travel through the air and settle on freshly cut wood. Against these, disinfecting the blade does strictly nothing: what stops them is the timing of the cut, how clean it is, and the absence of any stub.

What genuinely does travel by blade is vascular bacteria picked up from infected living tissue, foremost among them fire blight (Erwinia amylovora), which strikes precisely the ornamental Rosaceae that are so widely planted, along with certain viruses and viroids carried in the sap. And one clarification contradicts a piece of advice that circulates widely: rose mosaic virus is not spread by secateurs, it is spread through grafting and propagation from already infected stock. Disinfecting between two healthy, dormant roses protects against absolutely nothing.

Two conditions must therefore both be met for the blade to matter: cutting into living tissue that is visibly affected, an oozing canker, a shoot wilted into a shepherd’s crook, an active lesion, and doing so during the growing season. Fire blight is inactive during dormancy, to the point that the risk of transmission is essentially nil then, which is exactly why sanitation pruning of suspect specimens is recommended outside the growing season. Photinia illustrates the other side of the problem: it is not the blade that spreads its leaf spot, it is the pruning itself, because it produces the flush of young leaves that the fungus colonises.

No product makes up for a dirty blade, and that is the step most often skipped. Organic matter, sap and sawdust, inactivates disinfectants: wiping the blade clean before treating it matters more than which product you choose. The disinfectants that have been studied, ethyl and isopropyl alcohol, diluted household bleach, household disinfectants, trisodium phosphate, all work under certain conditions, but not equally. The reference trial was not carried out on pruned trees, its authors noting that testing disinfectants that way was not practical, but on young detached apple and Asian pear fruit, nicked with a scalpel: in that setup, rubbing alcohol failed to stop the transmission of fire blight, while three products came out tied for the best result, so bleach was not the only one that worked. The dilution runs counter to the most widely repeated advice: it was the one in ten solution that failed most often, while the undiluted product, or one diluted to one in five, with the blade soaked for at least a minute, held up. And no disinfectant tested gave complete protection, the authors say so explicitly: disinfection reduces the risk, it does not eliminate it. None of this is harmless either: bleach corrodes tools, which then need rinsing and oiling, something that does not apply to every product tested; it irritates the airways and must never be mixed with any other product, and there is no homemade preparation that can be presented as a risk free disinfectant. The step that actually pays off lies elsewhere: a sharp blade that slices instead of crushing, a cut made at the right distance from the bud, and diseased stems carried out of the garden rather than dropped on the compost heap.

Rejuvenating a shrub that has been left to its own devices

The diagnosis is the same whatever the species: a bare base, thick grey stems with cracked bark, flowering that keeps climbing higher and getting sparser, a dead interior starved of light. This is not a disease, it is the mechanical result of never being renewed. Old stems consume a great deal, flower poorly, and hog the light that would otherwise let young stems break from the base.

The method that works spreads over three seasons. In the first year, at the end of dormancy, remove around a third of the oldest stems, cutting them close to the ground, between 5 and 15 cm, along with anything dead, spindly or lying on the ground. In the second year, remove half of the remaining old stems, and prune the rest normally according to the species’ flowering wood. In the third year, remove the last of the old stems. After three seasons, the clump is made up entirely of young wood, without ever having lost either a full year’s flowers or its full canopy of foliage.

Spreading the work out rather than cutting everything at once answers three reasons, only one of which is about looks. First, keeping some foliage: the roots of an established shrub are sized for a given leaf mass, and removing all of it at once suddenly cuts off what feeds them. Second, keeping some flowers every year. Third, and this is the serious reason, you often do not know whether the shrub will regrow from old wood; spreading it out runs the test on a third of the plant instead of the whole thing, and the first year serves as a full scale trial.

Cutting everything back at once remains an option on species that clearly tolerate it, done at the end of dormancy: buddleia, hazel, elder, dogwoods grown for their coloured stems, hornbeam, beech and yew. Cut it all back, water through the following season, and accept one flowerless year followed by very dense regrowth that will need thinning out. Box behaves differently: it does regrow well from old wood, but slowly, and cutting it back all at once costs it several seasons of shape.

Two diagnostic mistakes come up again and again. The first concerns lilac: a good number of the stems rising from the base can come from the rootstock rather than the variety, with different foliage and vigour, and they will never produce the expected flowers. Remove these at the base instead of counting them as renewal growth. The second concerns species that do not regrow from old wood: for these, there is no such thing as a rejuvenation prune at all, and an old, bare specimen is one to replace. Cutting back just to see amounts to killing the plant to confirm something you could have known before you ever picked up the secateurs.

Finally, a shrub that has been cut back restarts with an oversized root system and explosive regrowth. Two follow-up moves are essential: water regularly through the first growing season, and thin out the regrowth the following year, because ten shoots from the same stump will produce ten spindly stems all crowding each other. A surface mulch of organic matter helps; a nitrogen fertiliser applied at the time of cutting back only makes the production of soft wood worse.

Common pitfalls to avoid

The care sheets to read next

Every plant mentioned has its full sheet: watering, light, soil, pests and a month-by-month calendar.

Browse every sheet in the catalogue →