Winter is a food problem before it is a cold problem, and there are only four answers: leave, shut down, stay active, or die and leave eggs. All four are committed to in advance, on day length — which is why a warm autumn can be a trap.
The question "how does an animal survive winter" has fewer answers than it looks. Cold itself kills relatively few things; what winter really does is remove food, and an animal can respond by leaving, by shutting down until the food returns, by staying active on stored or cached supplies, or by not surviving at all and leaving eggs that will. Almost every temperate species does one of those four. The genuinely interesting part is when the decision gets made. None of these strategies can be started once winter arrives — a bird needs weeks of feeding before it can cross a desert, and an insect must arrest its development before the first frost. So the commitment happens in advance, and it is made on day length, because day length is the only environmental variable that is exactly the same in a warm year as in a cold one. That was a superb bet for as long as day length predicted temperature. It is a worse one now. Warming has moved the conditions without moving the calendar, and long-term British records covering more than ten thousand seasonal events show the schedule is not simply sliding earlier as a block: different trophic levels are advancing at different rates, so the gaps between species and the food they depend on are opening. Which strategies suffer most follows a simple rule — the ones that committed earliest and can least be reversed.
Developed record · 70% complete · reviewed 2026-08-11
What this page covers
A comparative subject rather than a group of organisms. Draws on temperate insects, birds, mammals and plants, where seasonal timing has been measured longest.
Often confused with: Hibernation, which is one strategy within this and not the whole of it; Migration, likewise; Sleep, which is a daily state and unrelated
Quick facts
The cue
Day length, because weather is not a reliable signal
Insulation is cheap and animals are good at it. A small bird in a British winter is not usually in danger of freezing; it is in danger of running out of daylight hours in which to find enough food to get through the following night. Reframe winter as a food shortage with a cold snap attached and the strategies stop looking like a miscellaneous list and start looking like a small set of answers to one question.
There are four, and almost every temperate species uses one of them.
Four answers to the same problem
Strategy
What it means
Examples
What it costs
Leave
Go where the food still is, and come back
Swallows, warblers, many hoverflies and moths
The journey itself, and weeks of feeding beforehand to fuel it
Shut down
Suppress metabolism until food returns
Ground squirrels, dormice, bats, hedgehogs
Fat laid down in advance, and total vulnerability while under
Stay active
Keep going on stored fat, cached food or whatever is still available
Foxes, wolves, tree squirrels, most corvids, honey bees
A high daily food requirement in the season with the least food
Leave eggs
The adult does not survive; the next generation overwinters as eggs, larvae or pupae
Most wasps and most annual insects
The entire adult population, every year
The last one is the most commonly misread, because it does not look like a strategy — it looks like failure. A wasp nest in September holds thousands of workers and by December every one of them is dead. That is not the colony losing. It is the colony spending itself: the year’s output was several hundred new queens, each of which mates and then overwinters alone in a crevice, and next summer’s nests are built by the survivors. The colony was always the disposable part.
Plants use the same four-way logic with two of the options removed. They cannot leave, so it is shut down — deciduous trees dropping leaves and going dormant — or leave seeds, which is what an annual does.
Words used here
Overwintering
Whatever an organism does to get from one growing season to the next. Not a synonym for hibernation.
Annual
An organism whose whole life fits inside one growing season, leaving seeds or eggs to carry on.
Animals commit to winter before winter arrives, using day length rather than weather
Well supported
Good evidence backs this, though some details remain open.
Seasonal transitions — dormancy, migration, moult, reproductive timing — are cued predictively rather than by current conditions, and photoperiod dominates because it is the only environmental variable that does not vary between years. Strategies differ in how far in advance they commit and how reversibly: behavioural adjustments can be reversed within hours, whereas developmental arrest or deep dormancy cannot. Climate warming shifts conditions without shifting photoperiod, and the resulting mismatch is most damaging for the strategies that commit earliest.
Who this applies to
seasonally timed strategies in temperate organisms
Studied in
Animalia, Insecta, Aves
You may have heard
“Animals hibernate or migrate when it gets cold”
Cold is the problem, not the signal. Waiting for cold would be too late — a bird needs weeks of feeding before it can fly to Africa, and an insect must arrest development before the frost arrives. So the cue is day length, which is perfectly reliable and says nothing at all about how warm this particular autumn is. That gap is where climate change does its damage.
Why we rate it this way, and what the caveats are
Well supportedModerate confidence
Photoperiodic cueing is established experimentally in many systems, and the differing rates of phenological advance are measured across ten thousand time series. Confidence is moderate because the commitment framework itself is NatureHQ’s synthesis across physiologically unlike mechanisms rather than a result anyone has tested as such.
How far it can be extended
Photoperiodic control of seasonal timing is documented across insects, birds, mammals and plants; the degree of commitment varies by mechanism.
Caveats
Comparing strategies across insects, birds and plants is conceptual rather than quantitative, and the commitment ordering is NatureHQ’s framing rather than a published ranking.
Temperate systems dominate; tropical seasonality works differently.
Photoperiod is not the only cue — temperature, food and rainfall all modulate it.
The mismatch evidence is British, and geographic generalisation is an assumption.
Still unanswered
How quickly can photoperiodic thresholds themselves evolve as seasons shift?
Which strategies are most exposed to mismatch, and is commitment the right predictor?
The physiological end of the commitment spectrum, in torpor and hibernation.
The obvious cue would be temperature, and almost nothing uses it as the primary one. The reason is that temperature is the variable that matters and a terrible signal — a mild fortnight in February carries no information about March. Day length is the reverse: useless as a description of present conditions, and a perfect statement of the date. It is identical in a warm year and a cold one, unaffected by cloud, and has not changed in millions of years. Given a choice between a signal about what is happening and a signal about what time it is, seasonal biology consistently chose the clock.
It has to be measured, which is a real physiological problem, and animals solve it by comparing light against an internal circadian rhythm rather than by counting hours directly. The threshold — the day length at which the response fires — is genetically set, varies with latitude in a clean gradient, and is heritable enough that it can shift measurably within a few decades. That last point is the hopeful one: photoperiodic response evolves, and it has been observed doing so.
Temperature does still get a say, as a modifier rather than a trigger. A cold spring delays what day length has already authorised; it does not authorise it. This is why the timing of a wood’s bud burst varies by a couple of weeks between years while never occurring in December, however mild.
The reason garden centres can sell poinsettias in flower every December is this same mechanism, run backwards: growers control the plant’s day length in a blackout house to fire a response the calendar has not reached.
The day length an animal responds to is itself an evolving trait
Well supported
Good evidence backs this, though some details remain open.
Critical photoperiod — the day length at which a seasonal response is triggered — is heritable and varies clinally with latitude across many insect and vertebrate populations. In the pitcher-plant mosquito Wyeomyia smithii, populations resampled across several decades show a measurable shift towards the shorter critical day length characteristic of more southern populations, consistent with selection under a lengthening growing season.
Who this applies to
demonstrated in one insect; latitudinal clines are generalDo not extend this beyond the taxa listed — the popular version over-reaches.
Studied in
Insecta, Wyeomyia smithii
You may have heard
“Animals will adapt to climate change”
Some will, and the evidence that they can is real — a mosquito has already moved its calendar. The question was never whether adaptation is possible but whether it is fast enough, and that depends on generation time and on how much variation a population happens to be carrying. A mosquito gets many attempts per decade. An albatross gets very few.
Why we rate it this way, and what the caveats are
Well supportedModerate confidence
The heritability and the clines are established across many systems. Confidence is moderate on the rate of change, which rests on a small number of resampled populations.
How far it can be extended
Latitudinal clines in critical photoperiod are widespread, but observed evolutionary shift over decades has been documented in very few systems.
Caveats
One well-studied insect carries most of the weight on the rate of change.
Showing that a threshold can shift does not show it will shift fast enough for any particular species.
Long-lived animals have far fewer generations in which to do it.
Still unanswered
Which taxa have enough standing variation in photoperiodic threshold to track a shifting season?
Bradshaw and Holzapfel, 2007 · Annual Review of Ecology, Evolution, and Systematics
Heritability of critical photoperiod and the latitudinal cline that gives the shift its direction.
How we know
Measuring the same mosquitoes decades apart, in a light cabinet
Seasons are shifting and organisms are responding. Is the seasonal timing mechanism itself evolving, or are animals simply reacting to warmer weather?
The difficulty is that observing an animal do something earlier in a warm year says nothing about evolution — a plastic response looks identical. This design removes the weather from the measurement entirely. Pitcher-plant mosquito larvae from populations along a latitudinal gradient in North America were reared in the laboratory under controlled day lengths, and the critical photoperiod — the day length at which they enter diapause — was measured as a property of the animal rather than of its season. Those values were then compared against the same measurement made on the same populations decades earlier.
What happened
Critical photoperiod had shifted towards shorter day lengths — the southern direction — and the shift was largest in the northern populations, where the growing season has lengthened most.
What it shows
That the seasonal clock is a heritable trait under selection and can move measurably within decades. Because the trait was measured under fixed laboratory light, the change cannot be a plastic response to a warmer summer, which is the explanation that would otherwise account for everything.
What it does not show
It does not show that adaptation will keep up in general. The pitcher-plant mosquito runs several generations a year and carries unusually well-characterised photoperiodic variation; a long-lived vertebrate has neither advantage. The correlation with warming is also inferred from the direction and geography of the shift rather than from any manipulation of climate, and one species is one species.
The controls — what makes this evidence rather than a story
Rearing under controlled artificial light, so the trait measured is the inherited threshold rather than a response to this year’s conditions.
A latitudinal gradient sampled throughout, giving an independent scale against which any temporal shift can be read: north-to-south variation is large and well established.
The same assay as the historical measurements, so the comparison is between like and like.
Multiple populations rather than one, so a shift at a single site cannot drive the result.
The axis that decides which strategies climate change breaks.
Strategies differ in something more useful than their mechanism: how early they commit, and whether the commitment can be undone. A bird that has started south can stop and wait out a cold snap. A caterpillar that has entered diapause has switched off its own development and cannot restart it until it has been cold for weeks, whatever the weather does in the meantime. Both are bets on the same forecast, and they are not the same size of bet.
Partly — the caches remain useful whenever winter comes
Migration
Weeks — fattening must precede departure
Departure can be delayed; the journey can be interrupted
Hibernation
Months — fat must be laid down through autumn
Arousal is possible and costs a large share of the winter budget
Diapause
Often a generation ahead, at a fixed sensitive stage
No — termination requires a chilling period first
Read down that column and the vulnerability ordering falls out without needing a separate theory. A strategy that commits late and reverses cheaply can absorb a bad forecast. One that committed a generation ago on a cue that no longer predicts what it used to has no move available. Diapause is the extreme case in both directions: it is the most reliable defence against a winter that behaves normally, and the least able to respond to one that does not.
Leaving, when the animal that arrives is not the one that set out
Words used here
Daily torpor
A short bout of suppressed metabolism lasting hours, used overnight by hummingbirds and some small mammals. The cheap, reversible end of shutting down.
Diapause is what most insects do instead of hibernating, and the difference is worth being exact about. Hibernation suppresses the metabolism of an animal that is otherwise developmentally finished. Diapause halts development itself — an egg that will not hatch, a larva that will not pupate, a pupa that will not emerge — and it is initiated by a cue rather than by conditions, often long before conditions deteriorate.
Sometimes the cue arrives in a different generation from the arrest. In several moths the day length experienced by the mother determines whether her eggs will enter diapause. The individual that stops developing never measured anything; the decision was made for it, weeks earlier, by an animal that is now dead.
Ending it is the part that surprises people. A diapausing insect does not simply resume when it warms up — it usually has to be cold for a sustained period first, and only then will warmth restart development. The chilling requirement is a safety catch: it makes the arrest immune to a mild spell in November, which would otherwise be a fatal invitation. It also creates a failure mode that only exists in a warming climate. A winter too mild to satisfy the requirement leaves the animal unable to terminate diapause on schedule, and some horticultural species now fail to emerge properly in mild years for exactly this reason.
An insect in diapause usually has to be cold for weeks before warmth can restart it
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Diapause is a programmed developmental arrest initiated by a cue in advance of adverse conditions, maintained hormonally, and in most temperate species terminable only after a sustained chilling period. Warmth applied before the chilling requirement is satisfied does not resume development. Entry, maintenance and termination are separately regulated, and the cue is frequently received at a life stage — sometimes in the previous generation — different from the one that arrests.
Who this applies to
temperate insects with a winter diapause
Studied in
Insecta
You may have heard
“A warm winter wakes insects up early”
Usually the opposite. Most diapausing insects cannot restart until they have been cold long enough, so a mild winter does not wake them early — it risks leaving them unable to wake on schedule at all. The safety catch that stops a warm November from being fatal becomes the problem when November stops getting cold.
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
Directly testable by manipulating temperature in the laboratory, and confirmed repeatedly across species.
How far it can be extended
Chilling requirements are documented across many insect orders, though the duration and the sensitive stage vary widely.
Caveats
Not universal: some species have shallow diapause with little or no chilling requirement.
The hormonal account is best characterised in a handful of laboratory species.
Field chilling is harder to quantify than laboratory chilling.
Still unanswered
How many species now experience winters too mild to satisfy their chilling requirement?
The separate regulation of entry, maintenance and termination, and the chilling requirement.
Plant seeds have the same catch under a different name. Stratification — the cold period many seeds require before they will germinate — is the same logic: do not sprout during an autumn warm spell.
Words used here
Diapause
A programmed halt in development, triggered in advance by a cue and usually requiring a period of cold before it can end.
Chilling requirement
The sustained cold an organism must experience before dormancy can be terminated. What stops a warm November from being fatal.
"Does it hibernate" is the question almost everyone asks, and for most familiar animals the answer is no. That is not a technicality. Assuming an animal is asleep somewhere all winter gets its ecology backwards, because staying active through the leanest season is a much harder thing to do than shutting down.
What familiar animals do
Animal
Strategy
The detail that matters
Red fox
Stays active
Hunts all winter, and mates in the middle of it — cubs are born in spring because the pregnancy runs through the worst months
Tree squirrel
Stays active, on caches
Sleeps through the worst weather but does not hibernate; it depends on food it hid in autumn
Ground squirrel
Deep hibernation
One of the most extreme hibernators known — the arctic ground squirrel holds a body temperature below zero
Wasp colony
Dies; queens overwinter
The whole nest dies. Mated new queens hibernate individually in crevices and start again
Honey bee colony
Stays active
Never dormant. The colony clusters and shivers, running on stored honey, and must keep the cluster warm all winter
Bumblebee colony
Dies; queens overwinter
Same pattern as wasps, and the reason a very large bumblebee in early spring is always a queen
Bear
Hibernation, by the metabolic definition
Metabolism to about a quarter of basal; body temperature drops only a few degrees, which is why the term was argued over
Wolf
Stays active
Larger prey is more accessible in winter, not less — deep snow favours the pack
Squirrel is the entry that shows why the question cannot be answered at the level of a common name. Tree squirrels and ground squirrels are both squirrels, and they sit at opposite ends of the whole gradient — one stays awake all winter on hidden food, the other runs at a few per cent of its normal metabolism for months. A single answer to "do squirrels hibernate" is wrong whichever way it is given.
Day length works as a cue only because it correlates with everything else. Warming is dismantling that correlation: spring conditions have moved earlier by weeks in parts of the temperate world, and the solstice has not moved at all. An organism reading the calendar is now reading a calendar that is running late.
If every species were late by the same amount, little would break — the whole schedule would simply have shifted. That is not what the records show. An analysis of more than ten thousand British seasonal time series across land, freshwater and sea found that sensitivity to temperature differs systematically between trophic levels, with plants and the invertebrates that eat them advancing faster than the predators above them. The calendar is not sliding; it is stretching, and the gaps that open are between species that need each other.
The textbook case is a woodland food chain three links long: oak bud burst, the caterpillars that eat the new leaves, and the tits whose chicks need the caterpillar peak. Oak advances fastest, caterpillars follow, and the birds — which must commit to laying weeks earlier — advance least. The peak arrives before the chicks need it. Some populations are keeping up and others are not, and which is which appears to depend on how much variation in timing the population had to begin with.
The commitment gradient predicts who is exposed. A generalist that adjusts its behaviour weekly can track a shifting season. A long-distance migrant leaving Africa on a photoperiod cue has no way of knowing that spring in the Netherlands came three weeks early. And an insect in diapause, committed a generation ago and requiring cold it may no longer get, is holding the least reversible position of all.
Species are not all moving earlier by the same amount, which is what does the damage
Well supported
Good evidence backs this, though some details remain open.
Analysis of over ten thousand long-term British phenological time series across terrestrial, freshwater and marine environments finds that sensitivity of event timing to temperature differs systematically between trophic levels, with primary producers and primary consumers generally advancing faster than secondary consumers. Warming therefore alters the intervals between interacting species rather than translating the whole seasonal calendar.
Who this applies to
British terrestrial, freshwater and marine systemsDo not extend this beyond the taxa listed — the popular version over-reaches.
Studied in
Animalia, Plantae
You may have heard
“Spring is arriving earlier”
True, and harmless if that were all. If everything moved forward together the year would simply start sooner and nothing would break. What the long records show is that the parts are moving at different speeds — leaves faster than caterpillars, caterpillars faster than the birds feeding chicks on them — so the calendar is stretching rather than sliding, and the gaps that open are between species that need each other.
Why we rate it this way, and what the caveats are
Well supportedModerate confidence
The dataset is exceptionally large and the trophic pattern is consistent within it. Confidence is moderate because it is one country, and because sensitivity to temperature is not the same as a demonstrated population consequence.
How far it can be extended
A single national dataset, unusually large. Whether the same trophic ordering holds elsewhere is an assumption rather than a finding.
Caveats
British data; the geographic generalisation is not tested.
Time series vary in length, method and taxonomic resolution.
A widening gap in timing is not by itself evidence of a population decline.
Still unanswered
Does measured asynchrony translate into measured population change, and in which systems?
The differing rates of advance across trophic levels, measured across ten thousand time series.
There is one clear observation of the cue itself evolving: a North American pitcher-plant mosquito shifted its critical day length measurably over a few decades, adapting the clock rather than the response to it.
Words used here
Phenology
The timing of recurring biological events — budburst, emergence, laying, migration.
Phenological mismatch
When two species that depend on each other shift their timing by different amounts, so the overlap they need shrinks.
Trophic level
Position in a food chain — plant, herbivore, predator.
Garner and Allard, working on tobacco that refused to flower, showed that the controlling variable was day length rather than temperature or nutrition, and named the phenomenon photoperiodism. The result came from agriculture and reorganised seasonal biology.
1936
Replication
Photoperiodic control extended to animals
Bird reproductive condition was shown to be controllable by manipulating day length in captivity, establishing that the mechanism was not peculiar to plants.
2002
Modern discovery
Diapause regulation synthesised
A review of insect diapause established entry, maintenance and termination as separately regulated processes under endocrine control, and put the chilling requirement at the centre of why the arrest cannot be reversed by warmth alone.
Photoperiodic response shown to be an evolving trait
A synthesis of latitudinal clines and selection experiments established that critical day length is heritable, varies geographically in a predictable gradient, and can shift within decades — making the cue itself a target of selection rather than a fixed constant.
Mismatch demonstrated as a difference in rates, not a shift
Over ten thousand British time series showed that phenological sensitivity to temperature differs systematically across trophic levels, so warming pulls interacting species apart rather than moving the calendar as a block.
Why it matters: It is the difference between populations adapting to a shifted season and populations running out of time. One clear case of rapid shift exists and it is not enough to generalise from.
What would settle it: Repeated measurement of photoperiodic thresholds in the same populations across decades, in more than the handful of systems where it has been done.
Does mismatch reduce populations, or only reduce breeding success?
Why it matters: Timing effects are much easier to measure than demographic ones, and a species can tolerate poor years without declining. The link from mismatch to population trend is assumed more often than demonstrated.
What would settle it: Long-term demographic data alongside phenological data in the same populations.
Is the commitment gradient actually predictive?
Why it matters: It is a framework assembled from physiologically unlike systems, and its main claim — that irreversible strategies fail worst — is plausible rather than tested across taxa.
What would settle it: A comparative analysis ranking strategies by reversibility and testing that ranking against observed population responses to warming.
How do tropical organisms time anything?
Why it matters: Day length barely varies near the equator, seasons are driven by rainfall, and rainfall is not predictable in the way photoperiod is. Nearly all of this subject is temperate.