Life history theory explains how survival, growth, maturation and reproduction fit together across an organism’s lifetime, and how those schedules can evolve under particular conditions. Its gold standard is a testable account of the alternatives: define the life cycle, measure the relevant rates, identify constraints and inheritance, and check whether the proposed explanation survives a change in environment or assumptions.
This guide helps you read a life table, distinguish lifetime offspring from population growth, investigate a reproductive trade-off and recognise when an attractive evolutionary story exceeds the evidence. The numerical cases are original, fictional teaching models. Their values are assumptions rather than measurements from a species; the purpose is to make the reasoning inspectable.
Choose your life history theory route
Follow survival, maturation, reproduction, trade-offs and their demographic consequences.
Route 1 · Chapters 1–3: Define and count the life cycle
Route 2 · Chapters 4–6: Compare timing and allocation
Route 3 · Chapters 7–9: Test offspring and environmental responses
Route 4 · Chapters 10–11: Interpret diversity and ageing
Route 5 · Chapters 12–14: Transfer, diagnose and continue
Chapter index
Define and count the life cycle · Chapters 1–3
Compare timing and allocation · Chapters 4–6
Test offspring and environmental responses · Chapters 7–9
Interpret diversity and ageing · Chapters 10–11
Transfer, diagnose and continue · Chapters 12–14
A life history records the timing and pattern of biologically important events. When does an organism begin reproducing? How much does it grow before and after maturity? How many offspring are produced at successive ages or stages? What fraction survives between those points? These questions turn a broad description such as “short-lived and prolific” into quantities that can be compared.
Imagine two fictional plants. Both produce sixty seeds during the observation period. One flowers in its first year and dies. The other grows for two years, flowers in the third and survives afterwards. The seed totals agree, but the schedules do not. Neither total tells us how many seeds establish, whether the surviving plant reproduces again or how descendants contribute to future generations.
Life history theory asks why such schedules differ and what their consequences are. A description identifies a pattern; a demographic model follows its consequences; an evolutionary explanation investigates how inherited differences interact with reproductive success. Keeping these tasks separate prevents a fitted population curve from being mistaken for proof of adaptation.
The word “strategy” does not imply conscious planning. In this subject it can describe an allocation pattern, a developmental response or a modelled set of decisions. A seed does not forecast the next drought. A model can nevertheless compare the consequences of germinating immediately and remaining dormant under specified conditions.
Start an investigation with a sentence that names the organism, population, time interval and measured event. “Plants in this marked cohort first flowered at these ages” is more useful than “this species lives fast.” It also makes the missing information visible: a flowering record may still lack survival, seed viability and establishment.
Use The Gold Standard of Evolution for the wider mechanisms of inherited change. Here the organising question is narrower: how do the parts of a lifetime combine, and what evidence explains their joint pattern?
Before calculating anything, decide what enters the account. A newborn animal, a newly emerged seedling and a seed entering a seed bank are different starting points. A survival probability has meaning only when its beginning, endpoint and interval are stated. “Twenty per cent survived” leaves the biological task unresolved until those boundaries are known.
Age is elapsed time from a defined origin. Stage is a biological state, such as juvenile, non-flowering adult or flowering adult. Organisms of the same age can occupy different stages; organisms in the same stage can have different ages. Size can be a useful state variable when future survival and reproduction depend strongly on size rather than age alone.
Original census example
Suppose a fictional population contains 100 adults immediately before a short breeding event. They produce 200 newborns, and no individual dies during that event. The immediate post-breeding count is 300. That change is a birth pulse, not evidence that the annual population growth factor is three. We still need the subsequent survival and recruitment events to compare equivalent annual censuses.
Now suppose only 40 newborns survive to the next pre-breeding census, and 70 original adults remain. The corresponding count is 110. In this one-interval account the ratio is 1.10. It is an observed model transition from the stipulated initial state, not automatically the long-run growth factor of every future population composition.
Writing the events in order prevents two common errors: omitting newborn survival because births were counted as recruits, and multiplying by that survival twice because it was already included in recruitment. Label a coefficient as births per adult or surviving recruits per adult; those quantities are not interchangeable.
The specialist article on changing census timing without changing the animal life cycle develops this bookkeeping carefully. For this root guide, the essential habit is to draw the life cycle before selecting the formula.
For a simple age-based cohort, let l(x) mean the probability of surviving from the initial newborn state to reproductive age x. Let m(x) mean expected newborn daughters produced at that age per female alive then. The daughter-only convention simplifies this teaching model; it assumes reproduction is not limited by mate availability. Other biological systems require other conventions.
Original cohort calculation
Begin with 100 fictional newborn females. Fifty survive to age one, twenty-five to age two and ten to age three. Thus l(1) = 0.50, l(2) = 0.25 and l(3) = 0.10. Suppose the females alive at those ages produce, respectively, 1, 2 and 3 newborn daughters each. Reproduction occurs at those stated ages, and there is no later reproduction.
The expected daughter counts are 50 × 1 = 50, 25 × 2 = 50 and 10 × 3 = 30. Together they give 130 daughters per 100 initial newborn females. Net reproductive output is R0 = 0.50 × 1 + 0.25 × 2 + 0.10 × 3 = 1.30 daughters per initial female.
The conditional survival from age one to age two is 25/50 = 0.50. It is not 0.25: that number describes survival from birth to age two. Confusing cumulative and conditional survival can make an otherwise tidy calculation count mortality twice.
The denominator in fertility matters just as much. If an observer reports 50 daughters per 100 original newborn females at age one, that already includes survival to age one. Multiplying it again by 0.50 would incorrectly reduce the contribution to 25. Write the denominator beside every rate before combining them.
Finally, 1.30 is an expectation under the model, not a guarantee that a particular female produces 1.30 daughters. Some may leave none and others several. We have also assumed a complete cohort record. Field observations need an additional account of missed encounters, uncertain ages and individuals that leave the observation area.
R0 adds reproductive contributions across a lifetime. It does not retain all the information about when those contributions occur. A population whose descendants reproduce sooner can grow at a different rate from one with the same lifetime output delivered later. To compare growth per unit time, the schedule must remain in the calculation.
Original single-age comparison
Assume a deterministic, closed, single-type population in a constant environment without crowding. Schedule A produces an expected 1.44 newborn replacements at age one and has no other reproduction. Schedule B produces the same 1.44 at age two. The expectations already include survival to reproduction; do not multiply by survival again.
For A, the growth factor per interval is λ = 1.44. For B, two intervals together must give the factor 1.44, so λ² = 1.44 and λ = 1.20. Both schedules have R0 = 1.44, but they have different growth factors per interval. A generation factor cannot simply be relabelled as an annual factor.
For several reproductive ages, the discrete renewal relationship is the sum of l(x)m(x)/λ^x over all reproductive ages, equal to one. Each term discounts a contribution by its time of arrival in the growing or declining population. For the preceding cohort, the equation is 0.50/λ + 0.50/λ² + 0.30/λ³ = 1. Its positive solution is about 1.157 per age interval.
This is a long-run quantity for the stipulated schedules, not a prediction that every adjacent count has that ratio. Initial age composition can produce temporary departures. Seasonal birth pulses can make newborn counts especially uneven. Density dependence, environmental change and random events require additional modelling rather than a confident extension of the same number.
For a deeper derivation and further counterexamples, continue to comparing reproduction schedules when lifetime offspring totals agree. The useful repair is to restore the ages, not to search for a more impressive name for the total.
A trade-off means that improving one component has a cost elsewhere under the relevant constraints. A finite resource budget can create such a constraint, but organisms do not all acquire identical budgets. Comparing individuals with different access to resources can hide the cost of allocating more of a given budget to reproduction.
Original two-budget exercise
Give a fictional organism ten resource units after unavoidable operating costs. If six units go to reproduction, four remain for maintenance. Increasing reproductive allocation to eight leaves two. Within that fixed budget, the arithmetic establishes a negative allocation relationship. It does not yet specify how either allocation translates into offspring, survival or fitness.
Now compare two organisms rather than two allocations by the same organism. One acquires ten units and divides them six to reproduction and four to maintenance. The other acquires twenty and divides them twelve and eight. The second invests more in both. A positive association between reproductive and maintenance investments across these individuals does not abolish the within-budget constraint.
Resource acquisition and allocation were separated explicitly in van Noordwijk and de Jong’s 1986 model. Brown’s 2003 scorpion study tested the associated expectation that variation in investment and allocation affects the visibility of offspring size–number trade-offs. These sources support examining the origin of variation before interpreting a correlation.
A useful study therefore measures plausible differences in condition, resource access, size and age. Where suitable, it compares responses within individuals over time or uses a justified experimental design. Statistical adjustment alone cannot guarantee that every relevant difference has been removed; the biological meaning of the measured variables still matters.
Avoid treating a resource budget as a literal universal currency. Time, nutrients, space and physiological capacity need not substitute freely for one another. Ten energy units cannot automatically repair a missing nutrient. An allocation model is helpful when its constraint corresponds to the mechanism being investigated, and misleading when it silently compresses several independent limits into one number.
More offspring now need not maximise reproductive contribution over a lifetime. The relevant cost might appear as reduced survival, slower growth, lower later fertility or changed offspring performance. Finding the cost requires following the appropriate later events rather than assuming that an immediate reproductive increase must have one particular consequence.
Original two-opportunity case
Consider two hypothetical schedules available to an adult. Moderate effort produces four recruits now, permits survival to a second breeding opportunity with probability 0.80, and produces five further recruits conditional on survival. High effort produces six now, permits survival with probability 0.40, and produces four later conditional on survival. There are no further opportunities in this model.
Expected total recruits are 4 + 0.80 × 5 = 8 under moderate effort, and 6 + 0.40 × 4 = 7.6 under high effort. The larger immediate brood loses on this particular lifetime-output criterion. If survival after high effort were 0.60 instead, its expectation would become 8.4 and the ranking would reverse.
That reversal identifies what needs evidence. It does not establish that a real organism can freely choose either schedule or that the assumed survival difference is caused by reproduction. An environmental difference could affect both. An inherited constraint could make some combinations unavailable. The calculation compares supplied alternatives; it does not discover the feasible biological alternatives for us.
Timing remains important. These totals are useful for an explicitly lifetime-based comparison, but they do not automatically rank long-run growth in an expanding age-structured population. A fuller comparison must retain the dates of the recruits and the demographic context. Changing the objective after calculating the answer is a subtle way to overclaim.
The word “cost” also needs an endpoint. A temporary drop in body mass may recover before it affects later reproduction. Conversely, an apparently normal adult count may conceal poorer offspring survival. A strong trade-off claim states the mechanism, the delayed outcome and the interval over which the proposed cost was measured.
An offspring-size question involves both a production budget and a relationship between investment and later performance. Larger offspring are not automatically better in every environment, and a larger brood is not automatically more successful. The comparison needs a clearly defined stage at which success is counted.
Original investment calculation
Suppose a fictional parent can allocate 120 resource units to a brood. Producing an offspring of size two costs two units, so the budget permits sixty. Producing size-four offspring permits thirty, and size-six offspring permits twenty. Assume survival to the chosen recruitment stage is 0.10, 0.30 and 0.40 respectively.
Expected recruits are therefore 60 × 0.10 = 6, 30 × 0.30 = 9 and 20 × 0.40 = 8. The middle option wins among these three supplied alternatives. The largest offspring have the highest individual survival, but their smaller number more than offsets that advantage relative to the middle option.
The phrase “among these alternatives” matters. We have not examined size three, size five, mixed broods or different total investment. We cannot call size four a universal optimum. We have also omitted parental survival and later reproduction, so the result concerns one brood’s recruitment rather than the complete parental life history.
Now improve survival for size-two offspring from 0.10 to 0.20 while holding the other assumptions fixed. Their expected recruitment rises to twelve. The preferred option changes because the performance relationship changed, even though the total budget stayed at 120. This is why an allocation argument needs the environment attached.
In real data, count the stage carefully. Eggs, hatchlings, fledglings, established seedlings and reproductive adults answer different questions. A large number at an early stage can coexist with low later contribution. Record parental investment and subsequent offspring transitions separately so that an apparent size–number pattern can be traced to the processes producing it.
A population can mature earlier because conditions changed, because its inherited composition changed, or through both processes. The observed trait alone does not distinguish them. Phenotypic plasticity means that a given genetic background can produce different phenotypes in different environments. A reaction norm describes that relationship over the environmental range examined.
Original common-environment reasoning case
Imagine population A matures at twenty days in one field site and population B at thirty days in another. Bringing descendants into a common environment produces averages of twenty-five days for both. This outcome weakens the claim that the original ten-day difference directly measured inherited divergence. It is compatible with a substantial environmental contribution.
If the difference persists, inherited variation becomes a stronger possibility, but the interpretation still needs care. Maternal effects, developmental carry-over and shared rearing conditions can persist beyond the initial comparison. Replication, appropriate generations and experimental controls determine which alternative explanations have actually been reduced.
Suppose each population is measured in both a cool and warm environment. A matures at thirty and twenty days; B at thirty-five and twenty-five. Both shift by ten days, while B remains five days later. Those invented values illustrate an environmental response alongside a between-population difference. If the lines cross, the ranking itself depends on environment; a single universal ranking would then discard relevant information.
The existing explanation of how the same genes can produce different traits in different environments provides a useful next route. The life-history application is to measure survival and reproduction alongside development, because earlier maturity can have different consequences depending on body size and later opportunity.
An observed plastic response is not necessarily adaptive. To support that stronger claim, compare the consequences of the response with relevant alternatives under the conditions in question. Likewise, evidence of heritable variation does not prove that natural selection caused the observed population difference. Drift, migration and history remain possible contributors until the evidence distinguishes them.
Predators, disturbance and food shortage can alter life histories, but “more danger means earlier reproduction” is too broad to serve as a general law. Which ages or sizes experience the added mortality? Does reduced population density release survivors from competition? Is mortality independent of the trait being compared? The answer can change the prediction.
Dańko, Burger and Kozłowski’s 2017 modelling study demonstrates why the type and age pattern of density dependence matter when interpreting extrinsic mortality. Its model results are conditional on those mechanisms. They do not support assigning the same response to every population exposed to an external hazard.
Original compensation example
Assume a fictional annual population begins with 100 breeders. Each produces four offspring, and one quarter of offspring recruit, giving 100 recruits. Now an external event reduces breeders to fifty. If offspring survival stays at one quarter, recruitment is fifty. The breeder reduction carries straight through to recruitment under that assumption.
Alternatively, suppose reduced competition doubles offspring survival to one half. Fifty breeders then produce 200 offspring and 100 recruits. The same loss of breeders gives a different recruitment outcome because the model includes compensation at a later stage. These figures are an accounting illustration, not evidence that compensation is complete in nature.
The two models make different measurement priorities clear. A breeder census alone cannot identify the mechanism. We need offspring production, subsequent survival and the relationship between density and those rates. If the relationship changes with habitat, an estimate from one location may not transfer to another.
Nor does compensation in population numbers mean that no evolutionary change occurs. Demography describes who survives and reproduces; evolution requires information about inherited differences among them. A population can recover its count while its composition changes, or change its count without demonstrated inherited evolution. Treat those as separate findings until a study connects them.
CHAPTER 10 OF 14
10. Use fast–slow patterns without forcing every trait onto one axis
Many comparisons describe a pace-of-life contrast involving maturation, reproductive timing and longevity. Such summaries can be useful for organising data, but a single label loses information. A relatively long-lived species can still differ from another long-lived species in how often it reproduces, how reproduction changes with size and how variable recruitment is.
Salguero-Gómez and colleagues’ plant analysis examined 418 species and identified both a pace-of-life dimension and a reproductive-strategy dimension. Read it with the 2017 correction, which corrects survivorship-curve symbols in a table without changing the reported conclusions. The practical lesson is to inspect the measured traits behind a summary category.
Original classification check
Consider two fictional plants that first reproduce at age three and typically survive for ten years. One produces a modest crop each year; the other produces a large crop in occasional favourable years. Calling both “slow” may summarise their maturation and longevity, but it does not tell us their exposure to a bad reproductive season or the timing of seed-bank replenishment.
A second pair might have the same average annual seed count but different variance and dependence on adult size. A research question about response to drought could require those differences even if a broad classification puts the pair together. Summaries should serve the question rather than determine which details are allowed to matter.
For cross-species work, also consider shared ancestry and measurement comparability. Closely related species can resemble one another partly because of their common history. Different field protocols can create apparent differences. A trait recorded over one breeding season is not necessarily comparable with a lifetime estimate from another study.
Avoid extending these biological labels into judgements about individual people, ability or worth. Even within biology, “fast” and “slow” are descriptions relative to a specified comparison. They are not a universal scale of success, sophistication or evolutionary advancement. Return to the underlying schedule whenever a label appears to settle more than the measurements can support.
Lifespan is an outcome; ageing concerns changes with age. A short-lived population can experience high mortality throughout life without a steep age-related increase. Conversely, a population can have relatively low early mortality followed by a strong increase later. A single maximum recorded age cannot distinguish these histories.
The comparative research Diversity of ageing across the tree of life reports varied age patterns of mortality and reproduction across organisms. That finding makes it important to measure the age trajectory instead of assuming that every species follows a human-like pattern. It also separates an empirical pattern from any particular proposed mechanism of deterioration.
Original survival interpretation
In a fictional cohort, 100 individuals reach age one, eighty reach age two and sixty-four reach age three. Conditional survival is 0.80 across both intervals. The declining survivor count alone does not demonstrate increasing mortality with age: the interval death probability is 0.20 each time.
Compare a second cohort with 100, eighty and forty survivors. Its conditional survival changes from 0.80 to 0.50. Here the observed interval mortality increases, although a biological explanation still needs comparable conditions and an account of uncertainty. If the second interval coincided with an unusually harsh year, an age effect and a period effect could be entangled.
Experimental evolution can test particular predictions. Stearns and colleagues’ fruit-fly experiment altered adult mortality while controlling population densities and reported changes in growth, reproduction and ageing. Its design illustrates why the selection regime and population conditions belong beside the result. It is not a demonstration that every environmental mortality increase has the same consequences.
When reading ageing evidence, ask whether individuals were followed over time, whether ages were known and whether selective disappearance changed the composition of the survivors. A late-life sample contains individuals that survived earlier filters. Its average performance need not describe the trajectory of every original individual. That distinction matters before explaining an apparently flat or improving population average.
A useful transfer check changes one assumption and asks which conclusions remain valid. Begin with the offspring-size example: a budget of 120, sizes two, four and six, and survival probabilities 0.10, 0.30 and 0.40. The expected recruit counts were six, nine and eight. Now suppose drought changes those probabilities to 0.02, 0.18 and 0.35.
Original changed-context answer
The revised expectations are sixty × 0.02 = 1.2, thirty × 0.18 = 5.4 and twenty × 0.35 = 7. The largest option now produces the most expected recruits among the alternatives. The budget arithmetic transfers unchanged, while the survival relationship and resulting ranking do not.
Suppose drought occurs one quarter of the time, and the original environment occurs three quarters. If the task is expected recruitment for a randomly selected single brood, the weighted values are 0.75 × 6 + 0.25 × 1.2 = 4.8; 0.75 × 9 + 0.25 × 5.4 = 8.1; and 0.75 × 8 + 0.25 × 7 = 7.75. The middle option again leads on that defined criterion.
Do not quietly reinterpret that arithmetic average as long-run evolutionary success in a fluctuating population. Temporal sequence, shared environmental shocks, density and storage across years can matter. As a separate scalar illustration, growth factors of 2 and 0.4 have arithmetic mean 1.2 but multiply to 0.8 over the two intervals. The corresponding geometric mean is about 0.894, indicating decline across that repeated sequence.
The repair is to identify what is being averaged and what the process actually does. Single-brood expectations add across alternative states; population growth factors multiply through successive times. Neither operation is universally the correct one for every life-history question.
For a plant whose size carries information across years, the specialist guide to integral projection models, continuous traits and kernel uncertainty shows how richer demographic structure can be retained. The extra complexity is useful when it represents a mechanism the simpler calculation leaves out.
Does the organism with the most offspring have the best life history?
Not from that fact alone. Identify the offspring stage, their survival, reproductive timing and future contribution, together with costs to the parent. A count of eggs and a count of established recruits measure different transitions. The appropriate comparison also depends on whether the question concerns one brood, lifetime output or population growth.
Does a negative correlation prove a trade-off?
No. It may be consistent with one, but environmental differences, age structure and sampling can generate associations. A convincing account links the proposed constraint to the measured outcomes. Conversely, a positive correlation does not rule out a trade-off when individuals differ in resource acquisition. Repair the inference by examining the source of variation.
Is earlier reproduction always favoured?
No. Earlier reproduction can alter growth, offspring performance and later reproductive opportunity. Its consequences depend on the feasible schedules and demographic environment. The single-age model isolates timing deliberately; it is not a claim that shifting reproduction earlier leaves every other biological quantity unchanged.
Can we infer evolutionary change from a shift in average maturity?
The shift is evidence of a changed phenotype distribution. Establishing inherited evolution requires additional information about inheritance and population composition. Environmental responses, migration and selective sampling should be considered. A common-environment comparison helps with some alternatives but should be interpreted in light of its actual design.
Why does a well-fitted model fail at a new site?
Check whether the census stage, age range, detection process, resource conditions or density relationship changed. A coefficient can remain numerically precise while no longer representing the new biological process. Re-estimating everything is not always necessary; first identify which relationship lost its justification and collect evidence for that link.
What is the fastest way to find an arithmetic mistake?
Write the unit and denominator next to each number, then follow one cohort through the events. Check that cumulative survival does not increase in a closed cohort, conditional probabilities remain between zero and one, and recruits have not been multiplied by the same survival twice. Fractional expected counts are legitimate; silently mixing stages is not.
For a first exercise, construct a fictional three-age life table and explain every denominator aloud. Calculate cohort contributions before using compact notation. Then change one survival probability and identify exactly which terms change. This is a better understanding check than memorising the names of several life-history strategies without being able to reconstruct one schedule.
For the next exercise, compare two allocation options using a stated budget and a stated performance relationship. Keep the resulting conclusion narrow: which option performs better, on which criterion, under which assumptions? Add a third option or change the environment. The point is to see whether the reasoning can be rebuilt when the original ranking fails.
For a research-reading exercise, choose one of the primary studies linked above. Identify the observation unit, comparison, measured outcomes and strongest alternative explanation. Separate a comparative association, an experimental result and a model prediction. Each can contribute valuable evidence, but each answers a different question and carries different limits.
For field or conservation applications, begin with existing, appropriately collected records. Avoid treating a classroom optimum as a recommendation to manipulate organisms or populations. A useful analysis states what information would be required for an actual decision: the affected life stage, realistic intervention effect, uncertainty, time horizon and ecological consequences. It also distinguishes changes in vital rates from changes in detectability.
The connected reading route begins with Evolution for mechanisms, continues through reproduction schedules and census timing, then extends into continuous-state population models. For a neighbouring application of evolutionary reasoning, Evolutionary Medicine examines vulnerability, population change and the limits of adaptive explanations.
The Gold Standard of Education explains why changed-context practice matters. How X Works offers broader mechanism-based routes, while The Gold Standard master connects the subject collection. Keep one concrete test at the centre of the learning: can you explain how the conclusion changes when the life cycle, environment or evidence changes?