The history of developmental biology explains how researchers learned to investigate the making of an organism: where its cells come from, how they acquire different identities, how tissues influence one another and how genes contribute to an organised body. Its central change was from describing successive forms to testing the processes that connect them.
The gold standard is to understand what a historical experiment actually distinguished. Separating cells, moving tissue, tracing descendants and changing gene activity answer different questions. This guide connects selected milestones with original worked cases, so that a reader can judge a developmental claim rather than merely remember a discovery date. All numerical examples are hypothetical teaching exercises, not historical measurements or laboratory instructions.
Start with Developmental Biology for the modern subject. The Gold Standard master connects the wider collection, while the Education parent supplies the practical learning test: can you use the explanation when the organism, measurement or experimental conditions change?
Choose your developmental history route
Chapters 1–3: Read the experimental question
Chapters 4–6: Test tissue interactions
Chapters 7–9: Follow lineages and genes
Chapters 10–12: Compare reprogramming and models
Chapter index
Read the experimental question · Chapters 1–3
1. Begin with a question that a sequence of pictures cannot answer
2. Understand why the treatment of a cell changes the question
Test tissue interactions · Chapters 4–6
4. Read the organiser experiment as a test of relationships
5. Learn what necessity and sufficiency can establish
6. Understand why a receiving tissue must be able to respond
Follow lineages and genes · Chapters 7–9
7. Follow a cell lineage without confusing ancestry and identity
8. See how genetic screens changed the scale of the question
9. Separate the identity of a body region from its mere presence
Compare reprogramming and models · Chapters 10–12
10. Read nuclear reprogramming as a test of retained capacity
11. Understand what induced pluripotency added
12. Choose the organism for the question, then state the transfer limit
Test transfer and continue learning · Chapters 13–14
13. Changed-context exercise: does a new signal create an organiser?
14. Build a reading practice that survives a different experiment
CHAPTER 1 OF 14 · Read the experimental question
1. Begin with a question that a sequence of pictures cannot answer
A series of embryo drawings can establish an order: one form precedes another. It cannot, by itself, tell us why the transition occurs. A fold might arise because cells change shape, because neighbouring tissue pulls, because growth differs across a surface or because several processes act together. A convincing history of development follows the methods that made such alternatives distinguishable.
Imagine three photographs of a fictional growing structure. Its length increases from two units to three and then four. A caption says that its cells are multiplying. That might be correct, but length alone does not demonstrate it. Existing cells could elongate, rearrange or take up more space. Counting cells and measuring their shapes would answer a more specific question than adding another photograph of total length.
Separate description, mechanism and evidence
Description names the change. Mechanism proposes the process that produces it. Evidence is the observation or comparison that supports the proposed connection. These categories belong together, but they should not collapse into one sentence that appears to explain everything while testing nothing.
A developmental timeline becomes useful when each entry includes the question that became answerable. An improved microscope might reveal divisions. A label might identify descendants. A transplantation might test whether a cell’s ordinary position determines its later behaviour. A genetic comparison might identify a component necessary for a particular pattern.
The distinction also helps prevent a misleading story of inevitable progress. Researchers did not start with today’s categories and steadily fill in their missing details. Their available specimens, instruments and assumptions shaped what they could ask. An experiment can be ingenious within its historical setting even when its interpretation later changes.
When reading a historical paper, first write a neutral account of the manipulation and observation. Only then write the author’s interpretation. Finally, ask whether the observation also fits a rival explanation. This order makes it easier to appreciate both the achievement and its limits.
The chapters below select experimental turning points rather than attempting a complete global history. Plant development, regeneration and comparative embryology have substantial histories beyond this selection. The purpose here is a coherent route through the reasoning that transformed studies of embryos into investigations of developmental mechanisms.
CHAPTER 2 OF 14 · Read the experimental question
2. Understand why the treatment of a cell changes the question
The late nineteenth-century comparison between Wilhelm Roux and Hans Driesch is valuable because superficially similar experiments were not equivalent. The NCBI developmental biology account describes Roux’s 1888 destruction of cells in early frog embryos and Driesch’s subsequent isolation of sea-urchin blastomeres. Destroying a cell while leaving it present is different from separating a living cell from its neighbours.
The reported outcomes helped expose the difference between a cell’s ordinary contribution and what it can produce under changed conditions. The lesson is not that every early embryonic cell behaves identically across species. It is that a conclusion about developmental potential depends on the manipulation, stage and surrounding conditions.
A fictional comparison
Suppose a two-part system normally produces a left and a right structure. In treatment A, one part is damaged but remains attached. The other produces only one side. In treatment B, the two parts are separated and each produces a smaller complete arrangement.
Treatment A does not automatically prove that each part contains instructions for only half the structure. The damaged material may still affect space, contact, signalling or mechanical conditions. Treatment B demonstrates a capacity under isolation, but it does not show that the isolated part would ordinarily exercise that capacity inside the intact system.
A good comparison would record what was removed, what remained, whether contact persisted and how survival was assessed. It would also compare equivalent stages. The same intervention performed before and after a developmental transition may have different consequences without either result being mistaken.
This is a general repair for an overconfident historical summary. Replace “the experiment proved cells are fixed” with a description of the observed outcome under the stated treatment. Then explain which additional comparison would test the broader claim.
Do not turn these historical manipulations into an independent practical activity. Their educational value can be explored through diagrams, published observations and fictional datasets. The reasoning is accessible without reproducing procedures on living embryos.
CHAPTER 3 OF 14 · Read the experimental question
3. Distinguish a fate map from a map of possibilities
A fate map records what a region or cell normally contributes to under specified conditions. It is not a catalogue of everything that region could become. This distinction is one of the most useful ideas a reader can carry from experimental embryology into modern cell biology.
Imagine a marked patch that normally contributes to an outer surface. Its fate map assigns it to that surface. If the patch is moved and later contributes to an internal structure, the original map has not necessarily been disproved. The map described the ordinary trajectory; the transplantation tested behaviour in a different context.
Worked case: what does a labelled percentage mean?
In a fictional dataset, a marked region produces 80 labelled descendants. Sixty are found in tissue A, sixteen in tissue B and four cannot be assigned confidently. The observed distribution is 75 percent in A, 20 percent in B and five percent unassigned.
It would be misleading to say the region is exclusively destined for A. It would also be misleading to exclude the four unassigned cells silently and report 60 out of 76 as though every descendant had been identified. That alternative denominator answers a different question: the proportion among classified descendants.
Now suppose the same region is moved and produces 40 descendants, of which ten enter A and thirty enter B. The change in distribution is informative, but it must be interpreted alongside the change in total recovery. Did the treatment alter survival, division or tracking success? A shift in percentages is not automatically a pure change of identity.
The strongest account preserves both numbers and conditions. It reports the ordinary distribution, the altered distribution and the unresolved causes of any difference. It does not ask a label to answer questions that the experiment never measured.
For a beginner, draw two separate diagrams: one labelled “usual descendants” and another labelled “outcomes under tested conditions.” This simple exercise prevents the common mistake of treating observed history as an unchangeable limit on developmental possibility.
CHAPTER 4 OF 14 · Test tissue interactions
4. Read the organiser experiment as a test of relationships
The organiser studies associated with Hans Spemann and Hilde Mangold changed the questions that could be asked about tissue interactions. The University of Freiburg’s historical account describes the transplantation work and its place in the history of embryonic induction. Spemann’s 1935 Nobel lecture is a primary historical route into the experimental reasoning.
The essential conceptual distinction is between a graft that simply develops into its own structures and a graft that influences surrounding host tissue. Those possibilities require evidence about the origin of the resulting cells. A second structure alone does not tell us which cells built it.
Ask who contributed what
Consider a hypothetical graft marked blue and a host marked yellow. After transplantation, an additional organised structure contains both blue and yellow cells. That observation supports host participation; it prevents the simple claim that the entire new structure grew only from the graft.
It still does not identify a particular signalling molecule. Nor does it establish that every host cell has the same responsiveness. The observation concerns an interaction under a specific developmental context. Molecular explanation requires additional experiments.
An appropriate comparison might place other tissue at the same site, place the same tissue at another stage or examine whether host tissue survives without the proposed interaction. These comparisons test different alternatives. A sham operation can help distinguish the effects of manipulation from the effects of the graft’s identity.
The word “organiser” is useful when tied to this evidence. It becomes misleading if it encourages a picture of a miniature commander that knows the complete body plan. Biological coordination does not require a cell that surveys the finished organism.
Historical credit should likewise follow the work rather than only the most famous name. Mangold’s experimental contribution belongs in an account of the transplantation evidence. A history that omits the practical achievement loses part of its explanation of how the knowledge was obtained.
CHAPTER 5 OF 14 · Test tissue interactions
5. Learn what necessity and sufficiency can establish
A developmental structure can depend on a component without that component being sufficient to generate the structure alone. Conversely, adding a component can produce an effect in a prepared system without proving that it is the only normal cause. These distinctions make historical experiments easier to compare.
Necessity is tested by asking what happens when the component’s relevant contribution is absent or impaired. Sufficiency is tested by asking whether introducing it, in a specified context, can produce the proposed outcome. Neither word should be used without stating the context and outcome.
Three fictional results
A tissue forms in 90 of 100 untreated cases. After a targeted intervention, it forms in 20 of 100. After a restoration treatment, it forms in 75 of 100. The intervention is associated with a 70-percentage-point reduction; restoration recovers 55 percentage points relative to the intervention group.
These results support a contribution from the targeted process, especially if the intervention and restoration are well controlled. They do not establish complete rescue to the untreated outcome. They also do not show whether the affected component acts directly on tissue identity, indirectly through survival or through another process.
Suppose adding the component to a different region produces an additional structure in 30 of 100 cases, while the corresponding control produces none. That is evidence of an effect in the tested setting. It does not mean the component can create the structure in any cell at any age.
The repair is to attach qualifiers to the experimental claim rather than to bury them in a footnote. State the developmental stage, receiving tissue and measured outcome. A precise conditional claim is more useful than an impressive universal claim that the data cannot support.
When reading older and newer papers together, use the same questions. What was changed? What remained available? What was restored? What outcome was counted? This makes the comparison about evidence, rather than about whether the technique sounds old-fashioned or advanced.
CHAPTER 6 OF 14 · Test tissue interactions
6. Understand why a receiving tissue must be able to respond
A signal can be present without producing a developmental response. The receiving cells may lack the relevant machinery, may already be committed to another trajectory or may require another condition. This responsiveness is often discussed through the concept of competence.
The practical consequence is that a failed induction experiment has more than one possible explanation. The candidate signal might be ineffective, but the recipient might also be unable to respond at the chosen stage. The experiment must discriminate between failure to send and failure to receive.
A stage-by-treatment comparison
Imagine an early tissue responds in 36 of 40 treated cases, compared with four of 40 controls. A later tissue responds in six of 40 treated cases and two of 40 controls. The response proportions are 90 and ten percent for the early group, and fifteen and five percent for the later group.
The treatment difference is 80 percentage points early and ten percentage points later. This pattern suggests that stage modifies the observed response. It does not, by itself, identify why. Differences in tissue health, delivery or exposure could also contribute.
A useful follow-up would check whether the treatment reached both tissues and whether an independent response remains possible in the later tissue. Merely increasing the treatment intensity might create damage without resolving the causal question.
The comparison also illustrates why pooling can mislead. If one study contains mostly early specimens and another mostly late specimens, their overall response rates may differ even when their stage-specific results agree. The stage composition is part of the evidence, not an incidental detail.
For historical reading, competence provides a bridge between tissue transplantation and molecular regulation. A receiving cell is an active participant with a history. Its response depends on its state, and that state is itself a product of earlier developmental events.
CHAPTER 7 OF 14 · Follow lineages and genes
7. Follow a cell lineage without confusing ancestry and identity
A lineage records relationships between cells through division. It answers where a cell came from. Identity describes its present characteristics or function. Position describes where it is. These three descriptions can be connected, but none is a substitute for the others.
The 1983 paper by Sulston, Schierenberg, White and Thomson, The embryonic cell lineage of the nematode Caenorhabditis elegans, provides a landmark primary reference for tracing embryonic ancestry. Its historical importance includes making a complex developmental sequence explicit enough to compare and investigate.
Count divisions rather than assuming doubling
Suppose four tracked cells are present at the beginning of a fictional interval. Three divide once and one does not. Seven cells are present at the end if none dies or leaves the observed population. The count increases by three because each division replaces one cell with two.
If two cells subsequently die, the population becomes five. A final count of five does not reveal that three divisions and two deaths occurred. The same count could arise through a different combination of division, movement and loss. A lineage record contains information that an endpoint total discards.
Now suppose two descendants look similar despite coming from different branches. Their resemblance does not erase their different ancestry. Conversely, two sister cells can develop different identities. A branching diagram is not automatically a diagram of functional similarity.
Tracking also has observational limits. If cells cross, leave the imaging volume or become indistinguishable, the record may contain uncertainty. A responsible reconstruction marks ambiguous links rather than making every branch appear equally secure.
For a learning exercise, create a small lineage with known divisions and losses, then give a partner only the final count. Ask how many histories could produce it. The exercise makes clear why temporal observation was a scientific advance rather than merely a more elaborate illustration.
CHAPTER 8 OF 14 · Follow lineages and genes
8. See how genetic screens changed the scale of the question
A genetic screen can begin with an altered developmental outcome and work toward the genes associated with it. This complements transplantation: instead of moving a piece of tissue, investigators compare organisms carrying inherited changes and examine the resulting patterns.
Nüsslein-Volhard and Wieschaus’s 1980 paper, Mutations affecting segment number and polarity in Drosophila, reported fifteen loci whose mutations altered larval segmental pattern. The abstract distinguishes spatial organisation at several scales. That provides a concrete historical example of using systematic differences in outcomes to organise questions about developmental control.
Why classification matters
Imagine a fictional screen finds three recurring types of defect: loss of a broad region, loss of alternating units and disruption within individual units. These categories suggest different levels of organisation. They do not immediately establish a complete causal sequence, but they give investigators more specific hypotheses than the undifferentiated label “abnormal development.”
A screen is also shaped by what it can detect. If researchers examine only visible external features at one time, they may miss internal defects, earlier lethal effects or later functional changes. Absence from the results is not proof that a gene has no developmental role.
Suppose 1,000 candidate lines are examined and forty show a selected phenotype. The observed yield is four percent of the screened lines. It is not four percent of all genes in the organism. Lines may affect the same gene, some changes may be missed, and the selected phenotype defines what enters the count.
The denominator therefore belongs in the explanation. Ask whether the unit is a line, an allele, a locus, an embryo or a scored feature. These are not interchangeable quantities, even when a summary uses the convenient word “mutant.”
The historical gain was partly systematic coverage: many perturbations could be compared under a shared scoring scheme. The intellectual gain came from connecting patterned differences with experimentally testable levels of developmental organisation.
CHAPTER 9 OF 14 · Follow lineages and genes
9. Separate the identity of a body region from its mere presence
A body can contain repeated units that differ in identity. Development therefore has at least two related tasks: producing an arrangement and assigning appropriate characteristics within it. A change in identity should not be confused automatically with the addition or deletion of an entire unit.
Edward Lewis’s work belongs to this history of genetic control. The NICHD account of its Nobel-associated research identifies his contribution to understanding the relationship between genes and body-segment organisation. Lewis shared the 1995 physiology or medicine prize with Nüsslein-Volhard and Wieschaus.
A comparison using a fictional repeated structure
Imagine a structure with six positions, normally labelled A, B, C, D, E and F according to their features. After a genetic change, the positions still number six, but the fourth displays features normally found at the third. The count is unchanged; the identity pattern has changed.
A second alteration produces only five positions while preserving the recognisable identities of those that remain. That is a different phenotype. Calling both “missing development” conceals the distinction that could guide further investigation.
There is also a measurement problem. If identity is assigned using only one visible feature, an intermediate or mixed state may be forced into an overly simple category. A stronger classification examines several relevant characteristics and records ambiguous cases.
The useful mechanism question is how a regulatory change alters downstream activities and tissue behaviour. Saying that a gene “makes a segment” can hide the processes between gene activity and anatomy. A gene does not act as a miniature sculptor; its effects occur through interacting molecular and cellular processes.
For students, a practical test is to redraw the fictional six-position system twice: once with a position removed, once with a position retaining its location but changing identity. If the diagrams look the same, the distinction needs further work before moving to molecular detail.
CHAPTER 10 OF 14 · Compare reprogramming and models
10. Read nuclear reprogramming as a test of retained capacity
A specialised cell behaves differently from an early developmental cell. One historical question was whether this difference necessarily involved losing the genetic information required for other cellular possibilities. Nuclear-transfer experiments provided a way to test retained capacity in a different cellular environment.
The 2012 Nobel scientific background describes Gurdon’s experiments and their importance for reprogramming. The critical reasoning connects the state of a donor nucleus with what it can support after transfer. It distinguishes the information retained from the way that information is being used in the original cell.
Do not erase the contribution of the receiving system
A successful outcome after transfer does not show that the nucleus acted alone. The receiving cell supplies a biological environment. The conclusion concerns what the transferred material can support within that environment, not what it would do without it.
Suppose a hypothetical transfer study starts with 200 attempted procedures. One hundred produce assessable early outcomes, twenty reach a later criterion and five pass a more demanding final test. The percentages are fifty, ten and 2.5 percent relative to all attempts. Among the hundred assessable early outcomes, the final success rate is five percent.
Each fraction can be valid if its denominator is stated. Selecting the most flattering denominator without explanation changes the apparent meaning. The historical significance of demonstrating a possibility is also different from demonstrating a reliable practical method.
Failures need care as well. A failed transfer may reflect technical damage, incompatibility, incomplete reprogramming or limitations in the source material. It does not automatically prove irreversible loss of information.
This chapter connects with History of Cell Biology. Both histories ask how experimental context changes the interpretation of cellular behaviour. The stronger account identifies the contribution of each part of the system rather than assigning the whole result to its most conspicuous component.
CHAPTER 11 OF 14 · Compare reprogramming and models
11. Understand what induced pluripotency added
Takahashi and Yamanaka’s 2006 paper, Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors, supplied a different route into reprogramming. Its reported factors were Oct3/4, Sox2, c-Myc and Klf4 under the study’s specified culture conditions.
The historical change was a more defined experimental approach to altering cellular state. This does not mean that identity is trivial to change, that every treated cell becomes equivalent or that any resulting cell population is automatically suitable for practical use.
A marker is evidence about a marker
Consider a fictional experiment in which 1,000 starting cells produce fifty colonies that pass an initial marker screen. Ten of those colonies then pass a more demanding functional assessment. The initial yield is five percent; the yield under the second criterion is one percent of the starting number.
Calling the entire fifty-colony group fully validated would merge two different evidential thresholds. Conversely, reporting only the ten later successes without the screening history would conceal how the group was selected.
A marker can be useful because it correlates with a state, but the relationship needs validation. A strong claim combines appropriate molecular, cellular and functional evidence. It also specifies whether the observations concern a single cell, a colony or a mixed population.
The same care applies to the word pluripotent. It refers to developmental potential under appropriate conditions; it should not be used as a synonym for a complete organism or for unlimited capability. A historical discovery about cellular state does not by itself establish a safe treatment.
For independent study, compare the logic of this chapter with nuclear transfer. What was changed in the donor material? What did the receiving environment provide? Which outcome established the claimed capacity? The comparison reveals continuity in the scientific question alongside a change in experimental approach.
CHAPTER 12 OF 14 · Compare reprogramming and models
12. Choose the organism for the question, then state the transfer limit
The history of developmental biology includes a succession of productive model systems, but it should not be told as a competition to identify one universally best organism. Different systems make different observations and interventions feasible.
An externally accessible embryo may support one kind of observation. A well-characterised lineage may support another. Genetic resources, generation time, tissue accessibility and established staging systems all influence which question can be answered convincingly.
A comparison with developmental time
Suppose organism A reaches a chosen landmark after ten hours under the stated conditions, while organism B reaches a superficially similar landmark after twenty. Comparing both at twelve hours would compare different developmental states. Equal elapsed time does not guarantee equivalent stage.
Now suppose a treatment delays A’s landmark to fifteen hours. At twelve hours, the treated group may appear to lack a structure that is present in controls. A later observation might reveal delayed formation rather than permanent absence. The distinction matters before proposing a mechanism.
This does not mean staging can solve every comparison. Two specimens can share one landmark while differing in other processes. State which feature defines equivalence, and avoid treating a convenient category as proof that every tissue is aligned.
The Zebrafish guide develops the importance of stage, genetic background and observed outcome. The Zygote guide helps separate early division, cellular organisation and later development.
Transfer across organisms should begin with a precise relationship: a conserved component, comparable cell behaviour or similar experimental response. Shared vocabulary alone is insufficient. A gene with a related sequence may act within a different developmental setting, and a similar shape may arise through different processes.
A useful learning habit is to finish every comparison with two sentences. First state what the model supports. Then state what remains untested in the target organism. This keeps the value of model research visible without turning it into an unrestricted claim.
CHAPTER 13 OF 14 · Test transfer and continue learning
13. Changed-context exercise: does a new signal create an organiser?
A fictional report introduces a candidate signal into a small embryonic region. An additional patterned structure appears in thirty of forty treated cases, compared with two of forty controls. The authors conclude that the signal alone creates an organiser and controls the entire body plan.
Before accepting that conclusion, separate the observed effect from the larger interpretation. The treated proportion is 75 percent, the control proportion five percent and the difference seventy percentage points. Those are strong descriptive differences in this invented dataset, but the calculation does not specify the mechanism.
Work through the alternatives
First ask which cells built the additional structure. If all its cells derive from the introduced material, the result differs from recruitment or redirection of host cells. A lineage label could help distinguish these possibilities if it remains reliable through the relevant period.
Second ask whether the result depends on the recipient’s developmental stage. If only an early tissue responds, the claim must include that condition. The effect would concern a signal acting within a competent system, not a signal carrying every requirement by itself.
Third ask whether the introduced signal changes survival or growth. Extra tissue might reflect expansion of an existing population rather than a new regional identity. Cell counts, identity measures and temporal observations would help distinguish those alternatives.
Fourth ask how the structure was classified. One marker or an approximate shape may not establish a complete organised axis. A stronger assessment would use several appropriate features and a prespecified scoring rule, including an ambiguous category when necessary.
Now change the evidence. Removing the candidate signal prevents ordinary development of the relevant structure, and restoring it recovers the outcome. This strengthens the case for a normal contribution. It still does not erase the role of other signals, the receiving tissue or the physical environment.
A defensible conclusion would say that the candidate affects the probability of the scored outcome under the tested conditions, with additional evidence supporting a role in normal development if the removal and restoration results hold. That is a substantial finding without claiming more than the comparisons establish.
CHAPTER 14 OF 14 · Test transfer and continue learning
14. Build a reading practice that survives a different experiment
A useful study of developmental history ends with a method for reading unfamiliar evidence. Select one primary paper and one historical account. Use the account for orientation, then examine the paper’s actual comparison. Do not assume that a later summary preserves every condition needed to understand the original result.
Create a short record containing the organism, stage, manipulated variable, comparison group, measured outcome and stated limit. Add a separate line for what remains unknown. This keeps the historical narrative connected to the experimental work.
Questions readers often ask
Is embryology the same as developmental biology? Embryology concerns embryos; developmental biology has a broader scope that includes processes of form, identity, growth and change across relevant parts of the life cycle. Their histories overlap substantially, so the terms should be interpreted in context rather than forced into a sharp historical boundary.
Did experiments replace description? No. An intervention is interpretable only when the ordinary sequence is understood well enough to recognise a difference. Careful description supplies the baseline, the staging system and often the unexpected observation that motivates a new experiment.
Does a fate map prove that a cell cannot change? No. It records an observed trajectory under specified conditions. A claim about alternative possibilities requires an appropriate change of conditions and evidence about the resulting outcome.
Does a successful rescue prove the proposed mechanism? It strengthens the interpretation when the restoration is specific and the controls are appropriate. It does not automatically identify every intermediate step or eliminate all indirect effects.
Why study an old experiment when newer molecular methods exist? Older experiments often isolate a logical distinction with unusual clarity. Understanding removal versus isolation, ancestry versus identity or necessity versus sufficiency improves the interpretation of modern data.
A practical route through the sources
Begin with the Roux–Driesch comparison for experimental design. Read the organiser material for tissue interaction. Use the Sulston and colleagues lineage paper to examine ancestry through time, and the Nüsslein-Volhard–Wieschaus paper to see how systematic genetic differences were classified. Compare nuclear transfer with the Takahashi–Yamanaka paper for two approaches to developmental potential.
For a written exercise, choose one claim and give it a deliberately misleading summary. Then repair the summary by restoring the missing organism, stage, denominator or comparison. Explain why the repaired version is scientifically stronger even if it sounds less dramatic.
The How X Works hub extends this style of explanation to other subjects. Within development, return to the modern root when you need the present-day mechanisms. History becomes useful when it helps you see exactly what an experiment adds, what it leaves unresolved and which observation should come next.