A phylogeny of scientific insight
From Mendel’s ratios to biological memory: a proposed course on the origins and history of hypotheses.
The missing subject
Science education is comfortable teaching justification. Discovery is harder to teach. By the time an experiment becomes a classic, the question can appear inevitable, while the change in representation that made it possible disappears behind the result.
I want a course about the origins and history of hypotheses—a phylogeny of scientific insight. Students should leave with knowledge of the basic discoveries that built modern biology, taste for scientific questions, and the courage and discipline to chase an Aha! without becoming sloppy.
The point is disciplined perception. A hypothesis does not arise from “observation” alone. It often appears when a mind finds a new representation for an old fact. Something the field already saw acquires a different significance. An exception becomes a map. A repeated pattern becomes a memory. A nuisance becomes a ruler.
My opening notes describe a thought trying different isomorphic molds as it moves into written language. The developmental perspective matters because the successive forms contain information that the final form may conceal. I want to bring that perspective into the way we teach science: reconstruct the change in view that made a hypothesis possible, then let students practice making such changes themselves.
The anatomy of an Aha!
Every historical case would be taught with the same surgical structure. What did the field see before? What was the anomaly or dissatisfaction? What representation changed? What analogy made the new representation possible? What technical trick made it testable? What false explanation died? What modern descendants inherited the same Aha!?
The repeated template is the soul of the course. It keeps the historical seminar connected to an experimental question. The proposed weekly structure combines a historical insight seminar, a paper autopsy or reconstruction of the problem, work on a common cell-size dataset or lab question, and an Aha studio.
Cell size would be the practicum. It is simple to recognize and surprisingly difficult to explain completely. Students can watch a cell grow and divide, then ask what it might be “waiting for.” A particular size? A duration? Something accumulated, something diluted, a memory of an earlier condition?
At first there would be no favored theory. Students would list several hidden rules that could explain division timing. They would learn to ask what observation would separate the rules before deciding which one feels most persuasive.
Ratios, exceptions, and rare events
Mendel’s case would begin with the representation of heredity as discrete, countable ratios. “Peas have traits” is not the insight. The interesting move is that visible proportions can imply invisible units. In the practicum, students would ask whether distributions of birth and division size can likewise suggest hidden rules.
Morgan and Sturtevant would supply the next move. Failed independence could have been treated as a nuisance. Instead, the pattern of exceptions became informative about spatial arrangement. Exception frequency can become distance. Variation in cell size, viewed through the right representation, might become the ruler rather than the noise.
Benzer’s work would shift attention again, toward what rare events can make visible. A selection can turn something otherwise difficult to detect into the principal thing one observes. Which rare cells are most informative for a model of division? Extremely small newborns, unusually large ones, or particular outlier lineages? Do they obey the same rule as the rest?
Those cases would give students three different ways to treat the same general difficulty: a visible quantity might indicate an invisible unit; a departure from expectation might encode a relationship; and rarity might be made informative through the structure of an observation.
History and control
Meselson and Stahl would introduce a further idea: if a process is invisible, mark the past so that the future carries evidence. Molecular history can become something physically distinguishable. The corresponding cell-size question is whether a division decision depends only on a present measurement or also on what the cell has experienced.
Jacob and Monod would make biology available as a representation of control. A gene can participate in the regulation of an activity, rather than simply providing a structural component. Students could then draw competing pictures of division: a size sensor, a timer, an integrator of growth, dilution of an inhibitor, accumulation of an activator.
The cell-cycle work associated with Nurse and Hartwell would offer the idea of making a dynamic process intelligible through interruption. A process that is difficult to disentangle while it runs may reveal structure when it stops at a particular stage. The emphasis would be on the conceptual move: what becomes visible when a changing system is held in a state?
Each historical case would return to the same question. What could a representation of the cell preserve that another representation would discard?
Reading, amplification, and indexing
Sanger would bring ordered sequence into the course. Once biology becomes readable as an ordered molecular text, differences that were formerly hidden become locatable. Students would ask which kinds of difference that representation helps them recognize, and which aspects of a dynamic system require something beyond a sequence.
PCR would introduce amplification as a change in what questions are practical. A rare molecular signal need not remain inaccessible merely because it is rare. In an imaging practicum, repeated observations of single-cell lineages offer a different form of increased access: a relation that is ambiguous in one encounter can become visible across many.
The Church and Shendure lineage of parallel assays would lead to indexing. The leap is not simply to read faster, but to give many entities stable addresses so they can be distinguished while examined together. Indexing converts separate experiments into a landscape.
The course itself could inherit that logic. Different cohorts could ask related questions under different conditions, with enough common structure to compare what they find. A growing collection of observations would become valuable because its differences have addresses, not because the collection is merely large.
Memory, instruments, and sufficiency
The CRISPR story would begin with unusual repeated sequences and the change in interpretation that made them intelligible as historical records. Sequence pattern became memory. That invites another cell-size question: might the hidden variable be history, and what would it mean to record the history relevant to a division decision?
The conceptual move from a natural recognition system to a programmable instrument would then be examined alongside the move from DNA change to a record of events. A natural mechanism becomes an epistemic instrument: part of biology is made into a way of learning about biology.
The studio would ask for an impossible experiment first. If money and tools were unlimited, what would students want to know about growth, molecular state, size, and division? They would then reduce that imagined experiment to a question that an available observation could address. The reduction is itself an intellectual exercise: what must be preserved for the answer to remain useful?
Yamanaka’s reprogramming work would provide a final change in strategy. Sometimes the productive conceptual move is to establish that a combination is sufficient and then ask what can be removed, rather than begin by assuming that one factor at a time will reveal the whole. The cell-size counterpart is a question about interacting explanations: how would we recognize a system in which more than one rule contributes?
The genealogy of my hypothesis
The final symposium would require students to reconstruct their own changes of mind. What did they first think controlled division? What did their observations break? Which model survived—sizer, adder, timer, or a hybrid? What hidden variable do they now suspect? Which historical Aha! helped them see it? What one observation should future students make next?
The final essay would be the genealogy of a hypothesis. It would include the failed representations, the surprising cases, and the question that became possible only after an earlier question failed. A conventional result is only one part of that account.
The ambition is a course that trains students in the historical anatomy of insight while giving them a real problem in which the gap between data and explanation remains visible. The lab must be humble. The intellectual ambition must be enormous.
Students would leave with foundational biology. I also want them to leave with the habit of looking at a biological fact and asking: “What invisible thing would make this obvious?”