Grandma Reversing Method

Experiment 1 — From the Conclusions Back to Their Conditions

A reconstruction of the reasoning underlying Einstein's 1905 theory

We shall begin, not with the assumptions from which a theory was constructed, but with the conclusions to which it has led.

This reversal of the customary procedure may appear unusual. Ordinarily, one introduces certain principles, derives their consequences, and finally compares those consequences with experience. Here we shall proceed in the opposite direction. We shall regard the completed theory as something already before us and inquire into the conditions without which its conclusions could not have been obtained.

The procedure is simple enough to be entrusted to Grandma.

She does not ask, at first, whether the theory is true. Nor does she attempt to invent another theory. She asks a more restricted question:

Given this result, what must have been true for the result to be possible?

We must immediately distinguish between two kinds of answer. A condition may be necessary for a result, in the sense that the result could not occur without it. Or it may be merely one of several possible conditions capable of producing that result.

To confuse these two possibilities would be to mistake a successful explanation for the only possible explanation.

This distinction will guide the entire investigation.

Einstein's 1905 paper, On the Electrodynamics of Moving Bodies, provides a particularly suitable subject. Its conclusions include the relativity of simultaneity, the transformation of spatial and temporal measurements, the composition of velocities, and the reconciliation of electromagnetic phenomena with the motion of observers.

We shall take these conclusions as given and examine their logical dependencies.

The investigation is therefore not an independent derivation of relativity. It is an examination of what can, and what cannot, be inferred by reasoning backward from an established theoretical construction.

I. The problem of different observers

Let us suppose that two observers, A and B, are moving uniformly relative to one another.

Each possesses measuring rods and clocks. Each records the positions and times of events according to the instruments available in his own system.

Now suppose that a physical event occurs at a point where the observers' paths intersect.

Observer A assigns the event certain coordinates. Observer B may assign it different coordinates.

Nevertheless, both are describing one and the same event.

We must therefore distinguish the event itself from the numerical description assigned to it.

The coordinates are not the event. They are a means of identifying it within a chosen system of measurement.

Grandma begins here.

She observes that the finished theory contains rules for transforming one observer's coordinates into another's. Instead of immediately accepting those rules, she asks why rules of this kind are needed at all.

The answer is that measurements performed in different reference systems must be related if they are to constitute descriptions of the same physical world.

Without such relations, the observations would remain separate collections of numbers.

We may therefore formulate our first inference:

A theory that describes physical events from different reference systems must provide a consistent means of relating those descriptions.

This is a necessary requirement of the intended theory.

But it does not determine the mathematical form of the relation.

Grandma has identified a requirement without yet identifying its solution.

That is an important difference.

II. The consistency of physical descriptions

We shall now examine more closely what is meant by consistency.

Suppose that observer A records the meeting of two particles.

Observer B, moving relative to A, also describes the experiment.

Their measurements of position, time, and velocity may differ. Yet their descriptions must be compatible with the occurrence of the same meeting.

It would be inadmissible for one description to predict that the particles meet while another, purporting to describe the same physical experiment, predicts that they do not.

Grandma therefore concludes that the relations between reference systems must preserve the identity of physical events and the consistency of experimental predictions.

This conclusion is independent of any particular theory of space and time.

For example, in classical mechanics, the Galilean transformations provide relations between coordinates in uniformly moving systems.

In special relativity, the Lorentz transformations perform the corresponding function.

Both are mathematical structures capable of relating descriptions of events.

Their physical predictions differ, but the general requirement of consistency does not, by itself, select one rather than the other.

Grandma now refuses a tempting conclusion.

She will not infer the Lorentz transformations merely because the theory under examination contains them.

To do so would be equivalent to assuming the very result whose conditions we are trying to reconstruct.

We have therefore obtained a second conclusion:

Consistency between observers is necessary, but consistency alone does not determine the transformation laws.

This establishes the first limit of our backward investigation.

We have discovered what the transformations must accomplish, but not yet what they must be.

III. The question of simultaneous events

The finished theory tells us that events simultaneous in one inertial reference system need not be simultaneous in another.

We shall ask what must be established before such a statement can have a definite meaning.

Consider two clocks situated at different places, A and B.

An observer standing beside clock A can compare the clock's indication with an event occurring at A.

Similarly, an observer beside clock B can compare the indication of clock B with an event occurring at B.

These are local observations.

But suppose we wish to determine whether an event at A occurred simultaneously with an event at B.

We cannot simply compare the two clock readings, for that would already presume that the clocks possess a common definition of time.

Thus, before distant simultaneity can be discussed, we require a procedure for relating the indications of spatially separated clocks.

Grandma recognizes that this requirement precedes the question of whether simultaneity is relative.

She asks:

How can two clocks at different places be said to agree?

One possibility is to exchange signals between them.

Yet this introduces a new difficulty.

If a signal travels from A to B, its travel time cannot be determined from the readings of the two clocks unless their temporal relationship is already known.

The proposed measurement therefore presupposes the very relationship it was intended to establish.

We must avoid this circularity.

A signal sent from A to B and reflected back to A can, however, have its total travel time measured by a single clock at A.

If the signal leaves A at time t_1 and returns at time t_3, the observed round-trip duration is

This measurement does not require prior synchronization of the two distant clocks.

But it determines only the total travel time.

It does not, by itself, establish how that time is divided between the outward and return journeys.

Einstein resolves this problem by introducing a synchronization rule according to which the two directional travel times are taken to be equal in the chosen inertial system.

The event of reflection at B is then assigned the time

Grandma observes that this is an additional rule.

It is not an independent measurement of the two directional travel times.

The distinction matters because the finished theory depends upon a definite way of assigning time coordinates to spatially separated events.

We may therefore conclude:

A statement about the simultaneity of distant events requires a specified procedure for relating distant clocks.

But Grandma refuses to conclude that this requirement, by itself, establishes the relativity of simultaneity.

That further conclusion depends upon the transformation laws connecting different reference systems.

Once again, backward reasoning has identified a necessary condition without uniquely determining the complete theory.

IV. The transformation of lengths and times

We next consider two further conclusions of the finished theory: the dependence of measured lengths and time intervals upon the observer's state of motion.

At first sight, these results may appear to concern the physical alteration of rods and clocks.

But before discussing their interpretation, we must determine what is meant by measuring a length or a duration.

Consider a rod moving relative to an observer.

To determine its length in the observer's reference system, the positions of its two endpoints must be recorded at the same time according to that system's synchronization procedure.

If the endpoints are recorded at different times, the rod's motion must be taken into account before their separation can be interpreted as its measured length.

Thus, a measurement of the length of a moving object depends upon a definition of simultaneity.

Grandma now recognizes a dependency.

The theory's treatment of length cannot be separated from its treatment of time.

This does not yet establish the Lorentz contraction formula.

It establishes that the measurement procedures must be specified before the formula can be interpreted.

The same care is required when discussing clocks.

A clock measures the succession of events along its own path. Comparing that clock with clocks in another reference system requires rules for relating the relevant events and their temporal coordinates.

Grandma therefore refuses to treat the transformation of length and time as two entirely independent questions.

They are connected through the procedures by which spatial and temporal measurements are defined.

But she also refuses a stronger claim.

The existence of this connection does not, by itself, establish the particular quantitative relations derived by Einstein.

For those relations, additional physical assumptions are necessary.

Our next inference is therefore:

The transformation of measured lengths and times requires a coherent definition of spatial and temporal measurement, but those definitions alone do not determine the transformation equations.

V. The composition of velocities

Let us now turn to another consequence of the theory.

A particle moves relative to observer A. Observer A moves relative to observer B.

How is the particle's velocity relative to B to be determined?

Velocity is defined through changes of position and time.

Consequently, the transformation of velocities depends upon the transformation of spatial and temporal coordinates.

Grandma can therefore work backward from the velocity-composition law to a more fundamental requirement.

Before velocities can be combined consistently between reference systems, the relation between their coordinate measurements must be established.

This explains why the relativistic composition of velocities is not an isolated addition to the theory.

It follows from the transformation structure.

But does the existence of a velocity-composition law require the Lorentz transformations?

No.

Classical mechanics also possesses a velocity-composition rule.

The mere requirement that velocities be consistently related does not select the relativistic formula.

Grandma therefore refuses to infer a particular limiting speed solely from the fact that velocities must be composed.

The requirement of consistent composition is general.

The specific law requires further physical constraints.

We may express the conclusion as follows:

A velocity-composition law must be compatible with the transformation of space and time, but the requirement of compatibility does not uniquely determine that transformation.

VI. The electromagnetic problem

We now return to the phenomenon with which Einstein begins his paper.

Consider a magnet and a conducting circuit.

When they move relative to one another under suitable corresponding conditions, an electric current may be induced.

In the customary description available to Einstein, the explanation depends upon which object is regarded as moving.

If the magnet moves and the conductor is stationary, the explanation invokes an electric field associated with the changing magnetic conditions.

If the conductor moves and the magnet is stationary, the explanation invokes the magnetic force acting upon charges in the moving conductor.

The observable effect can be the same, although the explanatory descriptions differ.

Grandma asks whether this difference in explanation necessarily corresponds to a fundamental difference in nature.

The answer is not immediately apparent.

Different mathematical descriptions may represent the same physical phenomenon without implying that the phenomenon itself has changed.

We must therefore distinguish a difference in representation from a difference in observable consequences.

Grandma concludes that a satisfactory theory should account consistently for the electromagnetic phenomena in both descriptions.

But she refuses to infer that the equality of the observed effects, by itself, proves the complete equivalence of all inertial reference systems.

That would extend a limited observation into a universal principle.

Nor can the magnet–conductor example alone determine the transformation laws for space and time.

Its importance is different.

It exposes a difficulty in the conceptual organization of the existing explanation and suggests that the distinction between electric and magnetic descriptions deserves closer examination.

We therefore obtain another necessary requirement:

The theory must provide mutually compatible descriptions of electromagnetic phenomena in different states of relative motion.

But the observation alone does not uniquely specify how that compatibility is to be achieved.

VII. Have we reconstructed Einstein's postulates?

We may now collect the results of our backward investigation.

Starting from the completed theory, Grandma has identified several requirements.

Measurements in different reference systems must be related.

Descriptions of the same physical events must be mutually compatible.

Comparisons of distant times require a synchronization procedure.

Measurements of moving lengths depend upon the treatment of simultaneity.

Velocity composition must be consistent with the transformation of space and time.

Electromagnetic phenomena must admit compatible descriptions in the relevant reference systems.

These are substantial requirements.

Yet none of them, considered alone, uniquely establishes the two postulates introduced by Einstein.

In particular, the requirement of consistency does not establish the Lorentz transformations rather than the Galilean transformations.

The requirement of clock synchronization does not independently establish the invariance of the speed of light.

And the magnet–conductor observation does not prove that no privileged reference system can exist.

The finished theory satisfies the requirements we have identified.

But this does not mean that those requirements uniquely determine the finished theory.

Here Grandma encounters the central logical difficulty of reasoning backward.

From the statement

we cannot generally conclude

If a particular set of assumptions produces an observed result, the result does not necessarily establish that those assumptions were the only ones capable of producing it.

Grandma therefore distinguishes between reconstructing the dependencies of a known theory and demonstrating the necessity of that theory.

The former has been our task.

The latter requires additional argument and empirical evidence.

VIII. The boundary of the first experiment

We began with Einstein's completed theoretical construction.

We asked what conditions had to be satisfied for its conclusions to form a coherent description of physical phenomena.

We found that the theory depends upon consistent relations between observers, defined measurement procedures, and compatible descriptions of electromagnetic effects.

But we also found that these requirements do not uniquely determine the particular theory Einstein constructed.

This is not a failure of the backward method.

It is the identification of its logical boundary.

Grandma has reached a point at which the available information no longer determines a unique continuation.

She does not fill the gap by introducing Einstein's postulates without acknowledgment.

She stops and identifies what remains undetermined.

The distinction is essential.

A method that always arrives at a predetermined conclusion may merely be reproducing what its operator already knows.

A method that identifies where further assumptions are necessary provides a more transparent account of its own reasoning.

We have therefore completed the first experiment.

It was deliberately not blind. We began with the published conclusions and examined their dependencies.

In the next experiment, we shall remove Einstein's postulates from the permitted starting assumptions and give Grandma only the magnet, the conductor, and the observed electromagnetic effect.

We shall then ask how far she can proceed before another assumption becomes unavoidable.

The object is not to make Grandma discover relativity. It is to discover exactly where the observations cease to determine the explanation.

Experiment 2 — Magnet and conductor: a restricted blind reconstruction

A narrated reconstruction of the reasoning, with the assumptions and stopping points made explicit. [1]

I. The magnet, the conductor, and the current

We now remove the completed theory from Grandma's desk. We also withhold Einstein's two postulates as premises. She is told only of the comparison with which the 1905 paper begins: in corresponding configurations, a magnet and a conductor can produce the same induced current when their relative motion is the same, although the traditional explanation changes according to which is described as moving.

The initial observation is more limited than it may first appear. It does not say that every arrangement of magnet and conductor gives the same current, nor that arbitrary acceleration or changes of geometry are irrelevant. It concerns corresponding arrangements with the same relative motion. Grandma insists on retaining those qualifications because the boundary of an observation is part of the observation itself. A broad principle cannot be extracted by quietly discarding its conditions.

One might immediately name electromagnetic induction and consider the matter settled. But naming a known law is not yet an explanation of why two differently described arrangements agree. Grandma asks which features of the apparatus determine the observable result and which belong to the observer’s choice of description. The distinction will determine how far the argument can proceed.

This is a restricted reasoning experiment, not a claim that the language model has forgotten special relativity. The model already knows the history and theory. Our control concerns which statements are allowed to function as starting assumptions.

Grandma looks first at the current. It is a measurable outcome. Something in the physical arrangement must transfer energy into the motion of charges around the circuit. But the mere presence of a current does not identify the mechanism uniquely. She refuses to read the cause directly out of the effect.

II. What is common to the two arrangements?

Suppose the magnet moves past the conductor in one description. In another, the conductor moves past the magnet with corresponding relative motion and otherwise comparable physical conditions. In the idealized comparison, the induced current is the same.

The comparison must be made with care. Relative velocity is one relevant feature, but the magnetic field configuration, conductor geometry, resistance, and manner in which the circuit is completed also affect the induced current. Grandma therefore does not say that relative motion is the only cause. She says that in the stated comparison, the measured outcome does not follow the difference between the two conventional verbal accounts.

This is a useful form of refusal. If we ignore everything except relative motion, we might announce a universal law before we have examined the experiment. If we refuse even the limited comparison, we learn nothing. Grandma keeps exactly the common experimental constraint and leaves the broader claim undecided.

Grandma asks what the arrangements share. Their relative motion and electromagnetic coupling are obvious candidates. Their velocities relative to a separately imagined laboratory need not be the same. The result can be described in terms of the relation between magnet and conductor without explicitly invoking absolute motion.

But she immediately corrects a possible overstatement. The observation does not show that relative velocity alone determines the current in every apparatus. Geometry, resistance, field distribution, acceleration, and other conditions matter. We are discussing corresponding experimental configurations, not an unrestricted theorem about every magnet and wire.

The surviving fact is narrower and stronger: in the compared arrangements, an observable electrical result need not reveal which body has been assigned the motion in the chosen description.

III. Does the difference belong to nature or to the explanation?

The classical explanations are different. When the magnet is described as moving, a changing magnetic environment gives rise to an electric field. When the conductor is described as moving, charges experience magnetic forces associated with their motion.

Imagine two notebooks describing the same apparatus. The first speaks of a changing magnetic field at a stationary circuit. The second speaks of charges moving through a magnetic field. If the notebooks predict the same measurable current, their differing intermediate vocabulary need not signify differing physical events. The descriptions may be alternative ways of organizing the calculation.

But there is an important distinction between compatible descriptions and interchangeable theories. A theory must not merely attach two names to the same effect; it must tell us how the relevant fields, forces, and measurements are related. Grandma has found a problem of representation. She has not yet found the mathematical relation that resolves it.

Grandma does not dismiss either calculation. Both may correctly describe their respective arrangements. Her question is why the explanatory division appears sharper than the difference in observed results. Perhaps the distinction lies partly in the representation, rather than in an essential division of nature.

This is a hypothesis about explanatory architecture. It is not yet a physical discovery. Grandma refuses the claim that two explanations must correspond to two fundamentally different phenomena merely because their mathematical expressions differ. She also refuses the opposite claim that every mathematical difference is only conventional. That would require demonstration.

IV. Changing the outer reference frame

Grandma now shifts from the current itself to the observer who describes the apparatus. To say that the magnet moves and the conductor rests is to adopt a reference system. Another observer can assign the motion differently.

We can sharpen the question by considering what would count as a contradiction. If two descriptions of one experiment predicted incompatible currents at the same terminals under the same conditions, the difference could not be dismissed as terminology. Conversely, agreement in one observable does not imply that all other observables agree. A successful explanation must preserve the particular measured relations, not just a general impression of symmetry.

Grandma therefore asks for a rule capable of carrying predictions between descriptions. She does not assume the rule is Galilean or Lorentzian. At this point, either name would import a solution from outside the evidence she has been given. The admissible conclusion is only that a consistent relation is needed.

A theory capable of treating both descriptions must provide rules relating its quantities between reference systems. Otherwise a change in descriptive standpoint might create contradictory predictions for one and the same experimental arrangement.

Grandma has discovered the need for a relation between descriptions. She has not discovered its formula. This is where a premature reconstruction of Einstein would go wrong: one could write down the Lorentz transformations because they are familiar, then announce that the magnet and conductor forced them upon us. They did not.

V. Could classical physics still describe the result?

Yes, at least within its relevant domain. Faraday induction and the magnetic Lorentz force already provide ways to calculate the electromotive effects in the respective classical descriptions. Their coexistence is precisely part of the asymmetry Einstein discusses.

A tempting shortcut now appears. Because the current depends on relative motion in the comparison, perhaps only relative motion can ever have physical meaning. Yet this would turn a local result into a claim about every conceivable experiment. One can conceive of theoretical frameworks containing a preferred structure that remains undetected by the particular induction experiment. The observation does not exclude them by logic alone.

Grandma asks a stricter question: what would a preferred frame have to do to contradict the measurement? Unless a distinct prediction is supplied and tested, its presence or absence is not decided by this case. She refuses to claim that an unobserved structure has been disproved merely because the present experiment does not require it.

Grandma asks whether the observation is incompatible with all theories that retain some preferred background structure. It is not. A hidden preferred frame can be introduced in a theory that reproduces the same measured effects. Whether such a structure is useful, testable, or dispensable is a further question.

She therefore refuses to convert the absence of an observed asymmetry into proof that no preferred frame can exist. 'Not required by this result' and 'logically impossible' are different statements.

VI. What must a transformation preserve?

Let two observers describe the same induction experiment. Their accounts may assign different field components and different motions to the apparatus. Yet their predictions must remain compatible with the actual current and other measurable outcomes, once the complete experimental circumstances are properly transformed.

This restraint also prevents the opposite error. Grandma does not declare a preferred frame necessary because two traditional explanations look asymmetric. An asymmetry in an explanation may arise from the choice of variables rather than from an asymmetry in the world. The task is to determine which differences survive comparison with measurement.

What remains is an economy problem, but not yet an economy verdict. If two descriptions are empirically adequate, a common formulation may be preferable because it eliminates arbitrary distinctions. Preference, however, must follow adequacy. A simpler explanation that fails the current measurement is no explanation at all.

Grandma concludes that a transformation framework must preserve empirical coherence. It need not preserve every individual numerical field component. The invariant at this stage is not a particular equation written in advance, but the requirement that the descriptions concern one physical arrangement and do not contradict its measured consequences.

This observation is enough to motivate searching for transformation laws. It is not enough to choose them.

VII. Can the transformation law be derived here?

Grandma tries. What information does she possess? A comparison of induced currents under corresponding relative motions. What information does she lack? Independent constraints on the behavior of clocks, lengths, light propagation, and the transformations between inertial frames.

The electric and magnetic quantities used by the two observers are not independent labels that can be rearranged at will. A transformation law must preserve the structure of the field equations and the measured effects within its domain. That requirement is much stronger than noticing that both arrangements produce a current. It brings us to a new question about the wider theory, one that the initial example cannot answer unaided.

Grandma can now identify what additional information would be useful: laws governing electromagnetic fields, the behavior of measuring devices, and experimentally supported relations between frames. She cannot honestly derive those ingredients from the induced current alone. The observation tells us where the problem begins, not where the full solution ends.

A single class of electromagnetic observations does not determine all those relations. There are more unknown structural features than the observation can fix. Any derivation that produces the complete Lorentz transformation at this point must have imported additional assumptions, whether openly or silently.

Grandma refuses to hide that importation. This is the experiment's first decisive stopping point.

VIII. What should be tested next?

Instead of inventing the missing transformation, Grandma asks which independent observations would discriminate among the surviving possibilities. Experiments involving the propagation of light, the motion of clocks, and the comparison of inertial systems could supply additional restrictions.

There is a subtle danger in reading the historical paper backward. Knowing Einstein’s later result, a modern reader may silently insert invariant light speed as the reason the induction example is troublesome. That may be a fruitful way to understand the completed theory, but it is not an inference licensed by the induction observation alone.

For this experiment Grandma therefore treats the two postulates as unavailable premises. This is a restriction on the argument, not a removal of the model’s learned knowledge. The distinction is necessary if we wish to describe the experiment accurately rather than give it the appearance of a blind historical discovery.

The direction of inquiry has changed. We no longer ask which explanation is attractive. We ask what further measurement would force competing explanations to disagree. This is a useful result of reversing the problem: the missing condition becomes a research question.

IX. The question of economy

Suppose two frameworks explain every relevant observation equally well. One contains a preferred frame that never affects any measurable prediction; the other does not. Grandma may prefer the second as a more economical description. But economy is a selection principle among adequate accounts, not evidence that the simpler account has been uniquely proven true.

Suppose Grandma is asked to choose between several complete theories that reproduce the induction comparison. She cannot choose solely from the current, because the theories have been arranged to agree on that observable. She must request a further discriminating observation or an explicitly stated theoretical criterion. This is where an investigation can make progress by declining to make a premature choice.

The refusal is constructive: it identifies the missing test. It does not tell us that all theories are equally good, only that the evidence presently admitted does not separate them. A new experiment may do so; a wider body of established results may already do so. Neither is contained in the restricted starting packet.

The order matters: adequacy first, absence of contradiction second, and only then economy. An elegant explanation that fails the experiment is not rescued by elegance.

X. Where Grandma stops

From the magnet and conductor, Grandma has recovered a real conceptual problem. She has identified the importance of relative motion, the possible dependence of explanatory language on reference frame, and the need for compatible descriptions. She has also identified what the observation does not provide: a unique space-time transformation or a proof of Einstein's postulates.

The experiment ends with a result more modest than a reconstruction of special relativity but more precise than a general statement about symmetry. We know the empirical comparison that motivated the question. We know that differing descriptions of that comparison call for a coherent relation. We also know that the comparison does not by itself select the transformation law or prove a universal relativity principle.

Grandma has followed the observation until its explanatory reach is exhausted. She has neither denied Einstein’s achievement nor smuggled it into her premises. The next experiment will admit a second historical consideration and ask how much additional ground becomes available.

This is not a disappointing result. It is a precise boundary. The observation constrains the space of acceptable theories without selecting a single theory. In the next experiment we shall admit another empirical motivation from Einstein's introduction and ask whether it reduces the remaining freedom.

Grandma has not found the answer. She has found the next question that deserves an experiment.

Experiment 3 — Before the postulates: two empirical starting points

A narrated reconstruction of the reasoning, with the assumptions and stopping points made explicit. [1]

I. Two difficulties instead of one

We now give Grandma more evidence, but not the solution. She retains the magnet–conductor comparison and receives a second historical motivation mentioned by Einstein: attempts to detect Earth's motion relative to a supposed luminiferous medium had not yielded the expected effect.

The new starting point is deliberately broader. Einstein’s introduction mentions the magnet–conductor asymmetry and unsuccessful attempts to discover motion of the Earth relative to the light medium. We may now consider both, but we must still separate what was observed from the theoretical interpretation attached to it. The two pieces of evidence do not enter the argument with identical logical force.

Grandma begins by asking whether a failure to detect a predicted effect is the same as a demonstration that the effect is impossible. It is not. A negative experiment constrains models that predict a detectable signal under the tested conditions. It does not automatically settle every possible model of the phenomenon.

These are not identical experiments. The first concerns electromagnetic induction; the second concerns the propagation of light and the proposed reference frame of the ether. Grandma's task is to ask what, if anything, their conjunction forces us to accept.

A caution is necessary. Einstein's statement that the laws of physics may take the same form in all inertial frames is not a third observation. It is the kind of general principle we are trying to assess. To give it to Grandma as an empirical premise would prejudge the inquiry.

II. What do the observations have in common?

In the induction comparison, the physical result does not display the asymmetry suggested by two traditional explanatory descriptions. In the ether-detection problem, the sought effect of motion relative to a privileged light medium was not observed in the experiments Einstein invokes.

The induction comparison concerns the relation between descriptions of one class of electromagnetic effects. The ether-drift question concerns whether an experiment can distinguish a state of motion relative to a hypothesized medium. These questions are connected historically, but their observations are not identical. Grandma keeps them separate long enough to understand what each contributes.

A theory might explain the induction comparison and still predict an ether-drift signal. Another might suppress the signal while giving an unsatisfactory account of electromagnetic induction. To satisfy both constraints is harder than to satisfy either alone. Nevertheless, the intersection of two constraints need not consist of only one theory.

Grandma sees a family resemblance: both raise questions about whether a particular state of motion can be distinguished by the anticipated observations. Yet she refuses to say that two puzzles with a common theme must have a single cause. Similarity can suggest a hypothesis; it does not prove one.

She therefore changes her outer reference frame from two particular phenomena to the requirements imposed on any theory intended to explain both.

III. What does a negative experiment mean?

A failed attempt to detect a predicted effect is evidence against the prediction under the conditions of the experiment. It does not logically establish that every possible underlying structure associated with the prediction is nonexistent.

Consider the form of a negative result. An apparatus is designed to respond differently if the Earth has a certain motion relative to the ether. The anticipated variation is not found within the experiment’s sensitivity and assumptions. What follows? The conjunction of the model, auxiliary assumptions, and predicted detectable variation has failed. It does not follow without further reasoning that every conceivable ether model has been refuted.

Grandma does not minimize the historical importance of such results. She asks instead for the exact logical object that has been tested. An experimental refusal is strongest when we know which prediction was refused, under what conditions, and with what uncertainty.

Grandma illustrates this with a simple question. If a measuring instrument fails to register a proposed signal, must the source be absent? Perhaps. But perhaps the proposed coupling was wrong, the effect was compensated, or the instrument was not sensitive to that feature. Which possibility remains depends on the actual experimental design and auxiliary assumptions.

She refuses to infer the nonexistence of all possible preferred frames from the mere absence of a particular expected detection. What survives is the constraint that any acceptable account must explain the negative experimental result.

IV. Does absolute motion need to appear?

The induction effect can be described in terms of the relation between magnet and conductor. The cited ether experiments did not supply the expected direct evidence of Earth's motion through a light medium. Grandma therefore asks whether absolute motion is needed as an observable input to account for these particular results.

It is possible to respond to negative evidence by adjusting auxiliary hypotheses. A proposed contraction of apparatus dimensions, for example, can alter what an ether-drift experiment predicts. Such modifications may restore agreement with a measurement. Yet the possibility of a repair does not establish the repair as true; it shows that the observation is compatible with more than one explanatory architecture.

Here Grandma sees why economy becomes relevant, but also why it must not be placed first. We may compare theories for simplicity only after asking whether they account for the admitted phenomena without contradiction. Otherwise a beautifully simple but experimentally false theory would win merely because it uses fewer assumptions.

It is not obviously needed. This is an argument for considering descriptions that do without it. But a hidden preferred structure could still be postulated if it produced no contradictory observable consequences. Grandma refuses to mistake explanatory dispensability for logical impossibility.

V. The appeal of a common principle

We may now consider a proposal: perhaps the same physical laws apply in all inertial frames. Such a principle would provide a broad organizing explanation for the absence of the anticipated motion-dependent asymmetries.

The relativity principle offers a powerful organizing statement: the laws governing physical phenomena have the same form in inertial reference systems. But this is a generalization about a wide domain of phenomena, not a direct transcription of one induction experiment or one unsuccessful drift measurement. Grandma can regard it as a proposed principle with considerable explanatory reach without pretending it has been logically deduced from the two examples.

This distinction also respects Einstein’s intellectual work. A physical principle can be an act of theoretical synthesis: a disciplined choice that organizes many observations and generates testable consequences. It need not be the only sentence that can be obtained by reversing a small number of observations.

Grandma finds the proposal attractive because it connects distinct problems with one rule. But she asks whether the observations actually entail the rule for every physical process in every inertial frame. They do not. A finite collection of observations cannot by itself prove a universal generalization without further assumptions.

Thus the relativity principle is a candidate generalization supported by the pattern of evidence, not a theorem obtained from these two historical motivations alone. Grandma records the distinction rather than concealing it.

VI. What can be said about light?

The second motivation concerns light, so perhaps the speed of light is the missing invariant. Grandma examines the inference carefully. The unsuccessful ether-detection attempts constrain certain possible directional or motion-dependent effects. They do not, by themselves, establish that the one-way speed of light is the same in every inertial frame.

The second postulate introduces a further question. A measured round-trip light speed, combined with appropriate assumptions, places severe constraints on a theory. But a statement about the one-way speed between distant points already involves a way of coordinating clocks. Grandma therefore asks whether the admitted observations include that coordination or merely invite it.

If we simply write down the constancy of the one-way speed of light at this stage, we have introduced a principle. That may be scientifically fruitful and historically faithful to Einstein’s construction. It is not, however, a result forced by the two empirical starting points alone. The difference between proposing and deriving must remain visible.

A one-way speed comparison also raises the question of how distant clocks have been synchronized. This issue will become the subject of the fourth experiment. At present, Grandma can say only that any proposed account of light propagation must agree with the observations actually made.

She refuses to introduce Einstein's light postulate as though it were a direct numerical reading of the historical null experiments.

VII. Could more than one theory survive?

Yes. Historically, Lorentz-type ether accounts and Einstein's formulation illustrate that distinct interpretive frameworks can be constructed with overlapping empirical predictions in important domains. Their conceptual commitments differ even where measured consequences coincide.

Now imagine the available candidate explanations as a set. The induction result excludes those that fail to reproduce the corresponding currents. The ether-drift results exclude those that predict an observable drift signal of the prohibited kind. The surviving set is smaller, but its size is not known merely from stating the exclusions. Grandma does not equate a narrower set with a unique element.

A theory can survive a test because it makes the right prediction for a deep reason, because compensating effects cancel, or because the test does not probe its distinctive features. Subsequent experiments and conceptual analysis are required to distinguish these possibilities.

Grandma is not required to settle all questions about the relative merits of these theories here. She needs only to establish that the two starting observations do not uniquely identify one explanatory architecture.

This is the underdetermination problem in concrete form. An observation may eliminate some hypotheses while leaving several others admissible. Adding evidence can narrow the possibilities without reducing them to one.

VIII. What additional structure would decide?

Grandma now asks what remains unmeasured or unspecified. To derive a definite transformation between inertial systems, one needs further conditions about space, time, symmetry, and the behavior of physical signals. Some of these may be stated as postulates; others may be constrained by additional experiments.

The outer reference frame has now shifted from an individual apparatus to the general architecture of physical laws. That shift is legitimate if it is announced. It would be illegitimate if the broader frame were quietly treated as something the initial observations had already proved. Grandma marks the shift because the interpretation of an observation depends upon the questions we permit ourselves to ask of it.

We can now say what the empirical motivations make attractive: a unified description in which arbitrary distinctions between inertial observers do not proliferate. We cannot yet say that no other coherent architecture exists. The search for economy operates among explanations that survive the empirical and logical requirements.

The essential requirement is transparency. If a new condition is introduced, it should be identified as such. We must not smuggle it into a definition, assume it during a calculation, and later present it as though the original observations had forced it upon us.

This refusal is especially important when the conclusion is already known to the investigator.

IX. Can economy choose the theory?

A theory that dispenses with empirically unnecessary structures may be simpler than one that retains them. Grandma is willing to use economy as a preference, but only after checking empirical adequacy and logical consistency.

This is the place to separate three judgments. An explanation may be compatible with the evidence. It may be preferable to alternatives because of coherence and economy. Or it may be uniquely required by the evidence. These judgments are not synonyms. Grandma will not convert the first into the third by adding confidence to the prose.

The 1905 construction has enormous explanatory and predictive success, established by a much wider scientific history than these introductory examples alone. Our restricted experiment does not challenge that success. It examines the inferential distance between the introductory observations and the principles used to construct the theory.

She therefore writes the order of evaluation plainly: adequacy, non-contradiction, then economy. If two surviving explanations agree with all relevant evidence, simplicity may guide a choice of representation or research program. It cannot transform a preference into a demonstration of unique physical truth.

X. What did Einstein add?

Einstein's two postulates supply a definite theoretical framework from which the familiar transformations and consequences can be developed. The magnet–conductor comparison and the unsuccessful ether-detection attempts help explain why such a framework was compelling. They do not, alone, deductively establish both postulates.

We conclude that the two empirical starting points exert real pressure upon the available theories. They help explain why a new architecture was worth considering. They do not, by themselves, yield a unique proof of the two postulates, a complete transformation law, or a full account of space and time.

Grandma’s stopping point is therefore informative. To continue she must admit further physical principles, definitions, and tests. In the fourth experiment she will investigate one such definition directly, beginning not with a magnet or an interferometer, but with two clocks.

Grandma has now identified the precise role of an added principle: it reduces the freedom left by the observations. She has also identified the limit of the reversal. Starting from these two motivations, one may approach the problem of relativity without uniquely reconstructing its solution.

There remains another question, however. Even before comparing inertial systems, how can measurements made at different places be said to occur at the same time? For that we shall put away the magnet and the ether and give Grandma only two clocks.

Experiment 4 — Two clocks: what can “the same time” mean?

A narrated reconstruction of the reasoning, with the assumptions and stopping points made explicit. [1]

I. Two clocks and one apparently simple claim

Imagine two identical clocks, A and B, placed at different locations. Both operate normally in their local surroundings. Someone announces that they show the same time.

Let two clocks stand at separated places A and B. Each is a perfectly serviceable clock in the local sense: an observer beside it can read it when a nearby event occurs. We are asked whether they show the same time. Grandma notices that the question sounds simpler than the operation required to answer it. A comparison between distant readings is not itself a local event.

The clocks could display identical numerals while being incorrectly coordinated; they could display different numerals while using a known offset. The visual agreement of their faces is not a physical definition of simultaneity. We must specify how their indications are related to events at different locations.

Grandma does not dispute the clocks. She asks what the announcement means. If a person stands beside A and sees twelve o'clock, that is a local observation. If another person stands beside B and sees twelve o'clock, that is another local observation. Have we thereby shown that the readings occurred simultaneously?

No. We have shown only that two local events involved the same numerical indication. To compare the times of distant events, we need a relation between the clocks. Grandma's first refusal is to treat equal numbers as proof of simultaneous events.

II. The difference between a clock and a time coordinate

A clock supplies a succession of local readings. It can assign times to events occurring at its own location. But a coordinate time intended to describe events throughout an extended reference frame requires more than one clock or some equivalent procedure for relating spatially separated events.

We might transport one clock to the other and compare them. But transporting a clock introduces questions about its behavior during motion, which cannot be answered by appealing to the conclusion we seek. Alternatively, we might exchange signals. That option seems simpler, but it too conceals an assumption about propagation time.

Grandma is not claiming that clocks cannot be synchronized in practice. She is examining what an account of synchronization must presuppose. The difference between performing a successful procedure and explaining why it defines a common time is precisely what the reversal is intended to expose.

Grandma now changes her outer reference frame. The problem is no longer whether either clock works. The problem is whether their indications belong to a common system of time assignment.

This is why Einstein's discussion of simultaneity begins with measurement operations rather than with an abstract declaration that time is relative. Before we compare theories of time, we must say how time comparisons are performed.

III. The tempting signal

Grandma proposes sending a message from A to B. The observer at A records the departure when A reads twelve. The observer at B records its arrival when B reads twelve-oh-one. Surely the difference gives the signal's travel time?

Suppose A sends a signal to B. Clock A reads a departure time and clock B reads an arrival time. Their difference looks like the travel time. Yet the subtraction has a physical meaning only if the offset between the clocks is already specified. Otherwise the unknown offset is mixed into the unknown propagation delay.

This is the circularity. To determine the one-way travel time by using distant clocks, we need their synchronization; to synchronize the clocks by subtracting the one-way travel time, we need to know that time already. Grandma refuses to disguise the circularity by treating the two readings as though they belonged to a single clock.

Not yet. That subtraction is meaningful as a one-way duration only if the clocks have already been related by an agreed synchronization procedure. The proposed measurement silently presupposes the very relation we are trying to establish.

Grandma refuses the circular argument. She does not deny that the signal travelled. She denies that the two remote clock readings, without an established relation, uniquely measure its travel time.

IV. A journey that returns

She changes the experiment. A sends a signal to B, where it is reflected, and A receives it again. Now the same clock at A records departure and return. No prior synchronization with B is needed to measure the total elapsed time.

A different experiment avoids this immediate difficulty. Send a signal from A, reflect it at B, and receive it again at A. The departure and return readings belong to the same local clock. If they are t₁ and t₃, the round-trip duration is T = t₃ − t₁. This is a locally accessible measurement of the complete journey.

Notice what has changed. We have not measured the two directional travel times separately. We have measured their sum. A single number cannot generally determine two unknown numbers. Grandma keeps the result intact and refuses the unwarranted division into equal halves.

Suppose the clock reads t₁ at departure and t₃ at return. The measured round-trip time is T = t₃ − t₁. If the total is two microseconds, Grandma knows that the signal took two microseconds to go out and come back, including any relevant reflection delay that has not been separately accounted for.

But how much of the total belongs to the outward journey? She does not know from this measurement alone. One microsecond each way is one possible allocation, not a separate observation.

V. One measurement, two directional times

Let the outward duration be u and the return duration be v. The experiment supplies the relation u + v = T. It does not independently determine both u and v. Even perfect precision in measuring T cannot turn one relation into two independent measurements.

For a concrete example, suppose the total journey takes two microseconds. It may be convenient to assign one microsecond to the outward journey and one to the return. But the round-trip reading itself only establishes that the two assigned durations add to two microseconds. The equality is not read directly from the local clock.

This elementary arithmetic is not an argument that nature permits arbitrary observable results. The propagation of light remains constrained by physical law and experiment. The point is narrower: the assignment of a one-way coordinate travel time between distant points requires a convention or additional structure beyond the single-clock round-trip reading.

Grandma has found an informational boundary, not a defect in the instrument. Repeating the same kind of round-trip measurement may improve precision, but it does not by itself remove the need for an additional convention or independent information about the directional times.

She refuses to fill the missing information with a convenient number and then call that number measured.

VI. Einstein's synchronization rule

Einstein chooses a definite rule within an inertial frame: assign equal travel times to the outward and return light journeys between A and B. If a light signal leaves A at t₁, reflects at B, and returns to A at t₃, the reflection at B is assigned the time t₂ = (t₁ + t₃)/2.

Einstein’s §1 gives a clear rule. The clocks at A and B are synchronized when the light-travel time from A to B is taken equal to the time from B back to A, in the selected inertial frame. If A sends light at t₁, it reaches B at t₂, and returns to A at t₃, the defining relation is t₂ − t₁ = t₃ − t₂.

Equivalently, t₂ = (t₁ + t₃)/2. Grandma observes the elegance of the rule: it makes distant time comparison operational. She also asks what kind of statement it is. It is a synchronization definition within a theoretical framework, not an independent measurement of the two one-way propagation times.

This provides a way to coordinate the readings of distant clocks. The rule is part of the definition of the time coordinate in the chosen framework. It is not an independent measurement of the one-way propagation time that preceded synchronization.

Grandma accepts the rule as a coherent construction, while asking whether it is compatible with the other physical requirements of the theory. She refuses both extremes: that the chosen allocation was directly observed, and that because a convention is involved, all observations about light propagation become arbitrary.

VII. What remains physically constrained?

The round-trip duration is measured by a local clock. The locations of emission and reflection are part of the experimental arrangement. Repeated experiments, the behavior of rods and clocks, and relations among multiple paths constrain any physically acceptable description.

One must be especially careful here. Saying that a synchronization convention is involved does not make all measurements of light conventional. A round-trip speed can be measured with one clock and a known path length; observable coincidences and local elapsed times remain subject to empirical tests. A convention assigns coordinates in a way that must remain compatible with those measurements.

Grandma therefore refuses two symmetrical exaggerations: that a one-way speed has been measured without a synchronization assumption, and that introducing a convention makes the whole theory arbitrary. Neither follows from the clock experiment.

A synchronization convention assigns coordinate times to distant events; it does not authorize altering local observations or the measured total travel time. Grandma therefore distinguishes freedom in describing an experiment from freedom to change its result.

This is a useful example of Nature by Refusal: the admissible descriptions may vary, but they must all respect the observational constraints that survive.

VIII. Does this prove relativity of simultaneity?

Not yet. Grandma now asks whether two events judged simultaneous in one inertial frame must be simultaneous in another. To answer, we need rules relating the spatial and temporal coordinates of the two frames.

We may now compare two inertial systems, each equipped with clocks synchronized according to its own specified rule. The same pair of spatially separated events may receive different temporal coordinates in the two systems. But the assertion that this occurs according to Einstein’s precise relativity-of-simultaneity relation requires the transformation laws connecting those systems.

The clock procedure tells us how to define simultaneity within a frame. It does not alone derive the relation between frames. Grandma again finds the boundary between a necessary conceptual preparation and a quantitative physical conclusion.

The clock experiment has shown why distant simultaneity requires a definition and a procedure. It has not determined whether the correct transformations are Galilean or Lorentzian. That distinction depends on additional physical structure.

Grandma refuses to leap from 'synchronization is required' to 'relativity of simultaneity has been derived.' The first is a requirement of the comparison; the second is a particular consequence of the relativistic transformation laws.

IX. A second route to a familiar boundary

Compare this investigation with the magnet–conductor experiment. There we needed rules relating descriptions from different reference frames. Here we need rules relating readings from different places. Both inquiries reach the problem of how local observations can be assembled into a coherent account.

There is a resemblance between the clock problem and the magnet–conductor problem. In one case we compare electromagnetic descriptions assigned from different states of motion. In the other we compare time readings assigned at different locations. Both require explicit relations between descriptions before numerical differences can be interpreted as differences in physical reality.

The resemblance is methodological rather than a proof that one case entails the other. Grandma may carry forward the shared requirement of coherent comparison, but she cannot use that resemblance as a substitute for the missing physical postulates.

But Grandma will not claim that the same mathematical solution has been independently derived twice. The shared result is methodological: a local observation does not automatically specify the relation needed to compare it with another observation. The relation must be defined, derived from further premises, or constrained by additional measurements.

X. What the two clocks have taught us

We began with a sentence that seemed self-evident: the clocks show the same time. Working backward revealed its hidden requirements. Local clock readings alone do not establish distant simultaneity. A signal's round-trip time can be measured with one clock, but that measurement does not uniquely divide the journey into two directional times. A synchronization rule supplies the missing coordinate relation within a coherent physical framework.

We have discovered why the apparently innocent phrase “the same time” requires careful treatment. Local readings are accessible; distant simultaneity needs a coordinating rule. Round-trip propagation can be timed locally; its directional decomposition is not fixed by that timing alone. Einstein’s synchronization rule supplies a coherent definition, after which further theoretical assumptions permit relations between moving frames to be derived.

Grandma stops at the boundary. She does not deny the successful theory that continues beyond it. She identifies the point at which additional premises enter. This is the value of the reversal: to make the structure of an explanation visible without pretending that every part of it is already contained in the first observation.

Grandma has not derived special relativity from two clocks. She has reconstructed a problem that any theory of distant time comparison must address. She has also located the precise point at which a convention or additional information enters the reasoning.

That is where the fourth experiment ends. Grandma does not need to know the answer in advance. She needs to know when the available evidence no longer permits an answer.