Past exam of the philosophy of science supplementary subject of the University of Oxford / 2021 (A14139W1)

Source: Public Oxford paper bundle. The paper's cover establishes Hilary Term 2021; the attachment filename uses 2020.

Scientists should take sceptical challenges seriously when they reveal unsupported assumptions or distinguishable sources of error. They need not obtain a proof against every conceivable deception before conducting an experiment. This position concedes the force of some philosophical scepticism while rejecting certainty as the normal standard of scientific warrant.

There are several importantly different challenges. Global philosophical scepticism questions whether an external world exists at all. The problem of induction questions the justification for extending observed regularities to unobserved cases. Local criticism questions a particular instrument, sample or inference. A response appropriate to the last challenge need not answer the first. Repeating a measurement with a second instrument can expose a faulty sensor; it does not prove that all experience is independent of a deceiving agency.

David Hume's challenge to inductive reasoning is especially relevant. A finite record does not deductively entail a universal generalization. Nor can the inference be justified simply by saying that past uses of inductive reasoning succeeded: that argument itself extends a past regularity beyond the observed cases. The possibility that an observed regularity fails tomorrow is not a contradiction. This identifies a real gap between deduction and empirical projection.^[ref]

For example, suppose a metal's measured resistance agrees with $R(T)=R_0[1+\alpha (T-T_0)]$ throughout a tested temperature interval. Both that expression and a rival with an extra term zero throughout the interval fit those measurements exactly, yet can disagree outside it. Additional measurements can discriminate specified rivals, but no finite collection excludes every possible future deviation. The scientific lesson is to state the tested domain, physical assumptions and uncertainty. A linear electrical resistance model can be useful without being a universal law; a phase transition is an independently intelligible reason why extrapolation might fail.

One might reply that successful prediction and scientific replication justify confidence. They certainly distinguish a tested model from an arbitrary guess under the ordinary standards of inquiry. They do not supply the noncircular proof demanded by radical philosophical scepticism. A Bayesian inference similarly makes the dependence on prior probabilities and likelihoods explicit; it does not deduce the correctness of those commitments from no premises. Acknowledging this limit is preferable to claiming that a mathematical updating rule dissolves the problem of induction.

The sceptic nevertheless owes us an account of the practical consequence. If two global possibilities give exactly the same observable consequences under every available intervention, their difference does not identify which laboratory procedure to change. A scientist can suspend judgement about that difference while comparing models that make different predictions. This is a defence of proceeding under fallible standards, rather than a refutation of global philosophical scepticism.

By contrast, local criticism can reorganize an entire research programme. If an apparent signal might be instrument drift, useful responses include independent calibration, reversing a controllable input, measuring an appropriate control and conducting scientific replication. Those procedures make the proposed error account answerable to evidence. Merely imagining that every control could also be deceptive adds no further discriminating test.

The defensible policy is therefore fallibilism: confidence proportionate to the testing and domain, openness to identifiable error, and no claim to absolute foundations. Sceptical arguments matter most when they improve the questions asked of a model. Their inability to be conclusively defeated does not make every model equally warranted or make scientific work irrational.

A distinction between physical constitution and perceptual response is plausible and scientifically useful. A division into wholly objective primary qualities and wholly unreal secondary qualities is much harder to defend. These are different interpretations of the primary and secondary qualities distinction.

John Locke attributes extension, solidity, shape, number and motion to bodies as primary qualities. Secondary qualities are powers to produce sensations through the constitution of bodies: a body's power to produce a colour experience is distinct from the experience itself. This distinction need not make the power unreal. Nor should the further resemblance thesis, that ideas of primary qualities resemble their causes, be confused with the more modest claim that bodily constitution explains experience.^[ref]

A modern model of colour perception shows what can be retained. For a nonfluorescent surface in fixed viewing geometry, a simplified cone cell response isHere $L$ is the incident spectral irradiance, $R$ the spectral reflectance, and $S_i$ an effective sensitivity for one of three cone cell classes. Neural processing and adaptation also matter, so these responses are not a complete identification of perceived colour. Nevertheless, the model separates the surface, illumination and observer rather than assigning every experienced variation to a changing surface.

It follows that one surface can produce different responses under different illumination. Conversely, distinct spectral reflectance functions can produce the same three responses for a specified illuminant and observer, a case of metamerism. Three weighted integrals do not in general uniquely determine a whole function. Thus identifying experienced colour with one intrinsic surface property loses both the dependence on the observer and the many-to-one character of the response.

The scientific benefits are concrete. A comparison of paints should specify illumination and observer conditions. A measurement intended to characterize a material can instead estimate its spectral reflectance, preserving information that a single colour match discards. In both cases the distinction tells an investigator what to hold fixed and what to vary. It organizes experiments and explains disagreement without requiring that one observer be hallucinating.

However, perceptual variability alone does not distinguish primary qualities from secondary qualities. George Berkeley points out that apparent extension and shape also change with viewpoint and perceiver.^[ref] A circular disc yields different projected shapes at different viewing angles. The stable object is inferred through a model of that variation; it is not simply copied into experience. Primary qualities therefore have no automatic exemption from the interpretative work involved in perception.

Nor is relationality equivalent to subjectivity. In special relativity, measured length depends on an inertial frame, but relations between such measurements are tightly constrained. Conversely, a disposition to affect a specified perceiver under specified conditions can be an objective physical fact. Classifying a property as relational does not settle whether it is real, measurable or explanatory.

The best distinction is consequently functional and revisable: distinguish the experience to be explained from physical properties and relations used to explain it. That retains the scientific benefits of separating constitution, environment and response. It does not establish an immutable list of primary qualities, a simple resemblance relation between experience and object, or a proof of scientific realism.

Old evidence challenges a simple account of confirmation as actual conditional updating, but does not refute every probabilistic treatment of the raven paradox. The strongest response distinguishes support from newly acquired information and makes the sampling procedure explicit.

Let $H$ say that every raven is black. Its contrapositive says that every nonblack thing is a nonraven. Since these statements are logically equivalent, evidence should support them equally. Combining this equivalence requirement with the unrestricted rule that every positive instance confirms a universal generalization appears to make discovering a white shoe support $H$. The tension is between that result and the judgement that such a discovery is irrelevant.^[ref]

Bayesian confirmation replaces an automatic instance rule with a comparison of likelihoods. For background $K$, positive prior probability for both $H$ and its alternative, and evidence $E$ of positive probability,The equivalence follows directly from Bayes' theorem. A statement's grammatical appearance as a positive instance does not determine either likelihood.

Consider an intentionally simple finite model. There are $r$ ravens and $M$ nonblack objects, with these counts fixed under both hypotheses. Under $H$ none of the nonblack objects is a raven; under the sole alternative $H_1$, exactly one is. Selecting uniformly among the nonblack objects and discovering a nonraven has likelihoods $1$ and $1-1/M$. Selecting uniformly among the ravens and discovering a black one has likelihoods $1$ and $1-1/r$. With equal prior probabilities, the respective posterior probabilities of $H$ areIf $M\gg r$, the first discovery supplies far less support. For $M=10^6$ and $r=100$, the increases over $1/2$ are approximately $2.5\times 10^{-7}$ and $0.00251$. This illustrates a possible resolution of the intuitive disparity, not a theorem that every nonblack nonraven always confirms $H$. More complex alternatives can reverse the comparison, and examining something already known to be a shoe need supply no relevant new information.^[ref]

Now suppose $E$ is already in $K$. Then $P(E\mid K)=1$, and ordinary conditioning givesThere is no further increase from learning the same fact again. Yet an already established observation can become an important reason to accept a subsequently developed theory. Clark Glymour's problem of old evidence exposes this mismatch between actual updating and scientific assessments of evidential support.^[ref]

One response evaluates the observation relative to an earlier or counterfactual background that does not contain it. That can represent how strongly a theory accounts for the observation compared with alternatives. It must specify that background rather than pretend to remove knowledge uniquely from a person's present beliefs. Another response concerns learning that a theory successfully accounts for the observation. That discovery can be new even when the observation is old. Modelling it requires attention to limited reasoning and changing hypotheses; an ideal agent already knowing every logical consequence cannot simply learn an entailment it previously overlooked.

Neither manoeuvre licenses counting the same evidence twice, or ignoring that parameters were selected to fit it. Neither establishes a unique complete theory of confirmation. Nevertheless, the old-evidence objection concerns the interpretation and completeness of a probabilistic framework; it does not invalidate the likelihood calculation for genuinely new sampling information.

The preferable solution combines a background-sensitive account of support with rejection of unrestricted instance confirmation. It preserves logical equivalence while allowing a white shoe to provide tiny support, no support or contrary support under different justified models. This is more explanatory than assigning confirmation solely from logical form, although assessing old evidence requires additional work beyond routine conditioning.

Explicit reference to laws is not necessary for every good scientific explanation. Laws remain indispensable in many explanations, but their occurrence alone is not sufficient either. The issue turns partly on whether a law means a contingent regularity of nature, any supporting generalization, or even a mathematical necessity.

The deductive-nomological model offers a strong positive case. Laws and circumstances entail the phenomenon, and at least one law is essential to the derivation. For example, Newton's second law and an inverse-square gravitational force, with appropriate initial conditions, yield predictions for an idealized two-body orbit. A universal relation unifies cases that otherwise look like unrelated observations. Carl Hempel and Paul Oppenheim explicitly develop adequacy conditions for such deductive explanations; their 1948 paper should not be read as a finished treatment of every statistical explanation.^[ref]

However, derivability does not always give the right explanatory direction. In a simplified geometrical optics model, a vertical pole of height $h$ on level ground casts a shadow of length $\ell =h/\tan \theta$ at solar elevation $\theta$. Given $h$ and $\theta$, one derives $\ell$; given $\ell$ and $\theta$, one derives $h$. The same geometrical relations enter both derivations. Yet the shadow does not normally explain why the pole was built to that height. The reverse calculation is a way to measure its height. The difference concerns which features produce or constrain which others, rather than deductive validity alone.

Mechanistic explanation is another challenge to a compulsory law-based format. A clock stopped because debris blocked a particular wheel, thereby preventing the mechanism from advancing. Inspecting the obstruction, removing it and observing restored operation can answer the causal question with considerable precision. The explanation describes organization and a relevant difference-making event without stating a universal law. It is more informative about this failure than the observation that every component obeyed Newtonian mechanics.

A defender of the covering-law model can reply that the causal account relies implicitly on regularities: debris that offers no resistance would not stop the wheel. This is a fair warning against treating a bare chronology as an explanation. But it does not establish that explanatory understanding requires identifying exceptionless fundamental laws, or that replacing the account with a complete microscopic derivation would improve the answer. Stable, tested relations in the actual operating regime may be sufficient.

An explanation can also turn on a mathematical constraint. Suppose an engineer asks why a connected network with four odd-degree junctions cannot be inspected by a route using every segment exactly once. Represent junctions as vertices and segments as edges of an undirected graph. Every intermediate visit pairs an incoming and an outgoing edge, so only the two endpoints of an open trail can have odd vertex degree. Four such vertices therefore exclude an Eulerian trail. The impossibility persists if segment lengths, construction materials or walking speeds change while the connections remain fixed. This is the obstruction in the Seven Bridges of Königsberg.^[ref]

The physical application requires that the graph represent the allowed routes correctly. Once that is granted, the explanatory necessity comes from parity, rather than a contingent physical law. Calling every mathematical constraint a law would preserve the original claim only by greatly widening its meaning. A critic may restrict “scientific” to causal explanations, but that restriction needs justification when applied mathematics answers a concrete engineering question.

Thus laws are powerful explanatory resources, and good explanations normally support relevant what-if judgements. Those judgements can be grounded in laws, mechanisms or mathematical constraints. Requiring explicit reference to laws imposes an unnecessarily narrow format; requiring some disciplined account of dependence is much more defensible.

Change is compatible with natural kinds, provided there is a defensible basis for tracking membership and explaining the changing properties. Requiring every property to remain unchanged would exclude nearly all physical objects. The important distinction is between variation within a kind, a transition out of it, and changes in the population that constitutes it.

Chemical elements offer a relatively sharp case. A carbon atomic nucleus has atomic number six, meaning six protons. Its isotope depends also on its neutron number; the ionization state of a carbon chemical atom can change without changing its chemical element. Temperature, position and velocity can also change. Classifying these objects together is useful because the relevant common constitution supports further inference, not because their entire property lists are identical at every moment.

A nuclear reaction that changes the proton number is different: it can produce a nucleus of another chemical element. This does not refute the original classification. It identifies a process under which membership changes. A theory of natural kinds should permit transformations between kinds rather than require its members to exist forever. The ordinary criterion also has a domain: an arbitrary mixture containing carbon is not itself a single carbon chemical atom.

Water makes a complementary point. A sample can undergo melting or boiling, changing density, rigidity and transport properties while retaining its ordinary chemical identification. The claim must be qualified: a bulk sample contains interacting molecules, and its detailed composition can include dissociation products. It need not consist at every instant exclusively of isolated identical water molecules. Nevertheless, changes of thermodynamic phase are explained through the constitution and organization of the material, rather than by treating every different appearance as a new unrelated substance.

These examples might suggest that some fixed microscopic essence is always required. Species put pressure on that inference. Individuals develop and differ; the distribution of inherited traits can change across generations through natural selection, genetic drift and gene flow. A demand that every member share an unchanging diagnostic appearance would misrepresent this history. It would also make a change in the frequency of one trait automatically create a new species, which is not how the relevant evolutionary processes work.

Homeostatic property cluster theory provides one alternative. Traits can tend to occur together because common development, shared ancestry and interactions within a population sustain their association. The traits need not be universal, and the maintaining processes need not freeze their frequencies. A population can change while retaining enough historical and causal continuity to support useful inference.^[ref] This is not a claim that every species concept gives the same boundaries. Reproductive relations, descent and ecological processes can motivate different criteria for different investigations.

An objection remains: if both members and their traits change, what prevents a natural kind from becoming a mere convenient label? The answer cannot be resemblance alone or an arbitrary chain of small changes. The proposed grouping must track processes that support reliable explanations and suitably qualified predictions. If the processes maintaining a cluster break down, or a lineage divides into independently evolving populations, continuing to use one label may cease to serve the relevant inquiry. Membership can be uncertain near such transitions without making every central case arbitrary.

There is also no requirement that all natural kinds have the same basis of unity. The atomic number criterion and a historically sustained population answer different scientific questions. Imposing the former structure on every biological grouping would mistake one successful form of classification for a universal metaphysical rule.

Changing properties therefore do not by themselves defeat naturalness. The scientific work lies in identifying which variations a kind tolerates, which transformations end membership, and why its remaining unity supports inference. A classification earns its status through that explanatory achievement, not through the fiction that nature is motionless.

A plausible demarcation assesses how a practice makes and tests empirical claims, while allowing unsettled boundary cases. An exact sentence-level test is less defensible than an account of sustained accountability to evidence. Pseudoscience should also be distinguished from honest scientific error and from activities that make no scientific pretension.

Karl Popper's falsifiability criterion identifies something central: an empirical theory excludes possible observations. A proposed relation between a controllable input and an instrument reading is testable only if some readings would conflict with it under the stated conditions. Collecting agreeable anecdotes is weak when an account can be made compatible with every outcome. Risky predictions matter because passing a test should have been difficult for a mistaken account.^[ref]

Yet falsifiability is not sufficient for good science. Someone could guess tomorrow's detector count and then advertise the guess as an established result. The numerical claim is falsifiable, but that does not supply calibration, a justified sampling model or a reliable research practice. Conversely, an investigation can be scientific even if the working model later proves false. The demarcation problem concerns the character of inquiry, not a guarantee of truth.

Nor does a failed prediction always justify immediate rejection of the principal theory. In a test represented by $T\wedge A\Rightarrow E$, the observation $\neg E$ yields $\neg T\vee \neg A$. The auxiliary hypotheses $A$ might include an initial condition, an instrument model or the composition of a sample. This Duhem problem is a logical limitation on what one unsuccessful test establishes.^[ref]

There is nevertheless an evidential difference between a justified repair and an evasive one. If a suspected detector defect is measured independently, and correcting it yields further successful predictions, the change increases the theory's exposure to testing. If a new invisible influence is introduced after each failure with no independent consequences, the repair preserves agreement at the expense of accountability. An auxiliary hypothesis is not disreputable merely because it is auxiliary; its evidential role matters.

Imre Lakatos therefore shifts attention to sequences of theories and their capacity to generate corroborated new results. A programme that repeatedly anticipates discoveries differs from one that continually explains away known failures.^[ref] This improves on judging an isolated claim, but it does not provide a timetable stating exactly how long an initially unsuccessful programme deserves support. New research may need time, and a previously fruitful programme can stagnate.

A workable standard consequently asks whether investigators specify discriminating predictions or comparisons, expose their methods and data to appropriate scrutiny, respond to contrary findings, and seek independent checks. These demands should fit the field. An astronomical claim can be tested by predicted observations and independent instruments even when the investigator cannot manipulate a star. A laboratory science can often use controlled interventions. Requiring all inquiry to copy one laboratory format would exclude legitimate evidence.

An objection is that these standards describe good scientific conduct rather than draw an exact boundary. That is partly right. But a usable demarcation need not be a mechanical classifier for every new idea. Persistent evasion of the standards while claiming scientific authority supports a judgement of pseudoscience; mixed or immature cases can remain unresolved. A single statistical mistake need not condemn an entire field, while a consistent refusal to acknowledge tests can.

The resulting distinction is plausible because it explains what the boundary is for: identifying practices that can correct error through public evidence. It permits criticism of established institutions as well as marginal claimants. Its price is contextual judgement, which is preferable to pretending that one syntactic feature settles every case.

Falsificationism is more plausible than a requirement of conclusive verification, but a fallible theory of testing should retain positive evidential support as well as criticism. The comparison changes if verificationism is weakened to mean empirical confirmability rather than proof.

First separate a methodological thesis from a thesis about meaning. Methodological verificationism says that scientific theories should be established by verifying observations. A stronger semantic version treats verifiability as a criterion of meaningful factual assertion. These positions are not identical. A theory might be meaningful and empirically significant even though no finite test can establish its unrestricted truth.

The central asymmetry is logical. For a universal claim $\forall xF(x)$, observing $F(a_1),\dots ,F(a_n)$ leaves unexamined cases open. A genuine counterinstance $\neg F(b)$ contradicts that universal claim. A finite sequence of favourable tests of an unrestricted physical law therefore differs from a conclusive deduction of it. This gives falsificationism a substantial advantage over strong verificationism: it makes scientific risk possible without demanding an unattainable final proof.

Karl Popper's methodological proposal is not to invent theories that are false, but to formulate conjectures with empirical content and subject them to severe tests. Surviving such tests amounts to corroboration, not conclusive verification.^[ref] A theory that prohibits nothing cannot gain comparable credit simply by retelling whatever happens.

However, the logical asymmetry does not make scientific decisions automatic. The alleged counterinstance depends on identifying the object, measuring the relevant quantity and applying the theory in an appropriate domain. In $T\wedge A\Rightarrow E$, failure of $E$ refutes the conjunction, not uniquely $T$. The Duhem problem therefore defeats a naive rule that every discrepant reading immediately falsifies the central theory. Independent tests of the auxiliary hypotheses can make rejection well justified, but that is additional evidential work.

Statistical models create another difficulty. For $n$ mutually independent trials, each governed by a Bernoulli distribution with success probability $0<p<1$, a specified sequence with $k$ successes has probability $p^k(1-p{)}^{n-k}>0$. An unusual sequence is therefore not a logical counterexample to this model. An appropriate statistical hypothesis test instead asks how surprising a specified discrepancy would be under the model, allowing controlled but nonzero error. The decision depends on the test protocol, sample size and alternatives. Recasting such testing as literal contradiction would erase the distinction between impossible and improbable outcomes.

The universal-claim argument also has a converse. An existential claim can sometimes be established by one well-authenticated instance, although no finite failed search proves that there are no instances anywhere. Therefore the logical asymmetry concerns forms of claims; it is not a proof that every scientific statement must have exactly the same testing structure.

A defender of weak verificationism can accept all this and require only that observations can increase support. Bayesian confirmation, for example, compares a theory's likelihood with those of alternatives. That avoids treating favourable evidence as a proof. It does inherit questions about prior probabilities and the problem of induction, but criticism does not eliminate those questions either. Scientists still need reasons to use one surviving theory rather than another when planning further work.

Thus falsificationism wins against conclusive verificationism by recognizing the openness of empirical theory and the importance of possible failure. It does not win a monopoly on rational scientific reasoning. A stronger account combines demanding tests, scrutiny of auxiliary hypotheses, appropriate statistical treatment and defeasible positive support. The main achievement is disciplined error correction, rather than replacing one promise of certainty with another.

It is implausible to deny every case of revolutionary theory change, but reasonable to reject the claim that such changes are instantaneous, complete or wholly disconnected from earlier science. Much turns on the strength of the proposed notion of revolution.

Thomas Kuhn challenges a picture of scientific development as the steady addition of true propositions. His contrast between normal science and revolutionary episodes concerns changes in accepted problems, methods and exemplary achievements as well as changes in propositions. A community can revise the framework within which it decides what needs explaining.^[ref] This is a historical proposal, not the claim that every important discovery is a revolution.

The transition from Newtonian mechanics to special relativity supplies a clear physical example of a deep conceptual change. Consider two spatially separated events, and let their differences in one inertial frame be $\Delta t$ and $\Delta x$. A Lorentz transformation givesFor $\Delta t=0$, $\Delta x\ne 0$ and $v\ne 0$, they are not simultaneous in the second inertial frame. By contrast, the Galilean transformation takes $t^{\prime }=t$, so simultaneity is invariant. The change concerns how time coordinates assigned by separated clocks enter physical measurement, not merely one corrected numerical constant. Albert Einstein begins by defining simultaneity through a clock procedure and then derives the transformation from his postulates.^[ref]

One reason to deny revolutions is precisely that the new account retains old achievements. For $|v|/c\ll 1$, and within regimes where relativistic corrections to the relevant measurements are negligible, Newtonian mechanics remains useful. Experimental records, instruments, mathematical methods and many successful predictions cross the transition. That continuity is real. It does not show that the physical interpretation of simultaneity remained unchanged. Numerical approximation in a restricted domain is weaker than identity of conceptual commitments.

A second objection is that major changes emerge gradually. Earlier theories and experiments prepare the transition; different researchers adopt it at different times; its consequences take decades to develop. This undermines a picture of one decisive moment transforming all science. It does not undermine a notion of revolutionary theory change as an extended reorganization of a field. Political revolutions are themselves extended events; the metaphor need not require an instantaneous rupture.

A third objection concerns rational comparison. If a revolution changes standards and meanings, perhaps rival theories cannot be compared at all. But this extreme conclusion is unnecessary. Rival accounts can share instrument records, limited predictions and mathematical limiting relations while differing about their interpretation. Explaining why those records agree in one domain and discriminate theories in another is itself a rational comparison. Conceptual change can make translation partial and difficult without making all evidence inaccessible across the divide.

These qualifications also constrain the historical claim. A local change in one field need not create a universal new worldview. Nor should every anomaly be described as a crisis: normal science routinely works on discrepancies without replacing its framework. Whether an episode is revolutionary should depend on evidence of the extent and character of reorganization, not on a dramatic label attached afterwards.

The defensible conclusion is therefore selective. Denying total, unprepared and irrational breaks may be historically plausible. Denying substantial transformations of fundamental concepts and research standards is much less so. The example of simultaneity shows how continuity of successful calculation can coexist with a change deep enough to merit the name revolution.

Theory change defeats an inference from scientific success to final, unrestricted truth. It does not show that successive theories make no progress in accuracy, explanation or warranted knowledge. Calling two theories false says too little about how their claims compare.

The strongest pessimistic argument is historical: theories once regarded as successful and well supported have later lost central commitments. If successful predecessors could misidentify what exists, the success of a present theory cannot by itself guarantee its full ontology. Larry Laudan uses the historical record to challenge a simple connection between success, reference and approximate truth.^[ref] The challenge is serious for scientific realism, particularly when it treats an entire accepted framework as a single package of warranted beliefs.

But the slogan conflates several claims: that an unrestricted theory contains a false assertion, that every one of its assertions is false, and that its successes were accidental. The first does not entail either of the others. A theory can be wrong outside its proper domain while accurately representing relationships within it. Establishing that domain is itself a scientific achievement.

The velocity-addition formula makes the difference precise. For collinear velocities, special relativity giveswhereas the Galilean transformation yields $u_{\mathrm{G}}=u+v$. Their difference isFor $|uv|/c^2\le \epsilon <1$,This is a controlled agreement, including the case $u+v=0$, where a relative-error formula would be undefined. At $u=v=0.6c$, the classical value $1.2c$ differs dramatically from $15c/17\simeq 0.882c$. At small velocities the same comparison explains why ordinary classical calculations work so well. The relativistic rule and its low-speed relation to the classical rule appear in Albert Einstein's treatment of velocity composition.^[ref]

The example does not prove that special relativity is the final account of nature. It shows that a successor can explain an earlier theory's restricted success and predict when it fails. One obtains a stronger account of the evidence than either blanket rejection of the older theory or unqualified belief in it. The surviving mathematical relation also need not preserve the older interpretation of absolute time.

Structural realism develops this possibility by locating continuity in interpreted relations rather than every proposed entity. John Worrall uses continuity between Augustin Fresnel's optics and later electromagnetism to motivate such a response.^[ref] Its attraction is discrimination: some tested structure can survive an ontological change. Its difficulty is specifying which structure deserves belief before the next change, rather than selecting the survivors afterwards. An equation without a physical interpretation and domain is too weak a candidate.

An empiricist can accept the improvement in prediction and systematic representation while withholding belief about unobservables. That shows that progress does not logically require scientific realism. The realist can reply that independent tests, novel successful predictions and quantitatively explained limiting relations support cautious belief beyond the immediately observed. Neither side should treat historical examples as a deductive proof of its complete position.

The original claim is therefore defensible only in a thin sense: successive unrestricted theories may each contain falsehoods. Used to deny progress, it is misleading. Theories can become more accurate, discriminate more possibilities and explain why earlier approximations succeeded. Theory change should make scientific confidence selective and revisable, rather than erase the distinction between a useful approximation and an arbitrary mistake.

The Bell theorem connects the philosophical problem of scientific underdetermination with a precise issue in quantum mechanics: whether local underlying causes can reproduce observed correlations. It shows how a dispute about interpretation can acquire empirical content, while also showing why an experiment does not select a unique ontology.

A shared cause can produce strong correlations without a signal between distant measurements. The philosophical question is whether this familiar explanatory strategy can reproduce all predictions for quantum entanglement. In a two-station Bell test, let $x,y\in \{0,1\}$ be local settings and $A,B\in \{-1,1\}$ the outcomes. A local hidden-variable theory satisfying Bell local causality takesMeasurement independence additionally requires $\rho (\lambda \mid x,y)=\rho (\lambda )$. The same ensemble of underlying states must therefore apply across the four setting pairs.

The resulting CHSH inequality can be derived without assuming quantum theory. First take a deterministic local hidden-variable model with responses $A_0,A_1,B_0,B_1=\pm 1$. For each $\lambda$,because exactly one bracket vanishes. Averaging against one normalized probability distribution givesStochastic local response functions give no escape: they can be represented by adding independent local random seeds to $\lambda$ and averaging deterministic assignments. The essential restriction is the common setting-independent ensemble and local dependence of responses, not a separate insistence that all nature be deterministic.

For two spin-one-half systems in a spin singlet state, ideal quantum measurements along unit directions $\mathbf{a},\mathbf{b}$ give $E(\mathbf{a},\mathbf{b})=-\mathbf{a}\cdot \mathbf{b}$. ChooseThe four correlations are $-1/\sqrt{2},-1/\sqrt{2},-1/\sqrt{2},+1/\sqrt{2}$, so $S=-2\sqrt{2}$. Thus quantum mechanics predicts an observable incompatibility with the specified local hidden-variable theory class. This is stronger than saying that two interpretations happen to fit the observations collected so far.

Actual Bell tests require more than the ideal calculation. Setting choices and distant readouts must have the relevant spacelike separation; discarding detected pairs selectively can create a detection loophole. The 2015 Delft electron-spin Bell test combined efficient readout, rapid settings and separated stations to address these issues together. The reported statistic was $S=2.42\pm 0.20$, with a test of the local-model hypothesis accounting for device memory.^[ref] This is statistical evidence under a specified protocol, rather than a logically infallible observation.

The philosophical conclusion must be equally precise. A violation challenges the conjunction of Bell local causality, measurement independence and the relevant experimental assumptions. It does not disprove every possible hidden-variable account, or establish that the observer's mind creates the outcome. Different theories can reject different premises while agreeing on the observable correlations. That remaining scientific underdetermination does not cancel the genuine exclusion of a well-defined class.

Nor does violating Bell local causality entail controllable faster-than-light communication. In the ideal spin singlet state, each local marginal is $P(A=+1\mid x,y)=1/2$, independently of the distant setting. This quantum no-signalling condition is weaker than the hidden-state factorization used in the proof. Conflating them turns a precise result into an unwarranted conclusion about signalling.

The example therefore illustrates both the power and the limit of empirical testing in philosophy of science. Mathematical clarification can transform an apparently metaphysical disagreement into rival observational constraints. Evidence can discriminate those constraints without determining every question about the reality represented by a successful theory.

Constructive empiricism offers a coherent account of accepting science without believing its full claims about unobservables. Its hardest task is to justify why observability should delimit the belief required by scientific acceptance. Its modesty is a strength, but not an automatic argument against every form of scientific realism.

For Bas van Fraassen, science aims at empirical adequacy. Accepting a theory involves belief in that adequacy and practical commitments to use, test and develop it. Empirical adequacy means getting the observable phenomena right, including future and unobserved phenomena, rather than merely fitting the data already collected. Acceptance need not include belief that every theoretical entity and process exists as described.^[ref]

The position nevertheless interprets theoretical claims literally. A claim about an electron is genuinely about an electron, not shorthand for a list of instrument readings. A constructive empiricist can suspend belief in the claim while believing the theory's empirical adequacy. This separates constructive empiricism from a semantic proposal declaring theoretical language meaningless, and from the claim that theories are simply calculation devices without truth conditions.

The observational distinction concerns what beings like us could observe in suitable circumstances, not just what has happened to be observed. A remote macroscopic object can be observable even if nobody has seen it. Conversely, observing marks in a detector does not by itself amount to observing whatever microscopic entity is invoked to account for them. Van Fraassen emphasizes that instruments produce new observable phenomena for theories to represent.^[ref]

This permits serious experimental work. In a Bell test, a constructive empiricist can accept predictions for settings, detector records and correlations while declining a full ontological interpretation of the quantum state. The theory remains exposed to failures in those observable predictions. Suspending belief about its underlying picture does not remove the obligation to test it or make every theory acceptable.

An attraction is economy of belief. The history of successful theories whose proposed mechanisms were abandoned cautions against inferring an entire ontology from empirical success. Moreover, agreement among observable predictions need not select one interpretation of the unobservable. Constructive empiricism makes room for this scientific underdetermination without denying that scientists can know much about observable reality.

The realist objects that the restriction is arbitrary. Ordinary perception itself depends on physical interactions and background assumptions; well-calibrated detection through several independent methods can be highly reliable. If both unaided perception and instrument-based inference are fallible, why should one warrant belief and the other only acceptance? A constructive empiricist can answer that scientific acceptance does not require belief in the additional explanation. That establishes an available epistemic stance, but it does not yet establish that belief in unobservables is unreasonable.

Vagueness and theory-ladenness sharpen this objection without producing an immediate contradiction. There may be no exact boundary between a barely visible object and an invisible one. Clear cases can remain despite uncertain margins. Scientific theories also inform estimates of human observational capacities. One can use their tested claims about what people discriminate without accepting every microscopic detail of their explanation. Still, this makes the observational boundary an empirically informed, revisable distinction, rather than an entirely theory-free foundation.

There is a further limitation: belief in empirical adequacy itself goes beyond the available observations. It includes what will happen in unperformed experiments. Constructive empiricism therefore does not abolish the problem of induction. Its claim is that commitment to observable regularities can be rational without the additional commitment to the complete truth of the explanation proposed for them. A critic must challenge that asymmetry, rather than merely point out that both involve inference.

The position consequently succeeds as a disciplined and nontrivial alternative to comprehensive scientific realism. It allows literal theories, ambitious predictions and experimental correction while separating practical acceptance from unrestricted belief. Its strongest defensible conclusion is that the scientific record need not compel belief beyond empirical adequacy; establishing that observability supplies the uniquely correct boundary of rational belief would require a further argument.