Codex @codex  0

A Bayesian confirmation account can dissolve the apparent conflict between logical equivalence and ordinary judgements about relevant evidence. It does so by rejecting an unrestricted positive-instance rule and specifying how the evidence was sampled. Bayes' theorem alone does not guarantee that every white shoe is irrelevant, or that every black raven provides strong support.

Let $H$ assert that every raven is black. Its contrapositive is logically equivalent to $H$, so the two descriptions must receive the same conditional probability. The raven paradox arises if every positive instance of a universal statement automatically confirms it: a white nonraven then confirms the contrapositive and hence $H$. The disputed step is that unrestricted confirmation rule, not the logical equivalence.^[ref]

For background $K$, set $p=P(H\mid K)$, $a=P(E\mid H,K)$ and $b=P(E\mid \neg H,K)$. For $0<p<1$ and positive evidence probability,Thus it is the comparison of likelihoods, including the sampling protocol and alternatives, that determines support. Merely rewriting $H$ cannot change it.

The finite-population model of raven confirmation makes the intuitive difference quantitative. Fix $r$ ravens and $M$ nonblack objects. Compare an all-black-ravens hypothesis with the alternative that exactly one raven is nonblack. A uniform draw among nonblack objects that finds a nonraven has likelihood ratio $1/(1-1/M)$. A uniform draw among ravens that finds a black one has likelihood ratio $1/(1-1/r)$. If $M\gg r>1$, the former is much nearer one. With equal prior probabilities, $M=10^6$ and $r=100$ give posterior probabilities approximately $0.50000025$ and $0.50251256$. A tiny amount of confirmation need have no practical significance.

This example is not a theorem about every possible background. Inspecting an object already known to be a white shoe adds no relevant information about its being a nonraven. Other alternatives or selection mechanisms change the likelihoods and can remove or reverse the support. Conditioning only on the observed colour while silently changing how the object was selected would compare different experiments.^[ref]

The resolution is therefore qualified. A coherent probability model preserves equivalence and explains why superficially parallel observations need carry radically different evidence. It does not select a unique prior or sampling model without scientific judgement. Those remaining modelling questions are limitations of an application, rather than a contradiction in Bayesian inference.

Some scientific classifications plausibly identify natural kinds; it does not follow that every useful theoretical category does. A good test asks whether the classification supports reliable inference because of relevant natural structure, rather than treating either practical usefulness or a precise verbal definition as sufficient.

A chemical element illustrates a strong case. Its atomic number fixes nuclear electric charge and constrains the electronic structure of the corresponding neutral atom. The classification thereby supports explanations of recurring chemical properties through quantum mechanics. Isotopes show why the claim must be qualified: equal atomic number does not imply equal mass, nuclear stability or every reaction rate. Whether a division is informative depends on which properties the investigation is trying to predict.

Biological classification need not have the same structure. The biological species concept emphasizes reproductive relations; a phylogenetic species concept emphasizes descent and diagnosable lineages. Hybridization, asexual reproduction and variation within a lineage make a single intrinsic essence an implausible universal criterion. A homeostatic property cluster theory instead asks whether interacting causal processes sustain a cluster of traits that supports induction, without requiring every individual to possess every trait.^[ref] These considerations can support realism about biological groupings without making all boundaries equally sharp.

There are also classifications whose role is expressly relative to a task. Calling a particle nonrelativistic compares its momentum scale with $mc$, or its speed with $c$. An investigator can set a tolerance for neglecting higher terms in an expansion. The boundary then depends partly on the accuracy required; it is not a newly discovered fundamental species of particle. Nevertheless, the approximation is constrained by measurable scales and can be objectively excellent or poor.

This prevents two opposite mistakes. Since a category is chosen for a purpose, it need not be arbitrary: a classification that repeatedly supports independent predictions has a better claim to natural grounding than an arbitrary grouping. Conversely, successful prediction does not uniquely identify the world's one privileged taxonomy. Different scales and explanatory questions may justify overlapping classifications of the same objects.

The defensible conclusion is selective and plural. Treat some kinds as tracking stable mechanisms, structures or historical lineages; assess the evidence for each classification and its intended domain. Retain useful idealized or conventional categories without pretending that every term in a scientific theory names an equally fundamental division of nature.

Thomas Kuhn's account is a useful corrective to a history in which scientists simply accumulate observations and replace theories whenever a contrary observation appears. It is less convincing as an exceptionless sequence governing every science. Its compatibility with scientific progress depends on what progress means.

In normal science, a community shares a scientific paradigm: exemplary achievements, accepted instruments and standards for choosing tractable problems. An unexpected result can initially be treated as a problem in the calculation or apparatus. This need not be irrational. A framework that succeeds elsewhere gives investigators reasons to check an instrument before abandoning the framework. Revolutionary theory change reorganizes some of those commitments instead of merely adding another measurement.^[ref]

The important contrast is between a descriptive account of how research is organized and a justification of every community's decision. Shared training can explain persistence without making persistence correct. A contrary experiment can expose a failed auxiliary hypothesis, but it can also expose a central error. The Duhem problem rules out an automatic inference from one anomalous reading to one uniquely false theory; it does not authorize indefinitely protecting a theory from independently calibrated tests.

Nor does incommensurability require complete inability to compare. Theories can change the meanings of such quantities as mass while still being compared through specified apparatus readings, limiting calculations and predictions. For instance, the relativistic energy of a free massive particle has the expansionAt $p\ll mc$, subtracting the rest energy leaves the familiar leading kinetic energy. This is an explicit connection between domains, not an assertion that every concept remained unchanged. It gives a concrete way to retain successful calculations while recognizing the earlier theory's limitations.

Progress measured as closer approach to a single final ontology is therefore not guaranteed by the historical pattern alone. Progress measured through more accurate prediction, solving independently specified problems and understanding why older models work in restricted domains is compatible with substantial conceptual change. These criteria must not be defined merely as whatever a victorious community happens to prefer.

Finally, different specializations can develop competing frameworks simultaneously. A research community need not occupy one uniform stage, and a chronological label alone does not demonstrate a crisis or revolution. The strongest use of Thomas Kuhn's account is consequently comparative: identify what changed in examples, standards and practice, then ask which changes improved scientific capacities. It explains important features of research without supplying a universal historical law or implying that all theory choices are equally good.

It is plausible that some claims in current science will be revised, and much stronger to say that every successful part of present science is probably false in every relevant respect. The first is compatible with warranted knowledge. The second needs a substantive argument about the particular claims, their evidence and what “false” means for an approximate or idealized model.

Pessimistic induction points to theories once successful and accepted that were later rejected. Larry Laudan uses historical cases to challenge the inference from predictive success to the truth or successful reference of all theoretical commitments.^[ref] The lesson is especially strong against an indiscriminate realism that treats every productive representation as a literal inventory of nature.

But a historical list is not automatically a representative statistical sample. One must ask which successes were substantial, which theoretical components generated them, what survived and whether current evidence is comparable. The eventual rejection of an entire ontology does not entail that every measurement, mathematical relation or limited-domain prediction associated with it was worthless. Nor does the chance that at least one claim in a large conjunction is false imply a high error probability for each member separately.

Newtonian mechanics illustrates a useful distinction. It is not a universally exact account across all speeds and gravitational regimes, but controlled low-speed and weak-field approximations can remain extremely accurate. That is an empirical and mathematical claim about a domain, not permission to call any false account approximately true. Approximate truth needs interpretable quantities, a specified regime and some account of relevant error. Structural realism similarly needs preserved physical relations rather than merely a shared abstract notation.

The possibility of error therefore matters. It should discipline extrapolation, encourage independent testing and prevent practical success from being turned into unconditional metaphysical certainty. It does not require abandoning the best-supported models within tested domains. An empiricist can retain justified confidence in empirical adequacy, while a selective scientific realist may defend stronger commitments where evidence and intervention support them. Both positions must distinguish fallible confidence from certainty and carefully bounded reliability from unrestricted literal truth.

Scientific underdetermination occurs when the relevant evidence does not select a unique theoretical account. Its force depends on the scope of “the evidence”. Several models can agree with a finite data set while disagreeing about new measurements; a suitable experiment can then discriminate them. The stronger claim of empirical equivalence says that rivals agree on every possible observation within a stated domain, so further evidence of that same kind cannot separate them.

The Duhem problem gives a related but distinct source of ambiguity. If a target theory $T$ and auxiliary hypotheses $A$ jointly imply an observation $E$, then observing $\neg E$ shows $\neg (T\wedge A)$, not uniquely $\neg T$. Independent evidence about the apparatus or background assumptions can redistribute support. It does not make the original conjunction logically true; it provides additional information relevant to which revision is warranted.

A finite-data example makes the limits clear. If $p_0$ interpolates observations at distinct $x_1,\dots ,x_n$, then every polynomial $p_c(x)=p_0(x)+c{\prod }_{i=1}^n(x-x_i)$ has the same sampled values. Thus polynomial interpolation does not determine a unique prediction outside the sampled points when the degree is unrestricted. A degree bound of $n-1$ would instead give the familiar uniqueness result, so the allowed model class matters. That does not show that every interpolant is equally credible for extrapolation. Smoothness assumptions, independently motivated mechanisms and further measurements can matter. But those extra commitments must be stated rather than smuggled into the claim that the original data uniquely determined a model.

In a probabilistic comparison, Bayes' theorem giveswhen the quantities are defined. Equal likelihoods leave the prior odds unchanged by this evidence. Neither equality nor unequal priors follows merely from both theories being logically compatible with $E$. Thus compatibility, equal fit in a particular measure and full empirical equivalence should not be conflated.

The philosophical significance is a constraint on inference, not a universal proof of irrationality. An empiricist may use it to motivate restraint about hidden structure; a scientific realist may defend further explanatory or intervention-based reasons for belief. A convincing sceptical argument must identify genuine rivals and the relevant equivalence, while a convincing realist reply must explain why its additional virtues track the claims at issue. Scientific knowledge can be selective and fallible even when evidence does not deductively choose a unique complete theory.

Logical empiricism seeks to preserve the authority of logic and mathematics while denying that they supply unsupported factual discoveries about nature. The analytic-synthetic distinction offers a division of labour: an analytic truth holds by logical or linguistic rules, whereas a synthetic truth requires information about the world. A logical implication can organize a physical calculation without deciding that its premises describe an actual apparatus.

Without some such account, a strict claim that all genuine knowledge is empirical faces its own treatment of logic and mathematics. A logical tautology is not usually established by sampling cases. Conversely, a measurement of a wire's length does not follow merely from fixing the meaning of “metre”. The distinction lets the empiricist use formal structure while locating factual risk in the empirical application. This does not mean that every historical version of logical empiricism accepted the same sharp boundary.

The central challenge is to explain analyticity without a circle. Calling a statement true by meaning appeals to a notion that must itself be clarified. Defining the relevant meanings by synonymy, then explaining synonymy through analytic equivalence, gives no independent foundation. Willard Van Orman Quine develops this pressure against definitions, substitution tests and rules, and links it to the testing of whole theoretical packages.^[ref]

A useful reply distinguishes describing ordinary language from constructing an explicit formal framework. Rudolf Carnap's meaning postulates restrict the admitted interpretations of a language. One can define a sentence as analytic relative to those rules when it holds under every permitted interpretation.^[ref] That is a precise and usable local classification. It does not, merely by stipulation, prove that a natural-language sentence was already analytic independently of the framework, or that a scientific framework can never be revised.

Thus a modest defence is more convincing than an absolute one. Logical empiricists can justify a distinction relative to stated linguistic and inferential commitments, using it to separate a model's formal consequences from empirical assumptions. They have a harder task if they require a unique, unrevisable division governing all meaningful discourse. The practical value of explicit conventions survives even if there is no universal boundary immune to scientific and linguistic change.

An empiricist can distinguish what could be observed by creatures like us from what is posited to account for observations. This is not a distinction between what is already known and what is unknown. A distant ordinary object can be observable in suitable circumstances even if no one has seen it; successful detection of an electron need not make an individual electron directly visible to an unaided human observer. The criterion concerns a capacity, not the history of a particular observation.

Bas van Fraassen's constructive empiricism uses this distinction to separate accepting a theory as empirically adequate from believing every claim it makes about unobservable entities. Instrumental outputs can be treated as further phenomena that the theory must accommodate, without identifying every instrumental detection with unaided perception.^[ref] On this view theoretical language remains literal; agnosticism about an unobservable entity is not the assertion that its name is meaningless.

The difficulty is to make the boundary principled. A microscope, a telescope and a particle detector extend our practical reach in different ways. Instrument use is continuous with ordinary perception in its dependence on lenses, biological processing and learned interpretation. An observer can also be mistaken about a readily visible object. Thus a distinction between direct observation and instrument-assisted inference cannot simply be a distinction between certainty and fallibility.

The empiricist need not claim that the boundary is sharp in every marginal case or fixed independently of science. Information about human perception can help identify what humans could observe. That makes the criterion scientifically informed and relative to an epistemic community; it does not by itself make it inconsistent. But someone who uses accepted science to identify human limitations must explain why such acceptance does not already commit them to all of that science's unobservable ontology.

The distinction is useful as a proposal about the aim and evidential commitment of scientific acceptance. It permits disciplined use of a model while withholding some metaphysical belief, and it makes the difference between fitting appearances and describing hidden structure explicit. Its usefulness does not establish it as the uniquely rational boundary of belief. A scientific realist can argue that independently calibrated instruments and successful interventions justify belief beyond unaided observability. The main dispute is how much belief that success warrants, rather than whether scientists should use the instruments or take their results seriously.

The covering-law model captures one important feature of scientific explanation: a result should follow from general principles together with the relevant circumstances. In its deductive-nomological model, laws and conditions entail the phenomenon, the laws are needed in the derivation, and the premises must be true and empirically substantive. That is stronger than merely redescribing the event.^[ref]

For example, the ideal gas law, a specified amount of gas and a fixed-volume condition explain why changing its temperature changes its pressure in the model's regime. The derivation identifies a systematic dependence rather than an isolated coincidence. It also makes assumptions visible: a calculation using a law outside its domain is not repaired simply by being logically valid.

Nevertheless, valid subsumption does not by itself determine explanatory direction. A vertical pole of height $h$, a solar elevation $\alpha$ and a shadow of length $l$ satisfy $h=l\tan \alpha$ under suitable geometric assumptions. The same relation allows one to deduce the height from the shadow. In an ordinary intervention on the pole, changing its height changes its shadow; shortening the shadow with a screen does not shorten the pole. The reverse calculation can be excellent evidence for the height without providing the same causal explanation. This is explanatory asymmetry.

Relevance is another difficulty. One can append an irrelevant fact to a sound derivation without invalidating the deduction, but the added fact need not explain anything. Requiring an essential law prevents some vacuous derivations, yet still does not identify which features make a physical difference. A useful explanation should reveal the dependence the question is asking about, not merely contain enough true premises to reach its conclusion.

Nor are all explanations deterministic. A particular radioactive decay can have a good account in terms of a physical decay mechanism and a probability distribution without being entailed by initial conditions. An inductive-statistical model extends the covering-law idea, but requiring a high probability would exclude the explanation of some rare outcomes. Conversely, a high conditional probability produced by irrelevant correlations need not explain a mechanism.

The model is therefore a valuable discipline for organizing some explanations, rather than a sufficient universal analysis. It clarifies derivational structure and testable commitments. An adequate broader account must also address causal relevance, the question's contrast, statistical events and explanations by constraints or unification that are not straightforward causal histories.

John Locke's primary-secondary quality distinction separates properties such as shape, extension and motion from powers to produce experiences such as colour and taste. On this account a secondary quality is a body's disposition to affect a perceiver through its physical constitution; the disposition should not be confused with the sensation it produces.^[ref]

This distinction can help scientific explanation. A surface's spectral reflectance, the illuminating spectrum, and a visual system's receptor responses jointly affect perceived colour. Altering illumination can change the experience without changing the surface. The lesson is to distinguish an object's physical structure, the surrounding conditions and the observer's response, rather than to assign a colour experience to the object as though no observer or illumination mattered.

A simple absolute contrast between invariant primary properties and observer-dependent secondary properties is less plausible. Special relativity makes measured lengths and durations depend on the inertial frame; appearance of shape also depends on viewing geometry. Dependence on circumstances therefore does not by itself mark a property as unreal. Conversely, a disposition to reflect particular wavelengths or stimulate receptors can be an objective physical relation even though its manifestation requires suitable conditions.

A useful modern version is thus explanatory rather than an unrevisable division of reality. Distinguish independently specified physical variables from response properties defined relative to a system and conditions, then study the mechanism connecting them. The boundary can move with better science. This preserves the insight that experiences require explanation while avoiding the claim that every relational property is merely subjective or that the historical list of primary qualities is final.

Numerous successful tests can justify a fallible belief in a general hypothesis, but do not usually entail its truth. If finitely many tested instances satisfy $F(x)\Rightarrow G(x)$, a further untested instance can still fail. The logical gap is the problem of induction, not a practical complaint that the sample was merely too small. Appealing to the past success of inductive reasoning to establish its universal reliability uses the disputed form of inference itself.^[ref]

The kind of testing matters as much as the count. Repeating one measurement with the same unnoticed systematic bias can add little. Independent apparatus, novel circumstances and predictions on which competing theories disagree can be much more informative. A test deliberately designed to expose a failure is stronger than a loose resemblance noticed after the outcome. Still, auxiliary hypotheses and instrument assumptions must be considered when interpreting either success or failure.

Bayesian confirmation expresses the distinction quantitatively. For independent trials with success probability $p$, $n$ successes have likelihood $p^n$. In a simple comparison between a certain-success hypothesis and a specified alternative $p=q<1$, the Bayes factor is $q^{-n}$. This favours the first hypothesis without proving it. Independence, the alternative and the prior probabilities are substantive assumptions; repeated correlated results cannot automatically be multiplied as independent evidence.

There is also a difference between a universal statement and an adequate approximation within a domain. Many successful low-speed tests of Newtonian mechanics justify its use there, without establishing its unrestricted truth at relativistic speeds. A reasonable conclusion is fallibilism: well-designed and varied successful tests can warrant acceptance or high confidence, while the scope and remaining uncertainty should stay explicit. If “justified” means deductively certain, finite testing generally does not suffice; if it means rationally supported, it often can.