A Historical Introduction to the
Philosophy of Science

The following is a summary of the thirteenth chapter of John Losee's book, A Historical Introduction to the Philosophy of Science (fourth edition), with some ancillary notes.

Philosophers discussed in this chapter: Paul Feyerabend (1924–98); Nelson Goodman (1906–98); Stephen Toulmin (1922—); Herbert Feigl (1902–88)

Is There a Theory-Independent Observational Language?

(p. 178) The logical reconstructionists make three assumptions:

1. Observation reports are true or false independent of the theories they support.
2. Observation reports provide proper tests for theories.
3. Terms in theories obtain their meanings from observation reports.

Contrawise, Feyerabend argued that observation reports get their meaning from theories (e.g., with a change in theory of colour observation, the property 'colour' changes from being an intrinsic property to being relational).

(p. 179) Achinstein illustrated how the observational/theoretical distinction depends on the context of the observation (e.g., observing muscle tissue unaided or with optical instruments). The context is the relevance of the accompanying effects (e.g., seeing smoke as fire).

The Duhem–Quine thesis further undermines the logical reconstructionists' strict observation-theory distinction:

1. There is no straightforward 'empirical content' of an individual observation statement.
2. Any statement can be saved from falsification with enough adjustments elsewhere in the theory.
3. There is no sharp distinction between synthetic statements (empirical) and analytic statements (non-empirical).

(p. 180) The logical reconstructionists' 'Safety-Net' analogy is mistaken as the supporting rods (observation reports) for the net are not independent of the net (axiomatic system) they support.

Doubts About the Covering-Law Model of Explanation

Hempel defended his view that a properly scientific explanation subsumes a phenomenon under general laws, either by deduction (DN) or statistical inference (IS).

(p. 181) Hempel conceded that deduction using 'indicator laws' and inference using statistical laws are not sufficient for scientific explanation (e.g., patient developing measles and recovering from a cold). But covering under a general law is necessary for explanation.

(pp. 181–2) Scriven objected that deduction from covering laws is not necessary either (e.g., explaining a bridge collapse). Hempel pointed out that leaving out laws in the explanation still presumes those laws in the explanation.

(p. 182) Salmon objected that the covering-law model fails to explain improbable events (e.g., developing leukemia from radiation exposure).

A Non-Statement View of Theories

Suppe rejected the logical reconstructionists' view of theories as collections of formal statements. He likened a theory to a proposition where a single proposition can be expressed by many different sentences.

(p. 183) For example, how quantum theory is equally 'expressed' by both Schrödinger's wave mechanics and Heisenberg's matrix mechanics.

For Suppe, theories are non-linguistic entities that:

• point to a class of phenomena
• model the phenomena using idealized systems
• make counterfactual claims

Giere thought these idealized systems had explanatory power by hypothesizing a structural relationship between the idealized model and the natural system (e.g. Kinetic Theory of Gases and Galileo's Theory of Falling Bodies).

Cohen pointed out how later developments increase the sophistication of the idealized model to better explain real systems (e.g. Newton's development of his model of planetary motion).

(pp. 183–4) Contra the logical reconstructionists, scientific explanation is not by deduction of empirical laws from theories. Sellars pointed out how theories explain why phenomena obey empirical laws the way they do (e.g., the kinetic theory explains why at high pressures gases diverge from the Boyle–Charles law).
[LA: Note how critics such as Suppe, Giere and Cohen rejected Duhem's (see p. 120) anti-realist view of theories.]

Goodman's "New Riddle of Induction"

(p. 184) Goodman attacked Nicod's Criterion (The Principle of Instance Confirmation, see p. 168) with the example of grue. 'All emeralds are grue' is as equally supported as 'All emeralds are green' by observation statements.

(pp. 184–5) Goodman noted that 'All emeralds are grue' is not a lawlike generalization; only accidental.
[LA: But it can be posited as a lawlike generalization.]

(p. 185) In 1953, Goodman suggested lawlike generalizations refer to predicates that do not involve a particular space or time. He rejected this demarcation as the 'grue' paradox can be restated without such involvement. Losee also notes that some scientific laws do involve reference to a particular space or time (e.g., Kepler's First Law).

Goodman solved his 'grue' paradox by restricting allowed predicates to those that have been used successfully in predictions in the past.

(p. 186) Thereby, Goodman elevated the importance of the history of a scientific theory to its degree of confirmation.

In 1964, Hempel conceded that the logical reconstructionist program of trying to show how theory confirmation is a function solely of logical/syntactical relations between observation statements and theoretical statements has failed.

Doubts About the Chinese-Box View of Scientific Progress

Feyerabend rejected Nagel's logical reconstructionist claim that empirical laws are reduced to higher-level theories.

Two examples:

1. Galileo's law that vertical acceleration of a falling body is constant is not deducible from Newtonian physics as distance between body and earth reduces.
2. Newtonian physics is not deducible from General Relativity Theory as meanings of 'length' are different in the two theories (semantically incommensurable).

(p. 187) Putnam suggested a patch of Nagel's reductionism: reducing theory need only approximate observational results of reduced theory.

Feyerabend countered that with Putnam's move, the logical reconstructionists have given up on explaining the actual history of theory replacement. As observation reports confirming a high-level theory are not theory-independent, theories are not objectively comparable (observationally incommensurable).

Whewell's tributary–river analogy, the logical reconstructionists' Chinese-box view and Bohr's Correspondence Principle all see scientific theory replacement as continuous.

(p. 188) Contrawise, Toulmin pointed to conceptual changes in what is seen as 'Ideals of Natural Order'. Natural order specifies what phenomena require explanation (e.g., Newton's First Law saw uniform rectilinear motion as not requiring explanation, contra Aristotle).

For Toulmin, an anomaly is a resistance to attempts at explanation via a 'natural order' schema (e.g., the motion of projectiles for Aristotelian mechanics). When anomalies mount, a fitter theory replaces it by revolution.

For Toulmin, the logical reconstructionist programme cannot explain theory change as the standards of intelligibility/reasonableness change with a change in theory.

(p. 189) Hanson saw revolutionary theory change as a shift in gestalt; a way of seeing the world.

Feyerabend and Feigl on the Death of Orthodoxy

Feyerabend saw the logical reconstructionist programme as useless to the living scientist as it cannot help her decide between competing theories. This is due to two false assumptions:

1. A theory-independent observation language is available to adjudicate between theories.
2. A theory can be consistent with all the known facts.

In reality, every theory has counter-instances.

(p. 190) Feyerabend recommended instead philosophers of science study the history of science.

Losse asks of Feyerabend what distinguishes philosophy of science from the history and practice of science. Feyerabend sought to diminish the specialness of science as a separate form of enquiry.

Feigl, on the other hand, sought to salvage the gains of the logical reconstructionist programme as explaining how theories are tested and compared. This is possible because:

1. empirical laws are deducible from theories
2. many empirical laws are stable and approximately true

(p. 191) Feigl conceded that empirical laws may be corrected by a high-level theory (e.g., an astrophysical theory may correct physical optics), but, none the less, there are thousands of stable empirical laws.

Feigl suggested that these relatively stable empirical laws are what test theories. Although the meanings of terms in empirical laws change with successive theories, as Feyerabend pushed, historically, theories are appraised by their absorption of laws.

Smart observed that Feigl's account ignored the history of biology that has no empirical laws; only generalizations that are restricted to the earth.

(pp. 191–2) Adjustments to these generalizations in biology do not necessitate revolutions. Ruse pointed out that the same is the case with generalizations in physics (e.g., laws of Kepler, Snel, Boyle, Ohm).

(p. 192) Beatty argued that there are no empirical laws in evolutionary biology because the same initial conditions and selective pressures can result in different types of organisms. This variability of outcomes is because of:

1. chance events (e.g., mutations, earthquakes)
2. the functional equivalence of different adaptions (e.g., larger body vis-à-vis burrowing)

Unlike empirical laws, biological generalizations do not support counterfactual claims.

Sober and Carrier responded that biological generalizations are at the higher level of supervenient properties. A supervenient property changes if and only if some lower-level property changes.

(p. 193) The supervenient property 'fitness' supervenes upon physical characteristics or behaviour and is measured by biologists to use in biological explanations (e.g., predator–prey, sex-ratios, anemia).

Brandon emphasized how supervenient properties do not figure in low-level biological laws, but in the schema: an organism is better adapted than a rival in a particular environment if and only if it is better able to survive.

Brandon connected the 'fitness' schema to biologies of specific organisms where:

1. biological entities are in 'chance set-ups with respect to reproduction'
2. levels of adaptedness differ in a common selective environment
3. adaptive differences are heritable

The 'fitness' schema, although not directly empirical, makes substantive existential claims and acts a directive principle for evolutionary biologists (much like Newton's second law and the law of conservation).

Questions to Consider:

1. Were the logical reconstructionists wrong in thinking there is a strict observation/theory dichotomy?
2. Is Goodman's 'grue' paradox an illustration of the Duhem–Quine Underdetermination Thesis?
3. Can a gradual process theory of scientific theory change be defended?
4. Are there low-level empirical laws in evolutionary biology? If not, does it matter?
5. Is evolutionary biology a genuine science? If so, how?

Copyright © 2022–3 Leslie Allan