Rational Justification of Quantitative Parsimony in Non-Additive Cases and the Scope of Its Applicability

Baker (2003) vindicated the scientist's preference for hypotheses that postulate fewer individual entities -- quantitative parsimony -- by arguing that in 'additive' cases, where qualitatively identical objects sum their individual contributions (paradigmatically, one neutrino rather than many to explain the missing spin in Beta decay), the more parsimonious hypothesis has greater explanatory power. He explicitly restricted the analysis to additive cases and acknowledged that non-additive contexts, such as postulating one or several unobserved planets to explain orbital perturbations, would require a substantively different approach; Nolan's broader challenge -- why quantitative parsimony is a virtue at all and how wide its applicability is -- was thereby only partially addressed. Jansson and Tallant (2017) offered a probabilistic rationale that reaches the Neptune case, while Vanderburgh (2014) pressed a dark-matter objection; the question remains open. The surviving problem is to determine whether, when, and why preferring quantitatively parsimonious hypotheses is rationally warranted outside additive cases, and to fix the scope of the virtue across theory choice.

Background

Occam's razor -- the methodological preference for simpler theories -- is a load-bearing but contested principle in scientific theory choice. Its ontological form, parsimony, comes in two strengths. Qualitative parsimony concerns the number of kinds of entity a hypothesis postulates; quantitative parsimony concerns the number of individual entities (tokens) of a given kind. Many philosophers have granted rational presumptions in favor of qualitative parsimony while denying any analogous presumption for the quantitative variant: adding a new kind of particle seems a genuine theoretical cost, whereas postulating two electrons rather than one has often been dismissed as no cost at all. Scientific practice, however, frequently defaults to the most quantitatively parsimonious hypothesis among empirically equivalent rivals: when the 'missing' spin and energy in Beta decay (radioactive emission of electrons) was explained in the 1930s, exactly one new particle, the neutrino, was postulated, even though hypotheses postulating two, ten, or more particles with suitably scaled properties fit the same data.

Nolan (1997) took the practice seriously and posed the challenge that frames the field: either there is a rational, truth-conducive justification for preferring hypotheses that postulate fewer individuals, or a pervasive piece of scientific methodology rests on aesthetic bias; and, in the former case, how wide is the justification's applicability? He examined the neutrino case and Avogadro's hypothesis (the resolution of equal-volume gas data in favor of fewer, heavier molecules) without endorsing a general rationale.

Baker (2003) gave the most influential answer. He focuses on additive cases: hypothesis families postulating collections of qualitatively identical objects that collectively explain the observed phenomenon by summing their individual positive contributions. In the neutrino case, compare hypothesis $H_1$ (each Beta decay emits one neutrino of spin $1/2$) with $H_{10}$ (each decay emits ten neutrinos, each of spin $1/20$). Both explain the observed missing spin of $1/2$. But no experiment of the period had found a 'missing' spin value such as $9/20$. $H_1$ explains this non-observation straightforwardly -- there simply are no undetected particles with spin $1/20$ or less -- whereas $H_{10}$, which postulates particles of spin $1/20$, must add an auxiliary claim (for instance, that neutrinos are always emitted in groups of ten) that degrades its syntactic simplicity. So in additive cases the more quantitatively parsimonious hypothesis has greater explanatory power, understood as the power to explain observed non-occurrences as well as occurrences, and the preference for it follows from the general preference for explanatory power. The mechanism depends on the explained quantity being discrete and quantized: Baker notes that continuous quantities, such as the missing mass-energy in Beta decay, do not yield the same advantage (Baker 2003, p. 255), and he explicitly restricts the analysis to additive cases.

The restriction is substantive. In non-additive cases the postulated objects may differ qualitatively from one another and their contributions need not sum: the gravitational perturbations of a planet's orbit could be produced by one unobserved planet, or by several whose attractions partially cancel and jointly mimic a single one; similarly, hypotheses about the multiplicity of unobserved dark matter (the mass postulated to explain galactic dynamics) need not decompose into summed identical contributions. Baker's set-over-subset mechanism has no grip on such cases, and he acknowledges that a rational justification there would require a substantively different approach.

Later work sharpened rather than closed it. Vanderburgh (2014) argued that Baker's account yields undesirable consequences when applied to astrophysical dark matter and that Baker had not succeeded in showing quantitative parsimony to be a theoretical virtue. Jansson and Tallant (2017) granted Baker only 'a limited success' in additive cases, rejected his analysis for failing breadth, and proposed a probabilistic rationale -- relativized to 'probabilistically additive' hypotheses and to particular episodes of theory choice -- which they apply to the neutrino, Avogadro, and, crucially, the postulation of Neptune, a non-additive case: several planets would have to orbit in specially restricted ways to reproduce the effect of a single planet, which is antecedently unlikely and so lowers the probability of the less parsimonious hypothesis. Sendłak (2018) disputed the clarity of the quantitative/qualitative distinction itself. Nolan (2023) argued that current debates about the structure of space and time (whether spacetime is discrete, 'gunky' -- divisible without limit -- or composed of points) need quantitative parsimony considerations to resist wildly unparsimonious rivals. Baker's own 2022 summary in the Stanford Encyclopedia of Philosophy still presents the extension to non-additive cases as 'an interesting question' on which Jansson and Tallant merely 'argue' positively -- the mark of an open problem rather than a settled one.

Problem Statement

Is there a rational, truth-conducive justification for preferring the more quantitatively parsimonious hypothesis in non-additive cases of theory choice -- cases in which the postulated entities may differ qualitatively from one another and their contributions to the explained phenomenon need not sum (as when one or several unobserved planets are postulated to explain a planet's orbital perturbations, or when hypotheses postulating different multiplicities of unobserved dark matter are entertained) -- and, correlatively, how wide is the rational applicability of quantitative parsimony across scientific and philosophical contexts, that is, what criterion delimits the class of theory-choice situations in which the preference for fewer postulated entities is warranted and why it fails outside that class? This is the residual left open by Baker (2003), whose explanatory-power justification is restricted to additive cases with quantized, discrete explanatory quantities, and it subsumes Nolan's general challenge to explain why quantitative parsimony might be thought a good thing and how far its use extends.

References

1. Alan Baker (2003). 'Quantitative Parsimony and Explanatory Power.' British Journal for the Philosophy of Science 54(2):245-259. DOI: 10.1093/bjps/54.2.245. https://doi.org/10.1093/bjps/54.2.245 (primary source)
2. Daniel Nolan (1997). 'Quantitative Parsimony.' British Journal for the Philosophy of Science 48(3):329-343. DOI: 10.1093/bjps/48.3.329. https://doi.org/10.1093/bjps/48.3.329
3. Lina Jansson and Jonathan Tallant (2017). 'Quantitative Parsimony: Probably for the Better.' British Journal for the Philosophy of Science 68(3):781-803. DOI: 10.1093/bjps/axv064. https://doi.org/10.1093/bjps/axv064
4. William L. Vanderburgh (2014). 'Quantitative Parsimony, Explanatory Power and Dark Matter.' Journal for General Philosophy of Science 45(2):317-327. DOI: 10.1007/s10838-014-9261-9. https://doi.org/10.1007/s10838-014-9261-9
5. Alan Baker (2022). 'Simplicity.' Stanford Encyclopedia of Philosophy, substantive revision of 16 May 2022, accessed 5 September 2026. https://plato.stanford.edu/entries/simplicity/
6. Maciej Sendłak (2018). 'On Quantitative and Qualitative Parsimony.' Metaphilosophy 49(1-2):153-166. DOI: 10.1111/meta.12286. https://doi.org/10.1111/meta.12286
7. Daniel Nolan (2023). 'Space, time and parsimony.' Noûs 57(4):763-783 (online first 17 June 2022; print issue December 2023). DOI: 10.1111/nous.12420. https://doi.org/10.1111/nous.12420
8. Sam Baron and Jonathan Tallant (2018). 'Do Not Revise Ockham's Razor Without Necessity.' Philosophy and Phenomenological Research 96(3):596-619. DOI: 10.1111/phpr.12337. https://doi.org/10.1111/phpr.12337
9. Samuel Schindler (2022). 'Theoretical Virtues: Do Scientists Think What Philosophers Think They Ought to Think?' Philosophy of Science 89(3):542-564. DOI: 10.1017/psa.2021.40. https://doi.org/10.1017/psa.2021.40
10. Simon Thunder (2026). 'Parsimony and internal complexity.' Philosophical Studies, published online 4 July 2026. DOI: 10.1007/s11098-026-02575-1. https://doi.org/10.1007/s11098-026-02575-1
11. James H. McIntyre (2026). 'Brute fact parsimony.' Philosophical Studies, published online 4 February 2026. DOI: 10.1007/s11098-026-02487-0. https://doi.org/10.1007/s11098-026-02487-0
12. James H. McIntyre (2026). 'Consciousness Multiplied.' Philosophy and Phenomenological Research, published online 13 July 2026. DOI: 10.1111/phpr.70146. https://doi.org/10.1111/phpr.70146