NOTE-191: Patterns All the Way Up

Contribution

The paper compiles fifteen propositions about how the sciences’ ontologies relate. They concern triangulation, the plurality of modality, scale-dependence, composition, levels, effective ontology, downward causation, coarse-graining and the agent-relativity of theories. Ladyman offers them as constraints that a naturalised metaphysics must “triangulate with” (p. 3). Along the way it restates, and in places revises, the Ladyman–Ross (ETMG 2007) programme: OSR about physics joined to rainforest realism about the special sciences via real patterns (fn. 3).

Key insight

Quine’s criterion (to be is to be the value of a bound variable in the single best theory) yields a “desert landscape” that is antirealist about nearly all of science. It also cannot be applied, because the sciences do not share one domain (pp. 1–2). Replace it with Dennett’s criterion, suitably modified. Whatever indispensably figures in projectible generalisations at some scale is real, and it is real only at the scales where it is effective. Modal structure is the general form of what is real, and ontology is scale-relative. Physics is special in range but not as a bottom level.

Assumptions

• Naturalism: metaphysics must unify science and answer “how does science as a whole hang together?” (p. 1).
• The semantic, model-theoretic view of theories: representation is map-like and mathematics is ineliminable (p. 2).
• Realism about modal structure is taken from Ladyman 1998 and ETMG (fn. 6) and not re-argued.
• The ETMG definitions of real pattern and PNC are presupposed. The real-patterns definition is referred to (fn. 5, thesis 15) but not stated. Its formal content has to be taken from ETMG or Ross 2000, neither of which was reachable here.

Key results

The paper’s results are theses, with illustrations.

• Criterion of existence (p. 2): “to be is to be a real pattern, where a real pattern is something that indispensibly figures in projectible generalisations that predict and explain the behaviour of the world.”
• Three positive heuristics from the Game of Life (pp. 2–3): emergent patterns are generally applicable; scientific representation simplifies or compresses; real patterns involve scale-relativity (“the ontology of gliders and eaters is only applicable when more than a few cells over a few iterations is considered”).
• Negative heuristics (p. 3): the five disanalogies listed under corrections.
• (1) Science expands ontology qualitatively and quantitatively and integrates domains.
• (2) Ontological commitment requires triangulation across measurement modalities, scales and domains. Perrin’s determinations of Avogadro’s number are the example (Poincaré 1913).
• (3) Kinds of causation, law and so on are plural. There is “no naturalistic motivation for eliminating modality from science”, because “there are no real things without projectibility”. Objects and their modal profiles are inseparable.
• (4) “The concept of natural modal structure is an abstraction which includes all of the above. In formal mode this modal structure is absolute, but in material mode it is domain and scale-dependent.” General realism is “commitment to the reality of the modal structure tracked by our best scientific theories”.
• (5) The relations between domains and scales have modal structure of their own: limits, affordances and constraints. Type-type or token-token reduction is implausible because higher-level descriptions are coarse-grained. Pressure is not identical to any microstate.
• (6) Composition in science is diachronic and dynamical and involves interaction. It is not mereological.
• (7) Dependence, determination and supervenience are regimented in science by mathematics, not by language.
• (8) Physics covers the widest range of scales, but all of it is domain-specific.
• (9) Materialism has succeeded as a research programme, which is why ETMG adopted the PNC. The PNC is quoted in full (see below) and then qualified.
• (10) The physics that unifies the special sciences “is not absolutely fundamental, and there is no single fundamental physics” (the Standard Model plus GR).
• (11) There is no single hierarchy of levels and no sharp boundaries between them. The idea of a single hierarchy “does not correspond to anything in scientific practice”. “In quantum mechanics more fundamental does not mean smaller because of entanglement.”
• (12) “All existing theories have an effective ontology of emergent entities. Effective ontology is scale-dependent.” ETMG is quoted: “[o]ntology is scale relative in respect of both space and time”. The scales include length, time, number of parts and number of interactions. The example is Maxwell’s definition of a “material particle”.
• (13) Downward causation “is not coherent, because there are no distinct levels”. Higher-level difference-makers are real because the explanandum is high-level.
• (14) Special-science ontology tracks statistical properties of low-level entities. Scientific real patterns are lossy. The ideal gas laws use three degrees of freedom against ~10²³, and the renormalization group reduces Hamiltonian degrees of freedom.
• (15) No theory describes the whole universe at all scales, because the agents using it cannot model themselves. “This is why the definition of real patterns in ETMG refers to perspectives.”
• PNC, quoted from ETMG p. 37 (p. 8): “If a metaphysical claim is to be taken seriously at a time t, it should, if true, show how two or more scientific hypotheses, at least one of them specific and at least one of them both specific and drawn from fundamental physics taken seriously at t, jointly explain more than the sum of what is explained by the two hypotheses taken separately.” Qualification: this “is problematic because it makes reference to fundamental physics. It might be modified to refer instead to ‘a more fundamental theory’.”

Claims

id	claim	strength	support
C1	Quine’s criterion is antirealist about most of science and cannot be applied across sciences with distinct domains	informal argument	iron-bar (continuum vs lattice) and plasma examples, pp. 1–2
C2	To be is to be a real pattern (indispensable to projectible generalisations)	assertion	presented as Dennett’s idea, taken as an alternative criterion, p. 2
C3	Natural modal structure is absolute in formal mode and scale-dependent in material mode	assertion	thesis (4), illustrated by the pendulum and asteroid cases
C4	There is no single hierarchy of levels, and metaphysics should not presuppose a fundamental level	informal argument	physics examples, theses (10)–(11), fn. 9
C5	Ontology is effective and scale-relative	informal argument	chemistry, vacuum, Maxwell, tectonic plates, thesis (12)
C6	Downward causation, as intervention of a higher level on a lower, is incoherent	assertion	thesis (13), two sentences
C7	Scientific real patterns are lossy compressions, unlike the Game of Life’s	informal argument	ideal gas, renormalization group, thesis (14)
C8	ETMG’s PNC should refer to “a more fundamental theory”, not to fundamental physics	assertion	one sentence, p. 8. No revised statement is tested
C9	”Relational structure is fundamental” was not intended as a claim that there is a fundamental level	assertion	fn. 9, the author’s report of intent

Method

The paper is a compilation of theses, each backed by examples from the history and current practice of physics, chemistry, biology and economics. It has no formal apparatus and makes no attempt to derive the theses from one another.

Concepts

• Real pattern: see C2. Ladyman notes it may be developed information-theoretically (Dennett, Ross 2000, ETMG), dynamically (Ismael) or statistically (L&R 2013) (fn. 5). ETMG’s definition is perspective-relative (thesis 15).
• Rainforest realism: realism about the special sciences’ real patterns, conjoined with OSR in the philosophy of physics (fn. 3). “For the purposes of naturalised metaphysics most of physics is a special science” (fn. 12).
• Scale-relativity / effective ontology: existence as an effective entity within a range of length, time, energy, numerosity and interaction scales (thesis 12).
• Triangulation: epistemic (multiple independent lines of evidence) and ontological (a real object manifests in more than one way) (thesis 2).
• Natural modal structure: the genus of causal, nomological, probabilistic and other structures (thesis 4).
• PNC: see Key results.

Connections

It is the sibling of LIT-184 (Ladyman & Lorenzetti 2023), which applies the same package to effective realism and is cited here as forthcoming (fn. 11). Its account of real patterns contrasts with LIT-151’s instantiation-based account and with Wallace 2003. It presupposes r1 and Ladyman 1998 for modal-structure realism (fn. 6). It builds on r4 (Dennett 1991) and flags that paper’s Game-of-Life exemplar as misleading. Ladyman & Wiesner, What is a complex system?, backs theses (6) and (14). The preprint cites it as 2020 in the text and as 2022 in the references, with the name misspelt “Weisner” in both places. It sits beside Ross’s OSR-for-economics argument (r3): games and equilibria as emergent real patterns (p. 2).

Bearing on the record

It is the record’s clearest primary statement of rainforest realism and scale-relative ontology. It is also the only one that gives the PNC verbatim and records its author’s retreat from it. Thesis (14)‘s coarse-graining and lossy-compression framing is the natural bridge to information-theoretic and complex-systems readings. It carries nothing for ML practice. Real patterns are sometimes invoked about neural-network representations. The same MIT volume has Rosa Cao’s chapter “The Scientist in the Machine” (per Crossref; not read), but this paper does not discuss them.

Limitations

• All fifteen theses are asserted and illustrated. None is argued against an opponent, and the paper says as much: “many of them might seem obvious” (p. 3).
• The real-patterns definition that does the ontological work is not stated. The paper points to ETMG for it (fn. 5, thesis 15), so a reader cannot check the criterion from this text alone.
• The PNC revision is one sentence. Whether “a more fundamental theory” still makes physics special, and so preserves the Primacy of Physics Constraint, is not discussed.
• It is a preprint. Differences from the 2026 published chapter are unverified.

Open questions

• What does the revised PNC, with “a more fundamental theory”, exclude that the original did not? Does it remain a demarcation criterion?
• If modal structure is absolute in formal mode, does that reintroduce a fundamental-level description in all but name?
• Can the information-theoretic, dynamical and statistical formulations of real patterns (fn. 5) be shown to agree?

Corrections to the seeded skim

• There is no dossier or skim, so these are corrections to the record’s existing accounts.
• The PNC is revised by its own author. NOTE-129 (on LIT-184) takes the Ladyman–Ross framework as given. The record nowhere states the Principle of Naturalistic Closure. This paper quotes it verbatim from ETMG p. 37 (see Key results) and then says it “is problematic because it makes reference to fundamental physics”. It proposes replacing “fundamental physics” with “a more fundamental theory” (p. 8). Any record statement of rainforest realism should give the PNC with this 2024 qualification. The companion Primacy of Physics Constraint is not stated in this paper. Its content appears only indirectly, in thesis (9)‘s claim that materialism has succeeded as a research programme.
• “No fundamental level” is agnostic, not a denial. LIT-184 (NOTE-129) records that Ladyman–Ross “deny there is one (fn. 9)”. This paper’s fn. 9 is more careful. ETMG “argues that there is not for all we know a fundamental level of reality (though there might be), and that accordingly our metaphysics should not presuppose that there is”. It also says the claim that “relational structure is fundamental” was read by commentators (McKenzie 2017) as fundamentalism, “but it was not so intended”. So the thesis is methodological agnosticism about a bottom level, joined to a positive denial that there is a single hierarchy of levels (thesis 11). “Fundamental” in OSR slogans means non-supervenient, not bottom-level, which NOTE-129 also records.
• Real patterns are not only compression. NOTE-038 (on LIT-045) glosses real patterns as “often glossed as compressibility”, and NOTE-129 as “a compression criterion relative to physically possible perspectives”. Here Ladyman says the ETMG definition is “articulated in information theoretic terms”. He says it “can also be developed in terms of the dynamics of phase spaces” (Ismael) or “in terms of statistical structures” (Ladyman & Ross 2013; Alexseev et al.) (fn. 5), and that ETMG “defines real patterns modally” (fn. 6). He also marks a difference from Dennett. Scientific real patterns are lossy, allowing “only approximate or probabilistic recovery”, whereas Game-of-Life patterns are lossless (thesis 14).
• The Game of Life is a misleading exemplar, in five stated ways (p. 3). It has a privileged fundamental level (the bitmap). Macrostates are “completely determined by and reducible to microstates”. Its cells do not genuinely interact. Its emergent entities are “not genuinely causal”. It is “one-dimensional”, with no triangulation across data sources. Any record account that explains rainforest realism by the Game of Life alone (see r4, Dennett 1991) should carry these disanalogies.
• Wallace’s use of real patterns differs from Ladyman–Ross on exactly one point. LIT-151 (Wallace 2021) defines real patterns by instantiation plus best precisification and says math-first ESR and OSR coincide “at every non-fundamental level”. This paper describes Wallace’s earlier use (Wallace 2003). There only higher-order entities are real patterns and “the wavefunction is understood in categorical rather than functional terms”. Ladyman–Ross instead “eschew the idea of a fundamental level and advocate a real patterns account of ontology across the board (including also merely notional or representational real patterns)” (p. 1). The record’s two accounts of real patterns (LIT-151 and LIT-184) therefore disagree about whether the fundamental level is itself a pattern. This paper states the disagreement.
• LIT-170 (French), LIT-173 (Morganti), LIT-196 (Chakravartty) and LIT-152 (Floridi) are not engaged. The paper does not discuss eliminative vs non-eliminative OSR, or quantum non-individuality, at all. For those, see r1 and LIT-045.