Contribution
A historical and philosophical paper. It gathers Paul Weiss’s scattered remarks on “system”, written mostly between 1962 and 1978, into one concept, and argues that this concept diagnoses four doctrines of modern biology: reductionism, genocentrism, determinism and mechanicism. Its claim to novelty is synthesis, not discovery: “my aim is to bring together his philosophical ideas about systems in a clear, digestible way so that they can be put to effective use in ongoing biological discussions” (§1). It adds a thesis of its own in §7: the four doctrines’ shortcomings are “mutually reinforcing consequences of a deeply entrenched habit of thought: the tendency to atomize reality and to treat wholes (or systems) as derivative composites, rather than as primary, integrated units”.
Key insight
For Weiss a system is defined by invariance of the whole against variation of the parts, and every other property Nicholson credits to it follows from that. The definitions he quotes all say so. In 1925 a system is “any complex which tends to retain its constancy as a whole relative to the outer world during and not withstanding [sic] the changes taking place in its parts” (§3). In 1971 it is “a complex unit in space and time so constituted that its component subunits, by ‘systematic’ cooperation, preserve its integral configuration of structure and behavior and tend to restore it after non-destructive disturbances” (§3). In 1973 it “retains a relatively stationary pattern of structure in space or of sequential configuration in time in spite of a high degree of variability in the details of distribution and interrelations among its constituent units of lower order” (§5). The mutual generation of parts, the triplet example, which is what brought the paper into this record, is not among Weiss’s definitions. Nicholson introduces it in §4 as the answer to “How does a system come into being?”
Assumptions
- Non-formal by design. The concept is to be “intuitive, non-mathematical”, articulated “philosophically, as opposed to mathematically” (§1). Its one formal element is Weiss’s inequality VS ≪ ∑ (va + vb + … + vn) (§5).
- “System” is an empirical concept. Weiss, for one, “treats the system as an empirical concept … it designates entities that we discover in nature” (§6), and the account is “derived empirically from experimental work” (§7). No experiment is reported in the paper, and the derivation is not shown.
- The doctrines are characterized by their strongest statements. Reductionism is “an explanatory strategy which prescribes that a whole can be fully understood by analytically decomposing it and then progressively piecing together all that can be learned from a study of its elements and their interactions” (§3), with Jacob and Watson as witnesses. Genocentrism is “the belief that genes are central to all organismic processes” (§4). Determinism is development as “an unfolding of pre-existing capabilities, an acting-out of genetically encoded instructions” (Medawar, §5). Mechanicism is “the machine conception of the organism” (§6).
- Weiss is read through Nicholson. Every Weiss passage is quoted, and the reconstruction is Nicholson’s: Weiss “never formally presented a comprehensive systems theory of the organism” (§1).
Key results
What the system concept claims (§§3–6)
Nicholson attributes to systems, in order of appearance:
- Interdependence of parts. “In a system, no part is truly independent of all the others (Weiss 1939: 111)” (§3), against an aggregate, whose parts can be changed or removed “without altering the condition of the others”.
- The whole partly determines the parts. “in a system, the character of the parts is at least partially determined by the character of the whole” (§3). Weiss 1969a: “[i]n the system, the structure of the whole determines the operation of the parts; in the machine, the operation of the parts determines the outcome” (§6).
- The “more” is epistemic. “The ‘more’ does not refer to any ontological, measurable quantity in the system. Instead … it refers to the epistemic necessity to supplement the knowledge obtained from the study of the parts by the knowledge required to appropriately describe the collective behaviour of the parts when these are considered as a coordinated unit” (§3).
- Self-regulation. The 1925, 1939 and 1971 definitions above.
- Mutual generation, and so no stepwise assembly (§4): “in a system the network of interactions between the parts is causally responsible for the generation of each part”. The triplet: “a triplet of interdependent processes—A, B, and C—each of which provides some of the conditions required for the production of the two others. This triplet cannot be assembled by simply adding A + B + C in a stepwise manner, given that no two members of the set—(A + B) or (B + C) or (C + A)—can exist or be maintained without the presence and contributions of the third member”. Cited to Weiss 1969a: 17–18 and 1973: 76. Nicholson names the modern term: “a form of closure (Varela 1979; Rosen 1991; Kauffman 2000; Moreno and Mossio 2015)”, and n. 18 separates Varela’s organizational closure, Rosen’s closure to efficient causation, Kauffman’s catalytic closure and Moreno and Mossio’s closure of constraints. The arch (Fig. 3) gives the corollary: such a structure is built over scaffolding later removed, and present cells arise only from cells, “omnis organisatio ex organisatione. Only systems beget systems” (§4).
- Constraint from above, freedom below. “the component activities have many degrees of freedom, but submit to the ordering restraints exerted upon them by the integral activity of the ‘whole’” (Weiss 1969a: 9, §5). Genes supply “‘boundary conditions’ that shape its developmental trajectory” (Weiss 1965, §5), and the cell’s bounds are “‘boundary conditions’ set by the dynamics of the system as a whole” (Weiss 1969a: 21, §6).
- The variance inequality. For parts a, …, n of S, with va the total variance of a’s parameters and VS the whole’s: “What we would invariably find, if we studied any system in this way, is that the sum of the individual variances of the parts is far greater than the total variance of the whole” (§5). “This is the only mathematical expression we find in Weiss’s theoretical writings about systems”, used “sometimes as a way of illustrating how one might go about empirically identifying systems in nature” (§5). It is claimed for every level of the hierarchy (§5).
- Hierarchy and turnover. Nested levels (Fig. 4), each more invariant than the one below; parts replaced continually by metabolic or generational turnover (§§5–6).
- Mechanisms as consolidated systems. Systems show “a tendency toward progressive consolidation (Weiss 1970a: 118–127)”: “Evolution tends to ‘freeze’ as much free systemic activity as possible into fixed sequences of reactions (i.e., mechanisms)” (§6).
The four critiques, and what each rests on
| doctrine | the paper’s verdict (§7) | what carries it |
|---|---|---|
| Reductionism (§3) | “analytical decomposition necessarily eliminates information about the organization of the whole that cannot be recovered from an inventory of the parts” | Argument and illustration. The chick-embryo homogenate (Fig. 2) and the telephone-directory and museum analogies show that a list of components omits organization. That is uncontested, and it is not the doctrine as §3 defines it, which includes the parts’ “interactions”. The §3 resolution, that the “more” is the knowledge needed to describe the parts’ collective behaviour, is one a reductionist who counts organization among the interactions can accept. The stronger claim, that the order “could never” be retrieved “from our knowledge of the parts alone” (Weiss 1977, §6), is asserted, not argued. |
| Genocentrism (§4) | genes are “indispensable participants in development and heredity, but not autonomous sources of organismic order” | Mostly argument about causal inference, with uncontested facts. The logical core: a knockout shows “necessary participation, never exclusive responsibility” (§4); “gene for” talk tracks differences, not formation (the eye example); apportioning a trait between genes and environment is like crediting a gas’s volume to temperature and pressure. The facts are standard: DNA is chemically inert and is replicated and expressed by existing machinery; Venter’s synthetic genome needed a host cell (§4). “Molecular animism” is a diagnosis of language. The origin-of-life scaffolding is a supposition: “it seems reasonable to suppose” (§4). |
| Determinism (§5) | “macroscopic regularity need not reflect rigid microscopic fixity” | The only critique with cited evidence. Weiss’s observation, redrawn as Fig. 5, that regional topology is conserved between stages while cell trajectories differ “even between closely related embryos developing under the same conditions”; later work on stochastic gene expression and cell-fate decisions (Elowitz et al. 2002; Raj and van Oudenaarden 2008; Losick and Desplan 2008; Balázsi et al. 2011). The paper shows none of the data: Fig. 5 is a schematic. The analogy with statistical thermodynamics is argument. The counterexample is conceded in n. 30: C. elegans, whose 671 embryonic cells can be tracked individually, with the reply that this is “the exception rather than the rule”. |
| Mechanicism (§6) | organisms are “self-maintaining, processual wholes whose parts are continually produced, replaced, and coordinated by the systems to which they belong” | Argument by contrast, with the facts of turnover. Machines are assembled from pre-existing parts, are explained by reversing assembly, and keep the same material parts; organisms develop from “an originally undifferentiated yet already integrated system” and turn their material over (§6). Turnover and the dynamic cytoskeleton (Weiss 1947; Dumont and Prakash 2014, n. 43) are evidence for the second half. The account of mechanisms as frozen systems dynamics is a hypothesis given no evidence. The claim that mechanicism “provides an ontological basis for the other three” (§6) is conceptual. |
The paper does not itself mark which critiques are argued and which are evidenced. It calls the concept empirical throughout (§§1, 6, 7).
The variance inequality, checked
Weiss’s inequality is stated loosely: for each part, “the cumulative record of these deviations” over “a range of physical and chemical parameters” is summed (§5), so the units of the two sides are not fixed. Read with one scalar quantity per part, and with the whole’s value either the sum or the average of the parts’ values, the variance identity Var(∑xᵢ) = ∑ Var(xᵢ) + 2∑ᵢ<ⱼ Cov(xᵢ, xⱼ) settles what it can show.
- Whole measured as an average: VS = (1/n²)(∑v + 2∑Cov). For n independent parts VS = ∑v/n² ≪ ∑v. Any aggregate satisfies the inequality, so it cannot tell a system from an aggregate.
- Whole measured as a sum: for independent parts VS = ∑v, and the inequality fails. It holds only if ∑ᵢ<ⱼ Cov(xᵢ, xⱼ) ≈ −½∑v, that is if the parts’ fluctuations compensate one another. That is a real mark of regulation, and it is what Weiss’s words “order in the gross with freedom of excursion in the small” (§5) would need.
A numpy check (100 parts, 20,000 samples, seed 0): independent parts gave VS = 0.010 against ∑v = 100 for the average, and VS = 101 against ∑v = 100 for the sum; parts constrained to a fixed total, plus a small common fluctuation, gave VS ≈ 25 against ∑v ≈ 99 for the sum, with the identity met to three decimals. So the inequality is a criterion only for an extensive whole, where it says the parts covary negatively. Neither Nicholson nor, as quoted, Weiss says this. The §5 conclusion, “what makes the whole more than the sum of its parts is that it varies less”, inherits the ambiguity.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Weiss defined a system by the persistence of the whole’s configuration while its parts vary, are disturbed or turn over | strong | four definitions quoted, 1925 to 1977 (§§3, 5, 6) |
| C2 | In a system the network of interactions generates each part, so it cannot be assembled stepwise; no two of the triplet A, B, C persist without the third | moderate | stated with Weiss 1969a and 1973 (§4); the triplet is an abstract example, and no biological triplet is exhibited |
| C3 | Present cells arise only from cells because their organization cannot be recreated de novo; the first cells arose over a biochemical scaffolding later lost | weak | the arch analogy (§4, Fig. 3); “it seems reasonable to suppose” |
| C4 | A system’s whole varies far less than the sum of its parts’ variances, at every level of the biological hierarchy | weak | asserted for “any system” (§5); no measurement given; as stated, satisfied by any aggregate measured by its average (my check above) |
| C5 | Developmental outcomes are reproducible while cell-level trajectories differ between embryos, so macrodeterminacy does not rest on microdeterminacy | moderate | Weiss’s observations as a schematic (Fig. 5) and cited work on stochastic gene expression (§5); the paper shows no data, and C. elegans is a conceded exception (n. 30) |
| C6 | A gene knockout establishes necessary participation, never exclusive responsibility | strong | logical point about what an intervention of that kind can show (§4) |
| C7 | Decomposition destroys organization that no inventory of parts records | strong for an inventory of parts, weak for parts with their interactions | the homogenate and the analogies (§3); the stronger reading is asserted |
| C8 | Mechanisms in organisms are consolidated, “frozen” systems dynamics, not primary components | weak | Weiss 1970a and an analogy with a stream that carves its own bed (§6); no evidence |
| C9 | The four doctrines are mutually reinforcing consequences of treating wholes as derivative composites, and mechanicism grounds the other three | moderate | conceptual argument (§§6, 7) |
Concepts
- System — Weiss’s, in the definitions under Key insight; contrasted throughout with an aggregate and with a machine.
- Aggregate — a whole whose parts can be changed or removed “without altering the condition of the others”, and whose character is fixed by theirs (§3).
- Molecular ecology — Weiss’s picture of the cell as a community of molecular “species” in niches they partly construct, whose order is “only the general statistical norm of the[ir] proportions, arrangement, and distribution” (Weiss 1949, §5).
- Macrodeterminacy / microdeterminacy — predictability of the whole process against predictability of its component events. Weiss preferred “determinacy” to “determinism” as empirical terms (n. 29).
- Molecular animism — Nicholson’s label for crediting molecules with ‘control’, ‘dictate’, ‘command’, ‘orchestrate’ of what the whole does (§4).
- Progressive consolidation — the hardening of repeatedly used systemic pathways into mechanisms (§6).
- Mechanicism — Nicholson’s word for the machine conception of the organism, chosen to keep it apart from other senses of “mechanism” (n. 38).
Connections
- Closure. Nicholson ties Weiss’s triplet to closure (§4, n. 18) and cites Moreno and Mossio 2015 (LIT-536, Unfetched), Rosen 1991 (LIT-1669, Unavailable), Varela 1979 and Kauffman 2000, the last two not held. He places Weiss among the precursors of the “autonomy school” (§7).
- Developmental systems theory (Oyama 1985; Griffiths and Gray 1994; Oyama et al. 2001): Weiss’s 1971 sentence that genes “do not ‘act’ as independent autonomous dictators, but … simply ‘interact’ as cooperative parts” is called “a remarkable anticipation” (§4). None is held here.
- Waddington (nn. 33, 36): the epigenetic landscape and homeorhesis.
- Bertalanffy (§2, nn. 9, 40, 41): the same city, the same opponent (the machine conception), and a different route, through general systems theory. Bertalanffy is quoted crediting Weiss with introducing the system concept in 1924.
- Process ontology (Dupré and Nicholson 2018, n. 45): §6 ends with Weiss’s processualism.
- Nicholson’s own earlier critiques of the machine conception (n. 39) and of genocentrism (Nicholson 2025, n. 17).
Bearing on the record
- TERM-070 and CLAIM-395: the triplet is a closure, and it is a closure of processes, not of constraints. The exchange’s description in LIT-1821 is right that the triplet is the record’s closure pattern, and needs two qualifications. First, Weiss’s members are “processes” that supply “conditions required for the production” of the others (§4), while Montévil and Mossio (LIT-582) separate constraints, which are conserved at the time scale of the process they act on, from the processes they constrain. The triplet becomes TERM-070’s closure only if each “condition” is read as a constraint in that sense; the paper does not say so, and n. 18 itself warns that the authors who write of closure “interpret” it “in slightly different ways”. Second, the triplet is stronger than TERM-070’s Definition 4: each member supplies conditions for both others, and “no two members … can exist or be maintained without the presence and contributions of the third”. In Montévil and Mossio’s terms that is a closed set with no closed proper subset, a strict closure, and more (every member depends on every other). CLAIM-395’s lemma that a closed set has at least two members is met with room to spare; its time-scale half has nothing to bite on, because Weiss gives no time scales.
- The “Why this is here” of LIT-1821 needs one correction. It says “Weiss’s system is defined by the network of interactions generating each part”. The definitions Nicholson quotes define a system by the invariance of the whole; the generation of parts is a further property, stated by Nicholson in §4 as the answer to how a system can “come into being”. The pairwise point stands as the exchange gave it.
- CLAIM-480: an interior that becomes a constraint, in the reverse direction. CLAIM-480 eliminates an interior and finds a constraint on the boundary. Weiss’s “‘boundary conditions’ set by the dynamics of the system as a whole” (§6) and the genome as “boundary conditions” (§5) are the same relation seen from the whole’s side: the whole’s dynamics constrain the parts. The triplet also illustrates CLAIM-480’s last clause informally. A three-way dependence (“no two members … can exist or be maintained without the presence and contributions of the third member”) that no pair of members carries is a relation among three that the pairwise relations do not state. But the paper gives no formal model, so this is an illustration, not a further instance of the computation.
- CLAIM-477: prior art for the recursion half. “those constraints are themselves realized and maintained by something” is Weiss’s triplet and his bounds “set by the dynamics of the system as a whole” (§6). The paper is an older statement of it, at the level of argument. It supplies nothing for the specification–realization half.
- CLAIM-478: progressive consolidation is the same idea at an evolutionary time scale. The stream that “later becomes constrained by the channel it has created” (§6) is a realization at one scale working as a constraint at another, which is CLAIM-478’s thesis. Weiss’s version is analogical and gives no mechanism.
- CLAIM-474: Weiss’s definitions sit at the individuation step, and his inequality is a candidate criterion of the kind QUESTION-115 and CLAIM-157 seek. Weiss’s system is individuated by persistence of configuration through turnover, and the variance inequality was meant as a way of “empirically identifying systems in nature” (§5). As checked above, it discriminates only for an extensive whole, where it detects compensating covariance among the parts. That makes it a statistic of the same family as CLAIM-157’s near-decomposability and QUESTION-115’s criteria, and a weaker one. Nothing in the paper separates the five relationships of CLAIM-474; Nicholson’s single “system” covers interconnection, dynamics and individuation at once.
- THEORY-072: neither supports nor undermines its demarcation. The paper never discusses dissipative structures. Weiss’s “whirl of fluid” appears only as a warning against naming a molecule its “spinner” (§4).
- THEORY-302: a parallel. Turvey’s coordinative structures, where a higher level fixes an equivalence class of states and the parts keep their freedom, are Weiss’s “many degrees of freedom” under “ordering restraints” (§5), stated for motor control. Neither cites the other.
- THEORY-465 holds that genes are reproducees with little reproductive autonomy; the §4 critique of “self-replicating” genes points the same way.
- Should produce: THEORY-1122 (the variance inequality, checked) and THEORY-1126 (macrodeterminacy without microdeterminacy, the one evidenced critique, Proposed).
- No instruction for machine-learning practice.
Limitations
- No new data, and the “empirical” concept is not derived here. The paper says the system concept is “derived empirically from experimental work” (§7); what it shows is Weiss’s statements and illustrations.
- The figures are schematics, several redrawn (Figs. 2, 3, 5, 6). Fig. 5 is the only evidence offered for the determinism critique, and it is a diagram.
- The doctrines are taken at their most extreme. Jacob 1973, Watson 1965, Medawar 1965 and Davidson 2012 are quoted; nobody’s current defence of a gene-regulatory-network account or of mechanistic explanation is engaged. The reductionism critique rebuts composition without interactions, not the doctrine as defined in §3.
- The concept gives no criterion for applying “system” beyond the definitions and the variance inequality, which is ambiguous as stated.
- The author says what he does not do: Weiss’s field concept is left out (n. 8), and “systems thinking is not the only route” to these critiques (§7).
- Weiss’s texts were not read here. What Weiss held is what Nicholson quotes, and the quotations are selected for the argument.
Open questions
- Can Weiss’s inequality be restated so that it separates systems from aggregates on its own: a fixed extensive quantity, and a test that the parts’ covariances are negative beyond what a conservation law or a fixed total imposes? Measurements of cell-to-cell variability against tissue-level variability would then be its test.
- Is there a biological triplet in Weiss’s strong sense, a set of processes of which no pair can be maintained without the third, documented as such? Minimal-cell work would be the place to look.
- How often is developmental macrodeterminacy achieved with variable cell trajectories, and how often with invariant lineages as in C. elegans? A comparative census of lineage variability would answer “the exception rather than the rule” (n. 30).