Viability, closure and memory

Towards an organisational, dispositional and multidimensional characterisation of living systems

Abstract

Defining living systems remains difficult because no isolated criterion seems sufficient. Metabolism excludes certain dormant forms when it is required as a permanent activity. Reproduction does not concern every organism taken individually. Darwinian evolution describes mainly the transformations of lineages. Regulation, memory and learning can appear in partial forms in non-living or artificial systems.

The present article therefore proposes neither an essence of living systems nor a universal boundary supposed to be already established. It develops an organisational, dispositional and multidimensional characterisation founded on the notion of constitutive closure.

Within this framework, an organisation presents a closure when some of its processes take part in producing or renewing the components, constraints and conditions required for their own continuation. That closure may be active or conserved in a latent form. Robustness, regeneration and recoverability are then distinguished as graded properties, which must not be confused with self-production itself.

Viability introduces an asymmetry between transformations compatible or incompatible with the continuation of the organisation. That asymmetry grounds a minimal normativity, without presupposing intention or conscious purpose. It is endogenous in its causal origin, but its observation depends on a scale, a temporal horizon and explicitly chosen indicators.

Adaptivity, memory and learning are described as possible elaborations of organisational autonomy. Evolution adds a distinct transgenerational dimension. The framework thus accepts that a single artificial organisation could be considered alive if it effectively realised a constitutive closure, even in the absence of a prior evolutionary lineage.

The cases of protocells, viruses, spores, artificial systems and erythrocytes nevertheless show that the boundary remains open. The proposal is therefore accompanied by a testing programme that explicitly distinguishes hypotheses of necessity from hypotheses of sufficiency.

Introduction

Defining living systems remains an open problem.

No criterion taken in isolation seems able to cover the whole set of recognised cases while excluding non-living systems that present certain similar properties.

Metabolism appears central, but becomes hard to require as a permanent activity when one considers spores, dormant seeds or certain states of cryptobiosis.

Reproduction plays a fundamental role in the continuity of living systems, but it does not characterise each organism taken individually. A sterile organism does not cease to be alive.

Darwinian evolution is an essential property of lineages, but it does not directly describe the present organisation of each individual.

Openness to exchange brings living systems close to many dissipative structures.

Regulation exists in artefacts.

Dependence on history appears in materials that display hysteresis.

Learning can be achieved by artificial systems.

Complexity itself is not enough. A machine can have millions of components, nested regulations and chains of dependence without producing or renewing the conditions of its own organisation.

The difficulty therefore does not consist in drawing up an ever longer list of properties.

It consists in understanding how certain properties are articulated and in determining which belong to the base of living organisation, which are elaborations of it, and which can appear independently in other systems.

The proposal developed here adopts an organisational approach.

It does not look for a particular substance of living systems.

It looks for a structure of relations.

The starting point is the possibility that an organisation is distinguished by a network of processes taking part in the production, renewal or reactivation of the conditions required for their own continuation.

This proposal is:

organisational, because it bears on relations of dependence, production and renewal;

dispositional, because certain capacities can remain temporarily latent;

multidimensional, because self-production, robustness, recoverability, adaptivity, memory and evolution are not different expressions of a single property;

situated, because its discriminating power depends on the scale at which an organisational unit can be identified.

This theoretical choice is not neutral.

Other approaches place heredity, replication or Darwinian evolution at the foundation of life. They might regard evolvability as the base and organisational autonomy as a derived property.

The framework proposed here follows the reverse path.

It first questions the conditions of constitution and continuity of a present unit. Evolutionary history is then introduced as an additional dimension that makes it possible to explain the transformation of lineages.

This orientation has an important consequence.

A single organisation, with no biological ancestry and no descendants, could be considered alive if it effectively realised the organisational conditions retained.

That consequence will be assumed rather than deferred.

I. The organisational base

1. From complication to integration

An organisation can be extremely complicated without being alive.

A clock, a computer, an automated factory or an engine can have many components, complex chains of dependence and sophisticated mechanisms of regulation.

The number of parts is therefore not enough.

Nor is the density of interactions.

In a cell, the membrane, the chemical gradients, the metabolic reactions, the mechanisms of synthesis and the regulatory networks do not function as merely juxtaposed modules.

They depend on one another.

The membrane contributes to maintaining internal conditions required for certain reactions.

The metabolic reactions produce or transform molecules required for renewing the membrane.

The systems of synthesis depend on the energy and resources supplied by metabolism.

Metabolic functioning itself depends on structures produced by those systems.

The processes are therefore not merely connected.

They take part in producing or renewing the network on which they depend.

This idea converges with the notion of autopoiesis developed by Humberto Maturana and Francisco Varela.

In its demanding formulation, autopoiesis does not designate simply a system that contributes to its own maintenance.

It describes a network of processes whose transformations produce the components that regenerate that network and take part in constituting the unit within which those processes can continue.

The cell nevertheless remains open to exchange.

It absorbs matter.

It receives and transforms energy.

It releases products.

It depends continually on its environment.

Its closure is therefore not material.

It is organisational.

Living systems can be materially open while presenting a closure of the network of dependencies that constitutes their organisation.

This closure means neither absolute independence nor self-sufficiency.

An organism imports resources, depends on a milieu and may require the presence of other organisms.

Closure designates the fact that certain processes and constraints become mutually necessary to the continuation of a single unit and take part in renewing the conditions that make that continuation possible.

2. Constitutive circularity

This way of describing living systems contains a circularity.

A condition is said to be constitutive when it takes part in the continuation of the organisation.

The organisation is itself identified from the relations on which its continuity depends.

The notion of a unit therefore refers back to the processes that constitute it, while those processes are recognised from the role they play in that unit.

This circularity must not be masked.

It lies at the heart of the organisational approach.

It does not, however, necessarily constitute a begging of the question.

A definitional circularity would be vicious if two terms were explained solely by one another, with no possibility of independent observation.

The circularity studied here is first of all causal.

The membrane allows certain reactions.

Those reactions take part in renewing the membrane.

The mechanisms of synthesis require gradients.

The gradients depend on structures produced by the mechanisms of synthesis.

The system therefore has no single centre from which all the other functions would be commanded.

Its unity emerges from a network of dependencies and reciprocal productions.

The circularity nevertheless does not disappear when it becomes empirical.

The researcher must choose a scale, select variables, propose candidate boundaries and decide which fluxes will be studied.

Organisational closure is therefore never simply « read off » reality without mediation.

It is formulated as a hypothesis from a provisional carving.

That dimension does not mean that the unit would be arbitrarily invented by the observer.

Dependencies, production fluxes and the effects of perturbations are not created by the analysis.

But their identification depends on a framework.

The unit is therefore both constrained by the real organisation of the system and constructed as an object of study.

The comparative method does not dissolve that circularity.

It makes it explicit and attempts to control it.

Several delimitations must be compared.

A candidate boundary is strengthened when it explains better than the alternatives:

the concentration of internal dependencies;
the fluxes of production and renewal;
the distribution of exchanges with the environment;
the persistence of a unit through time;
the capacity to reconstitute certain relations after perturbation.

Closure thus constitutes a causal and organisational hypothesis, not a boundary given in advance.

3. Dependence, production and regeneration

A perturbation can reveal that a set is strongly coupled without demonstrating that it is self-producing.

In an engine, removing the oil pump can cause the degradation of several components.

A cooling failure can cause a general overheating.

Cutting the fuel supply stops the whole system.

The components are interdependent.

Effects propagate.

Yet the engine does not produce a new pump when one is removed.

It does not spontaneously rebuild the elements required to restore its organisation.

Reciprocal dependence is therefore a weaker property than closure.

It shows that several elements need one another.

It does not show that the network produces or renews the elements on which its continuation depends.

Three questions must be distinguished.

The first is:

Does the absence of an element affect the rest of the system?

This question tests dependence.

The second is:

Does the system normally take part in producing, transforming, renewing or maintaining that element?

This question tests organisational self-production.

The third is:

After a perturbation, can the system restore the element, its function or a functional equivalent from its own dynamics?

This question tests regeneration.

These three properties are not equivalent.

A system can be strongly interdependent without producing its own components.

It can renew several components while being very fragile.

It can preserve in latent form the organisation required to resume its processes without displaying substantial current self-production.

Closure must therefore not be confused with an unlimited capacity for repair.

A cell deprived of its entire genome need not be able to recreate it in order to be recognised as alive.

A decapitated organism need not regenerate its brain.

The requirement bears on the normal role of the constitutive processes and on the fluxes through which they take part in producing or renewing the organisation.

Regeneration after perturbation then describes an additional and graded capacity.

4. Self-assembly, autocatalytic coupling and closure

The term « closure » will henceforth be reserved for the strong sense.

It will designate neither every self-assembly nor every autocatalytic loop.

Three phenomena must be distinguished.

4.1 Self-assembly

Molecules can spontaneously form a structure through their physical properties.

Lipids placed in an aqueous medium can organise into bilayers or vesicles.

A boundary can thus appear without any internal network having produced, maintained or regulated it.

Self-assembly does not yet constitute a closure.

It describes the emergence of a structure from the properties of the components and of the medium.

4.2 Autocatalytic coupling

A chemical network can contain reactions whose products favour other reactions of the same network.

Some molecules then take part in producing other components that in turn support the functioning of the whole.

An autocatalytic coupling appears.

That coupling may be an important step towards autonomy.

It is nevertheless not enough to demonstrate the existence of an organisational unit.

The network may depend on local conditions it does not maintain.

It may have no individuation of its own.

It may not renew the constraints required for its continuity.

4.3 Constitutive closure

A closure appears when several processes and constraints mutually take part in producing or renewing the conditions that make their own continuation possible.

The system's boundary, where there is one, must not merely be present.

It must be functionally integrated into the network.

Its composition, its upkeep or its renewal depend at least in part on processes belonging to the system.

In return, that boundary contributes to the conditions required for those processes to continue.

The essential distinction is therefore not:

natural structure
contre
artificial structure.

It is rather:

a boundary produced mainly by the spontaneous properties of the medium

and

a boundary integrated into a network of production and renewal that takes part in the continuity of the unit.

A lipid vesicle can self-assemble without being alive.

It becomes a more serious candidate for a form of autonomy when an internal network transforms, recruits or renews components of its membrane, while that membrane maintains the conditions required for the network to function.

5. Operational individuation and the gradual production of boundaries

The status of the boundary must be made precise.

Every known living organisation has a form of individuation.

But that individuation does not always correspond to a simple, perfectly closed membrane.

It can take the form of:

a cell membrane;
a multicellular envelope;
a tissue interface;
a network of relations able to maintain a functional unit;
a dynamic boundary controlling exchanges and coordination.

An operational individuation is necessary in order to attribute processes to a system rather than to its environment.

It makes it possible to distinguish, at least provisionally:

internal fluxes;
external inputs;
the processes that produce or renew the organisation;
the conditions simply supplied by the milieu.

The existence of an operational individuation is therefore a requirement of the framework.

The degree of production of that boundary, by contrast, can be graded.

Part of the boundary may result from spontaneous phenomena.

Another part may be transformed, maintained or renewed by the network.

The question does not consist in requiring that every component be entirely manufactured from nothing.

Organisms use external precursors.

Autonomy is not material independence.

One must rather determine what share of the organisation of the boundary depends on the processes of the system and to what extent that boundary takes part in the continuity of those same processes.

The production of the boundary is graded.

Operational individuation remains necessary.

6. Active closure and conserved closure

Closure must not be understood solely as an activity observable at every instant.

Some living forms suspend a large part of their exchanges and processes.

A spore can remain inactive for a long period.

A seed can wait for favourable conditions.

Some organisms can enter states of cryptobiosis characterised by an extreme reduction of measurable activity.

Two situations must be distinguished.

An active closure corresponds to an organisation in which the constitutive processes currently produce or renew the components and constraints required for their continuation.

A conserved closure corresponds to an organisation whose activity is strongly reduced or suspended, but whose structures preserve the capacity to reactivate those processes within a specified class of compatible environments.

The capacity for reactivation is not a property independent of every environment.

A spore is not viable under any conditions whatever.

Conserved closure must therefore be understood as a relational disposition:

an organisation has a conserved closure relative to a defined set of conditions under which its constitutive processes could be reactivated.

That disposition belongs to the system through the state of its structures, but its actualisation depends on a relation with an environment.

Conserved closure must not be confused with recoverability.

It designates the persistence of a latent organisation.

Recoverability designates the range of perturbations from which an organisation can restore certain processes or relations.

A spore can display very low current productive activity while preserving a strong capacity for reactivation.

An active protocell can produce several of its components while being destroyed by a minimal perturbation.

These profiles are different.

Conserved closure and recoverability are therefore ontologically distinct.

They nevertheless share the same epistemic regime.

Both describe dispositions.

Their existence cannot always be observed directly.

Their assessment rests on:

test conditions defined in advance;
explicitly specified environments of reactivation or restoration;
a duration of observation;
measurable criteria of resumption;
acceptance that a failure reduces the plausibility of the disposition without demonstrating an impossibility under all conceivable conditions.

The conceptual separation therefore does not remove their common difficulty of testing.

II. Viability, robustness and recoverability

7. Viability as a domain of continuation

A living organisation cannot pursue its processes under any conditions whatever.

Some variations remain compatible with the continuation of its functioning.

Others reduce its possibilities.

Others destroy the relations required to pursue or reactivate its closure.

This dependence defines a domain of viability.

The domain of viability corresponds to the set of conditions under which the organisation can continue to produce, renew or reactivate its constitutive processes.

This notion implies no intention.

Living systems must not be described as a system that « wants » to survive.

They do not necessarily pursue a project.

Viability describes a relation between an organisation and the conditions under which its closure can be maintained or reactivated.

But every far-from-equilibrium structure depends on certain conditions.

A flame requires a fuel, a supply of oxygen and certain thermal conditions.

A cyclone depends on energy gradients and particular atmospheric conditions.

Dependence on limited conditions is therefore not enough to distinguish living systems.

The difference proposed here lies in the articulation between viability and closure.

In an organism, the conditions of continuation concern a network that itself takes part in producing or renewing the unit.

8. Robustness, regeneration and recoverability

Robustness, regeneration and recoverability must not be confused with closure.

Robustness designates the capacity of a system to continue functioning despite a perturbation.

Regeneration designates the restoration of a structure, a function or a coupling after an alteration.

Recoverability designates the set of states or trajectories from which a restoration remains possible under defined conditions.

A system can be robust without being strongly regenerative.

It can absorb a variation thanks to reserves or redundancies without rebuilding what has been lost.

A system can be barely robust yet able to restore its organisation after an interruption.

A fragile protocell could present a real closure and a low recoverability.

A spore could present little current activity, a conserved closure and a substantial capacity for reactivation.

A culture maintained in a chemostat could produce all its components while depending strongly on precise external conditions.

Living systems must therefore not be defined by a high value on every dimension.

They have an organisational profile.

That profile may include:

closure;
activity or latency;
robustness;
regeneration;
recoverability;
adaptivity;
memory;
evolvability.

These dimensions can reinforce one another without being identical.

9. The passage from latency to death

The distinction between conserved closure and death cannot be reduced to a universal thermodynamic threshold.

Biological death does not necessarily correspond to a single quantity of energy, a general level of entropy or a precise structure valid for all organisms.

It concerns the irreversible loss of the relations required for the organisation to resume.

But irreversibility depends on:

the type of organism;
its state;
the functions considered;
the environmental conditions;
the interventions admitted;
the temporal scale.

A cell can lose a function while preserving a viable organisation.

An organism can suffer substantial damage while maintaining certain capacities of repair.

A structure can look intact although mechanisms indispensable to any resumption have been destroyed.

The passage from latency to death will therefore not be defined here by a universal threshold.

It will be assessed operationally.

An organisation will be considered as still possessing a conserved closure when structural and functional markers justify the hypothesis that a resumption remains possible within a defined class of environments.

That hypothesis is strengthened by a successful reactivation.

It is weakened when:

the structures indispensable to resumption are destroyed;
attempts at reactivation fail under biologically justified conditions;
the required repair mechanisms are absent or irreversibly impaired;
no indicator of organisational restoration is observed during the announced period.

A successful reactivation constitutes a strong validation.

A failure does not demonstrate an impossibility under all imaginable conditions.

It reduces the plausibility of a conserved closure within the experimental framework adopted.

The boundary between dormancy and death can therefore remain uncertain in some cases.

That uncertainty is not removed by the concept.

It is recognised as a limit of empirical access to latent capacities.

III. Normativity and organisation

10. An asymmetry without teleology

The transformations an organisation encounters are not functionally equivalent.

Some preserve its possibilities of continuation.

Some reduce its margins.

Some make a new organisation possible.

Some destroy the processes required for any resumption.

That asymmetry constitutes here a minimal normativity.

It designates neither a moral value, nor an intention, nor a conscious purpose.

It expresses a relation between an organisation and the conditions that modify its possibilities of maintaining or reactivating its closure.

The term « endogenous » must be made precise.

It does not mean that the system invents a value.

It means that the functional consequences of a transformation follow from the constitutive relations of the system itself.

A change of temperature, of concentration or of membrane integrity can really affect the continuity of a cell independently of an observer's gaze.

But our access to that normativity depends on a framework of analysis.

The biologist chooses:

a unit;
a scale;
a temporal horizon;
indicators;
experimental conditions.

Two levels must therefore be distinguished.

Normativity is ontologically endogenous when the asymmetries studied follow from the real dependencies of the organisation.

Its study is epistemically indexed because their observation depends on a delimitation, a duration and chosen variables.

The measurement framework does not produce the norm.

It conditions the access we have to its different dimensions.

11. A multidimensional normativity

A transformation does not necessarily produce a globally favourable or unfavourable effect.

It can improve one capacity while reducing another.

An adaptation can increase thermal resistance but slow growth.

A stress response can preserve certain functions in the short term while increasing a long-term cost.

A modification can improve individual survival while reducing reproductive capacity.

An evolutionary innovation can increase performance in one environment and increase vulnerability in another.

Biological normativity must therefore not be assumed to be scalar.

It can be described as a set of dimensions.

For a system (S), at a time (t) and over a temporal horizon (H), a normative profile can be represented by:

[ _S(t,H) = ( C_1(t,H), C_2(t,H), , C_m(t,H), R_1(t,H), R_2(t,H), , R_n(t,H) ) ]

where:

(C_i(t,H)) represents a capacity of functioning or of continuation retained for the analysis;
(R_j(t,H)) represents a capacity of restoration or recovery;
(H) specifies the horizon over which the effects are assessed.

The components may correspond, depending on the system studied, to:

the integrity of a boundary;
the maintenance of a gradient;
energy production;
the synthesis of components;
the capacity for repair;
reproduction;
tolerance to a constraint;
the restoration of a function after perturbation.

A transformation can improve some components and degrade others.

There is not necessarily a single value making it possible to rank all trajectories.

Two states can be partly incomparable.

One state can dominate another when no dimension retained is degraded and at least one is improved.

But when a state improves some capacities and reduces others, any global comparison requires an additional weighting.

That weighting does not necessarily follow from the organisation itself.

It may depend on:

the scale chosen;
the temporal horizon;
the scientific question;
the level considered — individual, population or ecosystem;
sometimes an external judgement.

A distinction must therefore be made between:

the normative structure of the system, constituted by the real asymmetries of its possibilities;

and

the evaluative aggregation of the observer, who may choose how to compare several dimensions.

Living systems can produce their own organisational constraints without producing a universal metric able to rank all their trajectories.

12. Two axes of regulation

The distinction between an assigned variable and a constitutive condition is useful, but it does not form a simple opposition.

A target may have been defined by a designer while becoming necessary to the continuity of the apparatus that regulates it.

Two axes must be separated.

Axis 1: the origin of the target

The target may be:

specified by a designer or a user;
issuing from the internal history of the system;
modified by learning;
inherited from an evolutionary history;
produced by a combination of internal and external constraints.

Axis 2: the present role of the target

The regulated variable may be:

not constitutive of the continuity of the regulator;
directly constitutive of its functioning, its renewal or its reactivation.

An ordinary thermostat generally maintains a target specified from outside and not constitutive of its own continuity.

A learning system can produce certain internal targets without those being necessary to its material maintenance.

A medical device can regulate a condition vital for a patient while remaining external to that patient's own organisation.

An autonomous robot could maintain an energy reserve, produce certain parts, repair components and restore its functions according to objectives initially specified by its designers.

Its targets would be historically assigned but operationally constitutive of its continuation.

The fact that a target was designed is therefore not enough to exclude a present autonomy.

Organisms are themselves the product of a history they did not choose.

The design history of an artefact does not on its own constitute proof of permanent heteronomy.

The criterion bears on the present organisation.

One must examine whether the targets are effectively integrated into a network that produces, renews and regulates the conditions of its own continuity.

13. First- and second-order autonomy

The capacity to maintain constitutive targets is perhaps not the maximal degree of autonomy.

A system can preserve variables necessary to its continuity without being able to modify the rules that define that continuity.

Two levels can then be distinguished.

First-order autonomy corresponds to the capacity to regulate, maintain or restore already established constitutive conditions.

Second-order autonomy appears when a system can durably modify:

the variables it regulates;
the priorities among several constraints;
the mechanisms that produce its targets;
the conditions under which it reorganises its own functioning.

A robot able to maintain an energy reserve according to a programmed set point could present a first-order operational autonomy.

A system able to produce new constitutive targets in response to its history, without each having been explicitly foreseen, would present an additional level of organisational plasticity.

That capacity is not a necessary condition of all minimal life.

Many simple organisms have limited regulations.

It may nevertheless represent a higher degree of autonomy.

The capacity to modify the conditions of change must not be confused with the mere capacity to change.

IV. Adaptivity and organisational relevance

14. From asymmetry to relevance

Two levels must be distinguished.

At the first level, certain transformations modify the possibilities of continuation of an organisation.

That asymmetry exists even if the system has no mechanism able to detect the transformation or respond to it.

At the second level, the organisation becomes sensitive to certain variations.

It then modifies its dynamics relative to the effects those variations exert on its continuity.

It is at that level that an organisational relevance appears.

A state can be unfavourable to an organisation without the latter being able to detect it.

A resource can be available without the system being able to exploit it.

A perturbation can approach a limit without triggering a response.

Organisational relevance appears when differences bearing on the possibilities of continuation effectively influence the dynamics of the system.

The progression can be stated as follows:

a transformation modifies certain possibilities of continuation;
the system has mechanisms sensitive to part of that transformation;
that sensitivity modifies its responses;
the responses in turn affect the possibilities of maintenance, reactivation or transformation of the organisation.

Organisational relevance therefore constitutes an elaboration of minimal normativity.

It is not simply a synonym for it.

15. Present contribution and evolutionary function

The word « function » can designate different realities.

A structure can currently contribute to the continuity of an organisation without having been historically selected for that effect.

Two questions must therefore be distinguished.

Present organisational contribution

A process currently modifies the possibilities of continuation, regulation or reactivation of the system.

That contribution can be studied by perturbation.

A process is inhibited or removed and its consequences are observed.

Evolutionary function

A trait was preserved in the course of evolution because some of its effects contributed to the reproduction or persistence of earlier lineages.

That claim requires an evolutionary history.

The two can coincide without being identical.

A structure may have been selected for one function and then used for another.

A property can contribute to present viability without having been selected for that effect.

An artificial organisation with no evolutionary history can present constitutive contributions without possessing a biological function in the historical sense.

Organisational relevance therefore describes a present causal relation.

It does not on its own allow us to claim that a response was selected to produce that effect.

16. Adaptivity as a graded capacity

An adaptive organisation does not merely persist as long as conditions remain compatible with its functioning.

It can modify some of its dynamics when its state evolves towards a reduction of its possibilities of continuation.

Thermal regulation, metabolic responses, certain repair mechanisms, physiological plasticity and several behavioural adjustments can be described within this framework.

The organism does not need to represent a danger consciously.

It is enough that certain variations trigger modifications able to preserve, restore or transform certain conditions of functioning.

Adaptivity must not, however, be imposed as a universal condition of all minimal life.

It can be rudimentary.

It can vary across organisms.

It can be strongly reduced in certain phases.

It can also develop through several degrees.

A simple cell can modify the expression of certain genes or adjust its metabolism.

A complex organism can coordinate several levels of regulation, mobilise reserves, durably modify certain structures and learn from its experiences.

These differences describe degrees of elaboration of autonomy.

They do not on their own establish a boundary between living and non-living.

17. Inhabiting constraints

The expression « inhabiting constraints » does not mean that living systems always turn difficulties into advantages.

Some constraints are compensated.

Others are circumvented.

Some become resources.

Some impose a reorganisation.

Others exceed the capacities of the system and cause an irreversible disorganisation.

Inhabiting a constraint means that an organisation can integrate part of the effects of that constraint into its dynamics of regulation, compensation or transformation.

A muscle does not respond to a resistance in isolation.

Its adaptation depends on relations with the nervous system, the skeleton, the circulation, the hormonal mechanisms and energy metabolism.

The response belongs to an integrated organisation.

Living systems therefore do not merely undergo every constraint passively.

They can sometimes incorporate it into a new way of functioning.

That capacity remains limited.

It depends on:

the resources available;
the diversity of possible responses;
the speed of the perturbation;
the intensity of the constraint;
the time required for reorganisation.

Constraint does not automatically create an improvement.

It reveals the margins of transformation the system still has.

V. Trace, memory and learning

18. From the physical trace to functional memory

Every transformation potentially leaves traces.

A rock preserves the effects of a pressure.

A material can keep the mark of a deformation.

A shape-memory alloy can recover an earlier configuration under certain conditions.

A physical system can display hysteresis and depend on its history.

Preserving a trace is therefore not enough to define a biological memory.

A trace becomes a functional memory when it fulfils several conditions.

It persists beyond the event that produced it.

It can modify a later response.

That modification contributes, directly or indirectly, to the regulation, adaptation, reproduction or continuity of the organisation.

Functionality does not come from an intention.

It depends on the relation between the trace and the constitutive processes of the system.

A molecular modification becomes functional when it influences a future response linked to the continuation, restoration or transformation of organisational dynamics.

The trace can remain local.

It does not need to be interpreted by the whole organism.

An epigenetic modification can transform the expression of certain genes.

An immune cell can preserve the effects of an earlier exposure.

A neural network can durably modify the probability of a response.

A tissue can preserve a structural transformation.

Biological memory can therefore be distributed across several levels.

History becomes functional when a past transformation modifies the way the organisation responds to a later situation.

19. From memory to learning

An organisation learns when an experience leaves a modification durable enough to transform the way it responds later.

The sequence can be described as follows:

Transformation

→ Trace

→ Preservation

→ Reactivation

→ Modification of the response

→ Learning

Not every memory necessarily constitutes a learning.

Some traces take part in maintaining a structure without producing an identifiable adaptive modification.

But when an experience durably transforms a future response, memory becomes a component of learning.

This approach makes it possible to envisage minimal forms of learning without a brain.

Cells, unicellular organisms or biological networks can modify their responses according to their history.

It must nevertheless be avoided to extend automatically the concepts of representation or consciousness.

A functional modification does not necessarily imply an internal image of the world.

It does not presuppose a subjective experience.

Minimal learning can exist in the form of a history-dependent reorganisation.

20. Artificial systems and memory

Some artificial systems can record a history and modify their future responses.

An adaptive thermostat can learn certain thermal properties of a building.

A computer system can adjust its predictions.

An artificial network can modify its parameters from earlier data.

These systems satisfy certain dimensions of functional memory.

The distinction therefore cannot rest on the mere presence of learning.

An artificial organisation can learn without producing or renewing the conditions of its own continuity.

Conversely, an artificial system could become more autonomous.

It could use its history to:

regulate its conditions of functioning;
produce or replace certain components;
modify its architecture;
transform its priorities;
preserve the processes on which its closure depends.

In that case, its artificial origin would no longer be enough to distinguish it from living systems.

The framework would have to assess it by the same organisational criteria.

VI. Continuity and transformation

21. Causal continuity under the constraint of closure

Living systems do not preserve their unity by maintaining the same components.

Molecules are renewed.

Cells appear, transform and disappear.

Tissues are remodelled.

Internal relations evolve.

Identity therefore cannot rest on an immutable substance.

A causal continuity is necessary.

Present transformations must remain linked to the processes and structures issuing from the earlier organisation.

But causal continuity is not enough.

A present flame is causally linked to the flame that preceded it.

A growing crystal extends an earlier structure.

Erosion progressively transforms a landform.

The continuity of living systems must therefore be understood as a causal continuity under the constraint of closure.

The transformations must remain linked to an organisation whose constitutive dependencies continue to be produced, renewed or reactivated.

A caterpillar becomes a butterfly through a profound transformation.

Continuity does not rest on preserving every structure.

It rests on a trajectory in the course of which the processes issuing from the earlier organisation take part in building a new configuration able to pursue an organisational closure.

Causal continuity is therefore not an escape from constitutive circularity.

It describes its temporal dimension.

The present framework does not claim to resolve the metaphysical question of numerical identity.

It seeks to specify the conditions under which a transformation can be described as the organised continuation of a biological trajectory.

VII. Discriminating cases

22. What the simple cases show

A flame presents fluxes, a causal continuity and a dependence on certain conditions.

It shows no network able to produce or renew the constraints required for its individuation.

A thermostat regulates a variable.

It does not generally renew the components on which its own continuity depends.

A memory alloy preserves the effects of a past transformation.

That trace does not take part in the continuity of a self-produced organisation.

These cases show that dissipation, regulation and dependence on history are not enough when taken in isolation.

They are not, however, the most demanding tests of the framework.

The difficult cases are those that already possess part of the organisational properties claimed.

23. Autocatalytic sets, chemotons and protocells

Autocatalytic sets constitute a more demanding test.

Some reactions produce elements that favour other reactions of the network.

An autocatalytic coupling can appear.

The network then takes part in producing several of its own components.

But autocatalysis does not guarantee the existence of an organisational unit.

The system may remain dependent on local conditions it does not maintain.

It may have no individuation of its own.

It may not renew the constraints required for its continuity.

The analysis must therefore determine whether the network produces only molecules that sustain reactions, or whether it also takes part in maintaining the conditions that individuate it.

The chemoton model proposed by Tibor Gánti combines several coupled functions:

a metabolic network;
a membrane boundary;
a system of transmission.

Its interest lies in the integration of functions which, taken in isolation, are not enough.

Artificial protocells pose a comparable difficulty.

A self-assembling vesicle is not necessarily alive.

A structure becomes a more serious candidate when an internal network takes part in renewing its boundary and when that boundary maintains the conditions of the network.

The question of the threshold remains open.

Is a very weak participation of the network in maintaining the membrane enough?

Is an almost complete renewal required?

What share of the organisation may result from the spontaneous properties of the medium?

These questions must not be settled after observing the desired classification.

The rules must be defined before the assessment.

The framework must measure:

the share of the boundary's components produced, recruited or transformed by the network;
the dependence of the boundary on those processes;
the dependence of the network on the properties of the boundary;
the renewal of the coupling over time;
the capacity to re-establish that coupling after a comparable perturbation.

The boundary between self-organised chemistry, proto-life and minimal life could be gradual.

Nothing guarantees that a single natural threshold exists.

24. Viruses

Viruses constitute another limit case.

A virion can remain inactive for a long period.

When it meets a compatible cell, its genetic information can take part in producing new viruses.

The capacity for reactivation is therefore not enough to distinguish a spore from a virion.

The difference bears on the nature of the network reactivated.

A spore preserves structures issuing from its own cellular network.

When conditions become favourable again, it reactivates metabolic and constitutive processes from that conserved organisation.

An isolated virion does not generally reactivate an autonomous metabolic network.

It depends on the energy, the systems of synthesis and many functions of the host cell.

It does not alone restore the organisational conditions required to produce new virions.

But that distinction does not close the debate.

A virus can be considered as a cycle comprising several phases.

The virion is a phase of propagation.

The infected cell becomes the site of an organisation profoundly transformed by viral processes.

Depending on the scale retained, the relevant unit may be:

the virion;
the complete viral cycle;
the transient organisation formed within the host cell.

The framework therefore does not necessarily lead to declaring all viruses alive or non-living in absolute terms.

It specifies the reasons for their borderline status.

25. The erythrocyte as a test of necessity

The hypothesis that every living entity possesses a closure must be confronted with difficult biological cases.

The mature mammalian erythrocyte is a particularly interesting candidate.

It is generally considered a living cell.

It maintains a membrane.

It retains an energy metabolism.

It regulates ionic exchanges.

It has a limited lifespan within the organism.

But it no longer has a nucleus.

It does not renew its genome.

Its capacities for protein synthesis and for replacing many components are strongly reduced.

Its organisation depends on structures produced during the earlier phases of its differentiation.

The erythrocyte therefore poses a real difficulty.

Several interpretations are possible.

The first consists in holding that it still possesses a minimal closure through the active maintenance of certain metabolic and membrane relations.

The second consists in describing it as a terminally dependent living cell, an important part of whose closure has been inherited rather than renewed.

The third consists in recognising that complete organisational autonomy belongs more to the cell lineage and to the organism than to the isolated mature erythrocyte.

The framework must not select the interpretation that protects it.

The erythrocyte is a pre-registered case for the test of necessity.

If no defensible form of active or conserved closure can be identified at its scale, then the hypothesis that closure is necessary to every entity recognised as living will have to be revised or restricted.

That case plays, for necessity, the role that self-organised chemical systems play for sufficiency.

VIII. Scale locality and individuality

26. The main domain of application

The framework has its strongest discriminating power when the organisational unit is relatively identifiable.

The cell is the reference case.

It has an operational boundary.

The production fluxes can be mapped.

The relations between membrane, metabolism and synthesis can be studied.

The effects of perturbations can be measured.

The framework also applies to multicellular organisms, provided one recognises that several levels of closure are nested within them.

Its extension to supra-organismal entities remains more exploratory.

A mycorrhizal network links several organisms.

An obligate symbiosis can make several species mutually dependent.

A holobiont associates a host with microbial communities taking part in its functioning.

A clonal colony can integrate several specialised units.

Siphonophores have elements whose autonomy is strongly reduced in favour of the whole.

Ecosystems organise fluxes, preserve traces of their history and reorganise after perturbations.

But their individuation is often harder to establish.

The present framework is therefore primarily a theory of autonomy at the cellular and organismic scale.

Its extension to supra-organismal levels must not be assumed.

It must be demonstrated separately.

27. Nested closures and levels of autonomy

When several units have closures of their own, their coupling can produce a collective organisation.

One must then determine whether a new closure genuinely appears at the higher scale.

A mere exchange between units is not enough.

A strong interdependence is not necessarily enough either.

A collective unit becomes a candidate for a higher level of autonomy when:

certain constitutive functions can no longer be attributed to a single sub-unit;
the sub-units take part in producing or renewing conditions required for the continuity of the whole;
the whole has an operational individuation of its own;
certain perturbations are regulated or restored at the collective scale;
removing the coupling destroys constitutive functions that exist only at the higher level.

These criteria make it possible to distinguish:

an association;
an ecological dependence;
a strongly integrated symbiosis;
a collective organisation presenting an emergent autonomy.

When several closures are superimposed, it is not always necessary to choose a single « true » unit.

Several levels of autonomy can coexist.

Biological individuality can be nested.

The question then becomes:

At which scale is each capacity produced, renewed and regulated?

IX. Evolution and the status of living systems

28. A living being without a lineage?

The framework adopted here places organisation at the base.

Evolution is not used as the initial criterion of the living individual.

A sterile organism remains alive.

A cell that no longer reproduces immediately can preserve an organisation, a metabolism and capacities of regulation.

But that remark does not entirely resolve the problem.

Many theories regard membership of a lineage capable of heritable variation as constitutive of the category « living », even when each individual does not reproduce.

The organisational framework presented here takes a different commitment.

It detaches, at least in principle, the predicate living from necessary membership of a prior lineage.

A single specimen, with no biological ancestry and no descendants, could be considered alive if it effectively realised a closure and maintained the conditions of its continuity.

That consequence is assumed.

It opens the possibility of an artificial life that has not emerged from an evolutionary lineage.

It distinguishes two questions:

What is a living organisation?

and

What is an evolutionary lineage?

These questions are linked without being identical.

The organisational approach seeks to explain how a unit produces and renews the conditions of its present autonomy.

The evolutionary approach seeks to explain how lineages inherit, vary and transform across generations.

A complete theory of living systems must articulate these dimensions.

But their articulation must not mask the choice made by the present framework.

29. Evolution as transgenerational historicity

Evolution adds a dimension that individual organisation is not enough to explain.

Reproduction allows certain structures, information and developmental constraints to cross generations.

Heritable variations modify the possibilities of development, regulation and adaptation of descendants.

Several temporalities must therefore be distinguished.

In the short term, regulation modifies certain responses.

Over a lifetime, plasticity and learning durably transform the organisation.

Across generations, evolution modifies heritable structures and the possibilities of future variation.

These levels interact without being equivalent.

Individual history is not to be confused with evolutionary history.

A modification acquired during a lifetime is not automatically transmitted.

A present organisational contribution is not necessarily a selected function.

An organisational theory therefore does not replace evolutionary biology.

It describes another dimension of the problem.

X. Conditions of testing

30. Necessity and sufficiency

The framework can be assessed under two distinct hypotheses.

Hypothesis of necessity

Every organisation recognised as living must present, at a relevant scale, an active or conserved closure allowing the production, renewal or reactivation of the conditions required for its continuity.

That hypothesis would be put in difficulty if a recognised living system presented no organisation of that kind.

The mature erythrocyte is a pre-registered adverse case.

The framework will not be able to dismiss its difficulty simply by asserting that a closure must exist at another scale.

The change of scale will have to be justified by observable dependencies.

Hypothesis of sufficiency

An organisation presenting a closure, an operational individuation and a continuity of its own constitutes a minimal form of living system.

That hypothesis must not be protected by an automatic reclassification of counter-examples.

Non-living controls must be defined before the analysis.

The expected results and the conditions of refutation must also be announced.

If a reference non-living system satisfies all the pre-registered criteria, the framework will have to recognise that its conditions are insufficient to reproduce the boundary under study.

A revisionary extension of the concept of life could then be defended.

But that extension would be a new thesis.

It could not be counted as a confirmation of the initial test.

31. Pre-registered panel and predictions

The following panel is a starting proposal.

The predictions bear on the expected presence of the organisational dimensions before detailed analysis.

System	Operational individuation	Internal production of constitutive components	Renewal of the boundary	Expected closure	Expected status
Mineral chemical garden	Partial	No	No	Absent	Non-living
Reaction-diffusion system	Weak or contextual	No	No	Absent	Non-living
Autocatalytic network in a flow reactor	Depends on the reactor	Partial	No	Absent or incomplete	Non-living or proto-organisation
Passive lipid vesicle	Yes	No	No	Absent	Non-living
Strongly coupled, self-lubricating mechanical system	Yes	No or external	No	Absent	Non-living
Growing bacterium	Yes	Yes	Yes	Present	Living
Eukaryotic cell	Yes	Yes	Yes	Present	Living
Viable spore	Yes	Suspended	Suspended	Conserved	Latently living
Minimal synthetic cell	Yes	To be measured	To be measured	Test case	Undetermined before analysis
Mature erythrocyte	Yes	Weak and partial	Weak	Test case for necessity	Undetermined

Predictions are not enough.

A decision rule must be fixed before the analysis.

A closure will not be attributed on the mere presence of a dependence.

It will require converging evidence showing that:

several components or constraints required by the unit are produced, transformed or renewed by processes belonging to the system;
those components or constraints contribute in return to the continuation of those processes;
the individuation of the system makes it possible to attribute those fluxes to the unit rather than to an external infrastructure;
the relations are maintained over time or reactivable according to a pre-registered protocol.

The exact quantitative threshold will depend on the methods of measurement.

It will have to be fixed before the final comparison and applied without modification to all systems falling under the same protocol.

No threshold may be shifted after observation in order to preserve the expected classification.

Conclusion

The proposal developed here does not describe several independent properties placed on the same level.

It distinguishes an organisational base and several additional dimensions.

The base rests on an active or conserved closure.

In an active closure, certain processes currently take part in producing or renewing the components and constraints required for their own continuation.

In a conserved closure, activity may be strongly reduced, but the organisation required to reactivate those processes remains present relative to a defined class of compatible environments.

This formulation contains a constitutive circularity.

The processes define the unit they help to renew, while the unit makes it possible to identify the processes constitutive of it.

That circularity is not removed.

It is controlled by comparing several delimitations, studying the production fluxes and testing the dependencies.

Closure must not be confused with self-assembly, autocatalysis, robustness or recoverability.

A strongly coupled system can transmit the effects of a perturbation without producing the elements on which its continuity depends.

A system can be self-producing yet fragile.

A dormant organisation can present little current activity while preserving a possibility of reactivation.

A robust system can absorb certain perturbations without producing its own components.

These differences must remain separate.

Because an organisation depends on limited conditions, the transformations it encounters are not equivalent relative to its continuation.

That asymmetry constitutes a minimal normativity.

It designates neither an intention, nor an absolute value, nor a conscious purpose.

It follows from the constitutive relations of the system.

Our access to that normativity nevertheless depends on a scale, a temporal horizon and chosen indicators.

The norm is endogenous in its causal origin.

Its study remains epistemically indexed.

That normativity is multidimensional.

A transformation can preserve certain capacities while reducing others.

Living systems do not necessarily produce a single measure making it possible to rank all their trajectories.

Adaptivity appears when certain differences affecting the possibilities of continuation effectively modify the responses of the system.

Organisational relevance describes that present causal relation.

It must be distinguished from evolutionary function, which requires a history of selection.

Memory appears when past events leave traces able to modify future responses.

Learning appears when that dependence on history durably transforms the way of responding.

Evolution adds a transgenerational dimension when certain variations become heritable and modify the trajectories of lineages.

The framework accepts that a single organisation, with no ancestry and no descendants, could be considered alive if it effectively realised a closure.

That consequence distinguishes present autonomy from membership of an evolutionary lineage.

In this perspective, a minimal candidate for living status would be neither a structure that assembles spontaneously, nor a merely autocatalytic network, nor a system whose components are only interdependent.

It would be a materially open organisation in which:

certain processes produce, transform or renew part of the components and constraints required for their continuation;
an operational individuation makes it possible to distinguish that organisation from its environment;
that individuation maintains conditions required for the network to function;
the network takes part, in return, in producing, maintaining or renewing the conditions of that individuation;
the constitutive relations can remain active or be conserved in a latent form;
robustness, regeneration, recoverability, adaptation and memory can vary along distinct dimensions.

This characterisation does not guarantee that a single boundary exists between the inert, proto-life and living systems.

It proposes a framework for studying degrees and forms of autonomy without reducing life to an accumulation of complexity.

Living systems do not persist by remaining identical.

They preserve a continuity when their transformations continue to produce, renew or reactivate the conditions that still make their own transformation possible.

Methodological appendix

Programme for testing the organisational framework

A.1. Purpose of the appendix

This appendix is not a proof of the framework.

It describes the minimal conditions of an experimental programme able to put it to the test.

It aims to avoid several confusions:

dependence and self-production;
compensation and regeneration;
activity and latency;
robustness and recoverability;
expected classification and a decision revised after observation.

A.2. Identifying the candidate units

The analysis begins with several possible delimitations.

No boundary is to be taken as established before the study.

For each candidate unit, the following must be specified:

the spatial scale;
the temporal scale;
the components included;
the fluxes crossing the boundary;
the processes assumed to be internal;
the functions ensured by the environment.

The delimitations will be compared by their capacity to explain:

the production fluxes;
the reciprocal dependencies;
the renewal of components;
temporal continuity;
restoration after perturbation.

A.3. Mapping the production fluxes

For each component assumed to be constitutive, the following must be determined:

its material origin;
the precursors imported;
the transformations carried out by the system;
its rate of renewal;
the processes required for its production;
the functions to which it contributes.

A dependence will not be counted as a production.

The fact that a component is necessary to the system does not demonstrate that the system takes part in renewing it.

A.4. Comparable perturbations

Perturbations must be defined before the experiment.

A comparison based on the same quantity of matter removed is not necessarily valid.

Commensurability will be primarily functional.

Perturbations may be normalised by:

the reduction of a fraction of a flux;
the inhibition of a proportion of the initial activity;
the loss of a fraction of functional surface;
the decrease of a measurable capacity;
the controlled removal of part of a distributed component.

The rule must be applied identically to the candidate and to the controls falling under the same protocol.

A.5. Measuring regeneration

Let:

(F_0) be the functional value before perturbation;
(F_{}(p)) the value immediately after a normalised perturbation of magnitude (p);
(F(p,t)) the value measured at time (t).

The fraction of function restored can be described by:

[ G(p,t) = ]

where:

[ F_0 F_{}(p) ]

The general interpretation is as follows:

[ G(p,t)=0 ]

indicates that no restoration of the lost function has been observed.

[ 0<G(p,t)<1 ]

indicates a partial restoration.

[ G(p,t)=1 ]

indicates a return to the initial functional level.

[ G(p,t)>1 ]

indicates a level beyond the initial one, which must be reported as an overcompensation and not artificially reduced to one.

This function does not by itself measure a closure.

It must be interpreted together with:

the normal production fluxes;
the spontaneous renewal of components;
the internal or external origin of the restoration;
the possible reconstruction of the organisational coupling.

A.6. Intrinsic and assisted recoverability

Intrinsic recoverability corresponds to a restoration produced by the dynamics of the system within a compatible environment.

Assisted recoverability involves an external intervention that:

replaces a function;
adds a structure;
directly repairs a component;
supplies a mechanism normally produced by the system.

The two must be reported separately.

An external reconstruction does not demonstrate an internal capacity for recovery.

A.7. Testing conserved closure

An inactive organisation will be assessed relative to a class of reactivation environments defined before the experiment.

The protocol will have to specify:

the physical conditions;
the resources available;
the signals of reactivation;
the duration of observation;
the expected markers;
the criteria of organisational resumption.

The mere resumption of an isolated signal will not necessarily suffice.

Reactivation will have to concern several processes linked to the resumption of closure.

The failure of a trial will reduce the plausibility of a conserved closure under the conditions tested.

It will not demonstrate an impossibility under every conceivable condition.

A.8. Multidimensional assessment

The normative profile will not be reduced to a single score without further justification.

For a system (S), at a time (t) and over a horizon (H), the following will be used:

[ _S(t,H) = ( C_1, C_2, , C_m, R_1, R_2, , R_n ) ]

The components will have to be defined before the analysis.

A global weighting will be applied only if its justification is explicitly distinct from the properties observed.

The absence of a global score is not a shortcoming when the dimensions are genuinely incompatible.

It may reflect the multidimensional structure of the organisation.

A.9. Criteria for levels of autonomy

Results may be described in terms of several levels.

Level 0: physical self-organisation

The system produces a form or a pattern without producing the conditions of its own continuity.

Level 1: autocatalytic coupling

The network takes part in producing some of its components without sufficient organisational individuation.

Level 2: minimal closure

The network takes part in producing or renewing constitutive components and in maintaining an operational individuation.

Level 3: regulated autonomy

The system modifies certain dynamics relative to its conditions of continuation.

Level 4: historical autonomy

Traces of the past durably modify future responses.

Level 5: second-order autonomy

The system modifies certain rules, priorities or targets that organise its own transformations.

These levels do not necessarily constitute a linear scale.

A system can present an uneven profile.

A.10. Decision rule and pre-registration

Before the final analysis, the protocol will have to announce:

the controls;
the predictions;
the variables;
the instrumental thresholds;
the minimal effects retained;
the durations;
the rules of perturbation;
the criteria of closure.

A closure will not be attributed on the mere presence of a dependence.

It will require converging evidence showing that:

several components or constraints required by the unit are produced, transformed or renewed by the system;
those components contribute in return to the continuation of the network;
an operational individuation makes it possible to attribute the fluxes to the unit under study;
the constitutive relations are maintained over time or reactivable under the announced conditions.

The quantitative thresholds will have to be fixed before the final comparison.

They may depend on the methods of measurement.

They may not be shifted after observation in order to preserve the expected classification.

A.11. Conditions of revision

The framework will have to be revised if:

a recognised living system presents no defensible closure at an empirically justified scale;
a pre-registered non-living system satisfies all the criteria without an essential function having been externalised outside the unit under study;
the distinction between self-assembly and closure cannot be operationalised without case-by-case adjustment;
the thresholds required for classification change systematically according to the expected result;
the selection of the scale entirely determines the verdict with no criterion allowing the delimitations to be compared.

A modification of the notion of life remains possible.

But a redefinition of the concept must not be presented as the confirmation of the test it replaces.

A.12. Scope of the programme

The protocol does not claim to provide an immediate universal measure of life.

It seeks to turn theoretical notions into comparable empirical questions.

Its objective is to determine:

what the system produces;
what it receives;
what it renews;
what it can restore;
what it preserves of its history;
at which scale those capacities form a unit.

The question therefore no longer becomes only:

Is this system alive?

It also becomes:

What form of autonomy does it realise, at which scale, through which processes and within which limits?