How biological codes break causal chains to enable autonomy for organisms

Autonomy, meaning freedom from exogenous control, requires independence of both constitution and cyber-netic regulation. Here, the necessity of biological codes to achieve both is explained, assuming that Aristotelian efficient cause is ‘formal cause empowered by physical force’. Constitutive independence requires closure to efficient causation (in the Rosen sense); cybernetic independence requires transformation of cause–effect into signal-response relations at the organism boundary; the combination of both kinds of independence enables adaptation and evolution. Codes and cyphers translate information from one form of physical embodiment (domain) to another. Because information can only contribute as formal cause to efficient cause within the domain of its embodiment, translation can extend or restrict the range over which information is effective. Closure to efficient causation requires internalised information to be isolated from the cycle of efficient causes that it informs: e.g. Von Neumann self-replicator requires a (template) source of information that is causally isolated from the physical replication system. Life operationalises this isolation with the genetic code translating from the (isolated) domain of codons to that of protein interactions. Separately, cybernetic freedom is achieved at the cell boundary because transducers, which embody molecular coding, translate exogenous information into a domain where it no longer has the power of efficient cause. Information, not efficient cause, passes through the boundary to serve as stimulus for an internally generated response. Coding further extends freedom by enabling historically accumulated information to be selectively transformed into efficient cause under internal control, leaving it otherwise stored inactive. Code-based translation thus enables selective causal isolation, controlling the flow from cause to effect. Genetic code, cell-signalling codes and, in eukaryotes, the histone code, signal sequence based protein sorting and other code-dependent processes all regulate and separate causal chains. The existence of life can be seen as an expression of the power of molecular codes to selectively isolate and thereby organise causal relations among molecular interactions to form an organism.


Introduction
Code biology tells us that ''the cell is a system that is capable of creating and conserving its own codes'' (Barbieri, 2012(Barbieri, , 2015)), with the implication that codes are essential and unique to life.Using an informational approach, this paper aims to reveal the essential roles that codes play in enabling living systems to act as cognitive autopoietic systems, thereby possessing the defining characteristic of life according to the organisational approach to biology (Mossio et al., 2009).The identity, autonomy and agency of biological systems are so special that they have led some physicists to claim that new laws of physics are needed to explain them (e.g.Walker and Davies, 2013;Walker, 2017), prompting a response that showed how recasting the process of living as information processing enabled these properties to be explained in established terms of physics (Farnsworth et al., 2013;Farnsworth, 2022).Two, in particular, of the five characteristics of organisms identified by Kauffman and Clayton (2006) are both necessary and dependent on codes.Firstly, organisms must perform at least minimal cognition (Bich and Moreno, 2016;Bitbol and Luisi, 2004), E-mail address: k.farnsworth@qub.ac.uk.
given a clear internal/external demarcation established by a selective boundary (see e.g.Luisi, 2003).Secondly they need evolvable selfreplication as defined in principle by Von Neumann and Burks's (1966) theory of self reproducing automata, recently described in biological terms by the ( , ) system of Hofmeyr (2021) and the interpretation of Rosen's (1973) (, ) system by Vega (2023).Here, codes will be identified as crucial for (a) isolating information from efficient causes at the organism boundary, transforming cause and effect into signal and response and (b) isolating template information as purely formal cause, protecting it from internal efficient causes and thereby enabling reproduction.Taken together, these facilities enable organism autonomy.

Codes match domains of information embodiment
Codes are strictly informational, entirely defined in terms of, and dependent upon, the communication of information.At first sight https://doi.org/10.1016/j.biosystems.2023.105013Received 29 June 2023; Received in revised form 25 August 2023; Accepted 26 August 2023 codes are derivative artefacts of communicating systems, rather than fundamental to physical reality.But a developing theory of physical reality inspired by Pattee's (2001) general/particular dichotomy identifies information as the particular configuration (out of all possible) of particles of matter in space and time and claims that by setting the coordinates from which physical forces originate (all emanating from fundamental particles), information constrains all actions within the universe (Farnsworth, 2022).By this, physical information, i.e. embodied as a spatiotemporal pattern of matter, directs the unfolding of the universe, gives objects their existence and constrains forces to enable physical work to be done.Things happen this way or that because of the relative positions of all the particles that interact at time  to cause the ensuing dynamics.If we consider the set of positions of all labelled particles (coordinates for the ith particle   () ∈ ()), then () embodies information and is particularised by that information and it is that information which constrains the unfolding dynamics to be what it is, rather than any other.Given this view, information is a fundamental aspect of physical reality.The communication of information is the (at least partial) transfer of a particular pattern from one set of particle coordinates () to another  ′ () (or from one pattern of a forcefield to another) and this is equivalent to correlation being established between the two sets (once communicated, () and  ′ () share mutual information).It is this transfer of information from one set to another that constitutes the macroscopic phenomenon of efficient causation (Farnsworth, 2022).Physically, this correlation arises from the mutual interactions among force fields surrounding and emanating from the particles (i.e. the vector summation of forcefields) and is therefore itself constrained by physical compatibility: the medium in which  ′ () is embodied must match that of (), else there can be no mutual interaction.For example a pattern in the concentration of molecules in a fluid, which constitutes a sound wave cannot transfer into a pattern of electrical charges in a silicon chip, without several intervening steps that perform the role of transducers to match patterns from one medium to another.It is here that cyphers and codes play their part.
Information, as pattern, can be embodied in a wide range of media, e.g.electrical and magnetic fields, pressure variation in a fluid, colour pattern, or the shape of a solid, or of a molecule such as a catalytic protein.Nature is arranged as a modular hierarchy so that higher level patterns are formed among lower level patterns, ranging from the distribution of galaxies in the universe, down to the probability distribution of elementary particles (Hazen, 2009).To that extent, information is embodied at different spatiotemporal scales and in different 'bases' (I will use the term 'basis' to avoid confusion with nucleotide base), each defined by the pattern (i.e.motif) that provides the elemental unit for forming a higher level pattern.For example the information embodied by the distribution of trees in a forest has the individual tree as its basis and writing has the individual character as its basis.In analogue electrical engineering, especially radio communication, the modulation pattern of a signal is formed in its carrier wave acting as the basis and in digital communications the binary bit is the basis, but predefined 'words' of 2  bits are motifs forming higher level bases.The most relevant examples are those of DNA and RNA where the basis is the codon and polypeptide where the basis is the amino acid.As linear sequences, RNA molecules can embody information through higher level patterns (sequence motifs) as well; most obviously the gene, but also the host of regulatory motifs, found from multiple sequence alignments.Further, RNA motifs form the syntax for a 'language' of secondary (and presumably, higher) structures (Komatsu et al., 2020).These structural motifs demonstrate information embodied in molecular shape (a concept easily generalised to proteins and small molecules acting as e.g.ligands) and represent a bridge between the purely informational purpose of nucleic acids (formal cause) and regulatory, or even enzymatic, functions (efficient cause).The set of structures into which RNAs and polypeptides can fold into secondary, tertiary and quaternary patterns, provides a set of bases for information embodiment that is equivalent to a set of component parts for a machine.In a language of physical forms, the semantic meaning is the formation of functional wholes, composed of the component parts.Putting together an engine is making a meaningful statement in this language; autopoiesis of a cell is both a physical fabrication and assembly (Hofmeyr, 2021) and the construction of a meaningful statement in the language of molecular form motifs.
Scale is more obvious: pattern matching (correlation) necessarily involves matching of spatiotemporal scale (which is why analogue radios have to be tuned).The ribosome is 'tuned' to the nucleotide scale, so it does not see variation at a larger scale and molecular variation at a smaller scale result in reading errors.This applies to phase as well: In the matching of RNA codons with tRNA anti-codons, the initiation factors are needed for the ribosome to achieve precise alignment, just as in cell signalling, receptor sites have to be precisely matched to ligand shapes before efficient cause can be enacted.Failure in this precision is equivalent to a mismatch of teeth in the cogs of a car's gear box, leading to failure in function and possible damage to the system.Translation to polypeptide in the ribosome includes elaborate mechanisms for coping with mutant tRNAs that may cause e.g.frameshift errors, which we are only beginning to fully understand (see e.g.Hong et al., 2018).
Finally, for correlation of pattern to transmit information or perform efficient cause, the media of embodiment must match, which is trivial if the medium does not change, but where it does (e.g. from sound to axon voltage pulses), the process must depend on one of relatively few physical phenomena having effects in both media.Mechanical forces obviously can generate variation in molecular concentrations of fluids, charged particles create electromagnetic waves when accelerating and we have various opto-electrical, mechano-electrical (e.g.piezoelectric) processes and quantum effects, notably the conformation change of opsin receptors in response to photon excitation (e.g.Ikuta et al., 2020) and those associated with mechanotransduction in cochlea hair cells (McPherson, 2018).
Each combination of medium, scale and basis constitutes a domain for information embodiment, e.g. the frequency and modulation mode for a radio transmission, the codon for DNA/RNA embodied information, the individual components of an iron bridge (where the information is embodied in the component forms and the rules of their assembly) and the electric field near the active site of an enzyme, where the basis is its shape at the atomic scale (Table 1).In this sense, a domain is the combination of medium, basis and scale in which information is embodied.Medium is the physical substrate, basis is the elemental unit of pattern and scale is the spatiotemporal extent of the basis.

Information domain as 'language'
This idea of domain in communications leads naturally to the term 'language', here used in the general sense of code theory.For this, we define both code and cypher as a mapping from one domain  to another  (the codomain in set theory language, or target in category theory) so that  ∶  → , where  is a subset of all the possible ordered pairs  ⊆  ×  that includes only those {, }, such that  =  (), where  ∈  and  ∈ .Then  and  are each languages with a set of symbols (each of  ∈  and each of  ∈ ) and we need the code  to translate between them.Having just argued that correlation of formal causes cannot occur across domains, it follows that if a formal cause in one domain is to effectively combine with another in a different domain, one of the two must first be transformed using the mapping  ∶  → , i.e. it needs translation.The difference between code and cypher is that in  ∶  → , for the cypher  is a single function, but it is a set of pair-wise elemental transformations   ∶   →   for the code.To effect code transformation, the set { } of pairwise transformations must be embodied as information in the translational apparatus: the transducer, or in case of adaptor molecules, each is a member of { }.The obvious and important corollary of this is that a formal cause in  is not effective in  and vice-versa.This is why (within reasonable force limits) communication channels at one frequency are unaffected by those at another frequency; electrical charge patterns in a silicon chip are unaffected by movement of the whole chip in space (e.g. on an engine piston) and the sequence of nucleotides in DNA is not altered by meeting amino acid sequences.1That is true unless there is a translation system between the two domains: a signal converter such as the heterodyne receiver circuit for radio communications; a displacement-transducer for the engine, and the tRNA/aaRS/ribosome complex for genetic translation, respectively.This prohibition can be used to advantage whenever a system requires a formal cause to remain ineffective, unless and until some particular conditions are met.Such a system only needs to embody the formal cause in a domain that is different from its other (physically involved) components, until its effect is needed, whereupon translation using transducers (adaptor molecules or some other physical arrangement) is deployed.That, of course, is what the genetic code and many cell-signalling pathways do, as and when required.
It is no coincidence that the mapping  ∶  →  used for code translation is the same as that used for efficient cause.Indeed, the action of the transducer, translator or transcriptor is a specialised kind of efficient causation.Crucially it is a kind that links one domain of formal cause embodiment with another.This special kind of efficient cause consists of an empowered formal cause, specifically a configuration of matter, which is at least in part a member of both domain  and .That is most clear in the case of a tRNA which has one part matching the domain of amino acids and another matching that of codons.A microphone, which transforms pressure fluctuations into electrical, is an assembly of material (a form) having a causal link to both the kinetic and electrical physical domains.Chemoreceptors in a cell's outer membrane can be interpreted as translators from an external molecular 'language' to the internal language of the cell.Such domain translators can, in principle, be turned on and off, opening and closing the link between domains, that way providing a potentially controllable link in the chain of cause and effect (see Section 5.1).Notably, this controllability is created by the organism itself and is entirely a part of the organism, independent of exogenous causation.For codes, the organism is needed to physically make and maintain the translator, determine its behaviour and give meaning to its output and in principle this is true for all semiotic translation processes (Kull, 1998(Kull, , 2020)).Semiotic systems are strictly information-based (formal cause), but unlike the formal constraints that are translated into a domain where they can directly enact efficient cause by constraining forces, semiotic information can only influence systems by providing information that correlates with internal control systems (Section 5.2).
This leads to identifying information domains with different classes of causality.Semiotic causality has only indirect 'causal power' enabled by pre-determined response within a receiver to a particular signal from the transmitter.Cybernetic causality arises when information directly affects other information by computation (a network of logical relations implemented as correlations within a common domain).Formal constraint occurs when information as pattern in the distribution of matter constrains, hence specifies, the direction and strength of physical forces.Formal constraint can act at the macro scale, e.g. in mechanical devices such as the Jacquard loom, or at the molecular scale e.g. in the operation of biological molecular machines, such as the ATP-synthase motor (see analysis in Farnsworth, 2022).Genetic information in DNA, transcribed into RNA operates through formal constraint (real physical cause), though it may be interpreted as semiotic (e.g.Deacon, 2021;Rohr, 2014).
In eukaryotes, additional information processing codes have greatly extended the range of computation available to the cell, particularly in the branching of possibilities for replication, leading to cell differentiation.The histone code (Jenuwein and Allis, 2001;Kühn and Hofmeyr, 2014), splicing code (Baralle and Baralle, 2018) and RNA polymerase II CTD code (Dieci, 2021) all translate from one domain to another using adaptor proteins and interact through logic operations.Adaptor proteins match the definition of transducers (see below) and do pass information from one domain to another, but in epigenetic control codes (e.g.control of transcription), their causality (Table 1) is strictly cybernetic.The fact that they are genuine codes (arbitrary pairwise mappings) ensures the causal isolation of their domain and codomain, enabling the separation and constraint necessary for information processing independent of physical, especially thermodynamic imperatives: these codes are the foundation of 'natural computation'.

Codes are embodied by transducers
With the exception of dissipating heat, energy is transformed from one form to another by an engine; obviously the combustion engine, but also e.g. the dynamo and the chloroplast are engines.By analogy, information (which must be physically embodied) can only be transformed from one domain into another by means of a transducer.Transducers necessarily involve either a cypher or a code and include adaptor molecules and coupled receptors (e.g .G-protein or enzyme-linked), as well as sensory cells such as those found in the cochlea and eye (see e.g.Stieve, 1983, for general review).The two domains may differ in medium, basis or scale, or any combination of these.The translation complex for transforming RNA information into a polypeptide chain exemplifies a transducer matching one basis to another.A microphone translates from one medium to another and heterodyne circuit from one scale (radio frequency) to another.An olfactory cell translates all three: medium, basis and scale from molecular shape to neuronal axon pulse.In information terms, most transducers perform a reversible linear transformation (i.e. it applies equally over the domain) and so can be represented as a cypher.This sort of transducer embodies little constraining information: just one mapping rule that applies equally to all patterns in the basis.These cypher transducers are essentially physical: their transformation is a causally inevitable force-mediated process, exemplified by the vibrating needle leaving its impression in hot vinyl to make a record.But by their structure, they can very substantially change the magnitude of forces acting in the receiving domain relative to the transmitting domain -a capability much exploited by organisms.
In contrast, code transducers embody particular and arbitrary constraints on the relation between one domain of information embodiment and another.One of the defining characteristics of code, as opposed to cypher, is that the set of transformations { } is arbitrary (Barbieri, 2012(Barbieri, , 2015(Barbieri, , 2019)), not inevitable, not thermodynamically determined, but established by selection (natural or otherwise) and the reinforcement wrought from repetition, both of which imply function (Farnsworth et al., 2017;Hazen et al., 2007;Mossio et al., 2009) and with it, teleonomy (Mayr, 1974), or even teleology (Mossio and Bich, 2017), especially of communications within biological systems (Frick et al., 2019).All of these characteristics are unique to life (plus human technology), so we can say that codes exist only because life utilises them and extant life could not exist without them.This mutual, reflexive dependence is consistent with the general organisational structure of living systems, summarised in Kauffman's (1986Kauffman's ( , 2006) ) concept of a 'Kantian whole', described in detail by models such as Hofmeyr's (2021) ( , ) system and Vega's (2023) analysis of Rosen's (1973) (, ) system and the chemoton of Gánti (1979Gánti ( , 2003Gánti ( , (translation from 1971))).All of those systems include template information, as opposed to the implicitly embodied information found in the vessicles of Segré et al. (2000Segré et al. ( , 2001) ) and Kahana and Lancet (2021), the minimal autopoetic micelles of Luisi and Varela (1989) (see also Bitbol and Luisi, 2004;Luisi, 2003) and Deacon's (2006) autocell, all of which do not.In other words, life requires both the 'function' and 'symbol' sides of Pattee's (2001) epistemic divide and these are 'bridged' by codes.Codes require the preservation of embodied information, both the code book { } and the symbols, during translation and this combination is summarised in the word 'template': a universal characteristic of living systems (see Section 4).
Coding is essentially an informational process in which physical forces play no necessary part.As such, coding transducers such as the first and second messenger pathways of cellular signalling (see e.g.Capra and Laub, 2012), transfer information independently of the physical forces which brought it.That is possible because efficient cause arises from the particular (informational) constraint of physical force and equivalently, efficient cause is the force-empowered action of embodied information (Farnsworth, 2022 -see Section 3 below).A coding transducer, embedded in a physical boundary (such as a cell membrane) maps the information component of an efficient cause into a different domain, leaving the force to dissipate, usually in temporary and reversible distortion of the transducer (and/or the physical boundary in which it is embedded).A cell signalling transducer is matched to the external domain by its receptor site and to the internal cellular domain by, e.g. its second messenger: in this, the force of ligand concentration is only sampled as one ligand binds to the receptor.The force magnitude in this case is proportional to the concentration of ligands, which is easier to conceptualise if they are ions, for which osmotic pressure is familiar.This phenomenon is clearer in the case of a mechanical pressure transducer which responds to an externally applied physical force with displacement or strain in a component that belongs to the exterior domain.This displacement creates a relatively tiny change in e.g.electrical charge distribution in the internal domain at the other end of the transducer, thereby stripping the information (embodied as displacement) from the force.For example, in a cochlea hair cell, displacement is transformed to a change in K + gradient across the cell membrane (McPherson, 2018).There is no necessary relationship between the change in the external domain and that in the internal domain.The link between physical displacement in a cochlea hair cell and a change of electrolyte gradient is arbitrarily embodied in the hair cell's structure and function and is contingent upon the cell's current state.The physical transducer effectively transforms physical cause into signal by changing the medium in which information is propagated, leaving force behind.In the case of coding transducers, even the information mapping is contingent.Coding transducers make a break in the otherwise necessary causal link between two parts of a physical system, since by this contingency, they permit causal branching and blocking (Ellis and Kopel, 2019;Farnsworth, 2018) (see Section 3.1).

Biological autonomy
At the simplest level, autonomy is the degree to which behaviour is under self-control, hence independent of exogenous control.It is the extent to which we can attribute the cause of any action of an agent to that agent, but is also the freedom of that agent to act in more than one way because without that freedom, its behaviour is fixed and can be entirely attributed to its structure and organisation.For any agent, we can ask (a) what is it able to do, (b) what will it do in specific circumstances and (c) what determines that it will do that?The first question is answered mainly by its constitution -the material and organisational composition which determines its range of capabilities.The second question concerns the scope of possible responses to different circumstances, ranging from there being only one thing it can do (e.g. a clock) to an indeterminate (perhaps infinite) set of responses.The last question directly addresses the degree of internal as opposed to external control.
Emphasising that autonomy is not absolute, but rather a continuum, Froese et al. (2007) identified behavioural and constitutive autonomy as two distinct dimensions of variation.Since behavioural autonomy concerns the degree of self-control for a system, I term it cybernetic independence and use constitutive independence for the degree to which a system is responsible for its own constitution.By definition, autopoietic systems are self-constituted, but because organisms follow an inherited template in their self-constitution, they create only partial constitutive independence to contribute to their autonomy.That is, the individual organism has capabilities for which it is not entirely responsible, because they were determined by its inheritance (indeed, autopoiesis more properly refers to life as a whole, rather than any individual organism).Boden (2008) defined autonomy using all three of the questions above: (a) the extent to which responses to the environment are mediated by inner mechanisms, (b) the extent to which the (inner) controlling mechanisms are self-generated rather than provided by an exogenous source and (c) the extent to which these inner mechanisms can be appraised and modified from within, to better suit the objectives of the system in the light of the current situation -i.e. are adaptable.The last of these implies a third axes of freedom for a cybernetic system: there must be more than one possible action and there must be at least one level of meta-control to determine which action is taken.In general every additional level of meta-control adds degrees of (controlled) freedom to the cybernetic system, so belongs in the same conceptual category as the range of possible actions and I term this concept cybernetic freedom (the scope for choice).
Autonomy therefore consists of (a) freedom from direct exogenous control, (b) self-constitution (autopoiesis) of the mechanisms of internal control and (c) adaptation of those mechanisms to suit a changing environment.These are attributes of all living cells and of every living organism and they are conceptually distinct.It is possible to have any one or two of them without the other.Each of them can be quantified by a ratio: (a) internal/external control, (b) endogenous/exogenous constitution and (c) flexibility/fixation (which can be quantified by degrees of freedom).This enables autonomy to be quantified by a position in a three dimensional space, where each of these ratios is an axis (Fig. 1).For illustration, a star is self-made to a limited extent, but has no cybernetic control, so is placed at zero on the two cybernetic axes.Automata and computers are entirely dependent constitutively, but have some independence of cybernetic operation and they range in degrees of freedom from very little (the automaton) to large (the universal computer).Organisms being autopoietic systems have significant Fig. 2. The hierarchical construction of cybernetic freedom from nesting cybernetic set-points, i.e. a nested hierarchy of homeostatic control systems.The set point is the elementary unit of cybernetic independence and is only made possible through the circular causal structure of constitutive independence which enables the system to maintain independent information (Farnsworth, 2017).
constitutive independence and being self-regulated have very considerable cybernetic independence, but they vary in cybernetic freedom from rather little (the microbe) to the great flexibility of behaviour seen in e.g.octopus.
Although these axes describe independent metrics of autonomy, there is an organisational dependence: cybernetic freedom implicitly builds upon cybernetic independence and cybernetic independence is only meaningful and possible for a system having a distinct identity and this in turn implicitly builds upon constitutive independence (Fig. 2).The reason is that for a system to have cybernetic independence, it must have at least one 'goal' to enact goal-directed behaviour (most elementally, a homeostatic set-point) and this necessarily implies a purpose and that implies a self-referencing identity for the system (Bich et al., 2016;Kauffman and Clayton, 2006;Miquel and Hwang, 2016;Weber and Varela, 2002;Farnsworth, 2018Farnsworth, , 2017)).
Self-referencing identity is already a given for organisms which are recognised as teleonomic systems (Mayr, 1974), though often interpreted more strongly as teleological (Mossio and Bich, 2017).Goals, leading to purpose and the biological concept of function, are only meaningful concepts when applied to systems that have a self-referencing identity.Collier (2002) argued that individuality in combination with self governance, provides ''independent functionality''.Indeed, self-governance makes no sense unless there is an identifiable self to self-govern.Since the systems we are considering are materially and thermodynamically open systems, the enclosing boundary that most fundamentally provides for individual identity is a cybernetic/organisational boundary.The clearest way to ensure one is through closure to efficient causation (clef ), whereby the efficient cause of every component of the system is another component of the system.Rosen's (, )-system (Rosen, 1973(Rosen, , 1991(Rosen, , 2000) ) and Hofmeyr's (2021) ( , )-system qualify as clef systems; the latter is a causal account of the living cell and, following decades of effort, the former has recently been identified as such (Vega, 2023).Constitutive independence ensured by clef may be an overly-strong requirement for cybernetic independence in general systems (that is an open debate in philosophy at present), but here we will concentrate on living organisms, for which it is well established as universally applicable (excluding viruses Farnsworth, 2021).If the autonomy of organisms (in its three aspects) is founded on clef, the question here is: to what extent and in what way do codes (and their associated transducers) enable clef ?To answer that we need to delve deeper into formal and efficient cause, to understand how codes can make cause-effect chains contingent at the organism boundary and how codes preserve the template information necessary for clef, within the cell.

A deeper look at formal and efficient causes
All efficient causes are informed causes (using Hofmeyr's (2018) language).In the formulation of Farnsworth (2022), efficient cause is composed of physical force-fields shaped by embodied information, the latter being identified as formal cause.
Hofmeyr (2018) interprets Fig. 3A (his Fig. 4B) as representing either formal () and efficient ( ) cause separately working together as 'efficient, formal cause' (his Fig. 5B), or formal cause acting as a selection () of a particular (  ): the 'informed efficient cause' (his Fig. 5A).In the latter case,  is considered a pluripotent operator, which acts as one of several possible efficient causes, of which one in particular is selected by informing it with the formal cause (information) , e.g. a carpenter ( ) makes a table () from materials () using a plan ().In the Farnsworth (2022) interpretation,  and  have analogous, but more general and physical meanings:  is the pluripotent action of physical forces and  is the particularisation of their action resulting from the particular spatiotemporal configuration of elementary particles from which the forces emanate (an idea inspired by Pattee's general/particular dichotomy Pattee, 1969Pattee, , 2001)).The configuration is embodied information (equivalent to the information needed to specify the configuration of matter in a coordinate space).It sets initial conditions and continues as the dynamic coordinate values of moving particles and therefore the shape and strength of the forcefield they  4B.In this case, formal and efficient cause are 'distinct entities'.In B, we take literally the idea that efficient cause is information-constrained (particularised) physical force (Farnsworth, 2022).Hence the information  is 'empowered' by force  to give the informed efficient cause, equivalently, the force-empowered formal cause of the material transformation  → .Fig. 4. The physical process of signal transduction.Dashed arrows represent information propagation as potential formal cause.Formal cause is empowered (thickened arrow) when it constrains physical force to become efficient cause.If the constraint is dynamic, i.e. components of the system move in response, then flow of energy (open arrow) is required to empower formal cause.The membrane resists the force of external efficient cause   E , but the transducer, in this case a receptor, permits propagation of information carried by the efficient cause, transforming it into formal cause  E .This is combined with internally generated formal cause  C using the logic module L (this is an approximation of what is strictly a sequence of conditional probability biases), which in turn requires  E and  C to be encoded in the same 'language', hence the need for translation by the transducer.The resulting formal cause  A (=  E conditional on  C ) informs an actuator system A (such as a molecular motor).That in turn empowers the actuator formal cause   with flow of energy to become a dynamic efficient cause   A .
produce in combination.In this sense  is information that constrains the physical forces by specifying the points from which they emanate.The resulting organisation of forces produces physical actions, such as molecular binding, cleaving and conformational change, from which we get molecular fabrication, assembly, disassembly and organised movement, such as the ratchet sliding of actin over myosin in muscles.
Generally, when physical forces are constrained by the configuration of particles, energy flow is organised to do physical work (as opposed to entropic processes such as free gas pressure), e.g. the constraint of chamber and piston in a steam engine, or the configuration of proteins in the muscle fibre.This is the empowerment of the particular configuration (information ) by the general and universal forces (energy flow) to enact efficient cause.As Hofmeyr (2018) says ''Without a plan the carpenter is unable to function; without a carpenter the plan remains unimplemented''.He generalises that to ''without being informed by formal cause the efficient cause has operator potential but no agency''.The Farnsworth (2022) interpretation starts with formal cause as pure information that has no agency until it is empowered by physical forces, hence efficient cause is the outcome of physically empowering formal cause.A plan, a blueprint, data stored in patterns of electrical charge (silicon memory chip), or nucleotide sequence, are formal cause with potential agency, waiting to be empowered by physical forces to become efficient cause (Fig. 3B).

Codes isolate formal cause to protect and selectively empower it
Removing the empowerment of force from efficient cause, leaves formal cause, the latent constraint that is embodied information.Separating this formal cause from physical force, protects and preserves it and is achieved by translating it into a different domain of embodiment.Practical examples of this separation include the data on a silicon chip 'memory stick' and the genetic program embodied in the relatively inert and protected double stranded DNA molecule (opened to efficient cause by separating the strands) and the paper tapes or punch cards of early computers, Jacquard looms and pianolas.All of these are embodiments of information which act as sources of formal cause (constraint) only within the special circumstances of being an input to a transducer that (a) transforms their information into a domain where it can be empowered by physical force and (b) protects it from the larger forces of efficient causes within the system that it constrains.
Isolation of formal cause from efficient cause effects a break in the causal chain.The sort of break made by a physical transducer is analogous to the break made in an electrical circuit by a transformer for alternating currents: electromagnetically embodied information is allowed to pass through, but the configuration of the transformer Fig. 5. Physical causal representations of semiotic processes. is the state of the source S and also the stimulus,  is the state of the receiver R and also the response.The communication of information between source and receiver constitutes formal cause  =  (|), only if the communication channel  (⋅) matches the information domain of R to S (and if the power of the communication signal is negligible at the receiver).A depicts a pure signal-response system in which there is no interpretation, so no sign:  is directly a function of , as in remote control.B shows a physical link between S and R, so the channel is empowered as .But the transducer transforms the domain of  from  to  and in so doing, removed S, disempowering the efficient cause.The receiver's response is contingent upon, or modulated by an internal signal (, ) making formal cause (|).To logically combine ( (|)) with (|) (a mediation process), these formal causes must be in the same domain (hence the translation of  (, ) by (⋅)).The mediation (combining) to place R in a new state  is an example of interpretation by R, so ( (|)) acquires the status of a sign.In C, the source broadcasts a signal carrying information as a sign-vehicle , which has a different effect on each receiver: this can be achieved by structuring the transducers of receivers 1 and 2 so that they only detect the signal destined for them.Internally,  becomes  1 ( ( 1 |)) 2 ( ( 2 |)), respectively.
determines the magnitude of forces appearing on the receiver side and there is no electrical connection between this and the transmitter side.The code transducer extends that by making even the information flow contingent (e.g. an analogue to digital converter).To claim a break in a physical causal chain is a radical departure from physical inevitability; it is the first sign of freedom in the universe, though it is strictly still determined at the fundamental scale of physical processes where causality occupies a central generative role according to causal set theory (see Bombelli et al. 1987, Yurchenko 2023).We might therefore refer to efficient causes as the macroscopic (relative to the Planck scale) appearance of physical causation and it is this that living systems are able to manipulate using code transducers.

Individual identity requires codes
The ability of code transducers to reduce efficient causes to formal causes at the organism boundary, frees organisms to create and apply their own internal constraints, enabling the causal organisation of life to be locally self-constraining.Using ↣ to denote constraint of a component's state, for a set of components , , , we have () ↣ (+) ↣ (+2) ↣ (+3), cyclically.This cycle of constraints is equivalent to a cycle of efficient causes (Montévil and Mossio, 2015), since efficient cause is the constraint of physical forces by formal cause, which in turn is equivalent to embodied information (Farnsworth, 2022).The explicit use of time is to emphasise diachronic causation that does not violate the mandatory order of causal sequence.The circularity of causes implied when time is not explicitly included has unfortunately irked a number of critics who point out that cause can never loop back on itself.Reciprocal causation, whereby a system constrains and even fabricates itself, or more formally 'closure to efficient causation' (clef ) describes the result of a set of strictly diachronically ordered interactions: it is the final net result of the interactions that forms a circle, but the way it is achieved is better described by a helix in spacetime.The net result, analogous to an equilibrium emerging from the helical dynamic, is that the system embodies its own causality because it is cyclical through time (the structure of an (, ) system) and this gives rise to diachronic identity, especially the distinction between 'inside' and 'outside' by which function, autopoiesis and autonomy become meaningful.With this inside/outside distinction, transducers at the boundary can match an external domain with an internal one, contingently conveying information, without necessarily surrendering the system (to which they belong) to external causation.An organism is then seen as a local causal helix (explicit time) or cycle (integrated over time) with local independence, despite being embedded in, and a part of, the universal causal set.In effect, the mutual constraints imposed by the local pattern of matter (formal cause) reduce the range of accessible states to a finite set, equivalent to reducing an unbounded set Z of state labels to an equivalence class: the ring of integers modulo , i.e.Z∕.This is analogous to reducing the numbers on a ruler to those on a 12 h clock face, which is of course cyclice.g. 26 ≡ 2 (mod 12).It is very simply achieved in a purely informational (cybernetic) system; for example a finite state automaton is by definition cyclical in this sense (see e.g.Albantakis and Tononi, 2015).The conceptual difficulty is with conceiving this arising in a physical system through mutual constraints, but the principle is no mystery if one accepts that embodied information is the source of constraints on physical forces that leads to efficient cause (and also the substrate transferability, which Deacon (2021) identifies as the foundation of diachronic identity).The most important application of this mutual constraint is in the formation of clef achieved by the cell, for which code to support template information is essential.As Deacon (2021) explains: ''the template molecule can, in effect, offload some fraction of system dynamical constraints onto a structure that is not directly incorporated into or modified by the dynamics''.That is an important insight: transferring constraints into a separate information system frees the (autocatalytic) system to replace components without disrupting the 'essential selective specificity', which otherwise must be maintained as information-constraint embodied in the autopoietic network.Use of a template molecule relaxes the complexity and evolvability limits affecting template-free systems such as the autopoietic micelles of Luisi and Varela (1989) and Deacon's (2006) autocell.

The need for a template
A system cannot contain a copy of itself.If we consider the system in information terms, its total embodied information consists of the coordinates of every particle within it.A complete copy of it would necessarily embody exactly that information, so a system containing a copy of itself would embody twice the information, which is twice as much as is possible.Further, since the copy would have to include a copy itself, ad infinitum (the infinite regress problem), we see that no system can contain a complete copy of itself.But to reproduce, a system must copy itself and there are essentially two solutions to the logical problem that poses.
The first is found in systems that do not use template information (e.g.Luisi and Varela, 1989;Segré et al., 2000;Deacon, 2006) to enable organisational invariance: the preservation of organisational information despite material turnover (Deacon's substrate transferability).These systems follow purely 'compositional inheritance' (Kahana and Lancet, 2021;Markovitch and Lancet, 2014;Segré et al., 2001).Models of proto-life (e.g. the Graded Autocatalysis Replication Domain simulations of Lipid World Segré et al., 2000Segré et al., , 2001) ) maintain duplicates of their component parts, but by the argument just given, they cannot hold a copy of their whole selves.Given only duplicates of components, the higher levels of organisation which form a coordinated whole from the parts are missing.That is why non-template replicators are limited in their scope for higher level functions such as cognition (Bitbol and Luisi, 2004;Luisi, 2003) and adaptation (Vasas et al., 2012), which depend on the higher levels of organisation.Nontemplate replicators have no independent formal cause to preserve their organisational information: all their embodied information has to be simultaneously functional (via efficient cause) and formal (the specifying information that is duplicated in autopoiesis), as it is in metabolism-first models of proto-life.Every component part has to be duplicated using chemistry that follows thermodynamic gradients for molecular fabrication, though it may be pushed up gradients using abiotic sources of free energy (Liu et al., 2020) within the autocatalytic system.That is sufficient for minimal fabrication of component parts, but it cannot address the assembly of the parts into an organised whole (fabrication and assembly in the sense used by Hofmeyr ( 2021)).When a system is so simple that it can form spontaneously through thermodynamic gradients and yet fulfil the requirements of an (, ) system (Piedrafita et al., 2010), compositional inheritance is sufficient for organisational invariance.But assembly into an organised whole, that is not spontaneously produced by thermodynamic gradients, requires a formal template.Needing a formal template implies the need for code, since encoding into a different domain is the only way to ensure formal cause remains formal -by isolating it from the rest of the system.Interestingly, Kahana and Lancet (2021) show a possible route to developing the components needed for template replication within compositionally replicating micelles using pre-biotic chemistry, i.e. prior to the need for formal template molecules.
Once proto-life began to exploit the tremendous catalytic power of (initially simple) proteins, their chemistry was no longer entirely thermodynamically directed, in particular the (nearly perfect) thermodynamic equivalence of amino-acid sequences makes information necessary to ensure the correct sequence is followed in fabricating polypeptides.Polypeptides certainly can reproduce themselves, as accumulating evidence has shown (see Bao et al., 2022;Frenkel-Pinter et al., 2020;Sibilska-Kaminski and Yin, 2021, for recent examples), ever since Kauffman (1986) demonstrated that in principle a reflexively autocatalytic subset of polypeptides can emerge from a sufficiently large set of interacting 'protoenzymes' (updated by Hordijk et al., 2022;Kauffman and Steel, 2021).Selection of a viable reflexive autocatalytic set from among all possible reactions, among all possible polypeptides, is equivalent to gaining Shannon information (reducing system entropy).This information is what is needed to specify the organisation of the autocatalytic set.As the reaction network complexity and the variety and individual size of polypeptides increases, the autocatalytic set becomes increasingly specific, improbable and rare.As its sophistication increases, there must come a point where the likelihood of a system is so low that without the 'crutch' of an independent source of information (a template molecule) to act as a working memory for its evolution, it cannot be found.Eventually, the information needed to specify cascades of enzyme mediated reactions and their timing and possibly branching to match a varying environment, is too much to contain in compositional inheritance.Thus the increasing adaptive sophistication of protein catalysis networks brought with it the requirement for separate formal cause (though the threshold for this requirement has not yet been found, it probably could be using the methods of Kauffman and Steel (2021) and Hordijk et al. (2022)).The parallel development of nucleotide-based autocatalytic sets (e.g.Igamberdiev and Kleczkowski, 2023), including certain ribozymes (Ameta et al., 2023), shows how abiogenesis could have developed through ''gradual evolution of autocatalysis into template-based replication'' (Pavlinova et al., 2023).

What the template does
Theories of abiogenesis remain somewhat speculative, what we do know is that all extant organisms use additional formal causes, in the form of template molecules, that provide organisational invariance at the higher levels of their organisation.As Deacon (2021) says ''the structure of the molecular template literally re-presents the topology of the dynamical network of interactions that functions to re-present and re-produce itself'' (my emphasis).Reproduction involves transforming the coded information from formal cause into efficient causebridging the epistemic gap (Pattee, 2001) between symbols and actions.The physical reason for this gap is that the template is embodied in a domain that is different from that of the rest of the system, hence the need for code-translation.Keeping the template in a different domain protects it from the efficient causes that could change the information (but allows special efficient causes to repair errors and maintain the integrity of the template); it keeps the information as formal cause and it keeps it external to the system, solving the logical conundrum of a system maintaining a copy of itself.The template constitutes embodied information that is replicated separately from the information embodied in the form of the system (including membranes, cytoplasm, ribosomes and all the proteins and other biomolecules).DNA is not so much a part of a cell as it is an operating system on an 'external drive', gifted by the mother to the daughter cell, passed from generation to generation, held in safe keeping by each generation, separate and different in kind from all the working parts of the cell.Nevertheless, it would be wrong to think of the template as pure formal cause.Since all information is physical (the pattern of spatiotemporal configuration of particles) pure formal cause cannot physically exist, for example the charge pattern in a silicon chip and the nucleotide sequence of DNA themselves carry a physical forcefield.But we can call template information 'formal cause' because the forcefields do not correlate with those of other components in the system, except for the code translation apparatus.
Having the information molecule in relative physical isolation is necessary for replication as a matter of fundamental principle, as Von Neumann's (1966) constructor theory explains.It represents reproduction as ( +  + ) + () where  is a 'fabricator', () is the 'blueprint' (formal template) of machine ,  is a 'blueprint copier' and  a controller.''Cellular life conforms to this arrangement by encoding  = ( +  + ) in the form of DNA which acts as formal cause, leaving the information embodied by  ++ [the functional proteins, etc.] to inform efficient cause' ' -Farnsworth (2022), based on Hofmeyr (2021).The DNA template must remain effectively exogenous to the self-replicating machine (the cell) and yet needs to be within the cell to perform its function.DNA does not have a copy of itself embedded within, but it does have a blueprint for making a device that can duplicate the DNA of a cell.Thus the DNA is not made by following its own instructions directly.It is made by the machine that the DNA specifies.Because a cell's DNA is not made by following its own DNA, it is ontologically exogenous -it had to appear in the cell by some means other than following the blueprint.In real cellular reproduction that is simply a matter of having two copies of the information embodied by the DNA -one on each of the complementary pairs of the double strand.Of course, each single strand is incorporated into one of the two daughter cells and each of those two cells then uses the DNA duplicating machinery to create a further copy for itself, resulting in a double stranded DNA for both daughters.Marvellous though it is, this trick of using the complementary pairing of the strands is not the essential requirement for solving the infinite regress problem in cellular reproduction.Hypothetically, an extra copy of a single information molecule (e.g. the single stranded RNA in some viruses) could be made by the parent cell and injected into a daughter cell, which would carry on from there.That information molecule would be exogenous to the daughter because its origin lies outside the daughter cell itself.The essential requirement for solving the infinite regress is actually having an information molecule at all.That is, Von Neumann's infinite regress problem is solved by embodying the blueprint information not in and among the interacting molecules of the cell, but in an ontologically isolated, 'information only', molecule.

How code isolates the template
An easy example of embodied formal cause needing translation to efficient cause by a code transducer is the control tape of a Jacquard loom or mechanical automatic piano (pianola).There, the (input) domain of the formal cause is the pattern of holes in a paper tape, specifically their location across the tape (x-axis) identifying the warp thread or note pitch, and along the tape (y-axis) identifying the sequence of weft threading, or notes played.This pattern of holes alternately releases and exerts tiny forces that are mechanically amplified by a constrained flow of energy in the mechanism.That mechanism embodies the (output) domain where efficient cause is realised: it is the set of three dimensional hard component parts and the way they are interconnected, i.e. it is solid structure.The constraint on flow of energy within this structure is provided by the particular assembly of machine components, each with a particular form: in total, the embodied information of the machine.Adding kinetic energy creates movement, subject to the constraints now jointly provided by the tape and the embodied information of the machine.This empowers the information from the tape to be the efficient cause of a specified action of the loom or pianola.In effect, the mechanism logically combines the information on the tape with that embodied by the machine to produce an effect that is movement constrained by the mechanism, contingent upon the further constraint provided by the pattern in the tape.In reading the tape, the machinery's embodied information constrains the much larger actuator forces within the machine to do nothing other than to correlate with information embodied by the tape.Thus, the relatively fragile tape survives the interaction intact and its puny forces enact relatively much larger efficient causes.
Similarly, mRNA is a relatively fragile molecule which relies on the information-rich structure of the translational machinery of the cell to enact the fabrication of polypeptides without being destroyed by the process.As mRNA bases thread through the ribosome and are met with charged tRNAs moving from A to E to P binding sites, a fairly complicated sequence of conformation changes in both ribosomal subunits constitute a constraining mechanism that is both firm enough to ensure repeatable accuracy and delicate enough to preserve the information molecules (see e.g.Yusupova et al., 2006;Jenner et al., 2010;Shoji et al., 2009).Obviously it is only in the context of ribosome and charged tRNAs that mRNA can be causal and there it acts as formal cause for the selection of charged tRNAs according to codonanti-codon pairing in a way analogous to the tape of the pianola, but the mechanism for transferring its relatively weak forces to the stronger forces (peptide bonds) that elongate the polypeptide is considerably more complicated.
Some strikingly simple and well known features arise from coding that enables template replication.Firstly, like the pianola tape, mRNA is a linear structure (notwithstanding its propensity to form secondary and tertiary structure) which makes it far easier to 'read' than if the information were embodied in e.g. the three dimensional shape of a folded protein.Also it is of indeterminate length, so able to embody any amount of information, whereas a (finite sized) protein molecule would set a specific limit.These features are strictly a consequence of coding, since only with code can a string of differences embody information that could correlate with the three dimensional structures of cellular biochemistry (Section 1.1).The way the mRNA molecule interacts with other molecules in the cell is very constrained and organised.That would be hard to achieve if a folded protein were used to embody the information: very likely parts of its complicated surface would have unintended biochemical activity.Polypeptides are in theory linear chains of difference, but spontaneously fold into three dimensional objects, much harder to untangle than a folded mRNA molecule.At the core of the genetic code is the simple fact that the information domain of nucleotides is a linear chain of differences, whereas the domain of cellular biochemistry is the three dimensional shape of molecular surfaces.
Secondly, mRNA is digital (Walker and Davies, 2013), i.e. it is the sequence of bases, not their chemical structure that embodies information and that gives it near immunity to corruption by interaction with other biomolecules (by analogy, digital communications systems are far less prone to noise and interference than their analogue counterparts, though when digital does go wrong it is usually with far more dramatic effect).Embodying information in the sequence of nucleotides protects it from corruption and can only work in conjunction with a code, which must therefore have co-evolved (Barbieri, 2019).Much the same can be said of DNA, which is kept effectively inert (especially in eukaryotes by histone binding as chromatin) until transcription is required, whereupon RNA polymerase in combination with transcription factors (another information rich structure) finds the correct promotor sequence and starts to fabricate mRNA by correlation with the DNA sequence, again amplifying relatively tiny forces and of course leaving the DNA intact.

Codes enable cybernetic independence
Now let us look in detail at how coding transducers, embedded in the cellular boundary (with a few intracellular exceptions, such as steroid hormones in mammalian cells), make chains of efficient cause contingent upon the internal state of the cell and thereby effect causal branching and even blocking.This process, which transforms causeeffect into signal-response is an essential requirement for cybernetic independence, since without it, every process within the cell would be a direct (linear causal chain) consequence of exogenous efficient causes.It is also the means by which formal cause, propagated over a distance from some remote transmitter system, may effect influence over the cell without the need for empowerment as efficient cause (i.e.semiotic influence -Section 5.2).

From cause-effect to signal-response: internally combining formal causes
Referring to Fig. 3B, we see that information (formal cause) is empowered by physical force to produce efficient cause for a material transformation.Let us apply this to an external efficient cause impinging a cell boundary, denoting it by    , to emphasise that it is the combination of force  and information  and is exogenous () to the cell (Fig. 4).
As explained in Section 3, the transducer in combination with the membrane strips  E from its empowering force by transforming it into a different domain, leaving effectively a formal cause ( E ) presented as a signal (e.g. a conformation change) at the internal end of the transducer.For this to then contingently affect the cell, it must be logically combined with internally generated information representing the cell's state ( C ), so the combination produces a signal ( A ) that carries information about both the external signal and the internal state.This logically combined information is empowered by force (using energy) to create the efficient cause of a cellular actuator (  A ).That might appear to imply that the cell's response is the result of two efficient causes (  E and   C ) acting simultaneously.There is a serious problem with such an assertion because allowing for two simultaneous proximal causes is overdetermination: a violation of the exclusion principle, a principle based on causal completeness: ''Every physical effect has a sufficient physical cause'' (see Moore, 2019).The answer here is that  C is not empowered to become an efficient cause, it remains formal and as such can be combined with the formal cause  E through logical operations on information and more generally through correlation (see Appendix for more detail).The cell does not allow efficient causes to combine simultaneously, it performs the combination in formal causes and then empowers the result (i.e. it is cybernetic).We already saw this in the description of the working of a pianola: ''Adding kinetic energy creates movement, subject to the constraints now jointly provided by the tape and the embodied information of the machine'' (Section 4.3).
Most biological examples are far more complicated than the scheme depicted in Fig. 4, especially the computation by molecular switches and homeostatic control systems, often interacting with gene regulation systems, collectively all represented by the logic unit L (e.g. the way stomata respond to water stress - Buckley, 2019).Probably the most literal example is that of cellular mechanoreception (Gasparski and Beningo, 2015) and mechanoresponse (Chen et al., 2004), where yeasts have provided a model for study as they adjust turgor pressure during polarised growth in response to pressure from hard boundaries (Mishra et al., 2022).This involves pressure sensing transducers, GTPase-based regulation (biochemically coded logic) and physical actuation in the 'polarisome complex' using actin filament nucleation and exocytosis (Tcheperegine et al., 2005) to result in ''force-induced actin polymerisation at the tip of the budding yeast, which regulates the assembly and function of the polarisome'' (Xie and Miao, 2021).
In general, the simultaneous joint affect of efficient causes is produced by combining their formal cause components in isolation from their empowering forces. 2 Biological systems are often referred to as 'information processing', which we can now interpret as meaning they combine formal causes (i.e.constraints) because that is the only way to generate efficient causes that are determined simultaneously by both exogenous and internal states.This capacity is made possible by the causal structure of biological systems and, crucially here, their use of codes.It is copied in our technology of machines and control systems, but it does not exist in natural abiotic systems.Once a system is able to process formal causes, it is capable of being a cybernetic system with computational complexity only bounded by the amount of information embodied by its form and structure, both digital (Albantakis et al., 2014;Hagiya et al., 2016) and analogue (Auffray et al., 2020).
2 Otherwise, efficient causes can also be combined sequentially -one contingent upon another -without violating the exclusion principle.

Codes reduce formal cause to semiotic influence
Autonomous systems are of little biological interest if they are causally isolated (as e.g.clocks are); organisms are sensitive to their environment, including other organisms, so that they can adapt and interact in ways likely to benefit their biological fitness.For that, they need to perceive attributes of their environment and interpret their perceptions as signs upon which either to act or not.That requires the effect of information to be strictly semiotic, in the sense that it has causal power, but no direct causal effect via constraining forces (formal constraint) or even cybernetic control via formal cause (Table 1). 3estricting perception to semiotic influence leaves organisms free to be autonomous in the senses of cybernetic independence and freedom (Section 2).To describe semiotic influence, we adopt the conventional language of 'source' and 'receiver' communicating information that could be interpreted as 'signs'.
Fig. 5A shows a source S producing a sign-vehicle (a 'representamen' in Peircean language) upon which a receiver R acts, but does not necessarily interpret.To appreciate this more deeply, we need some new notation: I use calligraphic script for formal cause and bold for physical states and forces, so bold calligraphic represents efficient causes.Let  denote a formal cause (it is a spatial coordinate set labelled ) and  be its efficient cause, i.e. the formal cause having been empowered by the forcefield X (the overbar notation ⋅ denotes a vector force-field).
We can write the efficient cause as  = {, X}.
Strictly, the sign vehicle only becomes a sign on interpretation by the receiver.More generally, when the source S is in some state , it produces the sign-vehicle () which is correlated with .Thus, the sign perceived by the receiver depends on the state of the source, via  which is the information that can propagate to the receiver.Since  is a subset of , we can more directly refer to the source state  from here on.
Only by special arrangement of both source and receiver (a communications channel), can the sign-vehicle become formal cause for the receiver.At the receiver, this formal cause carrying information about  informs the receiver's cybernetic system (which constrains the states of the receiver ) so that some part of  correlates with .The sign-vehicle is therefore a formal cause () only if the communication channel permits it and its effect is restricted to a subset of the response repertoire of the receiver.The communication channel is a matching of information domain from source to receiver, represented by a mapping  (⋅), which matches  to  and  to .Because a sign is relational, depending on the receiver's perception of the sign-vehicle, it is a function of both  and .Perception is effectively the correlation between  and  made possible by matching the domain of  to that of .Thus the sign-vehicle is a formal cause  =  (|), where  (|) denotes a directional correlation 4 of  with .This in turn defines relational information as  (|), where  and  are interpreted as embodied information in the source and receiver, respectively.That is  informs  to the extent that  correlates with .The causing of this correlation originates with  and occurs in , which is why it is directional.In this case, then, we have remote control by formal cause (the sign-vehicle) for which a communications channel is necessary to match R with S, but we do not yet have sign because interpretation was not necessary.
If, physically,  (|) were to be empowered with a forcefield emanating from the embodied information (configuration, or state) , then it would be the efficient cause , which acts on  to change it to some new state (configuration)  ′ (e.g. a person S tugging on a kite string to change the configuration of the kite R).In that case the efficient cause would be the mapping  ∶  →  ′ , where  is {, S} = { (|), S}.That highlights that efficient cause is a consequence of both source and receiver systems, even if the source has physical (efficient causal) control over the receiver, as in a kite string.The result of any efficient cause is the consequence of both the cause and the object to which it applies (e.g.pressing on a block of butter is different from pressing on a piano key).However, in semiotic communications the physical forcefield of empowerment is to be found in R, not S (Fig. 4).The efficient cause is the mapping  ∶  →  ′ , where  = {|, R}, i.e. the formal cause is an internally created logical combination (or correlation) between the formal causes of the receiver state and what is now properly the sign.Because this involves combining exogenous with endogenous formal causes (| can be read as  conditional on , with  =  () and  =  (|)), the sign-vehicle no long acts as control, but rather is conditional and therefore subject to interpretation.Thus, in the semiotic case, S has no causal control over R, but it does influence R and the semiotic mapping {|, R} ∶  →  ′ is strictly directional from source to receiver.
Fig. 5B illustrates the resulting semiotic influence of  on R and explicitly shows the physical {, S} being stripped of its force S by a transducer  ∶  →  that transforms  =  (|) into ( (|)), the language of internal cybernetic computation.In an elementary example, the computation is a logic operation that combines ( (|)) with an internal formal cause (|), where  is the state of some internal variable (e.g.lactose concentration in a cell) and  translates this into the same computational logic.The response of R is therefore contingent upon the value of , not wholly determined by  and if this rudimentary self-determination of response to the sign-vehicle is considered an elementary form of interpretation, it raises  to the status of a sign.If that is not sufficient, then in a clearer example of mediation, if (|) was formal cause for a (molecular) switch determining the threshold for ( (|)), which may be e.g. a second messenger, then internal control would determine whether the sign was acted upon or not.For example, MAP (mitogen activated protein) kinase pathway switches, such as the Ras protein involved in growth factor signalling (and much more), regulated by e.g.guanine nucleotide exchange factors (GEFs) and GTPase activating proteins (GAPs) which respectively stimulate and deactivate G protein coupled-receptors, collectively act as controllable transducers (Osaka et al., 2021).That is the beginning of internal computation run on the cellular biochemistry, which may be arbitrarily complicated, but the 'special arrangement' that first enables  to influence  is the translation and disempowering of  by the transducer, e.g. a cell surface receptor, without which we would only see direct control of R by S. The downstream processing of ( (|)) and its consequences, dependent on , then enable autonomy as defined in Section 2. This process could be referred to as transforming cause and effect into sign and response.
Autonomous systems are free to interpret signals in their environment as it suits them (either by design, or adaptation).In Fig. 5C, we see a source S broadcasting a sign-vehicle  which is interpreted differently by two different receivers.We can imagine this as e.g. the stripes of a wasp that indicate danger to R 1 and food to R 2 .Again, a transducer is needed to transform the sign-vehicle into the domain of each receiver's internal logic.At minimum, each has a different domain, so requires a different transducer configuration:  1 ( ( 1 |)) for  1 and  2 ( ( 2 |)) for  2 (e.g. two radio receivers, each tuned into a different frequency).In a biological example, each transformed sign will be logically combined with other formal causes within the receiver (as in Fig. 5B) to give a different interpretation.This could include correlation with a memory of being stung by a wasp, or with a search image for a food item.Both of those internal conceptions are formal causes, embodied as e.g.patterns in the firing of neurons on a neural network.The important point here is that S does not control any of the receivers: it does not exert efficient cause, nor even formal cause (since that is generated by each receiver internally).Instead, we see only semiotic influence, since the responses are entirely determined by the receivers, which only utilise the broadcast information for internal computation.
Taking Peirce's (1998) definition of a sign as anything that is determined by a source, which acts as an interpretation (the interpretant) in some other system, then the formal cause  =  (|) is not a sign unless the receiver interprets it.The 'interpretant' is normally considered equivalent to the meaning of the sign, so a sign is that which communicates a meaning from a source to a receiver.The meaning cannot be the sign-vehicle, i.e. what Peirce termed the 'representamen'; the term 'sign' is reserved for an interpreted sign-vehicle.The signvehicle consists of embodied information, which may be embodied as part of the source, or free from the source, though generated by it.The information embodied in the sign-vehicle is always distinct from the meaning, which must be information embodied by the receiver alone (following the definition above).Interpretation is the formation of meaning from information.Meaning is relative and specifically relative to a self -an individual with identity (e.g. for me, that sign means X).A system can only be a self if it has the property of clef (which admits all whole live organisms) - (Farnsworth, 2018(Farnsworth, , 2022)).Meaning is internal to the clef system and consists of information that has the function of informing a decision: it is embodied as an internal state in a multi-state subsystem that informs a causal branch point in the global internal state of the system (Farnsworth and Elwood, 2023), consistent with Kull's (2018) definition of meaning-making (sign process) in terms of choice.Since meaning requires the possibility of choice, it can only exists within a free autonomous system (Section 2).
It should be emphasised that every physical configuration is potentially a source of sign for an autonomous system.Any configuration constitutes available information, with which other configurations may correlate.That should be no surprise, since every physical configuration has an associated forcefield shape which is always available to influence (by its physical action) any other.But semiotic influence strictly depends on the receiver system having within it a configuration specifically arranged (by design, evolution or learning) to correlate with the disempowered (pure formal) information.For example, the black and yellow stripes of a wasp can only produce a reaction in another organism if that organism embodies information configured to correlate with the signal produced in its detector when perceiving the coloured pattern (i.e.there is a neural network excitation pattern matching the stripes of a wasp, using a neural code).A kite responds to the efficient cause of tugging on its strings (efficient cause), whereas a remote controlled model aircraft responds to the formal cause of instructions relayed to it by a carrier wave, detected by a tuned circuit and translated into a physical pattern with which a designed component can correlate (e.g. the voltage supplied to an actuator such as a motor), but the work of interpretation is pre-programmed into the form of the system.Many cell signalling systems are similarly determined by their form, so the cell itself need not perform an act of interpretation: such remote control is formal causation.Transmitted instructions can only be formal cause for a system if it has a translation system transforming them into control information, just as a hormone is only causal for cells that have appropriate receptors and responses.This necessarily involves transducers: the embodiment of either code or cypher.However for semiotic influence, there must also be interpretation, which at minimum is the logical combination of exogenous with endogenous information and that requires transformation of information by coding into a cybernetic domain.Thus every case of semiotic influence requires a communication channel that is the embodiment of code.

Conclusion
Codes are a quintessential part of biological systems, since it is only by coding information that (a) Von Neumann's infinite regress problem can be solved, enabling closure to efficient causation and self-reproduction; (b) cause-effect links can be transformed into signalresponse links, freeing agents of exogenous control (cybernetic independence) and (c) causes can be logically combined to form unboundedly elaborate cybernetic systems enabling the cybernetic freedom of organisms.Accordingly, the autonomy of organisms would not even in principle be possible without their use of codes.It has been possible to come to these conclusions using a new understanding of efficient and formal causes based on a mechanistic explanation of causation rooted in physics and the appreciation that information is embodied as particular pattern in the distribution of (sub-atomic) particles (Farnsworth, 2022).Even if one is not prepared to accept that view of physical process, it would be hard to deny that biological codes play a fundamental and essential role in enabling life to exist, based on their property of transforming information from one domain to another, thereby enabling the separation of different kinds (or sources) of information and combining them in well controlled ways when required.

Declaration of competing interest
The manuscript herewith submitted presents original work that is not published or being considered for publication elsewhere, in whole or in part.No use was made of AI systems in writing any part of the work.
Publication of this work is explicitly approved by all the authors and tacitly by our employers.If accepted, it will not be published elsewhere in the same form, in English or in any other language, without the written consent of the copyright-holder.
The authors have no conflicts of interest, including, financial, personal or other relationships with other people or organisations, either now or within three years of beginning the submitted work, that could inappropriately influence, or be perceived to influence, our work.The work was desk-based and was not supported financially.
intended receivers and written text must be such that the words match patterns already present in the brains of their readers.In short, formal causes only combine to effect efficient cause when they are embodied as patterns in a common domain.

Fig. 1 .
Fig.1.The three primary axes of autonomy and freedom, showing with examples that these are effectively continuous variables (not presence/absence) and although hierarchically dependent, they are empirically independent (orthogonal) variables.

Fig. 3 .
Fig. 3. Causal deconstruction by graph theoretic relational diagrams.A shows formal cause () and efficient cause ( ) combining to 'informed efficient cause' -from Hofmeyr (2018) Fig. 4B.In this case, formal and efficient cause are 'distinct entities'.In B, we take literally the idea that efficient cause is information-constrained (particularised) physical force(Farnsworth, 2022).Hence the information  is 'empowered' by force  to give the informed efficient cause, equivalently, the force-empowered formal cause of the material transformation  → .

Table 1
A variety of domains that can carry information, each resolved into attributes of medium, basis and scale.The last column identifies the form of causality applicable to information in the corresponding domain.Formal constraint is physical causality; cybernetic is purely informational and semiotic is arbitrary code with no direct causal power (see text for explanation).