The creativity of cells: aneural irrational cognition
Handling Editors: Laura Bennet & Denis Noble
The peer review history is available in the Supporting Information section of this article (https://doi.org/10.1113/JP284417#support-information-section).
Abstract
Evidence of cognition in aneural cells is well-establish in the literature. This paper extends the exploration of the mechanisms of cognition by considering whether or not aneural cells may be capable of irrational cognition, making associations based on coincidental similarities and circumstantial factors. If aneural cells do harness such semiosic qualities, as with higher-level creativity, this might be how they are able to overcome old algorithms and invent tools for new situations. I will look at three examples of irrational learning in aneural systems in terms of semiotics: (1) generalisation in the immune system, based on viral molecular mimicry, whereby immune cells attack the self, which seems to be an overgeneralisation of an icon sign based on mere similarity, not identity, (2) the classical conditioning of pea plants to trope toward wind as a sign of light, which seems to be an association of an index sign based on mere temporal proximity, and (3) a pharmaceutical intervention to prevent pregnancy, using a conjugate to encrypt self with non-self, which seems to be an example of symbol use. We identify irrational cognition easily when it leads to ‘wrong’ outcomes, but, if it occurs, it may also lead to favourable outcomes and ‘creative’ solutions.
Every creature, no matter how simple, must have a sense of ‘aboutness’ or intentionality … However system-derived and non-conscious, they are not machines, and every living thing must be capable of negotiating a sometimes surprising world. This means that every organism has something like ‘mind’ … most of your mind and my mind and certainly the most creative parts, are non-conscious… (Wheeler, 2014a; 402).
Introduction
In ‘On having no head: Cognition throughout biological systems’, František Baluška and Michael Levin (2016) review the literature on aneural cognition in single-celled organisms, plants and animal tissue, all of which exhibit abilities for memory, learning, decision-making and goal-direction. ‘Cognition’, they argue, need not involve neurons per se. Bacterial biofilms use potassium ion channels to propagate long-range electrical signals (Nunes-Alves, 2015). In animal body tissue, all cells have ion channels, and most have gap junctions that function as electrical synapses forming bioelectric networks (Levin, 2012; Pezzulo & Levin, 2015; Tseng & Levin, 2013). Even the idea of plant neurobiology (Baluška et al., 2004, 2005) can be cogently advanced. Cells of all kinds appear to be capable of intelligent behaviour that emerges from dynamical networks that propagate signals and alter connections in response to the environment – very like neuronal activity. In fact, as Baluška and Levin conclude, neural tissue seems to have merely improved upon ancient mechanisms used by all living systems generally.
In this paper, I will not continue with this already well-established argument that decision-making in cells prefigures decision-making in animals with brains (see also Levin, 2022). Instead, I will focus on the extent to which simple organisms and cells may also be capable of what might be called ‘irrational’ cognition, that is, making subjective associations and connections. Irrationality is here not meant in a necessarily negative sense. Through history, irrationality has been linked conceptually with creativity and the ability to discover solutions that are not available through what we typically think of as ‘rational’ means.
Most of the work analysing aneural cognition focuses on cells and brainless organisms behaving in ways that make sense for them in terms of surviving and thriving. Relatively few instances in the literature on aneural cognition look specifically at examples of living systems (i.e. single cells, collections of cells or plants) learning to behave in ways that would not maintain or enhance health in normal natural conditions. However, some very interesting studies have been done, and we might learn something about the fundamental nature of cognition from these examples. Saigusa et al. (2008) trained Physarum polycephalum to associate a unit of time with a cold shock. Shirakawa et al. (2011) trained P. polycephalum to respond to cold as a sign of food. Such examples of associative learning may be considered the best proof that new learning has occurred, insofar as the organism cannot be simply drawing upon behaviours encoded in its genome. Furthermore, in these cases of unusual behaviour, the normal pathways are not simply blocked, they are co-opted and continue to function, albeit in new ways (Alexander et al., 2021). We cannot say it is a mere failure of a signal pathway or an arrested cycle, but a reinterpretation and repurposing of a pathway.
Is aneural semiosis a better descriptor than aneural cognition?
I seek to add to the literature by describing the physical and relational semiosic mechanisms underlying what we call cognition. I argue that learning is based on sign relations, that is, detecting qualities of things being identical/similar to, connected with/near to, and associated with other things or effects. We may say that the use of flexible (and even sometimes erroneous) signs is how old algorithms are overcome and innovations emerge. In particular, I will investigate irrational semiosic mechanisms in terms of (over)generalisation and (mis)association, which can lead eventually to novel encryption. In my C. S. Peirce-inspired semiotics, these processes are defined as using icons, indexes and forming symbols, respectively.
What are the indications that aneural semiosis underlies what we are calling aneural cognition? As Anthony Trewavas (2003) argues, for example, the evidence for cognition in plants is found in the fact that they are capable of communication with other plants and with insects, they can store experience in memory, they can anticipate, and they even show ‘intentions’. These, I note, are all semiosic processes involving factors that stand for other factors. When a plant communicates with other organisms, it does so by means of signals, which represent other processes or situations. Memories form as traces or signs of external factors. If a neutral factor in the environment can function as a sign of a useful factor, we can say that detection of the neutral factor allows the organism to anticipate the useful factor. When factors function as signs of useful processes, we can say they are about the other useful factor, and ‘aboutness’ is the aspect of intentionality. While the brain may be the only organ in which complex semiosis emerges, we should look for the primitive precursors of these processes in cell types other than neurons.
Semiosis is, by nature, potentially fallible because the objects that signs represent (e.g. to the interpreting signal pathway) may not, in fact, exist. Consequently, in addition to rationality, human mentation is also characterised by irrationality and unconscious dream-like thought processes that tend to make associations based on rare stochastic resonances, coincidentally similar qualities and circumstantial associations. Are single-celled organisms, body tissue and plants equally capable of making irrational associations? If so, as with artistic creativity in humans, this might be the means by which they are able to invent entirely new tools for new situations for which natural selection has not prepared them. A theory of creativity (learning/inventing new relationships) seems to require semiotics. Semiosis is the key to understanding both how new lawful relationships are discovered and later, if needed, overcome.
Irrational human memory and learning
an effect of long ‘conversations’ over time dominated by certain memories—morphologies/patterns of semiosis (similarity-in-difference/metaphors) which are encoded and read (in protein expression) in one particular way until environmental changes (new messages) produce pressures to find a new metaphor (a new expression) with sufficient similarity and difference to allow the great dialogue of life to shift (to a better adapted/more useful expressive form), and thus to continue.
Integral to this discussion is biosemiotics, which investigates semiosic processes below the level of consciousness and throughout all living systems, including neural and non-neural systems. This paper will draw from the corner of that field cultivated so well by Wheeler that investigates biosemiosis in terms of poetics.
Metaphor is a figure of speech that does not just compare dissimilar things, as a simile does (an arthritic hand looks like a gnarled tree branch), but transforms the one into the other, and ‘gnarled branch’ is used instead of the word ‘hand.’ Metaphor links arbitrarily similar things, and we can say that metaphor is an over-generalisation of icon signs. In this paper, I will be examining molecular mimicry as a biological metaphor.
Metonymy is a figure of speech that replaces one thing with another, the substitution being something nearby the replaced object; for example, a place name is used to indicate the official that presides there, such as ‘10 Downing Street’ for the British Prime Minister. Metonymy links arbitrarily proximate things, and we can say that it is a conditioned association of an index sign with an object. Metonymy may be thought of as a looser form of synecdoche, whereby a part stands for the whole, such as when ‘head’ is used to mean the person who is the ‘leader.’ A part standing in for a whole is a true index, whereas something standing in for something else that is merely nearby might be called a conditioned index. I will be looking at instances of Pavlovian conditioning as a form of metonymy.
A symbol links things by rule/convention, encrypted in habit, a rule book or a dictionary. A symbol is a purely arbitrary association whose originating contextual indexical or iconic associations are no longer relevant to its function. For example, the originating context of the phrase, ‘to kick the bucket,’ as a humorous, euphemistic code for ‘dying’ is unknown, but most folk etymologies speculate that the phrase is somehow indexical to advent of death throes (Etymonline, 2003). Perhaps cadaveric spasms often upset the washing bucket at the foot of the sick bed. Whatever its origin, the phrase has come to mean ‘to die’, and users of the phrase have ceased to think of it in literal terms. A symbol can thus be known as a ‘dead metaphor’ or ‘dead metonymy’ because it no longer refers to the originating iconic or indexical relationships that brought two things or circumstances together. Below I will analyse the use of a vaccine conjugate to encrypt the meaning of (the effect of) one type of protein with another in terms of symbol creation.
Although poets use figures of speech and other rhetorical devices self-consciously and deliberately, their effects are rendered in the audience at the subconscious level. Biosemiotics has been useful in articulating the processes whereby, in humans, subconscious meanings are created and reinforced in the brain as the semiosic flow of similar and proximate information (Alexander & Grimes, 2017). Such self-organisation does not necessarily flow in an obviously logical manner, and it does not require repeated reinforcement to be learned. Propaganda is an excellent example of communicators taking advantage of the way people process information rapidly without really thinking. Subconscious processes help us learn and adapt quickly, but since they can be irrational, these same mechanisms can allow us to fall into dysfunctional habits of thought. Analysing how these ‘poetic’ generalisations and associations in the subconscious work may also help us understand pathological processes in aneural living systems. Pathological processes, instead of being anomalous, may give us information about the flexible semiosic nature of cognition generally.
Creative learning does not come through reward and punishment
chance as random genetic change, followed by deterministic machinery and culled by natural selection, as imagined by Neo-Darwinism, seems incoherent. This is because it lacks the essential element of inventiveness which is characteristic of all the responsive and developmental processes which we see all around us in the world of the human and nonhuman living.
A semiosic model of learning reveals the active agency of living systems. Poetic semiosic processes are self-generated via interpretation that can cause changes in the way signal pathways function by making associations based on physical similarities or spatial/temporal proximities. In this way, living systems can sometimes learn new things instantly without repeated reinforcement.
In our own experiences, we may have noticed, for example, that rote memorisation and training cannot explain so-called ‘flashbulb memories’ that recall the arbitrary details of the context when we learned new information. Semiosic association can occur extremely rapidly. Those who listen to audiobooks in the car may have noticed that – if compelled to search through the recording to find where one left off, the day or week before – when re-listening, one can suddenly picture where one was on the road when listening to that part of the recording. This means an instantaneous connection was subconsciously made between an arbitrary context, the location on the road and spoken words. We may call this metonymy, an arbitrary index, or a coincidental contiguity. We cannot call this a conditioned symbol because there was no conditioning. If this information is emotionally charged, as with the case of flashbulb memories, we may be able to recall those irrelevant contextual details for a long time (Conway, 2013).
Also similarly, people with ‘space–time synaesthesia’ are able to recall apparently meaningless details of any date on the calendar by means of positional memory (Parker et al., 2006). When we observe these kinds of memory, we may be getting a glimpse of the semiosic way that neurons tend to self-organise.
How do humans memorise information? As Bower (1970) observes, before writing was practically possible, people used narratives and poetry to help them remember. Using rhyme, rhythm, temporal narrative and various other poetic devices, such as metaphor and metonymy, can greatly aid memory. With the method known today as the ‘mind-palace technique,’ memories can be formed via arbitrary associations. Ancient poets memorised long narrative poems by associating the lines in each stanza with the various random objects in the rooms of a palace. (A ‘stanza’ is a paragraph of a poem, and the term means ‘room’ in Latin.) Later, the poets imagined themselves walking from room to room looking at the objects, and, with these cues, they could recite the long poem from memory. Using poetic mnemonic devices does not require multiple repetitions as straight rote learning does.
Such poetic logic, linking things arbitrarily similar or arbitrarily nearby, seems to prevail at the subconscious level. As Harvey (2013) has noted, synaesthesia studies have contributed to a better understanding of how the subconscious works in a cross-modal manner, which is suppressed in the conscious thought of people without synaesthesia. Synaesthetes are better able to recall arbitrary facts because numbers or letters may have a unique colour, texture or shape for them, which they can readily recall (Tammet, 2007; Terhune et al., 2013). This type of memory formation does not conform to what we may think of as typical logical operations.
If the poetic logic of similarity/proximity provides the structure for new memories, this may help explain how rapid learning occurs without extended trial and error periods. Rapid learning may be compared to insights or epiphanies. The systemic reorganisation of neural activity, key to insight, is associated with gamma waves and the anterior superior temporal gyrus (Dietrich & Kanso, 2010), which, significantly perhaps, is linked to language use, understanding of literary themes and metaphors, and getting jokes (Jung-Beeman et al., 2004) which require making connections between disparate things.
How does local flexibility scale to affect network population dynamics?
I argue that the flexible, poetic sign action that exists in the lower-level processes must enable the rapid re-organisation and formation of new semiotic habits. The physical properties of sign-vehicles – insofar as they may be coincidentally similar to or coincidentally proximate to a component of a signal pathway – may lead to substitution and radical change in the pathway.
Lower-level semiosic flexibility in a group of cells, for instance, would enable a more naturally robust and efficient emergent behavioural pattern than would precisely correct local interactions (Alexander, 2018; Alexander & Grimes, 2017). As individuals interact with neighbours, they are subjected to similar situations and are likely to make similar generalisations and associations with regard to their own internal signal pathways. Such novel interpretations can spread easily and/or occur independently and simultaneously. This self-organisation would be, in part, a thermodynamic process, insofar as each individual state change could flow efficiently to the next potential state. Information flow (not selection) would produce an emergent diffusing wave of signals, which, in turn, would further harmonise those individuals co-producing the next wave. This might manifest as brain waves in neural assemblies (Tsuda et al., 2004), collective cell migrations, or the organisation of extracellular matrices and various other behaviours guiding growth and repair (Yeomans, 2023), as well as coherent but unpredictable behaviours such as starling murmurations (Pismen, 2021; 5−19).
Bacigalupi (2022), building off Kuramoto & Nishikawa (1989) and Saigusa et al. (2008), describes the next level of the process; further entrainment emerges as interfering, oscillating signal pathways come into phasic alignment as they synchronise with a similar regularity in the environment, thereby acting as a sensor of that regularity. Here interpreting a sign is forming a resonance between diverse and noisy internal states and external conditions (see also Bacigalupi and Favareau, this issue).
Once these relationships are formed, they can be reinforced if they have positive feedback effects. Stuart Kauffman's (2000) notion of the ‘adjacent possible’ or Donald Favareau's (2015a) ‘relevant next,’ to describe the flow of state changes or decisions, contribute to this understanding of how cells collectively organise very fluid and flexible emergent holistic patterns that are never quite the same but are always very similar. Self-organisation on its own may be robust enough to constrain behaviour or the forms produced by such processes may later be encoded by a selection process into an organism's instinctual memory.
Although a rational engineering approach may describe the goal-directed biological processes of morphogenesis, for instance, as performing ‘error minimisation’ toward a ‘set point’, instead, I stress that, in living systems, goal states are emergent from the overall closed-loop dynamics and cannot be an externally specified value, the deviation from which the system tries to minimise (Denizhan, 2023). Instead of minimising error, I would guess that errors, that is what I am here referring to as diverse and divergent sign readings, are being harnessed.
This type of harnessing would be different from the harnessing of stochasticity described in Noble and Noble (2018). In that case, the immune system is observed to trigger hypermutations producing a diversity of proteins, and this is followed by a selection process that gradually refines antibodies for the purpose of finding a good fit to neutralise a foreign antigen. In the case of the harnessing of semiosic chance similarities, random variation is not needed for creative change. Instead, a living system responds to qualities that can be used differently in different contexts. Such variation is always already available and need not be produced by random mutation.
The harnessing of semiosic chance would allow rapid simultaneous change in the group without perfect synchrony, and the overall effect would be of greater regularity among randomly interacting parts in a system, as similarity provides enough of a constraint to dampen difference without eliminating it. Error minimisation, on the other hand, where less and less flexibility is tolerated, can lead quickly to over-coherence and loss of the ability to adapt (Bacigalupi, 2013).
biologically instantiated sign relations interlocking with and reinforcing one another, and by so doing, providing directionality towards and away from other sign relations in the network, through the dynamic emergence and canalization of semiotic pathway biases and constraints. Such ongoing semiodynamic re-adjustment enables new scaffolds and new pathways within and between scaffolds to arise, increasing semiosic capacity exponentially.
When nuance from the language of biosemiotics is added to the language of cybernetics applied to biology, the role of agency is brought to the foreground. The flexibility of lower-level icon and index sign action, the diversity of the nevertheless constrained options available, is what allows living systems to smoothly and spontaneously form new organised habits or algorithms, which may eventually be encoded.
Examples of irrational or poetic semiosis in anueral biological systems
To support my argument that irrational semiosis exists and probably underlies new learning, I will try to illustrate with three examples. The first is an example, from Nunez-Castilla et al. (2021), of a pathological condition that arises from an icon sign reading. The molecular mimicry of certain virus protein conformations can result in the immune system being confused by the mimicry and launching an unhealthy autoimmune response. The second example, from Gagliano et al. (2016), is a laboratory manipulation of an index sign. Pea plant seedlings, trained to recognise wind as a conditioned stimulus, trope toward wind as a sign of light. The third, from Talwar et al. (1976), is a pharmaceutical intervention whereby a new symbol is created with the use of a conjugate to encrypt self with non-self, leading to a desired autoimmune reaction. In these three cases, existing signal pathways are repurposed by means of ‘poetic’ semiosic mechanisms that create new meanings, new signs, via a chance similarity or forced proximity. Aptly, Wheeler defines ‘biosemiosic chance’ as ‘a habit-upsetting happenstance’ (2014b; 373).
Irrational icon sign, molecular mimicry in viruses and autoimmune disease
Sign generalisation is a form of semiosic learning that occurs when a different but similar factor elicits the same response that the learned (or innate) sign elicits. The ability to generalise previous learning and apply it to novel contexts makes a living system flexible and able to act intelligently without having to go through training. Molecular mimicry is a type of sign generalisation about appearance, shape, colour or pattern.
An adaptive advantage of molecular mimicry is illustrated when past exposure to a pathogen results in full or partial immunity to another pathogen with a similar antigen; this is known as heterologous immunity (Agrawal, 2019). In mid-2020, for example, numerous studies reported T cell reactivity against SARS-CoV-2 in 20% to 50% of population with no known exposure to the novel virus (Doshi, 2020).
However, in rare cases, such lack of precision with regard to the target of the immune response may instead lead to disease. A disadvantage of molecular mimicry is that similarities between virus pathogen antigens and human proteins may provoke an autoimmune response, leading to transient or chronic disease (Albert & Inman, 1999) and this has proven to be the case with SARS-CoV-2. While the likelihood of developing autoimmune disease is, in part, affected by both genetic differences and the general health of the patient, molecular mimicry can lead to immune disorder in any patient who suffers an acute infection (Ermann & Fathman, 2001). Thus, the risk of molecular mimicry-induced immune dysregulation highlights the importance of early treatment.
The SARS-CoV-2 virus gains entry into host cells via the Spike protein, which is one of its main antigenic proteins. Early investigations comparing protein sequences of Spike epitopes – the part that conforms with the shapes of the infected host's antibodies, B cells or T cells – predicted multiple chances for cross reactivity (Kanduc, 2020; Khavinson et al., 2021; Lyons-Weiler, 2020; O'donoghue et al., 2021).
Defining molecular mimicry as a match of at least five identical consecutive amino acids, where at least three amino acids are surface accessible and have a high structural similarity, Nunez-Castilla et al. (2021) found that a TQLPP motif in Spike and human thrombopoietin share similar antibody binding properties (Fig. 1). This molecular mimicry could trigger thrombocytopenia, an autoimmune condition observed in COVID-19 patients (Bhattacharjee & Banerjee, 2020). As Hodcroft (2022) shows, in the Gamma variant of SARS-CoV-2, a mutation changed TQLPP to TQLPS, and in an Omicron variant, nine nucleotide deletions resulted in the loss of 60% of the TQLPP motif. Those infected with such mutated variants did not suffer as severely as those infected with variants closer to the original.
Another SARS-CoV-2 Spike motif, ELDKY (Fig. 2), well-conserved across numerous beta-coronaviruses (common cold viruses), binds to a human neutralising antibody (Pinto et al., 2021); thus, past exposure to certain beta-coronaviruses could result in a milder SARS-CoV-2 infection. Unfortunately, the ELDKY motif is also shared in multiple human proteins, one being protein kinase 1, which is involved smooth muscle contraction and cardiac function. This molecular mimicry can also lead to thrombocytopenia or other blood clotting disorders found in some severely ill COVID-19 patients. The ELDKY motif is also found in human protein tropomyosin (Fig. 2), whose cross-reactive activity with Spike is linked to known COVID-19 complications involving myocarditis (Nunez-Castilla et al., 2021).
In many cases of acute COVID-19 disease, especially in patients infected with early variants of SARS-CoV-2, the immune system – whose function is, in part, to eliminate self-tissue that is diseased – began to over-generalise self for non-self, based on a similarity between a small portion of the respective proteins, and turned the immune system on the patient's own tissue. This situation illustrates the fact that the intelligence of the immune system can be tricked into ‘irrational’ responses by misreading something as a sign of something else. Once the incorrect interpretation of the similar molecular structure of self-tissue (as an icon sign) has been made, the signal pathways cannot easily break the new bad habit; all habits tend to be self-reinforcing by nature.
Irrational index sign, associative learning
An index sign can be defined as a by-product of something and therefore it objectively points to that object. Smoke is an index of fire; a rising plume of smoke may indicate to someone that there is fire and a response is required. Another example of an index is found in the way that P. polycephalum may detect sugar diffusing from a concentrated source and may pursue increasingly stronger concentrations. To follow a gradient is to use an indexical sign.
When an organism is confronted with a new index sign, it may have no way of establishing whether or not the sign is physically linked to what it represents or only proximate to it in time or space. What appears to be smoke may be steam rising, but the fire department might be called to investigate nevertheless.
Any coincidence that occurs with some regularity may be mistaken as an index. This is the case with conditioned stimuli. The formation of such index signs is a kind of Pavlovian conditioning, during which unrelated but proximate factors in the environment can become associated, and one can become an index sign pointing to the other (or its object). The ability to associate a coincidental factor in the environment with an unconditioned stimulus (an inherited sensing ability that triggers a useful response) could have, in most cases, strong reproductive advantages. In natural environments, factors that frequently appear together may be causally linked. Thus, it would be advantageous to make such new associations.
I offer an example of pea plants learning via Pavlovian conditioning that illustrates irrational semiosic learning. Since plants have neither brain stem nor nerve cells, how a plant learns must depend upon some kind of structural re-organisation of its semiosic infrastructure. In a classical conditioning experiment, Gagliano et al. (2016) trained pea plants to grow toward or away from a blowing fan as a sign of light. (Note that Markel (2020) made an attempt to replicate Gagliano et al. (2016) and found the outcomes insignificant; however, Markel made a number of changes to the experiment, which may explain the difference in findings.) In the experiment, potted seedlings were put in a dark room with a Y-shaped jointed tube capped over each pot. One group of plants (A) were equipped with a fan in one side of the Y-tube and a light in the opposite side. Another group of plants (B) had both a fan and a light in the same side of the Y-tube. For a 3-day training period, while the seedlings were mostly deprived of light, the fan was turned on for 30 min before the light was turned on, after which time, the fan ran for 30 more minutes and the light stayed on for 1 h. When the light was on, the seedlings, stressed by poor light conditions, readily grew into the side of the Y-tube that was lighted. Each day of the training period, the set-up was switched in both groups. The Y-tube was removed and rotated 180 degrees, such that fan/light sign alternated between left and right sides, so that the plants had to switch sides in order to follow the light.
Next, during the testing period, the fan was turned on as per usual, but the light was not. A significant number of plants grew into the side of the tube in which they anticipated the light would be, based on the last training session. In group (A), 69% of the seedlings grew into the part of the Y-tube that did not have the fan. Thus, it appeared that for a significant number of plants in this group, wind functioned as an index sign that the light would be found in the opposite side. In group (B), 62% of seedlings grew into the side of the Y-tube with the fan. Thus, it appears that for a significant number of plants in this group, wind functioned as an index sign of where the light would be found. In both cases, the seedlings seem to anticipate the appearance of light by growing either away from or toward the wind, depending on their training, in search of light. According to the researchers, the experiment shows that plants are able to ‘encode both temporal and spatial information and modify their behaviour under the control of environmental cues’ (Gagliano et al., 2016; 3). We might refer to these plants as ‘Pavlov's peas’ because they responded to wind as if it were light, just as a sound of a bell had come to mean food to the dog in Pavlov's experiment.
Although the exact physiological/molecular mechanisms that allow plants to forage better are unknown, auxin signalling systems are likely involved (see Halliday et al., 2009; Telewski, 2006). The plant hormone auxin (indole acetic acid) is a key regulator of multiple aspects of plant growth, development and troping. It is implicated in cytoskeletal reorganisation and sensing light, as well as in sensing spatial orientation and temporal patterns (Chen & Yang, 2014). We can speculate that when different multi-step signal pathways are activated at the same time (one for detecting spatial position and others for growth/troping), they may interfere with each other if they compete for a specific molecule, use molecules that mimic each other or produce by-products that are used by each other. Such kinds of interference can create relationships, such that a change in one signal pathway can cause a change in the other signal pathway, possibly switching it on or off. (To explain how Physarum polycephalum can associate cold with food, Shirakawa and Sato (2013) offer a cybernetic circuit-like model of an activation/inhibition switch created by the interaction of two signal pathways (one sensing temperature, one sensing nutrients, both linked to movement). However, the model depends on protein production triggered by genes, and it is unclear how such functional proteins could have been selected. A model that depends upon signal pathways using constitutively available molecules would make more sense to explain the rapid learning to solve an entirely novel problem.)
In any case, it is clear that the pea plant's response of turning toward or away from wind is not under genetic control, since troping with respect to wind direction in this way would not be advantageous outside of this laboratory experiment. The plants were tricked/trained to act as if wind direction were an arbitrary but true index of light. The fact that the plants can be trained to turn toward or away from wind as a sign of light reveals that the response is not merely mechanical, but a learned association. Also the fact that the set-up was switched such that the plants had to re-learn the association each day, and when tested, tended to trope according to the sign relation learned the previous day, reveals the associative semiosic nature of this memory.
We may compare the arbitrary nature of this learned index sign to a similar indexical association made by magnetotactic bacteria (Magnetospirillum magnetotacticum), which synthesise nano-sized magnetic iron particles to use as an internal compass to swim up or down to reach favourable conditions with relatively low oxygen concentrations (Blakemore, 1975). The magnetic field exerts a torque on the cell which passively aligns the micro-organism with the field (Zhu et al., 2014). The indication of North is conversely positioned (at the flagella end or opposite end) in the bodies of bacteria in opposite hemispheres. Therefore, the constraint is not merely mechanical and these bacteria species independently evolved the ability to associate North with different movements. The needle's properties function as a regularity against which a differential sign for up or down evolved in the lineage (Denizhan & Karatay, 2007) if not in the individual. (Individual magnetotactic bacteria exposed to opposite hemisphere conditions do not learn to reverse their index rule that aligns them with the magnetic field. However, as Denizhan and Karatay (2007) note, after a few weeks and numerous generations, the lineage can learn the new rule via a selection process. This may be contingent upon daughter cells having to synthesise a new magnetic needle in altered conditions (Kasama et al., 2006).) Similarly, it is because the researchers in the pea plant experiment were able to train the seedlings to move in a specific but arbitrary direction, not of mechanical necessity, that they can claim new learning did take place.
Irrational symbol, encryption in a pharmaceutical intervention
It has been established that biological processes make use of encrypted symbols or codes to perform rational functions. For example, Marcello Barbieri (2008) argues that transfer RNA acts as an encryption device between an anticodon with three nucleotides and the corresponding amino acid. We may say that tRNA is like the arbitrary code of a cypher wheel, as there is no natural affinity between anticodons and amino acids. The structure of the tRNA acts as an adapter between materials that do not spontaneously bond together. Code Biology researchers (Barbieri & Pinz, 2022) have identified over two hundred arbitrarily encoded relationships, for example splicing codes, signal transduction codes, actin codes, cell migration codes, etc.
Barbieri emphasises the fact that new codes do not originate by the copying errors of previous codes, but in the origin of new rules for encrypting/decrypting a code, and therefore Darwin's idea of random variation and selection cannot account for the encryption process itself, only of the selection of a code once it has formed.
Such evolutionary processes may be impossible to observe. Generally, we can only recognise new codes after they have been formed and selected. But we do have an example of an artificial biological code created by means of an encryption device, which may be compared to Barbieri's concept. Such a new code will not be genetically fixed, but will serve as an example of a code being learned for use during that individual's lifetime or experience.
Pharmaceutical interventions can employ the strategy of encryption to more or less trick the body. Since the 1970s researchers have been developing anti-fertility vaccines. In one case (Talwar et al., 1976), the hormone human chorionic gonadotropin (HCG) is linked to a tetanus toxoid by a conjugate. When injected with the product, subjects develop antibody responses to the tetanus and the HCG by association. HCG is a necessary component for a successfully developing embryo. If a vaccinated subject becomes pregnant, the embryo will be spontaneously aborted. The conjugate here plays the role of an encryption apparatus. The body learns to mis-associate part of itself (HCG) with a toxic substance. This autoimmune response is a qualitatively different process from the normal process of an antibody response. Thus, we can say that learning has taken place.
With this example, we can confirm that biological systems can learn new encrypted codes, that is, new symbols, arbitrarily relating one thing to another. In this case it is irrational because it triggers an autoimmune response.
Conclusions
Poetic form and biological form stand in a homologous relation: it is the regularity of form (or habit) which makes creative evolution possible as formal disruption affording the possibility of new meaning (or functions)…To use Stuart Kauffman's (2013: 1−24) example, what was simply a ‘fish jawbone’ becomes understood as ‘mammalian ear.’
As Dietrich and Kanso (2010; 822) note, ‘Creativity is a cornerstone of what makes us human, yet the neural mechanisms underlying creative thinking are poorly understood.’ In this paper, I have tried to describe new learning in terms of semiosic processes that involve physical similarities of signs and the physical proximity of signs and/or signal pathways (such that they can associate with or interfere with each other). Understanding new learning in such physical terms, makes it easier to imagine how cognition does, in fact, occur in simple organisms, body tissue and plants, not just in organisms with complex neural tissue.
Biography
V. N. Alexander is a novelist and Director of Art-Science Programs at Dactyl Foundation in New York. She works in the field of Biosemiotics, focusing on the issue of creativity and she is widely published on Vladimir Nabokov's theory of butterfly mimicry, exploring the physiological processes that constrain wing pattern formation resulting in resemblances between species. She is currently exploring Alan Turing's work in reaction–diffusion modelling to understand how spatial and temporal neural patterns may become associated with different functions.
References
Additional information
Competing interests
The authors have no competing interests.
Author contributions
V.A.: Conception or design of the work, drafting the work or revising it critically for important intellectual content, final approval of the version to be published and agreement to be accountable for all aspects of the work.
Funding
No funding was received for this work.
