Introduction

Philosophers and biologists have become increasingly occupied with questions concerning plant behaviour and signalling. Specifically, they want to know what such processes say about plant ‘sentience’. Do plants have subjective experiences? As more is learned about the ways plants interact with their surroundings the clearer it becomes that plants live lives vastly more intricate than their seemingly sessile nature lets on. Although plants are not motile like animals they can twist and turn in place and exhibit directed growth. Plants grow with great precision towards sources that are beneficial to their persistence and away from those that are not. What makes this most intriguing is that tropisms, as such behaviours are called, as well as the release of volatile chemicals to fight off herbivores (Taiz et al. 2018, pp. 705–706) and altruistic behaviour towards kin (Dudley and File 2007; Pennisi 2019) are all done without a nervous system. And yet, it seems almost as if plants are making informed ‘choices’ about where they are ‘headed’, as if they are thinking, feeling entities similar to animals. The behaviours they express, the ways in which they grow make sense to us and invoke, especially when watching time-lapse footage of plants, a sense of goal-directedness, a sense of intention. The question now is whether such behaviours and the structures underlying them can be taken as evidence in favour of plant sentience.

Unsurprisingly, there is little agreement on the matter. In one camp are those who argue that we have neglected to see plant behaviour and physiology as indicators of consciousness. In this camp terms usually reserved for brain-having entities are being extended to plants: some argue that plants are intelligent (Baluska and Mancuso 2009; Garzón and Keijzer 2009; Calvo Garzón and Keijzer 2011; Calvo 2016, 2017; Calvo et al. 2020; Calvo and Trewavas 2021), that plants are cognitive agents (Calvo and Friston 2017; Segundo-Ortin and Calvo 2019; Linson and Calvo 2020), that they have a mind (Gagliano 2017; Maher 2017), that they are sentient (Pelizzon and Gagliano 2015; Calvo et al. 2017), and that plants are conscious (Trewavas and Baluška 2011; Baluška 2016; Trewavas et al. 2020; Calvo et al. 2021).

In the other camp, we find those who argue against these more expansionist approaches. There are those who argue that plants are not intelligent (Firn 2004), are not cognitive agents (Adams 2018), and are neither sentient nor conscious (Tye 2017; Ginsburg and Jablonka 2019, 2021; Taiz et al. 2019; Godfrey-Smith 2020; Hamilton and McBrayer 2020; Mallatt et al. 2021).

Confusingly, there is little conceptual clarity within the literature concerning concepts like cognition and consciousness and there is some interchangeability or strong semantic overlap. Take the following quotes as examples of this: “The scientific pursuit of nonhuman consciousness and sentience, understood as the capacities to be aware of the environment and to integrate sensory information for purposeful organismal behavior, has been a research priority for decades” (Trewavas et al. 2020, p. 216). Here sentience and consciousness are understood simply as the capacity to take in and use information to behave adaptively, a definition that overlaps with this one for cognition and intelligence: “(…) we put the emphasis on agency and adaptivity instead of conceptual competence, and we conceive of cognition primarily as intelligent behavior—that is, as the capability of organisms to actively interact with the environment in adaptive, flexible and sophisticated ways so as to maintain their systemic autonomy.” (Segundo-Ortin and Calvo 2019, p. 70).

In this paper, to minimise conceptual confusion, I will be using the term ‘sentience’ to refer to ‘phenomenal consciousness’, the ‘what it’s like’ of plants (Nagel 1974).

The extent to which scholars working on sentience agree that plants are sentient can be gleaned from the 30 commentaries to a recent article in the journal Animal Sentience by Segundo-Ortin and Calvo (2023) in which they argued for plant cognition and sentience. They present a two-pronged argument for plant sentience. First, they argue that plant behaviour indicates cognition. Second, they present evidence that the mechanisms underlying such behaviour is functionally equivalent to the animal nervous system. From this they maintain that we should take seriously the notion that plants too might be sentient i.e., that it is like something to be a plant.

Of the 30 commentaries to their article, only 2 were expressly in favour of the article’s views (Henning and Mittelbach 2023; Rouleau and Levin 2023), 12 either did not state their position or were on the fence (Booth 2023; Brooks Pribac 2023; Burgos and Castañeda 2023; Carls-Diamante 2023; Harnad 2023; Ivanchei et al. 2023; Milburn 2023; Pessoa 2023; Plebe 2023; Tiffin 2023; Vallverdú, 2023; Yilmaz 2023), and 16 were either opposed to the views addressed in the target article or highly sceptical of them (Baciadonna et al. 2023; Bennett 2023; Birch 2023; Broom 2023; Carranza-Pinedo 2023; Correia-Caeiro and Liebal 2023; Damasio and Damasio 2023; Dolega et al. 2023; Dung 2023a; Gutfreund 2023; Mallatt et al. 2023; Mastinu 2023; Robinson et al. 2023; Solé, 2023; Struik 2023; Ten Cate 2023). If we take these commentaries as a representative sample of the opinions of the field of consciousness research, then the field is at least sceptical of the idea that plants are sentient.

This paper does not set out to answer the question of plant sentience. Instead, I will present a diagnosis of the debate as it currently stands and outline a way forward. I will argue that we should move away from trying to use physiological similarities and some forms of behavioural similarities as a way to infer sentience. Instead, we should adopt the theory-light approach by Birch (2022) which searches for clusters of abilities facilitated by consciousness. The future of plant sentience research should focus on constructing plant-appropriate tests for dissociation between conscious and non-conscious processing.

The paper is structured as follows: In part 1 I present the arguments in favour of plants being cognitive agents. I then present evidence to the contrary and argue that the type of cognition plants supposedly have is not indicative of consciousness. Part 2 gives a short introduction to the nervous system and in part 3 I discuss the field of research known as plant neurobiology which argues that plants possess the functional equivalent of a nervous system. I present arguments in favour of this approach and then, in part 4, turn to a list of dissimilarities. I set aside the issue of defining the nervous system and argue that we have no way of settling the issue of whether the supposed plant equivalent to the nervous system is tied to conscious experience. Thus, neither plant cognition nor plant physiology, as currently understood, provides evidence of plant sentience. In part 5 I propose a way forward. I argue that the specific physiological and behavioural similarities pointed to by the proponents of plant sentience as evidence in favour of plant sentience should be abandoned and instead propose that we turn to paradigms which attempt to provide evidence of sentience by uncovering the presence of both conscious and unconscious processes. I argue that this paradigm should be adopted by those studying plant sentience. The focus of future research should be to try to construct plant-appropriate versions of dissociation paradigms.Footnote 1

Plant cognition

The first of two sources of evidence taken to speak in favour of plant sentience is the idea that plants are cognitive agents (Calvo and Lawrence 2022). The cognitive capacities of plants is understood to be evident through their behaviour which includes communication with other plants, recognition of kin, decision-making, risk sensitivity, anticipatory behaviour, learning and memory, foraging and competition, mimicry, numerosity, and swarm intelligence (Segundo‐Ortin and Calvo 2022; Segundo-Ortin and Calvo 2023). Many have suggested that certain liberties are being taken when some of the above-mentioned behaviours are attributed to plants. To make this point we needn’t run through every aspect of plant behaviour. Instead, I will mainly focus on the claims regarding communication, learning, and anticipatory behaviour.

Communication

One of the ways plants are said to communicate is via volatile organic compounds (VOCs) that are carried through the air an picked up by neighbouring plants of either the same or other species giving rise to adaptive interactions:

“Plants speak the silent language of scents. They do it through their leaves, shoots and roots, and of course through their glowers and fruits; trees discharge them into the open air even through their barks. Virtually all plants have mastered the tricks of chemical talks, synthesising and releasing into the air many different volatiles (…). We may picture each volatile itself as a building block in the vocabulary of plants, with ‘words’ being made up of many different organic compounds (…).” (Calvo and Lawrence 2022, p. 83).

Such compounds can be released when plants are subject to attacks from herbivores. Neighbouring plants can pick up the VOCs and elicit their own defence mechanisms. Is this a form of communication? Correia-Caeiro and Liebal (2023) argue in the negative. The minimal definition of animal communication requires intentionality meaning that, at minimum, the sender intends to produce a response in the receiver (2). They argue that plants have not been shown to send signs intentionally:

“S&C [Segundo-Ortin and Calvo] suggest that communication between plants occurs when they release Volatile Organic Compounds (VOCs) during a stressful event (e.g., being eaten). This allows other plants to detect the VOCs and trigger defence mechanisms. However, being able to detect VOCs in the environment is not communication; it is “behaviour-reading” at best, or, more parsimoniously, it is just “environmental-cue-reading”.” (3).

Given this, it seems we are not justified in claiming that plants communicate with one another unless we adopt a very abstract definition of communication understood as the ability to interpret cues from other organisms in order to produce adaptive behaviour irrespective of intention.

Learning

Another capacity plants are said to possess is the capacity to learn (Segundo-Ortin and Calvo 2023). The question of plant learning has a rather fraught history (Adelman 2018). There have been several studies trying to show that plants learn by habituation (Abramson and Chicas-Mosier 2016). Habituation is a type of non-associative learning where an entity, by repeatedly being presented with a stimulus decreases its response to that stimulus. One of these studies was done by Gagliano et al. (2014) on the plant Mimosa pudica which concluded that they were able to learn by habituation. This study has since been questioned because the data allow for an alternative explanation, namely that what seemed like habituation might be the result of motor fatigue (Biegler 2018).

There have also been studies of plants learning by association since the 60’s with mixed results (Adelman 2018). Associative learning occurs when an entity associates two stimuli, such as a bell and food in the case of Pavlov’s dogs, using one, e.g., the bell, as an indication of the other, i.e., the food. Gagliano et al. (2016) conducted the most rigorous experiment since those done in the 60’s with a clear positive result, namely that plants learn by association.

Others have since tried to replicate the study, but replication has failed (Markel 2020a; Ponkshe et al. 2023).Footnote 2 This has led several researchers to argue that the evidence for learning is inconclusive (Loy et al. 2021; Mallatt et al. 2023) or that it has not been demonstrated:

“Existing studies on plant associative learning face several major problems, some of which are: inappropriate experimental designs, lack of replicability, lack of proper control groups, high variability in the results reported, frequent use of small sample sizes (< 15 subjects per group) and inadequate knowledge about what a CS and a US could be for plants. Although some of these issues could be correctly tested experimentally, the general picture emerging from the studies to date support neither the inference that plants are capable of simple associative learning nor that they exhibit higher order cognitive processing (e.g., numerosity) and sentience.” (Baciadonna et al. 2023, p. 4).

These assessments should, however, be made with caution precisely because the study of plant learning is still waiting to mature. There simply is not enough evidence to make an assessment either way.

Anticipatory behaviour

Another indicator of cognition which plants are said to exhibit is anticipatory behaviour understood in the sense of ‘acting ahead of time’, thereby indicating that plants are predicting entities (Calvo and Friston 2017; Calvo and Lawrence 2022; Segundo-Ortin and Calvo 2023).

Two examples are given as evidence for this behaviour. The first concerns the reorientation of the leaves of some flowers during the night in what seems like an anticipation or prediction of where the sun will rise the next day, thereby maximising photosynthesis (Calvo and Lawrence 2022, pp. 74–75). Does this behaviour indicate anticipation? Mallatt et al. (2023) argue that another parsimonious explanation can be given of the behaviour, namely that the plant has an inherent timer and “a command to ‘reorient to the opposite direction of the previous sundown’ (…).” (6). Of course, this does not prove that plants do not anticipate where the sun will rise but it does provide a less demanding explanation of the process.

The second example given as evidence of anticipatory behaviour occurs at the root level of the young pea plant Pisum sativum (Segundo-Ortin and Calvo 2023, p. 9). The plants were grown with their roots in different pots with variable nutrient regimes, some being unchanging and some changing their amount of nutrients (Shemesh et al. 2010). Plants developed greater biomass in richer patches but, interestingly, they developed more biomass in patches with increasing nutrient levels even when they also had roots in patches that were more nutrient rich. This leads one of the researchers of the study to conclude: “These findings demonstrate that rather than responding to mere absolute resource availabilities, plants are able to perceive and integrate information regarding dynamic changes in resource levels and utilize it to anticipate growth conditions in ways that maximize their long-term performance.” (Novoplansky 2016, p. 63). This study is supposed to show that plant anticipate which conditions will become favourable in the future. Again, Mallatt et al. (2023) argue that these experiments have a much more parsimonious interpretation, namely, that the Pisum roots have a command to respond to increasing nutrients (6). As with the reorientation of leaves during the night, the science is not settled. Further, the example of the young peas, as opposed to the reorientation of leaves at night, can be framed in terms of reactivity to a stimulus rather than proactive behaviour setting it apart from animal anticipatory behaviour which is often based on a mental map (Mallatt et al. 2021, p. 466).

Thus, for both examples given for anticipatory behaviour, a more parsimonious explanation can be provided thereby putting into question the original interpretation of the experiments.

Cognition?

Given the above, are we still justified in claiming that plants are cognitive entities? The question of how widely cognition is distributed in the biological world has given rise to what Adams (2018) has called a ‘cognition war’. On one side we find those who advocate for a view of cognition which defines cognition first and foremost from human entities (Adams 2018), on the other, we find those who argue that nonneural entities like plants and even bacteria are cognitive entities (Segundo-Ortin and Calvo 2019), what have been called the anthropogenic and the biogenetic approaches, respectively. Some have argued that we should refrain from trying to find the necessary and sufficient conditions for cognition, claiming that cognition is not a natural kind (Allen 2017). Others argue that, instead of discussing whether plants fit into a certain definition of cognition we should treat the debate on plant cognition as being one of hypothesising about what cognition might be (Colaço, 2022). Further still, Lee (2023) argues that we should refrain from dichotomous thinking, such as demanding that plants either are or are not cognitive agents and instead adopt a degrees view of cognition whereby plants satisfy some aspects of cognition to some degree.

For our purposes we can set aside the question of whether plants are real cognitive agents. What we can say, given the above discussion on the supposed cognitive capacities of plants, is that if our definition of cognition is to include both plant and animal behaviour then it must be a definition of cognition that does not necessitate that the entity exhibit communication, learning, nor anticipatory behaviour or at least, only exhibit those in ways that are defined more abstractly than were they just to encompass animals. I take this to be an uncontroversial claim which still allows for overlap in some capacities of plants and animals.

Importantly, the proposed cognitive behaviour of plants is meant to shift, in a positive direction, our beliefs in the proposition that plants are sentient: “The interest in plant sentience emerges from observations of cognitive capacities in plants. Cognition in plants can be inferred from changes in their behaviors that improve their chances of survival.” (Segundo-Ortin and Calvo 2023, p. 4). For plants to be taken as cognitive agents cognition must be understood more broadly than if cognition were a feature only of animals. However, we do not know what the appropriate level of abstraction is if the behaviour is supposed to be tied to consciousness. This means that we cannot make a proper assessment of the extent to which the list of differences and similarities between animal and plant cognition speaks for or against plant sentience when we do not know whether plants are sentient.

To conclude this section, given the scant evidence of behavioural similarity (communication, learning, and anticipatory behaviour), it seems that the view that plant behaviour is cognitive is somewhat tenuous. However, while it may be unclear whether plants exhibit cognitive behaviour, we cannot, from this behaviour, settle whether it speaks for or against the notion that plants are sentient. The behaviours pointed to, on their own, are not indicative of sentience. Because of this, one might assume, the structural aspect of plants, i.e., the processes that give rise to their behaviour, bears an especially heavy burden of proof. The belief from the proponents of plant sentience is that the combination of the type of behaviour outlined above with the structural features which I will outline below, give us reason to think that plants are sentient. Before turning to the field of research which studies the supposed functional equivalent to the animal nervous system called plant neurobiology, I will briefly sketch the animal nervous system.

The nervous system: an interlude

Several theories of consciousness see the nervous system as a necessary condition for sentience (Ginsburg and Jablonka 2019; Godfrey-Smith 2020) and some that a functionally equivalent system justifies inferring sentience (Tye 2017).Footnote 3 There are good reasons to connect sentience or the having of feelings with a nervous system. In humans (and other animals), adaptive responses can be accompanied by a phenomenal experience: the instant removal of a hand in contact with a hot stove, thirst, hunger etc. Damasio and Carvalho (2013) argue that feelings have a neuronal basis because a nervous system is tied to behaviour and certain feelings are correlated with certain behaviours. And so, while a nervous system might not be sufficient for sentience it is intimately tied to the instances in which we have subjective experiences. And because of this, the argument goes, the nervous system is necessary for feelings.

For now, I will set aside the question regarding nervous systems being necessary for consciousness and instead give a brief description of the nervous system which we will use to understand the claims regarding functional equivalence made by the proponents of plant sentience between the animal nervous system and the vascular system of plants.

So, what is a nervous system? Briefly, a nervous system is comprised of the sum of all the neurons in an organism (which are sometimes clustered into what we call brains) and controls the behaviour, physiology, and development of an organism (Jékely et al. 2015, pp. 2–3). Neurons are electrically excitable cells that transmit chemical and electrical signals to one another making up an intricate information transmission highway.

Chemical signals are sent across small gaps between neurons called synapses. The presynaptic neuron releases chemicals into the synapse which are received by the dendrites of the postsynaptic neuron. This can result in a voltage increase past a critical threshold (thus becoming positively charged) which results in the postsynaptic neuron firing producing an action potential. This action potential results in the triggering of the release of chemicals in the postsynaptic neuron sending chemicals to the neurons it itself is connected to. The chemicals released into the synapse are called neurotransmitters and depending on the type of neurotransmitter, the postsynaptic neuron can either become excited (increasing positive charge and thereby likelihood of action potential) or inhibited (decreasing positive charge/increase negative charge and thus decrease likelihood of action potential). These neurotransmitters include serotonin, dopamine, norepinephrine, acetylcholine, and glutamate which increase excitability and GABA which decreases excitability.

Importantly, there is still disagreement as to why the nervous system evolved. Recently, the question of both the function and the evolutionary origins of the nervous system have re-emerged in scholarly debate (Keijzer et al. 2013; Jékely et al. 2015; Jékely et al. 2021). Here we will look at the three most discussed functions of the nervous system. The three views are the Input–Output view (IO), the Internal Coordination view (IC), and the Reafference View.

The input–output view (IO)

Keijzer et al. (2013) present what one might call the ‘received view’ of the function of the nervous system citing a definition of the nervous system taken from a current neurobiology textbook: “Nervous systems enable organisms to receive sensory information from their external environment, process this information and regulate neurosecretory and motor systems.” (Keijzer et al. 2013, p. 67). This view of the nervous system Keijzer et al. call the ‘Input–Output View’ (IO). In IO, the function of the nervous system is to produce adaptive behaviour (regulate neurosecretory and motor systems) by receiving an input (sensory information) and interpreting the input (process information) (Keijzer 2015, pp. 316–317).

The internal coordination view (IC)

The internal coordination view,Footnote 4 proposed by Keijzer (2015) and Keijzer et al. (2013) argues that the initial function of the nervous system was not to perform input–output processes but was an adaptation for the coordinated movement of bodies of a specific size. It is clear that coordinated movement does not require a nervous system. Bacteria and other microscopic organisms move around perfectly well without nervous systems. Instead, they use flagella or cilia. But, Keijzer argues, once organisms are of a certain size and are constituted by a certain number of cells a new system for joint movement is required. Cilia and flagella simply cannot propel organisms above a certain size forward. Enter nerve nets:

“We hold that the fundamental problem here was not so much to act intelligently—a problem that had already been solved in various ways without a nervous system—but to act as a single multicellular unit. In this story, nervous systems did not evolve initially to provide a more efficient information processing device. Nervous systems arose as a source and coordinator of patterned activity across extensive areas of contractile tissue in a way that was only loosely constrained by sensor activity.” (Keijzer et al. 2013, p. 68).

On this view, nervous systems enabled the contraction of muscle tissue, which enabled coordinated movement of larger organisms. It is important to stress two things about this view. Firstly, the view emphasises muscle tissue. The idea is that nerve nets connect to and enable the contraction of muscle tissues which in turn allows for motile movement. In addition, the information processing, felt or not, is done not in relation to the outside world directly but indirectly by the ability or inability to make contractions. This means that an organism need not sense an obstruction like, say, a rock in its environment, it only needs to sense that it cannot move its body in a certain direction. The organism is in a sense blind to the world but can ‘see’ the world indirectly by its internal restrictions: “When the external substrate obstructs bodily movement, the animal would sense the obstruction as a disturbance of patterning within the Pantin surface [the surface that can be contracted] even without any dedicated external sensor. Internal sensitivity suffices to generate a basic form of active touch.” (Keijzer 2015, p. 326).

The reafference view

Jékely et al. (2021) present another candidate view of the function of the nervous system called the Reafference view. Reafference is defined as the effect an organism’s own actions have on what it senses as opposed to exafference which are effects sensed by the organism due to external events (Jékely, Godfrey-Smith and Keijzer, 2021, p. 1). In short, an organism moving its own body changes what it sees (reafference) whereas light from the sun might cause an organism to squint its eyes (exafference). In addition, the ‘reafference principle’ states that: “self-initiated action evokes sensory effects that are correlated with these actions and, therefore, can be predicted and used.” (Jékely et al. 2021, p. 2). It seems that the movements, the contortions, and deformations of an organism, can themselves be used to make predictions about the world. For example, if almost every time an organism performs a specific deformation of its body objects appear closer to it, then the deformations of the body can be used to predict the distance to that object. A loop is created. This example stresses perception but the form of reafference presented is more general and takes any form of sensing (touch, sound etc.) that is the result of bodily changes as a case of reafference. Importantly, reafference is not a mechanism, it is a ‘feature of sensory episodes’ (Ibid). It is noted that many of the animal’s senses will be affected by the animal’s movement resulting in both reafferent and exafferent sensory episodes.

It is unsettled whether any of these theories capture why nervous systems evolved or what they are for. What is important is that there is no agreed-upon functional definition of the nervous system making claims of functional equivalence troublesome. I will now turn to the arguments from the proponents of plant sentience that plants have a system functionally equivalent to the animal nervous system.

Plant neurobiology

The first body of evidence which should increase our belief in the notion that plants are sentient was the behavioural plasticity exhibited by plants, evincing, so the argument goes, cognition. The second body of evidence for plant sentience comes from the structure underpinning their behaviour, a structure taken to be a functional equivalent of the nervous system.

“Although it is undoubtedly true that plants do not have neurons (and synapses) that could give rise to a ‘brain’ or a ‘nervous’ system, they respond electrically to many different environmental factors. Plants possess cells capable of electrical signaling and transmission; that is, cells that are functionally equivalent to animal neurons.” (Segundo-Ortin and Calvo 2019, p. 68, my italics).

The seeming similarity between the animal nervous system and the vascular system of plants lead to the emergence of the field of plant neurobiology (Baluška et al. 2005; Brenner et al. 2006):

“Plant neurobiology is a newly initiated field of research aimed at understanding how plants perceive their circumstances and respond to environmental input in an integrated fashion, taking into account the combined molecular, chemical and electrical components of intercellular plant signaling. (…) the goal of plant neurobiology is to illuminate the structure of the information network that exists within plants.” (Brenner et al. 2006, p. 413).

In this section I will provide the evidence given from those working within plant neurobiology for the functional equivalence between the plant vascular system and the animal nervous system. In Sect. 4 I will present arguments against the functional equivalence.

Plants have within them two major networks reaching from root to stem, to leaves to flower. One of these systems is called the xylem. It consists of dead cells and transports water and minerals from the roots up throughout the plant. The xylem is a one-way highway. Water only moves up, not down. The other system is the phloem and, like the xylem, is spread throughout the plant body. This system transports sap (sugar) made through photosynthetic processes in the leaves to the areas of the plant that need the nutrients i.e. it moves resources from sources to sinks. The parts that receive sap might be the roots or it might be developing leaves which themselves are incapable of photosynthesising until they reach a certain stage in development. Once they do so they too can start photosynthesising chipping in to feed the whole system. What is important here is that the phloem is made of living cells, not dead cells like in the xylem, and that the sap can move both up and down the plant body through the phloem.

But, the phloem does more than transport sap. If it did not it wouldn’t be much of a functionally equivalent system to the nervous system. The phloem, because it is made of living cells, can send electrical signals and regulatory molecules up and down the plant in relation to external stimuli. This system is a complex network of vascular tissues (Calvo et al. 2017, p. 2861) allowing for long-distance signalling which produce behavioural responses to the environment (Calvo and Trewavas 2021) and the plant ‘neuron’ that allows for this signalling is assumed to be what are called sieve tube elements i.e. the elements that make up parts of the phloem (Calvo et al. 2017, pp. 2862–2863). Such regulation of behaviour is exemplified by the many tropisms of plants such as phototropism, geotropism, thigmotropism, and hydrotropism. Take the case of phototropism. Phototropism occurs when a plant grows towards a light source optimising light absorption. Auxin, a growth hormone, elongates the cells on the side of the stem that sit in the shade. Because of this, the cells on one side of the plant grow longer than those on the other. This in turn bends the direction of growth towards that specific light source (Taiz et al. 2018, pp. 535–536). In addition to producing the growth behaviours, i.e., tropisms, the phloem also plays a key role when the plant is set upon by hungry herbivores resulting in the production of volatile molecules as a defensive measure (Kumar et al. 2020, p. 967).

The phloem enables plants which consist of multiple cells to function as one coherent individual: “The phloem is (…) one huge and ramified cable-like compartment spanning the whole plant body, connecting all plant organs into one unified huge axon-like super-cell. Intriguingly in this respect, phloem tubes provide a low-resistance medium allowing rapid spreading of plant-specific AP [action potential] throughout the plant body. This rapid electric signalling integrates the whole plant body into physiological and cognitive unity, allowing vascular plants to act as individualities having both plant-specific agency and cognition.” (Baluška and Mancuso 2021, p. 3).

To strengthen the analogy between the phloem and the animal nervous system further, plants also, it is argued, have the functional equivalent of gap junctions, the channel connecting the post and presynaptic cell membranes. As noted, electrical signals can travel long distances via the phloem but also short distances between plant cells via what are called plasmodesmata. Plasmodesmata are small channels that traverse plant cell walls meaning that two adjacent cells will have small gaps in the walls between them through which electrical and chemical signals can be sent. This small gap between cells which allows for cross-cell communication is then taken to be functionally analogous to gap junctions: “Thus, whilst plants lack gap junctions (pervasive intercellular channels coupling animal cell cytoplasm), plasmodesmata serve as functional analogues, facilitating direct symplasmic connection between adjacent cells.” (Lee and Calvo 2023, p. 10).

The proposed equivalence does not end there. As stated, the cells that allow for electrical signalling do so by action potentials just as in neurons (Baluška 2010; Lee and Calvo 2023). In addition, there are receptors for neurotransmitters: “[T]here is evidence that neuronal-like receptors for neurotransmitters glutamate and GABA are expressed in plants. As in neurons, GABA decreases, and glutamate increases, the excitability of plant plasma membrane.” (Calvo and Trewavas 2020, p. 215). In fact, many of the neurotransmitters found in animal nervous systems are present in plants: “Chemicals known to play a role as neurotransmitters in animals, such as acetylcholine, glutamate, dopamine, histamine, noradrenalin, serotonin, and g-aminobutyric acid (GABA), are also found in plants.” (Segundo-Ortin and Calvo 2023, p. 16).

Summed up, the argument for functional equivalence between the vascular system of plants and the animal nervous system goes like this: plants have an internal system, the phloem, that is extended throughout the plant body (like the nervous system). This system sends electrical signals and regulatory molecules throughout the plant body in response to external stimuli enabling adaptive responses (like the nervous system). The phloem consists of cells, the sieve tube elements, that send electrical signals throughout the plant body and do so using action potentials (just like neurons). They send the same neurotransmitters as are common in animal nervous systems through gap junctions in the form of plasmodesmata (just like neurons and synapses). Thus, we have a system, the argument goes, that is functionally equivalent to the animal nervous system because it is extended throughout the plant body, consists of electrically excitable cells which communicate electrochemically using neurotransmitters across gap junctions in order to produce adaptive behaviour.

What I will do in the next section is present some of the criticisms levied against the alleged functional equivalence made in plant neurobiology by other plant biologists and philosophers. I will argue that, just as in the case of cognition, even though there are many differences and similarities between the plant and animal nervous systems, we cannot come to know whether these differences and similarities count in favour of or against the background claim of plant sentience. This is fundamentally because the very strategy employed to infer sentience by looking for functional similarities without a commitment to some theory of what consciousness does will inevitably leave any purported equivalence mute.

Functional equivalence?

It may come as no surprise that not everyone accepts the idea that plants possess a system that is functionally equivalent to the animal nervous system and therefore do not accept that plants are sentient. Several biologists and philosophers have advanced arguments for why we should abandon this notion of functional equivalence (Alpi et al. 2007; Struik et al. 2008; Taiz et al. 2019, 2020; Hamilton and McBrayer 2020; Draguhn et al. 2021; Ginsburg and Jablonka 2021; Mallatt et al. 2021; Robinson et al. 2023). To them, the supposed plant nervous system and the animal nervous system are simply too dissimilar to warrant any claim of sentience in plants by analogy:

“Plant cells do share features in common with all biological cells, including neurons. To name just a few: plant cells show action potentials, their membranes harbor voltage-gated ion channels, and there is evidence of neurotransmitter-like substances. Equally, in a broader sense, signal transduction and transmission over distance is a property of plants and animals. Although at the molecular level the same general principles apply and some important parallels can be drawn between the two major organismal groups, this does not imply a priori that comparable structures for signal propagation exist at the cellular, tissue and organ levels. A careful analysis of our current knowledge of plant and animal physiology, cell biology and signaling provides no evidence of such structures.” (Alpi et al. 2007, p. 136).

Below I will point to a few dissimilarities flagged by the opposition. Whether these dissimilarities speak against the possibility of plant sentience I will discuss at the end of this section.

Signalling speed

One of the large differences between plant cells and neurons is their signalling speed (Plebe 2023). Mallatt et al. (2021), mention up to five differences in signalling in animals and plants one of them being the speed at which action potentials travel: “Plant action potentials travel more slowly than those of animals, 0.04–0.6 m s−1 versus 0.5–100 m s−1, respectively, and with long refractory periods between successive action potentials.” (Mallatt et al. 2021, p. 462).

This point is noted by Lee and Calvo (2023) in their analysis of the similarities and differences between plant cells and neurons. Yet, they note that some of the differences in signal propagation are not as large when discussing the Mimosa and the Venus flytrap. While this is the case, these particular plants are outliers among the tracheophytes meaning the analogy would only work for this subset of plants and even if this were the case the action potentials are not system-wide: “In the specialized case of the Venus fly trap the refractory periods are much shorter. But these action potentials are restricted to the local organ and therefore cannot contribute to any system-wide neuronal information processing.” (Robinson and Draguhn 2021, p. 6).

Chemical differences

Even though there are several chemical similarities, it has also been noted that there are large chemical differences:

“Electrical activity in plants is powered by transport of H+ and that of animals by transport of Na+. (…) in plants, typically 50–70% of the free energy generated by the plasmalemma H+-ATPases goes into the electrical component (the membrane voltage), the rest into the pH gradient. In animals, it is the other way around: roughly 80–90% of the electrochemical gradient generated by the Na+ /K+- ATPases goes into the chemical gradient of Na+ (and K+), and only a small fraction goes into the voltage difference across the membrane.” (Mallatt et al. 2021, p. 462).

Initially, we might be convinced by this argument but we can counter it by proposing that the focus on the chemical differences, H+ and Na+, is actually a material, not a functional, difference and as such does not touch on the overall argument for functional equivalence.

That there are material differences between plants and animals (let alone their signalling systems) is obvious and, given functionalism, it seems we can leave out arguments for non-equivalence based on chemical differences. Or can we? Recall, in the previous section on the phloem, one of the arguments that spoke in favour of a functional equivalence was the use of neurotransmitters in both plants and animals, specifically the same neurotransmitters GABA and glutamate which had the same effect on both neurons and their supposed functional equivalent, sieve tube elements, decreasing and increasing excitability, respectively. Why would such material (chemical) similarity speak to functional equivalence while the material dissimilarity between H+ and Na+ can be dismissed? If we allow for a chemical-level description of the systems in question we cannot simply tally only those that speak in favour of equivalence, neglecting those that do not.

To counter this, those who argue for equivalence could state that the equivalence is not in the materials GABA and glutamate per se, but in their effect i.e., decreasing and increasing of excitability, respectively. This manoeuvre elevates the level of description to that of effects and functions from that of materials. If this is the course of action taken then all material similarities, when mentioned in connection with functional equivalence, should be accompanied with a disclaimer stating that they have no bearing on the equivalence advocated because if they do then the argument for material dissimilarity re-emerges. But if this move is taken then a rather important dissimilarity can be introduced: “Despite the presence of glutamate and GABA in plants there is also no data to show that they act as neurotransmitters.” (Robinson and Draguhn 2021, p. 8).

Bi-directional signalling

One key aspect of the plant nervous system is that it, like the animal nervous system, is said to exhibit bidirectional signalling:

“An illustration of the way plants respond selectively to salient features of the environment, proactively sampling their local environment to elicit information with an adaptive value, is provided by the hierarchical deployment of distinct vascular cell populations, encoding expectations in plants, and functionally analogous neural architectures in the case of animals, with cross linked and bidirectional (forward and backward) communication pathways.” (Calvo et al. 2017, p. 2859).

Not only is it important to establish bidirectional signalling in plants because that is what animal nervous systems do, many theories of consciousness take the integration of information to require either feedback or recurrent processing (Northoff and Lamme 2020, p. 571). Again, if such processes could be shown in plants this would, if not count in favour of the possibility of plant sentience at minimum not count against it. However, no bidirectional communication has been demonstrated:

“Such integrative electrical signaling [feedback/recurrent processing] is easily recorded among neurons in the brains of conscious humans as well as in brains of other mammals performing the same mental tasks, but it has never been detected in plant phloem or any other part of a plant. That is, forward signals are documented but feedback signals have not been found. Such integrating signals have merely been hypothesized to occur.” (Mallatt et al. 2021, p. 464, italics in original).

It is somewhat troubling that the advocates of plant cognition and plant neurobiologists argue for functional equivalence on, among others, the grounds of bidirectional signalling when such signalling has only been hypothesised. It may well turn out to occur in plants but as the equivalence is currently being sold, it does sound like it has already been demonstrated.

To sum the point of contention: those who argue against functional equivalence claim that the signalling speeds of plants and animal vascular systems are dramatically different, that not only are there numerous chemical differences, there is also little evidence that the chemicals function as neurotransmitters in plants, and that bidirectional signalling in plants has yet to be established.

So, is there functional equivalence?

Do these differences matter for functional equivalence? One proposal is to broaden the definition of nervous systems to include plants (Miguel-Tomé and Llinás, 2021). Others argue that, if it is similarities we are looking for then the plant vascular system is much more similar to the animal blood-carrying vascular system than the animal nervous system:

“Instead, plant vasculature is better compared to the blood-carrying vascular system of animals. Both these systems function to transport water, nutrients, mineral ions, and hormones throughout the body, and both conduct electrical signals over distances. (…) Both plant and animal vascular systems use electrical signaling to regulate the hydrostatic pressure. Thus, xylem and phloem cells are arguably more analogous to the endothelial cells of animal arteries, which also lack synapses, than to the neuronal cells of the nervous system. (…) Importantly, vascular networks, unlike neuronal networks, are not directly involved in the mental functions of animals. Plants and animals both have vascular networks, but only animals have neuronal networks.” (Robinson et al., 2023, p. 5).

Just as in the case of cognition, I will set aside the question of what the definition of the nervous system should be and whether plants can be said to ‘really’ have such a system. Instead, I will present two points which both lead to the same conclusion, namely, that questions about functional equivalence are mute without the adoption of a given framework from within which equivalencies can be made meaningful.

The first point is identical to the one made regarding cognition. We need not accept some definite definition of the nervous system but only note that there are differences between the plant vascular system and the animal nervous system and therefore, for plants to have a nervous system we must accept that our definition of nervous systems, whatever it is, needs to be broadened. In fact, proponents of plant neurobiology are aware of this, here talking about the differences in action potentials between plants and animals:

“It is worth pausing at this point to flag the fact that considerations of phylogeny help us to respond to a potential worry: given the molecular diversity of APs, their range of functions, as well as the fact the complex picture of APs within neural systems (such as their absence in some cells and complexity in others), one might question whether APs are even a useful cross-taxa category. Whilst we cannot fully address this point presently, we hope to have shown that what we have been referring to as APs across animals and plants share a sufficiently robust ‘generic structure’, and enough interesting molecular and functional similarities, to warrant preserving the category.” (Lee and Calvo 2023, p. 18).

The second point, related to the first, is more fundamental having to do with drawing functional equivalencies in a ‘theoretical vacuum’. When we have no framework within which we can ask questions both the similarities and the differences between the two systems in question are meaningless. We, as it were, need a frame of reference. Without one we simply do not know which level of abstraction, which level of description we should adopt before we search for equivalencies. To illustrate this point in simple terms imagine two factories, one producing cars, the other producing boats. At one level of functional description the two factories are functionally equivalent; they both produce transportation vehicles. Yet, at a more fine-grained level of description, say at the level of maritime transportation, the two factories are not functionally equivalent as one factory is for the production of vehicles that can travel across bodies of water, the other is not. There is no privileged level of functional description which wholly captures the function of either factory. The same goes for the plant nervous system and the animal nervous system. Yes, there are differences and similarities, but we cannot judge to what extent any of them matter because we do not know what the proper level of functional description is for consciousness. This approach to inferring consciousness outside the human case is fundamentally flawed, what Birch (2022) called the theory-neutral approach. In what follows I will argue that we should refrain from the theory-neutral approach and the theory-heavy approach and start to move towards what Birch calls the theory-light approach to consciousness.

A way forward?: The theory-light approach

The current strategy for inferring plant sentience is likely to fail, not because of the scant evidence for cognition or similarities to the animal nervous systems but because the approach is fundamentally theory-neutral.

Birch (2022) distinguishes between three approaches to consciousness: theory-heavy, theory-neutral, and theory-light. I will discuss each in turn and argue that advocates of plant sentience cannot make headway because their approach is theory-neutral, an approach which, for reasons I will discuss, should be avoided. I will propose adopting the theory-light approach as a future direction for the study of plant sentience.

Theory-heavy approach

On the theory-heavy approach we first construct a complete theory of human consciousness and then use this theory to test for consciousness in non-human entities. The problem with this approach is what Birch calls the dilemma of demandingness. Any theory of consciousness we adopt such as Global Workspace Theory (GWT) (Carruthers 2019) or Higher-Order Thought (HOT) (Lau and Rosenthal 2011) does not tell us how much we can weaken the demands for consciousness while still maintaining what is sufficient for consciousness: “strong sufficient conditions for consciousness will not get us very far in making inferences about cases other than humans who can report their experiences, if they get us anywhere. Yet, as we formulate increasingly weaker conditions, the evidence from humans that they amount to a sufficient condition becomes weaker, and the positive case for animal consciousness becomes correspondingly weaker.” (Birch 2022, p. 6). Carruthers comes to a similar conclusion arguing that we cannot, ever, come to know whether non-human animals are conscious: “While our concept of phenomenal consciousness is all-or-nothing, the minds of other animals will only resemble the human global workspace more or less closely, to some or other degree. Moreover, because that concept is a first-person one, it can find no determinate application when applied to minds that are significantly unlike our own. The upshot, I have argued, is that there is no fact of the matter whether animals have phenomenally conscious states or not.” (Carruthers 2019, p. 163). The theory-heavy approach has no in-built method of escaping the human-centricity at its core. To avoid the dilemma of demandingness and the somewhat disheartening conclusion that some take to follow from it we might circumvent theory altogether.

Theory-neutral approach

The theory-neutral approach does just that. This approach, which I claim is implied in the debate on plant consciousness, takes as its method for uncovering consciousness in non-human entities the discovery of behavioural, functional, and anatomical/physiological similarities to humans and infers the presence of consciousness from these similarities, from analogy.

As mentioned earlier, the problem with analogy is that we have no way of knowing what level of description is appropriate to adopt when searching for consciousness. How should we define cognition or the nervous system? The more abstractly we define them the more entities reside under the terms. But how can we then come to know whether something is not conscious? This problem Birch (2022) calls the ‘inescapability of theory’ which essentially amounts to the problem that arises as soon as one argues that there is some aspect which defeats the analogy, e.g., if one claims that an entity is lacking some behavioural, functional, and/or physiological similarity, then that entity is not conscious. But on what grounds could we rule them out? And, further, on what grounds do we rule them in?

Here, one might argue that the theory-neutral approach can escape this worry simply because the human body is the theory. What I mean here is that the human body, the behaviours, functions, and physiology of humans constrains the set of possible similarities. This, however, is not a viable solution. To illustrate this point, take the story, perhaps apocryphal, about the two ancient philosophers Diogenes and Plato:

Plato was once asked to give a definition of ‘man’ to which he gave the following answer: “Man is a featherless biped.” One day, while Plato was giving a lecture at his academy, the philosopher Diogenes ran into the academy releasing a plucked chicken into the crowd declaring “Behold! Plato’s man.”

All would agree that Plato’s definition was too broad since it allowed plucked chickens to be labelled ‘man’ and thus, by extension, should lead them to be treated as such (whatever that meant in Plato’s timeFootnote 5). But chickens are not humans, without or without plumage. What this story nicely illustrates is a predicament in sentience research. It is not the case that we humans have a complete functional and physiological description of ourselves to which we can compare other entities in the world in order to infer whether or not they are good candidates for being sentient. In fact, we have fast-and-loose functional descriptions of ourselves, specifically of our nervous system, and, by extension, what makes us sentient. This is a product, I believe, of only a limited time in which the selection pressures on such descriptions have demanded specification. This stronger selection pressure, this tightening of especially the functional descriptions of ourselves, arises from comparative investigations of sentience. Finer-grained descriptions are suddenly demanded when we see that our current descriptions allow disparate phenomena to reside under the same heading. This is illustrated by statements of the following kind: “If that is what you mean by X, then this animal/plant/robot also has X.” The question then becomes one of biting bullets or returning to the definitional drawing board, the dilemma best captured in this quote by Dennett “[O]ne person's reductio ad absurdum is another's counter-intuitive discovery.” (Dennett 1995, p. 709).

Diogenes clearly meant his plucked chicken to function as a reductio thereby indicating an error in the definition given by Plato. Not so for those who believe plants are sentient. To them, the functional equivalence drawn between plants and the animal nervous system are a ‘counter-intuitive discovery’ and they are thus incentivized to retain aspects of the current functional description of the nervous system to maintain the equivalence.

On the assumption that the ‘inescapability of theory’ rules out the theory-neutral approach thereby demanding some theoretical commitment and that the ‘dilemma of demandingness’ rules out the theory-heavy approach Birch (2022) suggests a middle path: the theory-light approach.

Theory-light approach

The theory-light approach makes a minimal theoretical commitment, namely that phenomenal consciousness has some effect on cognition, that is to say, phenomenal consciousness is not epiphenomenal. The minimal theoretical commitment Birch (2022) calls the facilitation hypothesis:

“Phenomenally conscious perception of a stimulus facilitates, relative to unconscious perception, a cluster of cognitive abilities in relation to that stimulus.” (p. 8).

The idea that consciousness plays a functional role is adopted by several philosophers (Cleeremans and Tallon-Baudry 2022; Ludwig 2023) but is not without controversy (Seth 2009). For this paper I will assume that consciousness can play such a role.

What the facilitation hypothesis states is that there is some difference between the way a stimulus is processed consciously as opposed to unconsciously. Further, it is important to stress that the approach is concerned with a cluster of abilities, not a single ability, the idea being that there are several consciousness-linked cognitive abilities which can be turned on and off together.

How might we gather evidence that phenomenal perception facilitates a cluster of cognitive abilities relative to unconscious perception? Birch presents three possible tests for doing just this: trace conditioning, rapid reversal learning, and cross-modal learning. Trace conditioning is a form of classical conditioning and can be demonstrated by separating two stimuli by a temporal interval e.g., hearing a tone and then receiving a puff of air aimed at your eye (Droege et al. 2021). Humans can learn to associate the sound with the puff of air only if they are conscious of both the stimuli and the interval between them. Rapid reversal learning occurs when the subject learns the relationship between two stimuli and then learns the opposite relationship when it is reversed (by the experimenter). Just as with trace conditioning, subjects were only able to learn the reversal when conscious of the cues. The third test for uncovering conscious facilitation is cross-modal learning. Here, learning is done across sense modalities such as, for example, a consciously experienced odour can be associated with a consciously experienced sound, or visual stimulus. This association, the hypothesis goes, should be easier to make when conscious of the stimulus. Each of these cases illustrate what Doerig et al. (2021) call paradigm cases of consciousness. By pitting conscious and unconscious processing against each other they ensure that we are dealing with consciousness and not other co-occurring processes.

While it is possible that one of these abilities could be done without consciousness, we should remember that we are looking for a cluster of abilities which will strengthen the claim in favour of attributing consciousness. This is an important point. Birch (2022) is clear that there is no litmus test for consciousness or a list of markers we can look for like on the theory-neutral approach. Instead, the notion of ‘cluster of abilities’ is a commitment to the view that there are several consciousness-linked abilities which robustly correlate in that they can be turned on and off together. Such a turning on and off can be done through what is called masking. Masking switches perception on and off in humans and finding a similar pattern in the entity being tested would provide further evidence of consciousness. In such cases where there is a cluster of abilities which are turned off by masking the best explanation would be that the entity is conscious: “If we gather all this evidence for our target species, we will have the evidential basis for a scientifically credible inference to the best explanation to the presence of consciousness in that species. We will have started with an empirically supported hypothesis about the cluster of cognitive abilities that is linked to consciousness in humans, found evidence of that cluster in the target species, and found the same pattern of sensitivity to masking.” (Birch 2022, p. 12).

The theory-light approach is not without fault as Shevlin (2021) has pointed out. He presents three criticisms of the approach which I will now turn to.

Criticisms

The first criticism concerns a limitation of the theory-light approach to determine which mental states of an entity are conscious. The tests presented above all focus on perceptual states and therefore do not aid in uncovering other mental states such as emotions, beliefs, and desires. Yet, this is only a problem for the entities which show no signs of conscious facilitation of perception. Whether plants are among the set of entities for whom consciousness does not facilitate cognitive abilities but are still conscious remains to be seen. It would be ill-advised to not perform such tests on plants simply because they might not pass it.

The second criticism, which is closely related to the first, focuses on the possibility of false negatives. Shevlin writes: “While capacities like trace conditioning when present may provide evidence of the presence of consciousness, it is not clear why we should take their absence as providing evidence of the absence of consciousness, especially when dealing with systems with quite different cognitive architecture from humans such as artificial systems or simpler animals.” (Shevlin 2021, p. 308). This is not a proper criticism of the theory-light approach as Birch (2022) does not argue that failing the cognitive tests speaks against the entity being conscious. Rather, the theory-light approach maintains that if we find evidence of facilitation then we have a stronger case of inferring consciousness. If we do not find evidence of facilitation the approach does not decree that the failure equals non-consciousness. A positive result will be an indication of consciousness while a negative result simply would tell us that the entity does not show signs of conscious perception, not that it shows no sign of consciousness full stop. Again, it is possible that no evidence of facilitation can be shown in plants and that plants are conscious.

The third criticism focuses on the possibility of false positives. The example Shevlin (2021) gives is that of an AI engineered specifically to be capable of trace conditioning, rapid reversal learning, and cross-modal learning. While Shevlin specifically focuses on AIs the criticism can be broadened to any engineered systems, biological or synthetic (Davies and Levin 2023). Of course, such entities might actually be conscious but what Shevlin is after is “some principled way of assessing whether consciousness might be absent even despite the presence of the relevant markers.” (Shevlin 2021, p. 308).

A possible solution to this problem is presented by Dung (2023a, b). He distinguishes between ‘natural’ and ‘ad hoc’ AIs. The ad hoc AIs are those engineered to game the tests: “If an AI is constructed with the aim of passing a certain test of consciousness, it is not surprising that it succeeds. We can explain this success by appealing to the fact that it was designed to pass the test. By contrast, it is surprising when an AI passes a test for consciousness even if it was not created with the purpose of excelling at this test. In this case, that the AI passes the test cannot be explained in recourse to the aim of making it able to do so.” (Dung 2023b, p. 6). This move allows us to meet Shevlin’s third criticism by pointing to the origin of the system in order to explain why it passes the test something we cannot do for systems not designed to pass the test.

Before returning to the question of plant sentience I will introduce another approach to identifying consciousness in non-human animals.

Double dissociation

A modification to the dissociation between conscious and unconscious stimuli revealed by tests such as trace conditioning, rapid reversal learning, and cross-modal learning is so-called double dissociation paradigms where conscious processing is shown to improve learning while non-conscious processing impairs learning (Crump and Birch 2022, p. 117).

A recent study tested for the presence of such dissociation in rhesus macaques and humans (Ben-Haim et al. 2021). The test performed was a spatial cueing test meaning subjects had to locate a stimulus that would be present at one of two locations on a screen. Before the target stimulus was presented an incongruent cue was presented, followed by a mask. The cue is incongruent because it sits at the opposite location of the target stimuli. So, if the incongruent cue occurred on the right you know that the target cue will appear on the left and vice versa. This incongruent cue was displayed either supraliminally (for 250 ms) or subliminally (for 17 or 33 ms). The hypothesis Ben-Haim et al. tested was that when the cue was supraliminal i.e., was consciously perceived, this would facilitate learning. Further, when the cue was subliminally perceived, i.e., it was processed by the subjects but not consciously, this would impair learning. The results were the same for both humans and the rhesus macaques. When they perceived the cue consciously, they were faster than average at locating the target stimulus. When the cue was perceived unconsciously, they performed worse than average. Therefore, it could be concluded that such double dissociation tasks show evidence of consciousness in non-human animals as they provide evidence of consciousness facilitating, in relation to unconscious perception, cognitive abilities.

Crump and Birch (2022) provide an argument against this conclusion as well as a way to overcome it. They argue that there is a confounder in the double dissociation test by Ben-Haim et al. (2021). The confounder is that supraliminal cues are stronger signals than subliminal cues meaning they are easier to learn. Therefore, it cannot be ruled out that the reason that all subjects, humans and monkeys, learned to locate the target stimulus faster when the cue was supraliminal was because of a difference in signal strength. If so, then the test would not be evidence of two types of processing, conscious and unconscious.

The solution Crump and Birch (2022) propose is to find the ‘subjective threshold’ of the target species. This is done by changing the stimulus duration continuously throughout the test. When (or if) a marked difference in task performance is identified this would be evidence of a threshold for subliminal and supraliminal processing. This threshold should then be searched for across several cognitive tasks. If found it would be evidence that once the same threshold is met a marked difference in task performance occurs. What best explains this difference in processing would be consciousness.

Back to the roots

I maintain that the theory-light approach and double dissociation paradigms stand as viable approaches to assessing consciousness outside the human case. But how widely applicable is it? This is a worry readily admitted by Crump and Birch (2022):

“Rigorous evidence does not come cheap, and no one said animal consciousness research is easy. But we might worry about species without a wide enough cognitive repertoire for this approach. Perhaps it will work for octopuses or bees, but what about snails, earthworms, or sea slugs? The systematic facilitation approach can only deliver serious evidence of consciousness in species with the requisite cognitive sophistication. We do not know whether consciousness itself is limited to such taxa, or just our proposed cluster of cognitive markers.” (120).

If the approach has limited applicability to snails and earthworms one might fear that it cannot hope to be applied to plants especially because the tests presented which purport to show dissociation between two types of processing are learning-focused. As I have argued, given the current state of evidence on plant learning we should be sceptical about the possibility of plants learning by association. So, one might worry, if plants do not have the capacity to learn by association, then there is little hope for the application of the theory-light approach to plants. Importantly, the jury is still out on plant learning and more rigorous experiments are no doubt underway. But, suppose it turns out that plants simply cannot learn by association then we might attempt to construct dissociation tasks that do not rely on this capacity. The task ahead will be one where we test for cognition on the species’ own terms. In order to show dissociation in plants we will need to be inventive and construct plant-appropriate tests for cognitive abilities which in humans are facilitated by consciousness. This will be far from easy yet it will give us a clearer indication of what needs to be done as opposed to finding endless similarities and differences between plant and animal nervous systems leading to an endless tug-of-war.

Incidentally, finding the physiological underpinnings of dissociative processes in plants and animals would allow us to find the proper level of functional description for ‘nervous systems’ in plants in relation to those processes. Thus, the theory-light approach allows us to sidestep the question of physiological similarity and instead focus on a specific kind of cognitive behaviour that we know is tied to consciousness in humans, and, as a result, allows us to find the appropriate level of functional description of nervous systems.

Conclusion

The studies on plant ‘cognition’ and their ‘nervous system’ are not for naught. They have produced doubt. Some researchers are suddenly unsure about the status of plants and this doubt is necessary to get researchers engaged in and to acquire funding for research into plant sentience. The question of plant sentience is one of those fascinating question where, whichever answer is true we will all be in awe. If plants are sentient, then we need to rethink much of our current understanding in neuroscience. How could such a vascular system, different in so many ways from our own nervous system, give rise to consciousness? If plants are not sentient, then we are witness to a self-maintaining entity capable of complex cognitive behaviour without the presence of consciousness.

What needs to be done now is construct a series of experiments where the logic of those experiments is equivalent to those of human experiments within the theory-light paradigm. What these experiments will look like time will tell. In the meantime, we should move away from the theory-neutral approach as it provides no clear way forward. Similarities and differences in both behaviour and physiology are endless and can forever be used to construct arguments in favour of either position. To avoid such a stalemate, we should apply the theory-light approach to plant signalling and behaviour.