Extended cognition in plants: is it possible?

The boundaries of cognition: an exciting field of debate

All organisms have evolved ways of sensing and adaptively interacting with the world. In wild, real-world environments, where circumstances are variable, and survival is uncertain, organisms do not simply react to the external impingements but actively explore their environment and modify their behavior in flexible, context-sensitive ways related to fitness components such as growth, reproduction and ultimately, survival. Organisms must interpret the environment to fit well into it. The choices, decisions and actions they make to deal with the varying demands and challenges they face reveal their underlying intelligence.¹ Processes like remembering, making associations, perceiving, learning, solving problems, deciding and acting that bring about such intelligent behavior define their cognition.²

Much has been discussed about the limits and bounds of cognition across taxa. Such discussions are shaped by how we define ‘cognition’ and what we count as ‘cognitive’. The classical view of cognition (or cognitivism) assumes that a large brain and a complex neural system are necessary to support cognitive capacities. The core idea is that cognition pertains to the specific properties of the brain. The brain is treated as the information-processor that operates by receiving data from the environment in the form of symbolic inputs that it manipulates into internal representations to produce a relevant output in the form of different behaviors and experiences. Given such premise, it is unsurprising that we have reservations about ascribing cognitive capacities to animals with tiny brains such as insects, and difficulties with the idea of cognition in organisms such as plants that lack these anatomical structures entirely. Yet, despite their miniature brains, insects display a complex behavioral repertoire mediated by sophisticated cognitive abilities.³ And with no brain at all, plants too exhibit a wide range of exquisite behaviors,⁴ which has recently revealed unsuspected cognitive capacities in these non-neural organisms (e.g. Gagliano et al.^5,6). The new experimental evidence casts doubts on the validity of the classical view of cognition centered on the brain, and instead, more closely aligns with theoretical post-cognitivist models that have proposed alternative ways of exploring cognitive processes and their limits beyond neural bounds.^7–12

From a post-cognitivist perspective, cognition takes place in the context of a real-world environment and inherently involves perception and action. The process of perception-action occurs in the context of environmental contingencies, and the resulting behavior can be more or less affected by external stimuli.^13,14 The extent to which the environment influences the organism depends on the organism’s capacity to sense the environment and the organism’s level of autonomy from it (i.e. an organism’s capacity of keeping its auto-organization despite the environmental fluctuations that act on it; Souza et al.¹⁵, Souza and Lüttge¹⁶). From this perspective, cognition is not exclusively confined to an organism’s brain or the nervous system. Rather, it can be distributed, when the cognitive capacity required for completing a task is spread among many individuals of the same species (e.g. the ‘hive mind’ of ants in a colony or the crew in an aircraft cockpit); embodied, when some cognitive control is devolved to parts of the body other than the central brain, thus spanning the cognitive process to all the body of the organism (e.g. the multifaceted camouflage of the common cuttlefish and complex bending movements of the tentacle performed by the common octopus); and even extended, where physical objects in the world form part of the cognitive system of the animal that built or manipulates them (e.g. nest-building weaver ants and web-building garden spiders). For a review of these ‘species’ of cognition, see Cheng⁹ and Lavoie et al.¹⁷ bearing in mind that the overall classification of cognitive ‘species’ provides us with an initial context in which to specifically design and interpret experimental work and that these ‘species’ are not mutually exclusive (i.e. they can and do co-occur). While a post-cognitivist take on cognition implicitly includes non-neuronal organisms (and have received extensive philosophical consideration; see Calvo & Keijzer^18,19 and Segundo-Ortin and Calvo²⁰, for specific plant examples), existing experimental studies have still solely focused on organisms with a brain and neural system. Hence, we need to explicitly include brainless organisms such as plants in our experimental investigations of cognition. Here, we will consider the possibility of extended cognition in plants.

The extended cognition from humans to spiders

Over two decades ago, Clark and Chalmers²¹ put forward the idea that cognition extends beyond the seeming physical boundary of the organism (i.e. embodied) into its environment (i.e. embedded) and involves objects that are not part of the body. This is known as the Extended Cognition hypothesis. According to this hypothesis, the environment plays an active role in driving the cognitive processes, which are not confined to the central nervous system or the body, and involve the manipulation of the environment by the organism to enhance its cognitive abilities. For example, in humans, the simple use of pen and paper to write a note or “sketch” a thought represents a way of extending our ability to remember beyond the brain, making manageable what may otherwise be information overload. More poignantly, our dependence on computers and smartphones speaks volumes about how much we offload our cognitive capacity to external devices.^22,23

The idea that cognition may not be restricted to the brain but extends beyond the boundaries of the physical body was received both favorably and with criticism. The critics, who espoused the idea that cognition begins and ends in the brain, were concerned by the fact that it seemed particularly difficult to empirically test and demonstrate that a given process is an inextricable part of the observed cognitive system.^24–28 Kaplan²⁹ proposed to resolve the issue by using the mutual manipulability criterion.^30,31 According to this criterion, if intervening in one of the components of a mechanism alters the mechanism’s overall behavior, and if by altering the whole mechanism the behavior of the component is also changed, then both are part of the same mechanism.³¹ According to Kaplan²⁹ if there is a relationship of mutual manipulability between the putative components of the cognitive system, then we can demarcate its bounds. For example, suppose that A and B are suspected to be part of the same cognitive process, where A is the wider phenomenon under study (cognition) and B the external object that is supposed to be part of this phenomenon. Manipulation of A should causally affect the behavior of B and vice versa. To further elucidate the mutual manipulability criterion and clarify how it is already used – although not explicitly – in neurobiological sciences, Kaplan²⁹ referred to experiments performed with monkeys by Talbot et al.³² They found that a mechanical vibration applied to the hand of a monkey activated a specific region of their brains invoking action potentials at a certain frequency. In a subsequent experiment performed years later, Romo et al.³³ demonstrated that by electrically stimulating this same brain area with the same frequency, monkeys reacted as if they were being stimulated on the hand. Due to the mutual manipulability of the cognitive system’s presumed elements, its causal relationship was established.²⁹ This example does not specifically describe a case of extended cognition, but rather a case of embodied cognition; however, it is important because it showcases how mutual manipulability can be applied as a good empirical approach for defining the bounds of a cognitive system (be it extended or not).

It is important to distinguish between genuine components of an extended cognitive system from what Kaplan²⁹ called “causal background conditions”. Sometimes, components of a system are related by simple causality and are not part of the cognitive process. For example, it is widely known that oxygen is required for normal brain activity. As a person is engaged in a cognitive task, certain areas of the brain will engage in greater activity and in turn, this increased activity might increase the consumption of oxygen for respiration. On the other hand, if the supply of oxygen to the brain is interrupted, cognitive processes are compromised. It would be tempting to say that the mutual manipulability criterion is satisfied here; oxygen would be considered part of the cognitive process and human cognition would extend to the oxygen surrounding us. However, a person engaged in a cognitive task does not alter the concentration of oxygen in the immediate surrounding. Under this scenario A affects B, but B does not affect A. Oxygen will always be oxygen and, unless this person is closed in a sealed room, it will be present in more or less the same atmospheric concentration. Besides, enhancing the amount of oxygen in the brain (by increasing its concentration in the air) will not make the person spontaneously engage in a cognitive task or activate the same brain areas. So, oxygen is a mere causal background condition, a necessary substrate for the cognitive process as is gravity for walking. Again, if we change gravity, we change our ability to walk on land. Yet (and fortunately), walking does not affect the Earth’s gravity. This is another example of a causal background condition that enables walking.

Extended cognition is different from embedded cognition in the sense that the latter emerges from the interaction of an organism with its environment, but the cognitive process remains firmly within the body, despite the essential role of the environment in the process.¹² For extended cognition, the cognitive process occurs within and beyond the physical body and includes the external objects that are manipulated by the organism.

The idea of extended cognition was received favorably by evolutionary biologists who worked on niche construction theory,^34–36 especially in the context of spider ecology.^34,37 Spiders present a vast array of complex behaviors and cognitive abilities that defy the relationship between brain size and behavioral complexity, challenging the hypothesis that complex cognitive skills have a straightforward relation with larger brains.³⁸ In their review article, Japyassú and Laland³⁴ argued that spiders extend their cognitive capacities to the webs they spin. They proposed that by transferring part of the information processing to their webs, spiders simultaneously avoid compromising their cognitive abilities and overloading their small brains. In effect, the spider plus its web would constitute the cognitive system. In support of their hypothesis, Japyassú and Laland³⁴ offered several cases in which spider relations with their webs met the mutual manipulability criterion. For example, the web of many spider species serves as a ‘filter’ for which information will reach the spider. This information reaches the spider codified in the form of vibrations of the threads caused by external agents such as wind or some captured insect.^37,39 By pulling more strongly the threads of a particular area of the web, the spider can focus its attention in the desired web section, enhancing its success in catching its prey. Normally, the spider’s attention is directed toward the most profitable web areas. By artificially manipulating the tension on the web threads, one can alter the foraging behavior of the spider, inducing it to pay attention to other areas of the web, even though these areas have previously been unfruitful.^39,40 The spider responses to the stimuli and its information processing are intimately conditioned to web tension, and not only to its central nervous system,³⁴ demonstrating that the manipulation of the spider’s web or its central nervous system modifies the cognitive capacities of the spider and its understanding of the world. These findings also demonstrated that the idea of extended cognition can be applied not only to humans but also other organisms, especially those who build niches for themselves. Here we discuss the possibility that brainless organisms like plants may also extend their cognition beyond their bodies.

Extended cognition: extending it to plants

Unlike most animals that can wander in search of food and flee when perceiving threat, plants are sessile organisms that live inextricably grounded in their environment. Like their animal counterparts, plants have evolved very refined and sophisticated sensory abilities to constantly perceive and monitor their surroundings^41,42 and perform astonishingly complex behaviors without a brain or any other centralized organ. How is this possible? Might it be conceivable that like spiders, plants “offload” (see Risko & Gilbert⁴³ on cognitive offloading) at least part of their cognitive process to the environment by extending their cognition beyond their bodies? If so, how could a plant extend its cognition? We propose primarily two ways in which a plant may do this: by means of their root exudates and by means of the micro-organisms that live associated with the roots.

Before presenting examples of studies that seem to satisfy the mutual manipulability criterion for extended cognitive processes in plants, it is important to make the clear distinction between the plant’s extended cognition from simple causal background conditions. The mutual manipulability criterion can help us make this distinction clear. Firstly, we consider the wider phenomenon we intend to demarcate, namely plant cognition. Then, we consider the components in the environment that could be part of the plant’s cognitive system because the plant can manipulate them in some way. If the cognitive state of the plant alters one of these components and the manipulation of the component alters the plant’s cognitive state, then according to the mutual manipulability criterion, both are part of the same system demarcated via an extended cognition process. For example, environmental temperature directly affects plant cognition, for changes in temperature can speed up or slow down metabolic processes and, by extension, the plant’s capacity of processing and responding to the stimuli perceived in the environment. In principle, a plant’s cognitive activity can alter the temperature immediately around the plant via the opening and closing of the stomata (a process which requires exquisite computing abilities; Peak et al.⁴⁴, Merilo et al.⁴⁵) and affect transpiration rates. However, these temperature alterations will not change the transfer of thermal energy from the environment to the plant. The mutual manipulability criterion is not properly met here: heat is a causal background condition and not part of the cognitive process. Let’s now consider cases where the mutual manipulability criterion for extended cognitive processes in plants might be satisfied.

Plants could extend their cognitive processes and amplify the perception of the underground world beyond the physical limit of their roots by actively modifying the rhizosphere and the roots influence zone. For example, the perception of obstacles in the soil is related to the accumulation of allelopathic exudates between the obstacle and the root.⁴⁶ This causes root growth to be inhibited in the direction in which the exudates are being accumulated. Falik et al.⁴⁶ found that the experimental removal of the exudates from the substrate prevented the plant from perceiving the obstacle and made its roots grow toward it as if not present. The relationship between a plant and its exudates meets the mutual manipulability criterion affecting plant cognition, and like the spider with its web, the plant plus its exudates constitute the cognitive system.

The cognition of plants could be extended beyond this first example to the associated bacterial community of the roots. Plants actively manipulate the microbiome in the rhizosphere for their own sake. For example, Huang et al.⁴⁷ have demonstrated that Arabidopsis has specific pathways for synthesizing triterpenes that modulate the community of bacteria that lives in its rhizosphere, which creates an Arabidopsis-specific bacterial community. However, the abundance and diversity of this community can be altered by the plant through the many different substances exudated by the roots, and the composition of the microbial community can encode memories in the soil, outside the plant’s body. In a study conducted by Yuan et al.⁴⁸ five generations of Arabidopsis infected by Pseudomonas syringae pv tomato were grown in the same pot, and the plants have altered the substrate microbiota by secreting several different substances that modulated the bacterial community. When the fifth Arabidopsis was removed from the pot and a sixth Arabidopsis was planted in this same soil, the plant became significantly more resistant to this pathogen.

Plants could also extend their cognitive processes via their association with the mycorrhizal fungi they live with. During the root colonization process, fungi penetrate the plant’s roots and establish an interface where many different molecules and signals are constantly exchanged.⁴⁹ The fungi enormously amplify the roots’ absorbing area and help the plants in the uptake of water and nutrients, such as phosphorus. In return, plants repay the fungi with photoassimilates.⁵⁰ Through the association with mycorrhizae, plants can amplify their own perception and perceive things that are out of their reach. Based on what the mycorrhizae perceive and absorb in a particular region of the soil, for example, a plant could make choices about the growth direction of its roots and predict nutrient patches, water, or even forthcoming stress well before the roots get close enough to detect these resources or cues themselves. Therefore, the fungal network could extend the plant’s perception of the environment and facilitate cognitive processes such as communication, learning and memory in plants.⁵⁰

Clearly, not all responses and outcomes are the result of, or require, cognition. Hence, for this plant plus fungus to be considered a cognitive system, the cognitive processes involved must satisfy the mutual manipulability criterion. According to this criterion, experiments designed to determine whether the fungus represents a true extension of the plant’s cognitive system will have to establish a two-way flow, where the fungus plays an active role in driving the cognitive processes of the plant (e.g. while locating nutrients in the soil or navigating a complex environment) and at the same time, the manipulation of the fungus by the plant enhances the plant’s cognitive abilities. To our knowledge, no study has been specifically designed to test the mutual manipulability criterion in the plant-fungus association. A recent study by Kiers et al.⁵¹ demonstrated that the two-way transfer of resources between plants and mycorrhizae is regulated by sanctions and rewards. By growing the fungus in a medium with low levels of phosphate, which effectively prevented the fungus from delivering the nutrient to the plant it had colonized, Kiers et al.⁵¹ found that the plant stopped rewarding the fungus with carbohydrates. Similarly, when the plant was prevented from transferring carbohydrates to the fungus, the fungus interrupted the delivery of phosphate. The behavior of the fungus affected the plant’s behavior, and vice versa. Although the study was not explicitly testing extended cognition in plants with the mutual manipulability criterion, its finding provides support to the hypothesis presented here that extended cognition in plants is possible and offers the promise of a model system where the mutual manipulability criterion in plants can be tested experimentally. It also points at something quite extraordinary; it suggests that the cognition of one organism, the plant, can be extended via an organism (rather than a non-living object), and one from a completely different kingdom, the fungus.

In conclusion, the available evidence suggests that plants extend their cognition beyond their bodies, forcing us to reconsider where a plant ends and its environment begins. Experimental work using the mutual manipulability criterion to explicitly test this hypothesis is urgently needed to better understand how cognitive systems define themselves and specifically, how extended cognition is expressed in non-neural systems, like plants. This idea could have important repercussions on how we understand the ecology and behavior of plants, and more generally, have wider implications on what we think cognition is and how we understand our own relationship to our body and the environment.

Funding Statement

AGP and GMS are granted with a scholarship from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq - Brazil). GMS is a CNPq research fellow. MG is supported by the Templeton World Charity Foundation (TWCF) under the Diverse Intelligences Initiative.

Acknowledgments

The authors thank the LACEV’s members for enthusiastic and insightful discussions, five reviewers for their constructive comments and M Depczynski for providing feedback and proofreading the final draft.

Disclosure of potential conflicts of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.