A Broader Conception of Memory

2,800 words

Header image taken from here.

Sign up here for the Blank Horizons mailing list.

Supplementary reading: If you are intrigued by the ideas in this article, I strongly encourage you to read “The Intelligent Plant,” a 2013 essay by Michael Pollan that was featured in The New Yorker. Also, this blogpost’s section on plant memory was primarily inspired by an essay titled “The Hidden Memories of Plants.”


You are everywhere

In “The End of Time,” a book that I have discussed before, the physicist and philosopher Julian Barbour argues that the past and future are illusory constructs. There is only an infinite landscape of Nows that coexist simultaneously with one another, but we can only move through the landscape one moment at a time. Barbour claims that we only believe in the past because of records that we access in the Now (the future can be conceived of as an inverted record); these carry evidence about other moments of time that seemed to precede this one, but in reality all moments are happening at the same time. It is only because the brain stores such sophisticated, seemingly accurate memories, which count as records, that it seems like the past really did happen.

As I reflected more on Barbour’s idea, I began to wonder: what exactly constitutes a record? How broadly can a record be defined?

Barbour provides some examples of records other than the brain, like fossils. But arguably, anything that we interact with and leave some detectable imprint on could be construed as a record. If I make a very large dent on the table where I am writing this blogpost, then millions of years from now, once I have died and the rest of the human race has been wiped off Earth, the table (or the remnants of it) will still carry a record of my interaction with it and my existence. Could we say, then, that the table has a memory of me?

Furthermore, when I die, my body will get decomposed by fungi in the ground, and those fungi will then get consumed by another mammal. If that mammal happens to be a cow, then the molecules that once constituted my physical body will now be inside its milk. Could we say that the milk has a memory of me? If “anything that I interact with” encompasses all that my physical matter encounters in its long journey through the cosmos, then could we say that everything will eventually have a memory of me? (1) To highlight just how widely distributed our matter becomes after sufficient time has passed, I will note that the air that I am breathing right now has, on average, at least one of the molecules that Julius Caesar inhaled in his last breath.

Many people are reluctant to accept such a radically liberal conception of memory since it would be akin to anthropomorphizing non-conscious, non-living entities. Of course, tables don’t have memories in the same way that we humans do; they can’t freely and vividly recall a mental image or video of a subjectively experienced event. One might think that the attribution of memory to inanimate things like tables and milk would amount to nothing more than a semantic contrivance. That is, only by distorting the established definition of “memory” could one come to justifiably believe that everything is capable of storing memories.

It is certainly true that this broader understanding of memory, for the most part, doesn’t yield any (immediate) practical implications. We certainly can’t predict the physical state of a table more accurately by projecting memory onto it. My motivation to claim that memory is much more universal than we traditionally think is primarily philosophical. However, my aim in this blogpost is not to explain the metaphysical underpinnings of redefining memory in the way that I have described. Rather, I will argue that it is scientifically meaningful – and useful – to ascribe memory to one class of insentient objects: plants. The case for “plant memory” demonstrates that the phenomenon of memory arises not from the activity of neurons in a brain but instead from a homologous structure that is inherent to all living creatures.


Plant memory

In the late 1920s, the notorious Soviet scientist Trofim Lysenko discovered a phenomenon that he later termed “vernalization,” in which plants appear to be capable of “remembering” cold weather and adjusting their behavior accordingly. In particular, if plants underwent a long and brutal winter while they were germinating, then they will flower earlier than they usually do. Through vernalization, the two-year growth cycle of crops like winter wheat can be shortened, such that they can be harvested in the same year that they are seeded.

The notion of “memory” didn’t appear anywhere in the initial literature about vernalization, and indeed it seems like the phenomenon can be explained in its entirety without imputing any powers of recollection to plants. Vernalization could simply be construed as an adaptive mechanism mediated by the genetic structure of plants. Many decades after Lysenko’s findings, biologists realized that the fixed genome of an organism – the genetic code inscribed in its DNA – isn’t the sole determinant of the genes that it is capable of expressing. Rather, the study of epigenetics led to the insight that gene expression can actually be modulated by changes in the environment. In the case of vernalization, cold weather induces the inhibition of a protein that normally represses flower-promoting genes in plants.

It may be more compelling to attribute memory to plants in light of evidence that warm weather cannot undo the effects of vernalization. If plants are merely adapting to their environment when they undergo vernalization, then a hot climate should cause them to return to their normal flowering schedule. However, plants seem to “remember” a cold winter even if they are subject to intense heat immediately afterwards; that is, as long as they have already undergone vernalization, they will flower earlier than usual, no matter whether the environment changes. Furthermore, plants can store the memory of cold weather for up to an entire year. After vernalization, they only flower when there is enough daylight in the sky. In an experiment conducted in the mid-60s, a scientist hid a plant from sunlight for nearly 11 months and found that it still blossomed when it was exposed to sufficient daylight at the end of that time period.

Nevertheless, one might argue that we don’t have to posit a “plant memory” in order to account for the persistence of vernalization in hot weather. Well-established biological and genetic mechanisms could underlie a plant’s accelerated flowering in warm conditions, even though it seems like a hotter climate would reverse or inhibit those mechanisms. The epigenetic “switch” that triggers a plant’s flowering process simply flips on as soon as a stretch of cold weather begins, and it continues to stay on even after that period of time comes to an end. The ascription of memory to plants, once again, seems more like a semantic choice than a scientifically motivated one.

However, we may need to update our understanding of plant biology – and our concept of “intelligence” in general – in order to accommodate two recent empirical findings that present strong evidence for animal-like memory and learning capabilities in plants. Several years ago, the evolutionary ecologist Monica Gagliano wondered whether plants could be trained to filter out irrelevant stimuli in the same way that animals learn to ignore signals in the environment unless they are sent by predators. Gagliano experimented with a plant known as Mimosa pudica, whose leaves contract as soon as they are touched (see Fig 1). When mimosas are dropped, their leaves also fold up. Gagliano created a device that would drop mimosas from a height of 15 centimeters. She tested 56 mimosas, dropping each one over 60 times. For the first few trials, the mimosas would shrink up, as predicted, but after four to six repetitions of the experiment, their leaves would, remarkably, remain open (see Fig 2). Did the plants “learn” that their survival wouldn’t be threatened if they were dropped?


Fig 1. The leaves of Mimosa pudica contract when they are touched. Gif taken from here.


Fig 2. Left: A mimosa plant on the first trial of Monica Gagliano’s experiments. Right: a mimosa after several trials. Animation taken from here.

Skeptics of Gagliano’s findings have pointed out that the mimosas may stop shutting their leaves simply because they were fatigued from getting dropped multiple times. Anticipating this objection, Gagliano shook the mimosas at the end of each trial and found that they closed up. The mimosas could “discern” that getting dropped wouldn’t imperil its safety, but they treated other stimuli with just as much caution as they did before the experiments. Furthermore, when Gagliano repeated her trials one month later, she found that the mimosas were still unperturbed when they were dropped. The plants had retained the memory of Gagliano’s experiments much longer than expected; memories, when they are stored by animals like bees, are considered to be “long-term” if they last for just 24 hours.

A few years later, Gagliano conducted a second experiment in which she demonstrated that plants could actually alter their behavior based on memories that they had developed about patterns in their environment. In particular, she tried to determine whether plants would learn to associate an irrelevant stimulus, wind, with a biologically significant stimulus, light, if the two always coincided with each other. Gagliano essentially aimed to find out whether plants could undergo classical conditioning in the same way that Pavlov’s dogs did.

To answer her research question, Gagliano grew garden pea plants at the base of a Y-shaped maze that she had constructed (see Fig 3). Plants were randomly assigned to one of two groups: in the first, light and wind were emitted from the same fork of the maze, and in the second, they were issued from separate forks. After a three-day training period, Gagliano turned off the light and found that plants in the first group grew towards the fork where the wind was blowing from. The plants in the second group, on the other hand, sprouted in the direction opposite the wind. In other words, the garden peas must have stored some memory that linked the light to the wind.

Screen Shot 2019-04-30 at 4.56.01 PM.png

Fig 3. The set-up for Monica Gagliano’s second experiment. Here’s the caption from the original paper: “During training seedlings were exposed to the fan [F] and light [L] on either the same arm (i) or on the opposite arm (ii) of the Y-maze. The fan served as the conditioned stimulus (CS), light as the unconditioned stimulus (US). During testing with exposure to the fan alone two categories of responses were distinguished. Correct response: Seedlings growing into the arm of the maze where the light was “predicted” by the fan to occur [green arrow; iii (corresponding to scenario i) and iv (corresponding to scenario ii)]; Incorrect response: Seedlings growing into the arm of the maze where the light was not “predicted” by the fan to occur (black arrow; iii and iv).”

If Gagliano’s results are real and reproducible, then what could be the underlying mechanism of “associative learning” in plants? We don’t know the answer yet, and this uncertainty is perhaps the most exciting aspect of Gagliano’s research. Gagliano herself has proposed that “in plants and other organisms that do not have a nervous system, modifications of the patterns of interactions between molecules and communication between cells can be stored in a way rather similar to neural networks.” She also suggests that epigenetic reprogramming plays a role in altering these patterns; “presumably, then, the mechanisms maintaining associative learning operate in plants as in other organisms on the basis of fundamental ‘rules’ that alter the flow of information by modifying the shape and connections within a network via epigenetic change.” Firstly, epigenetics may not be responsible for most of the associative learning or memory storage that takes place in plants; as several Australian plant biologists stated recently, “epigenetic memory is likely a relatively rare event.” If epigenetic changes truly did regulate plant memory, then it should be the case that memories get passed down from one generation to the next. However, experiments have failed to demonstrate that a plant will be better at conserving water, say, if its “parents” experienced drought. Secondly, Gagliano’s language is noticeably vague; what exactly are these rules, what exactly qualifies as a “pattern of interaction,” and which patterns are relevant to associative learning? She doesn’t know yet, but future insights into these questions will yield a massive leap forward in our understanding of memory. Hopefully, we will come to realize that memory is not a phenomenon that is localized to brains or nervous systems, but rather one that is much more universal in character.



Many “plant neurobiologists” – scientists who, in Pollan’s words, “argue that the sophisticated behaviors observed in plants cannot at present be completely explained by familiar genetic and biochemical mechanisms” – have attempted to identify the mechanisms that govern plant intelligence by searching for structures in animal nervous systems that are homologous to those found in plants. In the article where they coined the term “plant neurobiology,” six pioneers of the field note that glutamate is both a key neurotransmitter in the animal brain and an important signaling molecule in plants. They say that the genes for glutamate receptors in plants are not too dissimilar to those in animals. However, they also acknowledge that we don’t know very much yet about the precise role of glutamate receptors in plants. (Then again, the article was written in 2006, so it’s possible that our knowledge has advanced quite a bit since then.) Although the scientists do list a series of functions that glutamate may serve, they don’t attempt to draw any connections between them in order to offer a systematic, coherent account of the chemical’s relationship to memory and intelligence.

Calcium-based signaling networks are, in my opinion, a more compelling candidate for the universal basis of memory. When a plant is repeatedly exposed to the same stimulus, its concentration of calcium progressively declines. Calcium concentration, therefore, could serve as the plant’s memory for the number of times that it has experienced a particular stimulus. Additionally, calcium concentration encodes the history of not merely one but multiple inputs that a plant receives from its environment. When water enters a plant cell, fluctuations in calcium levels will be lower if a lot of water had left the cell in the past, but higher if the cell’s water supply had been relatively stable. Calcium serves a relatively similar role in the animal nervous system. At least at the level of neurons, learning is modulated by a mechanism known as synaptic plasticity, in which synapses, which connect neurons to one another, change the strength of their electric potentials depending on past stimuli. Residual calcium mediates synaptic facilitation, the variety of plasticity in which synaptic potentials become stronger rather than weaker. When one stimulus triggers the firing of a neuron, calcium levels will increase significantly. Once the firing has stopped, calcium will return to its initial concentration, but not immediately. If the stimulus is repeated quickly, then calcium will increase again before it has had a chance to drop to its normal levels. Thus, synaptic facilitation is associated with an overall boost in calcium concentration.

I am skeptical of the notion that one atom or molecule is the substrate for memory. What is it that sets calcium or glutamate apart from other chemicals? The fact that they appear at relatively high levels in plants and animals does not really explain why they are uniquely capable of endowing organisms with memory. Despite its ambiguity, I actually prefer Gagliano’s idea that a pattern of interactions regulating the flow of information is, in some way, responsible for memory. Calcium does have a very sophisticated signaling network associated with it, and it essentially serves as a huge pattern of interactions that encodes and decodes biological information. It may be promising, therefore, to explore the homologies between the plant and animal versions of this network.

I will conclude by proposing an entirely speculative notion about a direction for future research. In an earlier blogpost, I discussed the concept of symmetry in physics, which, to paraphrase Wikipedia, is a feature of a system that remains unchanged after a transformation is applied to it. Could we somehow use the mathematical framework of symmetry to analyze the aspects of the calcium-based signaling network that are invariant between – and common to – plants and animals? This idea may be totally misguided, given my very superficial understanding of complex physics, and my limited knowledge impedes me from making any further progress on it. I’ll let the experts decide whether it is a thought worth entertaining.


(1) This line of questioning was inspired by a lovely animated short that I recently watched.

Sign up here for the Blank Horizons mailing list.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s