Plants Do Math
In The Highest Frontier, Jenny and her classmates study Arabididopsis plants with a nervous system. Fanciful–but what if plants can already do math?
So what math problem do plants have to solve? How to survive the night without light for photosynthesis. That may not sound too hard–after all, we animals fast through the night until “break-fast.” But for many smaller animals, that’s not so easy. Some rodents have to eat their own weight’s worth of food in a day. And bats have to consume enough insects by night to survive through the next day.
What’s amazing about plants is that somehow they seem to “know” just how much energy to store from photosynthesis to get them through the night. That’s important because whatever energy they don’t have to store can be used to growth their leaves and outcompete their neighbors for space in the sun. Apparently, Arabidopsis plants have a circadian timer (a mechanism timed to measure the daylength) that allows just enough starch storage to get the plant through the night.
But how does this timer work? What is the molecular basis? Researchers hypothesize that the plant has a way to compute the ratio of the amount of starch left to the amount of time left till dawn. To test this hypothesis, they seek mutants that have predicted defects. For instance, one mutant retains too much starch by dawn, more than it should need. Another mutant was found to be unable to adjust its starch levels to a change in night length. We still don’t know what the mutant proteins do, and we’re probably a few mutants away from an answer. But if plant cells can conduct arithmetic division, that might suggest clues as to how our own brains do it–something we really don’t know either.
Comments are closed.
Ultraphyte 
fascinating–thanks, joan.
Color me skeptical. Firstly my hackles always go up when I read “X does math”. Math is about symbol manipulation, so obviously plants cannot do this. Even humans can’t do much math without tools.
So consider the issue of timing. A simple model is the classic hole in a bucket with water leaking out. We can solve the timing issue with math, but we are essentially solving the physical model. We could try trial and error to fill the bucket with various levels of water so that it drains exactly overnight.
The reported experiment showed that shock night lengths were adjusted for and that genes associated with circadian rhythms appeared necessary, but that doesn’t mean that some sort of math or “counting” was going on. Far more likely is that the plant cell has a sensor for starch (possibly measuring starch production from sugar) that interacts in some way with the circadian proteins to affect the rate of starch degradation. That might be nothing more than affecting a transcription factor.
At the very least, I’m happy to see the authors do not characterize the effect as “Arabidopsis does math”.
You are right to examine the experiment skeptically.
I wonder though how we are defining what is or is not math. If the plant has a sensor with a graded output depending on the amount of starch, doesn’t it count the starch? Then if the circadian system measures time, and produces an output that can be related to the output from the starch, is that a form of division?
It doesn’t take that much intelligence to “count.” In my younger days, I built a computer of plastic parts that could count to seven. It could also add and subtract any numbers from zero to seven. As I recall, it could divide four by two.
What version of the experiment would convince you that the plant can “count”?
The reason this is interesting is that, to understand how our own brains count, ultimately we must find some kind of neuronal chemistry that compares sensory inputs and does something analogous to my plastic computer.
Counting is rather quantized. Would you still call it counting if the system had to discriminate between volumes/concentrations? Would you consider your reflexes counting when you hit ball with a bat? When we talk about the brain acting as a Bayesian computer, does that mean neurons are somehow doing computations?
You probably know that the outputs of a neural network basically define the output function based on the input. I don’t think that would count as doing math, even if an artificial neural network was doing exactly that in a computer.
To answer your question about a version of the experiment that would convince me counting was happening, I don’t think any in this experiment. Let’s instead take something more discrete. Should the number of cell divisions a cell undergoes be considered counting, even if we can determine some discrete physical changes, like telomere shortening? I don’t know if you saw the Science paper a few months back about engineering DNA logic gates? Very impressive work because the cell could be made to count (I think) like your tinker toy computer, with discrete events. That level of mechanism would put me on the fence.
As for human counting, what goes on at the neuronal level is interesting. As you know, we don’t count particularly well using memory. We can recite sequences, which are easy to model as hetero-associative memories. We can count up to about 9 in terms of distinguishing numbers of objects in memory, and we can use visual patterns for larger numbers. But ask someone to subtract 2 numbers (e.g. change for a purchase) and you will find most people cannot do that today (test your students for confirmation). But give them the affordance of coins and it becomes quite easy. I don’t think teh brain does anything like your computer.
If you recall my post The Bacterial Computer, yes, bacterial gene expression can do everything that computer circuits can do. And yes, bacterial circuits are used as a model system for investigating neurons.
Can bacteria “count” cell divisions? Controversial; we are studying this in our laboratory now. Another interesting case is the nematode worm C. elegans, which develops a body with a precise number of cells, always the same. So this developmental system might be said to “count.” Or you might call it a pattern replicator.
I think we should define ‘doing math” as at a minimum being able to add or subtract arbitrary pairs of numbers. This removes the set of processes that are not flexible, like cell division. I would also like to remove analog processes that mimic maths. That even includes analog computers that need to be configured for a particular problem, because that implies functional inflexibility. Are there any biological processes that would pass this test of counting that you know of?
To me, counting suggests planning: I have X but if current events suggest that I will need X+, what (if anything) do I then do about this shortfall?
Otherwise your counting example is similar to the horse (Clever Hans) that could do arithmetic: the trainer by his actions told the horse when to stop clopping his hoof.