The Evolutionary Role of Non-Working Ants

By observing the Shiwakushikeari ant species for over two years, evolutionary biologist Eisuke Hasegawa found that approximately 20% of the worker ants do not work.¹

But there is a lot of work to do for the ants: gathering food, caring for eggs, larvae, and the queen, and repairing the nest. However, his research team observed that these ants just lick their bodies or wander aimlessly around, doing nothing.

It's easy to dismiss their behavior as stupid and lazy, but ants are remarkable when it comes to survival. They, with an estimated 20 quadrillion individuals, outweigh all wild birds and mammals combined and have thrived for over 100 million years. This makes us wonder; having non-working ants must be an evolutionarily advantageous trait for the whole colony and individual ants living inside.

Backup Workforce for Unpredictable Crises

Hasegawa hypothesized that the existence of the non-working ants makes it easier to respond when a crisis occurs. He points out that ants live in highly unpredictable environments, citing several examples such as when food like cicadas is found and workers need to be mobilized for collection, or when the nest is damaged and needs to be repaired.

When everyone works at the same time, they become tired at the same time. This results in a period when no one can work, preventing tasks from being completed. And some things cause fatal damage to the colony if not completed.

So the existence of the non-working ants doesn’t lower the colony's survival due to laziness; rather, they are probably valuable beings whose absence could jeopardize the colony's survival. Perhaps, over generations, an ant species with this adaptation mechanism survived.

Response Threshold Model

But how are the ants dispatched for work? In their colony, there’s a queen, but they don’t give commands. A queen specializes in laying eggs, more like reproductive machines than leaders.

The secret is in their response threshold model.² It’s like each individual has a built-in "activation level" for doing work: if the need (stimulus) is high enough, they’ll start working on it.

Imagine a group of people working in the same café. How soon they start to feel annoyed by noise and put on headphones depends on the individual. People who are sensitive to noise have a low response threshold to sound, so they react quickly and put on headphones. Others with a higher threshold can stay without headphones even when it gets pretty loud.

By the same token, some of the ants have a low response threshold to specific tasks, like foraging or nest repair, so they start working as soon as the need arises, while others with a higher threshold remain inactive until the situation becomes more urgent.

The key is that this threshold varies from one individual to another. You can imagine the Gaussian distribution like below where the ants are distributed based on their response threshold.

The author translated the image from the source.

The response threshold model works well when new tasks arise. Ants with a low response threshold are usually the first to respond. But if they are already busy, the stimulus for the new task keeps increasing. When it gets strong enough, even ants with higher thresholds will respond and take on the task. Or in case the working ants become tired, stimuli begin to build up too. As the stimulus level rises and crosses higher thresholds, the ants that have not been working start to work. This prevents a stagnation of labor.

I was wondering how exactly work is assigned to the ants with different response thresholds; here is my hypothesis. Let's say we assign a value to each task. For example, watching one egg is 1, and carrying food as big as yourself is 5. There are two ants of the same size that notice the food that needs to be carried. The food is almost the same size with them. One has a threshold of 10, and the other has 2. So the former ant ignores the task, and the latter starts carrying food. It's a bottom-up approach where individual ants work independently as they notice tasks.

Initially, I tried to model a situation where there is work left that is worth 100 at a given point, and thought about how that work can be distributed. But this top-down approach won't work because there's no way for the ants to know the entire workload and communicate it to the ants in the colony.

Non-Working Behavior is Not Fixed

Hasegawa also discovered that an ant species of Shiwakushikeari's behavior is not something fixed to 20% of non-working ants. Interestingly, when separating working ants from non-working ants and creating respective colonies, about 20% of the ants stop working even there!³ Conversely, when they only take out the non-working ants, the working ants will still properly emerge from among them.

To visualize this, you can see how a response threshold distribution changes as you separate working ants from nonworking ants.

Sourced from Eisuke Hasegawa's talk.

You can see that the distribution of thresholds remains within each group, even if you create groups of the working ants and the non-working ants.

Strategic Laziness

Because each ant has a different reaction threshold, they can flexibly send out the right number of workers for each task, without needing a boss.

What looks like inefficiency might be nature’s way of building resilience. In ant colonies, and perhaps in human networks too, it’s not always optimal to have everyone working all the time. The non-working ants weren't being lazy or selfish; they were contributing to the survival of their colony by staying as a backup. A little slack might be precisely what keeps the whole system from collapsing when an unpredictable crisis hits.

1. Edward O. Wilson of Harvard conducted a similar study using colonies of a Pheidole ant species, comparing ordinary worker ants and soldier ants. He discovered that while soldier ants usually do not perform tasks, when they made up more than 80% of the colony, they began engaging in duties such as childcare, tasks usually done by ordinary workers. When the proportion of soldier ants was below 80%, they did almost no such work. Hasegawa conducted his research using the ant species Shiwakushikeari, in which all worker ants have the same form. This post focuses on Shiwakushikeari, but similar behaviors are observed in other species.

2. This model was first proposed in the 1990s by researchers like Gordon, Bonabeau, Theraulaz, and Deneubourg.

3. Some of you may have noticed already. This phenomenon, where working ants inevitably stop working when they extract only those who work, is called the 80/20 rule or the Pareto principle in economics.