Selective tool use in ants

January 9, 2017 • 9:30 am

You’re probably aware that tool use, once considered a uniquely human phenomenon, has now been documented widely in primates and birds. You may not know, though, that it’s also been seen in some insects, as in wasps that use pebbles to close off their burrows after laying eggs. Ants, too, have been seen to use debris from the environment to transport liquid food back to the nest; the ants that do this have a non-expandable crop, so they can’t simply suck up the stuff and then regurgitate it to their nestmates.

Now, in a new paper in Animal Behaviour by István Maák and his colleagues (link and free download below), we find that two species of the funnel ant Aphaenogaster not only use tools to suck up liquid food, but are selective in which tools they use. The authors collected colonies of A. senilis and A. subterranea, and presented the colonies in the lab with either straight honey or a mixture of honey and water placed in plastic plates on a piece of aluminum foil. The two species were treated differently: A. subterranea was given both foods (separately), while A. senilis was given only diluted honey (1 part honey to 3 parts water).

Each species was then presented with a variety of “tools” around the food with which they could use to soak it up or dip it (like a chip into guacamole) before taking it back to the next. The tools also differed between the two species. These diagrams of the experimental setup show what was given to both species. First, a specimen of A. senilis (a queen):


The experimental setup for A. subterranea:


A photo of the setup from the paper and the time course of how the ants behaved:


As you see, they were offered five items to use as “tools,” including sponges, which of course the ants had never encountered. These tools differed substantially in their ability to soak up or hold the two foodstuffs. As the table below shows, sponges were by far better at soaking up the honey/water solution, while leaf fragments (used like chips, I suppose), were by far the best for transporting straight honey. The expectation was that the ants would glom onto the substance that was best at soaking up the liquids, as that allows for more efficient transport (remember, ants are strong for their size).


Here, from a New Scientist blurb on the paper, is a photo of ants dropping tools into the food:

Photo credit: J. Coelho/CC-BY

In this species the ants didn’t really prefer the best “tool”: soil grains were put into the honey-water, and pine needles for the honey despite their poor ability to soak the stuff up. For transporting food to the nest, this species used soil grains most frequently for honey-water, and sponges for honey, despite the observation that both of these are not so great for dipping into and soaking up the viscous honey. Conclusion: this species isn’t using the best possible “tool” to transport the food.

The setup for A. senilis was different: these anbts were offered only honey-water, and were given twigs, strings, parafilm (a waxy, stretchable substance used to cover stuff in the lab), bits of paper, and artificial foam (it’s not clear why there was a difference in setup):


And the soaking properties. Sponges and paper were by far the best at sucking up the honey-water:

screen-shot-2017-01-09-at-8-21-26-amIn this case, the workers did behave “optimally”, though it took some time for them to figure it out. Over the ten trials, they apparently learned to both put paper and sponges into the solution more frequently, and transported them to the nest more frequently.

In one of the two species tested, then, the ants chose the optimal tools, and their behavior wasn’t fixed; they apparently learned somehow which tools were best.

Why did A. subterranea use suboptimal tools? The author suggest that there may have been problems handling some of the tools, as well as a lack of familiarity with things like sponges and paper, which aren’t present in the natural environment. But that doesn’t really explain why one species used the right tools and the other didn’t.

The main lesson of the paper, though, is that ants do use tools (we already knew that from other studies), that they can recognize foreign objects as tools, that at least one species can learn to select those tools that allow it to transport food back to the next more efficiently.

The only remaining consideration is this: are the ants really using tools? The authors say, yes when using the definition given below, and I agree. If chimps chewing leaves to make sponges for soaking up water our of crevasses counts as tool use, then so does this. And, in reality, the strict definition doesn’t matter: what is important is a tiny-brained animal can figure out how to solve problems by using bits of the environment. (References omitted in quote below, emphasis mine).

. . . we think that the behaviour of the ants meets one of the most important criteria that define tool use in foraging contexts, i.e. the use of an external object to affect the position or location of another object or substance. Obviously one could argue that nest-building materials affect the position of other nesting materials, which is why some authors object to the idea of tool use and would perhaps like to see either a more restricted use of the term or its complete abandonment for a wider term such as construction behaviour. Although we understand this position and the reasons for it, given the above considerations, we still prefer to refer to the behaviour of the ants as tool use, or the similar denomination of object use.

h/t: Nicole Reggia, Jo S.


Maák, I., G. Lőrinczi, P. Le Quinquis, G. Módra, D. Bovet, J. Call, and P. d’Ettorre. 2017. Tool selection during foraging in two species of funnel ants. Animal Behaviour 123:207-216.

49 thoughts on “Selective tool use in ants

      1. You can’t get updates unless you make a comment and check the checkbox. “Sub” is short for “subscribe”.

  1. The collective intelligence of ants never ceases to amaze. I tried to eradicate a nest using one of those liquid ant baits. At first they swarmed around it, but I observed that after many were killed the rest started to avoid the bait.

    1. Or was it that the ones who originally avoided it for some unknown unintelligent reason were the (only) ones left to observe afterwards? 🙂

    2. I am a little more cynical about this sort of thing (though a strong admirer of ants). It could be that the smell of all the decaying bodies or alarm pheromones were repellent to the ants.

  2. WOW, I’d never thought I’d read the words ants and tool use in one sentence.
    Makes you wonder, when will ants take over the world? 😉

    1. After the nuclear holocaust, it’ll be cockroaches. But there are so many subterranean species of ants that some will join the cockroaches I’m sure.

    2. Black fire-ants (Solenopsis richteri) can adapt its tool use. When provided with small containers of sugar water, they float and feed on the surface, but when researchers reduced the surface tension (so the ants could no longer float), they started depositing sand grains on the inside of the container leading out of it. “We knew some ant species are able to use tools, particularly in collecting liquid food; however, we were surprised by the remarkable tool use displayed by Black Fire ants,” said senior author Dr. Jian Chen, an entomologist at the U.S. Department of Agriculture’s (USDA) Agricultural Research Service (ARS). “This exceptional tool making skill not only reduced the risk of drowning, but also provided a larger space for them to collect the sugar water.”
      “The sand structures were found to be so efficient that they could siphon almost half of the sugar water out of the containers in just 5 minutes.”
      “Our findings suggest that ants and other social insects may have considerable, high cognitive capabilities for unique foraging strategies,” Dr. Chen said.
      Dorymyrmex bicolour exhibit tool use by dropping small stones on their competitors, particularly Myrmecocystus workers, to discourage them from leaving their nests to forage.
      They may even block the Myrmecocystus nest entrance. Bulldog ants (Myrmecia pyriformis) have an especially clever trick up their sleeves. Under cover of darkness, a group of ants will attack the nests of large Avondale spiders. These social spiders live in colonies of up to 300 arachnids with each one often ten times the size of the attacking ants. Both forage the same terrain for small insects to eat. The ants forcibly evict the spiders and then, rather than use the nest themselves or leave it for others to use, the ants fill up the nest cavities with twigs and leaves and whatever debris they can find, making them unusable to the spiders. This complex behaviour goes way beyond tool use and, as Bulldog ants are regarded as a very primitive type of ant, is amazing in its own right.

  3. I’m just blue sky-ing here for fun, but here’s a thought on how this ‘ant learning’ could be simulated by an unintelligent, non-learning process.

    1. Initial random or biological-set distribution selection of substance to dip.
    2. At nest, every ant is genetically programmed to try and get the same amount* of honey ‘off’ its dipping implement before making another trip (*again, subject to biological variation).
    3. This means the ants who randomly picked the more efficient implements get done at the nest faster, and thus make more trips per unit time.
    4. Because ants leave pheromone trails, over many trips this will cause more and more ants to travel the ‘successful’ routes to the optimal dipping implements.

    Some thoughts on this idea:
    (a) This doesn’t explain the difference in species behavior. Though if hypothesized step #2 was true for one species and not for the other, that would explain it.
    (b) You could test this idea by wiping down the arena floor at regular intervals. If this idea is correct, that should wipe out the ‘learning’ effect because you would be removing the ability of the pheromone trails to reinforce the optimal paths.

      1. Ha! My experimental biology efforts are limited to armchair web posting and telling my kid “here – eat this.” 🙂

        Another possibility I thought of after posting would be if each worker ant gets fed based in part on the amount of honey it brings back (maybe evaluated by smell), and the ones who get fed better exude more pheromones. In truth I have no idea whether either idea is even biologically reasonable. But the point is that any mechanism that results in the ‘successful’ ants leaving a stronger trail of pheromones would do the trick.

        Regardless, I forgot to say that I completely agree with Jerry. Somewhat deemphasized by my terse step #1 is the point that these ants select objects to gather honey. They’re ‘tool users’ in the broadest sense of the term, even if the ‘learning’ observed in this experiment isn’t really learning.

    1. I think that is a very good idea, and it reads like it would develop the same effect. One could model this in a computer with random automatons that have only a few algorithms: Follow trails. Follow trails that are stronger. Pick up something at the end of the trail. Be attracted to food. Bring back food.

      1. Lots of such modeling has been done. One of the more interesting simulations I saw was a central ‘nest’ with a food source moving around it in a circular path. When the movement of the food is relatively slow (compared to the speed the ants can move), the ants’ random search results in a ‘chase’ pattern; they follow the food around. But if you increase the speed of the food so they can’t possibly catch it, they still find an optima via random search; they’ll make a straight line path to one part of the circle and back, but over time all the ants will come into a time synchronization with the moving food source, so they all arrive at the spot right when the food passes it. Almost like a seasonal migration.

  4. This is an interesting result, though I’m somewhat troubled by the claim that some ants chose “sub-optimal” tools. If the tools chosen by the ants aren’t the ones the researchers think they should have chosen, all that really says is that the ants have different criteria for optimality than the researchers do.

    Ability to soak up food is at most half the problem. You also have to be able to carry the laden tool and get the food back out of it at the other end.

    Ants may prefer the pine needles (for instance) because they can grip one end and dip the other while keeping their mandibles out of the sticky stuff, even if that means they can carry less per trip.

    (Note also that the paper is behind a paywall and is not free for non-subscribers.)

    1. You are absolutely right. Unlike the other species, Aphaenogaster senilis, A. subterranea selected those tools that had the most optimal handling properties. That is why workers largely ignored e.g. leaf fragments (contrary to the fact that these have very good soaking properties), because these could not be handled easily by the ants.

      Btw, the term “suboptimal” tools is not used in the paper, neither it is claimed that the behaviour of this species would be “suboptimal”.

  5. Given the good ideas above for non-learning patterns that maybe could do the same thing, a simple variation of the experiment is to mix the tools, forcing the ants to choose the tool over just choosing a scent trail to follow. That oughta help sort out if they are learning an optimal tool versus just following a stronger scent trail.

  6. I have not read the paper, just what is provided here. Could it be that these ants abhor floods, whether rainwater or artificial floods of sweetened water. Their instinct may be to pile soil, duff, leaves or whatever small bits are available to stanch the perceived flow of water. I am thinking that that is what they may be doing. Then after they have followed that instinct, they notice that the bits have nutritional value, and a food transport instinct takes over. This hypothesis would be a possible good one if the arena were quite close to the nest. I’m saying that this may have nothing to do with tools.

    1. Have they put distilled water in the tray in the arena? What about saline of different concentrations or something noxious like alcohol? Would the ants ignore these or would they pile stuff into the tray? Ants are constantly moving solid substrate materials (soil) and their approach to unwanted stuff in their nest is to move it out or cover it. Could the tool idea be a delusion?

      1. Nope, it could not. It is true that foraging tool use in these ants may have evolved from the anti-drowning behaviour you mentioned, but here are some facts that prove that the dropping-retrieving behaviour can be considered as a true tool use:

        1. These ants do not always drop debris into non-food substances, but they always do that when they encounter liquid food sources.

        2. Some species never drop debris into non-food substances.

        3. In some cases, workers do not drop the carried tool item, only plunge it into the food while holding the tool with the mandibles, and than they carry the food-soaked tool back to the nest right away.

  7. Ok, but they werent conscious right? I just get creeped out that if these little guys could figure out how to use tools, maybe we should consider not crushing them…

    1. I have read that Chimpanzees sharpen sticks and use them to hunt bushbabies. And Gorillas are known to throw rocks (not to mention their poop) in aggression.

  8. To me, this is far more impressive for the fact that ants are blind so cannot see what the others are carrying. Does this imply some form of communication to ‘tell’ them to use materials with a certain texture or smell?

    1. Yes, they communicate through scent.

      Though its obviously not foolproof. There’s a species of ant-eating spider that will kill an ant, load the corpse on top of its body, and then go down into the nest (to eat the eggs I guess) because the scent of the dead ant fools the live ones.

Leave a Reply