Ants can learn lots, fast

by Tomer Czaczkes

Based on research for the paper “T.J. Czaczkes and P. Kumar. In press. Very rapid multi-odour discrimination learning in the ant Lasius niger. Insectes Sociaux.”

Most people are shocked to hear that ants can learn. While the readers of this blog are probably not surprised by this, quite how good they are might come as a surprise – it certainly surprised me! In our recent study, Pragya Kumar and I found that Black Garden ants (Lasius niger) can learn at least two (most likely three) odour-sugar associations, having only experienced each combination twice. They can learn one association in just one visit.

A curious Lasius niger forager sporting a stylish blue gaster mark. Painting ants helps us follow our trained ant over multiple training visits.

The joys of comparative psychology:

Let’s unpack that: we were exploring discrimination learning, where the ant has to learn that one odour (e.g. rose) means very sweet sugar water, another (e.g. lemon) quite sweet, another (e.g. lavender) slightly sweet, and another (e.g. blackberry) just a teeny bit sweet. If the ant learns successfully, she should prefer rose to lemon, and lemon to lavender. They should never prefer blackberry when one of the other options is available. So, we let individually-marked ants up a bridge to find a drop of flavoured sugar, then let her go back to the nest, and when she came back for a second visit she found another drop with a different taste and a different sweetness. After she’s experienced each combination once or twice, we give her a choice on a Y-maze: for example, does she follow the arm which smells like lemon, or the arm which smells like lavender? By the way, the methods are all published in another Insectes Sociaux paper (Czaczkes, 2018). And if you have a 3D printer, you can print your own mazes too.

The 3D printer Y-maze we used, with added walls (not used in this experiment). Why not print yourself one?

Ants made mostly correct decisions, even if they only experienced each taste/quality combination once. Clever ants! So… why did I say that ants can learn “at least two” combinations? Well, we simply can’t be sure of more.

Consider our situation: Rose > lemon >  lavender >  blackberry. Firstly, they didn’t prefer lavender to blackberry. So, we’re down to three. Now, imagine that the ants never learned the second worst smell (lavender). What would her decisions look like? It would still prefer lemon > lavender, because lemon is better than nothing. So, we’re down to two we can be sure of. Now, we’re pretty sure they learned lavender, but we just couldn’t prove it in this setup!

This, to me, is the joy of comparative psychology – every experiment is like a logical puzzle, where evidence builds up, ruling out alternative explanations, until you run out of alternatives (or evidence). I admit, it’s hard work – or, at least, I find it hard. Sometimes I feel my brain creaking under the pressure. But it’s also very rewarding, when you’ve lined up your evidence, and can knock the alternatives out. Even finding the logical holes is fun, as happened in this experiment!

A lucky marked L. niger forager enjoying a drop of high quality sucrose solution. Yum!

Back to the ants   

So, L. niger ants can learn quite a lot, and fast. Why is that interesting? Firstly, it’s perhaps surprising, given how small their brains are. But enough e-ink has been spilled on this topic. More practically, this opens the door to performing complex training regimes and tests. A lot of psychological research involves asking the question: “which option do you prefer, A or B? How about C or D?”. Because we cannot simply ask animals, we have to train them to associate each option with a cue, and then see which they prefer. So, for example, we can ask if ants like to gamble by teaching them that lemon is a risky smell, but rose is a safe smell. Being able to quickly train ants to complex option sets can open the door to a much deeper understanding of how ants think, what they like, and how they make decisions.


Czaczkes T.J. 2018. Using T- and Y-mazes in myrmecology and elsewhere: a practival guide. Insectes Sociaux 65, 213-224

Simple behaviors – collective outcomes. How individuals affect group decisions.

By Stamatios Nicolis

Based on research for the paper “Nicolis, Pin, Calvo Martín, Planas-Sitjà and Deneubourg. In press. Sexual group composition and shelter geometry affect collective decision-making: the case of Periplaneta americana. Insectes Sociaux

Gregarious arthropods such as cockroaches, constitute the most widespread social species in the animal kingdom. Yet, as far as collective decision-making is concerned, most of the literature is focused on eusocial insects such as ants or termites, which are complex societies. In comparison, cockroaches of the species Periplaneta americana have (as far as we know) no division of labor. During the night, each individual of this species searches for food and during the day, the individuals are at rest. While the first activity is solitary, the second one is collective and implies interactions between individuals. Moreover, resting often happens in shelters that provide protection against predators and whose selection depends on their shape and/or their physical characteristics such as the level of darkness or the temperature.

The aim of the study was to look at the influence of the shape of the shelters and of the sexual composition of groups of individual cockroachesin the decision-making outcome, and to highlight the interactions at work. Despite their less sophisticated modes of communication, our objective was to demonstrate that some combinations of environmental and group-related factors could lead to different collective behaviors that are reminiscent to those observed in eusocial insects.

To do so, we performed lab experiments in which three different group compositions were faced with three different environments. Groups either constituted of 16 males, of 16 females or of 8 males and 8 females. As for the environments, the three different groups had to choose between two identical horizontal shelters where individuals can only settle on the floor of the shelters, two identical vertical ones where cockroaches can settle on the walls and the floor, or one vertical and one horizontal shelter.

Experimental set-up consisting of an arena and two dark shelters (a) along with an example of organization within a vertical (b) and a horizontal (c) shelter.

For the nine conditions we showed that the sheltering behavior implies different levels of interactions between individuals and therefore different levels of collective choices depending on the different group compositions tested. In particular, for the symmetrical geometrical conditions (two vertical and two horizontal shelters), the social inter-attraction between individuals in homogenous groups of males are weaker than the ones in homogenous groups of females or in heterogenous groups of males and females. Formulated differently, males have difficulty reaching consensus (all individuals within the same shelter), which is not the case for the groups of females or mixed-groups.

While it is known that these insects prefer to stand vertically, we expected to observe different collective choices between the two symmetrical (two horizontal or two vertical shelters) conditions, but we didn’t. Yet, in the asymmetrical case (one horizontal vs one vertical shelter), the choice towards the vertical shelter was clear for the three group compositions, although the homogenous groups of males and of females had a weaker selection than the mixed groups of males and females. This result may sound puzzling considering the results obtained for the symmetrical conditions, but could be explained by the fact that homogenous populations of males, while having weaker social interattractions between them (as compared to females), are counterbalanced by a higher preference for the vertical shelter. Conversely, females have stronger interattractions between them and a lower individual preference for the vertical shelter. As for the mixed groups, the strong male-female interattractions coupled with the vertical position preference for the males led to a stronger selection of the vertical shelter as compared to the homogenous groups.

Finally, in terms of internal organization within the vertical shelters, no sexual segregation was found to be at work. But we showed that the tendency to stand in the vertical part of the shelter is amplified with the sheltered population for groups of females and for mixed groups but is independent for groups of males.

The results obtained thanks to experiments in a simple device show that the coupling between individual responses to the environmental heterogeneities and the network of interactions between individuals can lead to a diversity of responses. While they contribute also to our understanding of how individuals of Periplaneta americana are interacting and how they collectively choose their resting place, the genericity of the mechanisms implied lead to the conviction that the phenomena presented are likely to be present in species presenting the same modes of organization.