Learning from a miniature brain: what bees reveal to neuroscience

► Read for you

Learning, remembering, generalising: the honey bee performs surprisingly complex tasks with a tiny brain. Martin Giurfa’s review shows what neuroscience experiments reveal about these abilities — while also reminding us why laboratory findings need to be interpreted with care.

1. Key points

• This review shows that the honey bee does not learn only through simple association: it can also solve certain ambiguous tasks and generalise what it has learned.
• Olfactory conditioning of the proboscis extension reflex remains one of the most robust models for linking behaviour, neurons, and memory.
• The mushroom bodies play an important role in complex learning, but interpreting them as the "centre of higher cognition" should remain cautious.
• The most spectacular results — zero, arithmetic, mental number line — exist in controlled protocols but remain debated.
• At the apiary, the main value is in better understanding orientation, foraging, landmark recognition, and behavioural plasticity, rather than in deriving a new beekeeping rule.

2. What the study shows

Fig. 1: CS-US associations in the honey bee brain.
(a) Scheme of a frontal view of the bee brain showing the olfactory (CS; in blue on the left) and sucrose (US, in red on the right) central pathways.

The CS pathway: Olfactory sensory neurons send information to the brain via the antennal nerve (AN). In the antennal lobe (AL), these neurons synapse at the level of glomeruli (Gl) onto local interneurons (not shown) and projection neurons (Pn) conveying the olfactory information to higher-order centres, the lateral horn (LH) and the mushroom bodies (MB). MBs are interconnected through commissural tracts (in violet).

The US pathway: this circuit is partially represented by the VUMmx1 neuron, which has its soma in the subesophageal zone (SEZ) and converges with the CS pathway at three main sites: the AL, the LH and the MB. CC: central complex. Adapted from Menzel and Giurfa (2001).

(b) Scheme of the localisation and distribution of CS-US associations in the bee brain. ORNs: olfactory receptor neurons; GRNs: gustatory receptor neurons. The dashed line between GRNs and VUMmx1 indicates that this part of the circuit is actually unknown. (Source: Giurfa, M. (2025))

This chapter summarises Martin Giurfa's review on learning and memory in the honey bee. (Image: Giurfa, M. (2025))

Question. Martin Giurfa's paper raises a central question: how far can the cognition of an animal go when it has a miniature brain, smaller than 1 mm³ and comprising roughly one million neurons? The review aims to show how the honey bee has become a major experimental model for studying learning, memory, and certain more complex forms of cognition (Giurfa, 2025).

Method. The article is a scientific review, not a single experimental study. It brings together several decades of work around two main types of protocol. The first concerns free-flying bees trained to choose visual targets associated with a sucrose reward: colours, shapes, patterns, positions, or virtual objects. The second concerns bees harnessed in the laboratory, in which the proboscis extension reflex is conditioned. In this protocol, an odour is paired with a sucrose solution; if learning succeeds, the odour alone eventually elicits proboscis extension.

Results. Olfactory conditioning allows the involved neural pathways to be traced precisely. Odours are processed from the antennae to the antennal lobes, then to the mushroom bodies and the lateral horn. The sucrose reward is represented, in the historical work presented by the author, by the VUMmx1 neuron, an octopaminergic neuron that projects to these same regions. This convergence provides an anatomical and functional basis for the association between odour and reward. The article also emphasises that learning modifies the neural coding of odours, notably in the antennal lobes and mushroom bodies.

The review then presents more complex forms of learning. Bees can solve so-called non-elemental discriminations, for example when two odours A and B are individually rewarded but their compound AB is not. Such a task requires treating the compound as more than a simple sum of its components. According to the work summarised by Giurfa, the mushroom bodies are particularly involved in these ambiguous discriminations.

The article also brings together results on visual categorisation, relational rules such as "same" or "different", spatial relations such as "above" or "below", and certain numerical performances. In these protocols, bees trained with given stimuli can sometimes transfer their learning to new stimuli. Giurfa presents these results as indications of a more flexible cognition than has traditionally been attributed to insects.

Interpretation. The author defends a strong interpretation: some results would be consistent with concepts, relational rules, a form of numerosity, and even simple operations on small quantities. This reading is stimulating but not the only one possible. Other researchers propose more sober interpretations based on generalisation, continuous visual cues, or inspection strategies. The value of the review is therefore twofold: it shows the genuine richness of learning in bees while opening a discussion on what one can — and cannot — call "cognition" in a miniature brain.

3. Critical appraisal

The review is rich and useful, but some statements should be read with caution.

Strengths of the study. The main strength of the article lies in the quality of the experimental framework presented. Olfactory conditioning of the proboscis extension reflex is a robust, reproducible, and highly informative model. It allows a simple behavioural response to be linked to identified neural pathways, neuromodulators such as octopamine, and changes in neural coding. For this part, the state of knowledge is solid.

The review also has the merit of showing that the bee is not limited to reflexes. Its foraging, orientation, and recognition behaviours rest on individual experience. This helps explain why a forager can learn a source, return to a site, generalise a landmark, or modify its behaviour when conditions change.

Limits. The review is written by one of the major actors in the field. This is a strength, since the author knows the protocols and their history intimately. It is also a limit: several central results come from his own research network or from close collaborations. The text foregrounds the impact of the discoveries more than the methodological controversies surrounding them.

Another limit concerns the move from the laboratory to the apiary. Many results come from isolated bees trained in mazes, virtual reality setups, or under highly controlled conditions. Such protocols are indispensable for understanding mechanisms, but they do not reproduce the full complexity of a colony in natural conditions.

Possible biases or confounders. Tasks described as conceptual or numerical can sometimes be solved by strategies simpler than those suggested by the words "concept", "arithmetic", or "understanding of zero". A bee may exploit visual contrasts, surface areas, dot densities, flight routines, or inspection sequences without necessarily manipulating an abstract representation comparable to a human's.

Likewise, pharmacological inactivation of the mushroom bodies shows that a brain region is necessary in a given protocol, but it does not always say precisely which neurons, which synapses, or which computation are responsible. A harder task can also be more sensitive to a general disturbance of the brain: as in a fatigued human, complex exercises often fail before simple ones.

What cannot be concluded. The article does not support the conclusion that bees "reason" like vertebrates, nor that they understand numbers in the human sense of the term. Nor does it allow drawing a direct beekeeping rule about hive placement, feeding, selection, varroa control, or productivity. It mainly enables a better understanding of foragers' learning abilities and of the mechanisms that may support orientation and resource choice.

Limits of transfer to the Swiss or European apiary. The context of the review is not that of a field trial in Switzerland or in temperate Europe. The practical implications must therefore remain modest. The article helps to observe and interpret bee behaviour more carefully, but it replaces neither the Swiss health recommendations, nor on-site observations, nor the rules of colony management.

4. What related studies show

Related studies confirm several mechanisms while qualifying the most ambitious claims. (Image: Giurfa, M. (2025))

Mechanistic support. Devaud et al. (2015) reinforce the idea that the mushroom bodies are necessary for certain complex olfactory discriminations. In their experiments, bees whose mushroom bodies were locally blocked with procaine failed in configural discrimination tasks while still being able to solve elemental discriminations. This study therefore supports a specific role for the mushroom bodies in handling ambiguity.

Boitard et al. (2015) add an important nuance: the GABAergic feedback into the calyces of the mushroom bodies contributes to reversal learning, that is, the capacity to modify a previously learned association. This result is consistent with the idea that the mushroom bodies do not only store an association but also handle situations in which the meaning of a stimulus changes.

Methodological complement. Computational models invite caution in interpreting these results. Peng and Chittka (2017) show that a relatively simple model of the mushroom body circuit can reproduce forms of learning that appear complex, such as positive and negative patterning. This does not diminish the value of bees' performance, but it is a reminder that simple neural architectures can produce sophisticated behaviour. In other words: the behaviour is real; the interpretation as "higher cognition" is a matter of the level of description, not a directly observed fact.

Limits for abstract concepts. Experiments on spatial relations and abstract rules should be read carefully. Guiraud et al. (2018) showed, using high-speed videography, that bees can turn a task interpreted as learning the "above/below" concept into a simpler discrimination, based on stereotyped movements and sequential inspection of the elements. This does not deny the learning abilities of the bee, but it reduces the scope of the conceptual interpretation.

Debate on numerical cognition. Work on zero, addition, subtraction, and the mental number line is stimulating but debated. MaBouDi et al. (2021) showed experimentally that, when continuous visual cues — such as edge length or spatial frequency — are placed in conflict with the number of elements, bees follow the continuous cues and not the number. In some tasks it is therefore more accurate to say that they discriminate visual magnitudes rather than numbers in the strict sense.

The debate is also open on the mental number line. Pitt et al. (2023) contest the interpretation of an innate left-to-right mental number line in bees and propose alternative explanations linked to visual lateralisation. Giurfa et al. (2023) respond that the experimental controls do support a spatial interpretation of numerosity. The reasonable conclusion is therefore caution: this is an active and interesting field, but not a settled one.

Ecological relevance. Other studies bring cognition closer to natural situations. Grüter et al. (2011) show that flower constancy depends on ecologically realistic rewards: foragers are not simply rigid; they adjust their flower fidelity according to reward quality. Dyer et al. (2008) show that bees can recognise complex natural scenes as potential landmarks. Bullinger et al. (2023) show that foragers apply the memory of landmarks from their home area — for example, linear ground structures — when displaced to an unfamiliar area. Navigation thus also relies on generalisable structural elements, which is consistent with the practical attention paid to landmarks around the apiary.

5. Take-aways for the apiary

For the beekeeper, the value lies mostly in refining observation, not in directly changing colony management.

• Observing bees as animals capable of learning helps to better understand orientation flights, regular returns to a resource, and site fidelity.
• Visual landmarks around the apiary deserve attention, especially when the immediate environment of the hives is changed significantly. The study gives no strict rule, but it is a reminder that orientation partly depends on learned cues.
• Flower constancy and resource choice should not be seen as fixed automatisms. Foragers can adjust their behaviour according to reward, experience, and the cues available.
• The results on "concepts" or "numbers" are fascinating but remain mostly relevant for basic research. They are not enough to ground a practical beekeeping recommendation.
• At the apiary, this work encourages more nuanced observation: a change in entrance behaviour should be cross-checked with other signs — weather, nectar flow, colony state, disease pressure — before being interpreted.

These elements do not replace validated beekeeping recommendations, particularly on colony health, varroa control, feeding, or overwintering. They mainly cast light on the behavioural plasticity of bees.

Read the original study

Giurfa, M. (2025). Cognitive neuroscience and miniature brains—Dissecting higher-order learning in the brain of honey bees. Comptes Rendus Biologies, 348, 249–264. https://doi.org/10.5802/crbiol.187

Further reading on ApiSavoir

• Mini Brain — Mega Performance
• Behaviour and cognition: what a miniature brain teaches us
• The individual intelligence of the bee
• Bees can add and subtract
• How do bees see?
• Practical Guide: 4.8.1 Entrance observation

Short glossary

CS / conditioned stimulus
A learned signal, for example an odour that predicts a sugar reward for the bee.

US / unconditioned stimulus
A stimulus that naturally triggers a response, here the sugar solution that elicits proboscis extension.

Proboscis extension reflex
The response in which a bee extends its tongue when it detects or anticipates a sugar source.

Mushroom bodies
Brain regions in the bee involved in learning, memory and some complex discrimination tasks.

Octopamine
A signalling molecule in the insect nervous system, often associated with the sugar-reward signal in learning experiments.

VUMmx1 neuron
A neuron studied in honey bees that is involved, in some conditioning protocols, in representing the sugar reward.

Non-elemental learning
A learning situation in which the bee cannot simply associate one signal with one reward, but must process a combination or ambiguity.

References

Boitard, C., Devaud, J.-M., Isabel, G., & Giurfa, M. (2015). GABAergic feedback signaling into the calyces of the mushroom bodies enables olfactory reversal learning in honey bees. Frontiers in Behavioral Neuroscience, 9, 198. https://doi.org/10.3389/fnbeh.2015.00198

Bullinger, E., Greggers, U., & Menzel, R. (2023). Generalization of navigation memory in honeybees. Frontiers in Behavioral Neuroscience, 17. https://doi.org/10.3389/fnbeh.2023.1070957

Devaud, J.-M., Papouin, T., Carcaud, J., Sandoz, J.-C., Grünewald, B., & Giurfa, M. (2015). Neural substrate for higher-order learning in an insect: Mushroom bodies are necessary for configural discriminations. Proceedings of the National Academy of Sciences, 112, E5854–E5862. https://doi.org/10.1073/pnas.1508422112

Dyer, A. G., Rosa, M. G. P., & Reser, D. H. (2008). Honeybees can recognise images of complex natural scenes for use as potential landmarks. Journal of Experimental Biology, 211, 1180–1186. https://doi.org/10.1242/jeb.016683

Giurfa, M. (2025). Cognitive neuroscience and miniature brains—Dissecting higher-order learning in the brain of honey bees. Comptes Rendus Biologies, 348, 249–264. https://doi.org/10.5802/crbiol.187

Giurfa, M., Thevenot, C., & Rugani, R. (2023). Reply to Pitt et al.: Evidence from bees is consistent with a biological origin of a left-to-right mental number line. Proceedings of the National Academy of Sciences, 120, e2306470120. https://doi.org/10.1073/pnas.2306470120

Grüter, C., Moore, H., Firmin, N., Helanterä, H., & Ratnieks, F. L. W. (2011). Flower constancy in honey bee workers (Apis mellifera) depends on ecologically realistic rewards. Journal of Experimental Biology, 214, 1397–1402. https://doi.org/10.1242/jeb.050583

Guiraud, M., Roper, M., & Chittka, L. (2018). High-speed videography reveals how honeybees can turn a spatial concept learning task into a simple discrimination task by stereotyped flight movements and sequential inspection of pattern elements. Frontiers in Psychology, 9, 1347. https://doi.org/10.3389/fpsyg.2018.01347

MaBouDi, H., Barron, A. B., Li, S., Honkanen, M., Loukola, O. J., Peng, F., Li, W., Marshall, J. A. R., Cope, A. J., Vasilaki, E., & Solvi, C. (2021). Non-numerical strategies used by bees to solve numerical cognition tasks. Proceedings of the Royal Society B, 288, 20202711. https://doi.org/10.1098/rspb.2020.2711

Peng, F., & Chittka, L. (2017). A simple computational model of the bee mushroom body can explain seemingly complex forms of olfactory learning and memory. Current Biology, 27, 224–230. https://doi.org/10.1016/j.cub.2016.10.054

Pitt, B., Casasanto, D., & Piantadosi, S. T. (2023). No clear evidence for an innate left-to-right mental number line. Proceedings of the National Academy of Sciences, 120, e2306099120. https://doi.org/10.1073/pnas.2306099120