Bumblebees acquire alternative puzzle-box solutions via social learning.

The astonishing behavioural repertoires of social insects have been thought largely innate, but these insects have repeatedly demonstrated remarkable capacities for both individual and social learning. Using the bumblebee Bombus terrestris as a model, we developed a two-option puzzle box task and used open diffusion paradigms to observe the transmission of novel, nonnatural foraging behaviours through populations. Box-opening behaviour spread through colonies seeded with a demonstrator trained to perform 1 of the 2 possible behavioural variants, and the observers acquired the demonstrated variant. This preference persisted among observers even when the alternative technique was discovered. In control diffusion experiments that lacked a demonstrator, some bees spontaneously opened the puzzle boxes but were significantly less proficient than those that learned in the presence of a demonstrator. This suggested that social learning was crucial to proper acquisition of box opening. Additional open diffusion experiments where 2 behavioural variants were initially present in similar proportions ended with a single variant becoming dominant, due to stochastic processes. We discuss whether these results, which replicate those found in primates and birds, might indicate a capacity for culture in bumblebees.

Non-numerical strategies used by bees to solve numerical cognition tasks.

We examined how bees solve a visual discrimination task with stimuli commonly used in numerical cognition studies. Bees performed well on the task, but additional tests showed that they had learned continuous (non-numerical) cues. A network model using biologically plausible visual feature filtering and a simple associative rule was capable of learning the task using only continuous cues inherent in the training stimuli, with no numerical processing. This model was also able to reproduce behaviours that have been considered in other studies indicative of numerical cognition. Our results support the idea that a sense of magnitude may be more primitive and basic than a sense of number. Our findings highlight how problematic inadvertent continuous cues can be for studies of numerical cognition. This remains a deep issue within the field that requires increased vigilance and cleverness from the experimenter. We suggest ways of better assessing numerical cognition in non-speaking animals, including assessing the use of all alternative cues in one test, using cross-modal cues, analysing behavioural responses to detect underlying strategies, and finding the neural substrate.

Honeybees solve a multi-comparison ranking task by probability matching.

Honeybees forage on diverse flowers which vary in the amount and type of rewards they offer, and bees are challenged with maximizing the resources they gather for their colony. That bees are effective foragers is clear, but how bees solve this type of complex multi-choice task is unknown. Here, we set bees a five-comparison choice task in which five colours differed in their probability of offering reward and punishment. The colours were ranked such that high ranked colours were more likely to offer reward, and the ranking was unambiguous. Bees' choices in unrewarded tests matched their individual experiences of reward and punishment of each colour, indicating bees solved this test not by comparing or ranking colours but by basing their colour choices on their history of reinforcement for each colour. Computational modelling suggests a structure like the honeybee mushroom body with reinforcement-related plasticity at both input and output can be sufficient for this cognitive strategy. We discuss how probability matching enables effective choices to be made without a need to compare any stimuli directly, and the use and limitations of this simple cognitive strategy for foraging animals.

Olfactory learning without the mushroom bodies: Spiking neural network models of the honeybee lateral antennal lobe tract reveal its capacities in odour memory tasks of varied complexities.

The honeybee olfactory system is a well-established model for understanding functional mechanisms of learning and memory. Olfactory stimuli are first processed in the antennal lobe, and then transferred to the mushroom body and lateral horn through dual pathways termed medial and lateral antennal lobe tracts (m-ALT and l-ALT). Recent studies reported that honeybees can perform elemental learning by associating an odour with a reward signal even after lesions in m-ALT or blocking the mushroom bodies. To test the hypothesis that the lateral pathway (l-ALT) is sufficient for elemental learning, we modelled local computation within glomeruli in antennal lobes with axons of projection neurons connecting to a decision neuron (LHN) in the lateral horn. We show that inhibitory spike-timing dependent plasticity (modelling non-associative plasticity by exposure to different stimuli) in the synapses from local neurons to projection neurons decorrelates the projection neurons' outputs. The strength of the decorrelations is regulated by global inhibitory feedback within antennal lobes to the projection neurons. By additionally modelling octopaminergic modification of synaptic plasticity among local neurons in the antennal lobes and projection neurons to LHN connections, the model can discriminate and generalize olfactory stimuli. Although positive patterning can be accounted for by the l-ALT model, negative patterning requires further processing and mushroom body circuits. Thus, our model explains several-but not all-types of associative olfactory learning and generalization by a few neural layers of odour processing in the l-ALT. As an outcome of the combination between non-associative and associative learning, the modelling approach allows us to link changes in structural organization of honeybees' antennal lobes with their behavioural performances over the course of their life.

A possible structural correlate of learning performance on a colour discrimination task in the brain of the bumblebee.

Synaptic plasticity is considered to be a basis for learning and memory. However, the relationship between synaptic arrangements and individual differences in learning and memory is poorly understood. Here, we explored how the density of microglomeruli (synaptic complexes) within specific regions of the bumblebee (Bombus terrestris) brain relates to both visual learning and inter-individual differences in learning and memory performance on a visual discrimination task. Using whole-brain immunolabelling, we measured the density of microglomeruli in the collar region (visual association areas) of the mushroom bodies of the bumblebee brain. We found that bumblebees which made fewer errors during training in a visual discrimination task had higher microglomerular density. Similarly, bumblebees that had better retention of the learned colour-reward associations two days after training had higher microglomerular density. Further experiments indicated experience-dependent changes in neural circuitry: learning a colour-reward contingency with 10 colours (but not two colours) does result, and exposure to many different colours may result, in changes to microglomerular density in the collar region of the mushroom bodies. These results reveal the varying roles that visual experience, visual learning and foraging activity have on neural structure. Although our study does not provide a causal link between microglomerular density and performance, the observed positive correlations provide new insights for future studies into how neural structure may relate to inter-individual differences in learning and memory.

How honey bees make fast and accurate decisions.

Honey bee ecology demands they make both rapid and accurate assessments of which flowers are most likely to offer them nectar or pollen. To understand the mechanisms of honey bee decision-making, we examined their speed and accuracy of both flower acceptance and rejection decisions. We used a controlled flight arena that varied both the likelihood of a stimulus offering reward and punishment and the quality of evidence for stimuli. We found that the sophistication of honey bee decision-making rivalled that reported for primates. Their decisions were sensitive to both the quality and reliability of evidence. Acceptance responses had higher accuracy than rejection responses and were more sensitive to changes in available evidence and reward likelihood. Fast acceptances were more likely to be correct than slower acceptances; a phenomenon also seen in primates and indicative that the evidence threshold for a decision changes dynamically with sampling time. To investigate the minimally sufficient circuitry required for these decision-making capacities, we developed a novel model of decision-making. Our model can be mapped to known pathways in the insect brain and is neurobiologically plausible. Our model proposes a system for robust autonomous decision-making with potential application in robotics.

In the natural world, decision-making processes are often intricate and challenging. Animals frequently encounter situations where they have limited information on which to rely to guide them, yet even simple choices can have far-reaching impact on survival. Each time a bee sets out to collect nectar, for example, it must use tiny variations in colour or odour to decide which flower it should land on and explore. Each 'mistake' is costly, wasting energy and exposing the insect to potential dangers. To learn how to refine their choices through trial-and-error, bees only have at their disposal a brain the size of a sesame seed, which contains fewer than a million neurons. And yet, they excel at this task, being both quick and accurate. The underlying mechanisms which drive these remarkable decision-making capabilities remain unclear. In response, MaBouDi et al. aimed to explore which strategies honeybees adopt to forage so effectively, and the neural systems that may underlie them. To do so, they released the insects in a 'field' containing artificial flowers in five different colours. The bees were trained to link each colour with a certain likelihood of receiving either a sugary liquid (reward) or bitter quinine (punishment); they were then tested on this knowledge. Next, MaBouDi et al. recorded how the bees would navigate a 'reduced evidence' test, where the colour of the flowers were ambiguous and consisted in various blends of the originally rewarded or punished colours; and a 'reduced reward likelihood' test, where the sweet recompense was offered less often than before. Response times and accuracy rates revealed a complex pattern of decision-making processes. How quickly the insects made a choice, and the types of mistakes they made (such as deciding to explore a non-rewarded flower, or to ignore a rewarded one) were dependent on both the quality of the evidence and the certainty of the reward. Such sophistication and subtlety in decision-making is comparable to that of primates. Next, MaBouDi et al. developed a computational model which could faithfully replicate the pattern of decisions exhibited by the bees, while also being plausible biologically. This approach offered insights into how a small brain could execute such complex choices 'on the fly', and the type of neural circuits that would be required. Going forward, this knowledge could be harnessed to design more efficient decision-making algorithms for artificial systems, and in particular for autonomous robotics.

Bumblebees Learn a Relational Rule but Switch to a Win-Stay/Lose-Switch Heuristic After Extensive Training.

Mapping animal performance in a behavioral task to underlying cognitive mechanisms and strategies is rarely straightforward, since a task may be solvable in more than one manner. Here, we show that bumblebees perform well on a concept-based visual discrimination task but spontaneously switch from a concept-based solution to a simpler heuristic with extended training, all while continually increasing performance. Bumblebees were trained in an arena to find rewards on displays with shapes of different sizes where they could not use low-level visual cues. One group of bees was rewarded at displays with larger shapes and another group at displays with smaller shapes. Analysis of total choices shows bees increased their performance over 30 bouts to above chance. However, analyses of first and sequential choices suggest that after approximately 20 bouts, bumblebees changed to a win-stay/lose-switch strategy. Comparing bees' behavior to a probabilistic model based on a win-stay/lose-switch strategy further supports the idea that bees changed strategies with extensive training. Analyses of unrewarded tests indicate that bumblebees learned and retained the concept of relative size even after they had already switched to a win-stay, lost-shift strategy. We propose that the reason for this strategy switching may be due to cognitive flexibility and efficiency.

Foraging Bumblebees Selectively Attend to Other Types of Bees Based on Their Reward-Predictive Value.

Using social information can be an efficient strategy for learning in a new environment while reducing the risks associated with trial-and-error learning. Whereas social information from conspecifics has long been assumed to be preferentially attended by animals, heterospecifics can also provide relevant information. Because different species may vary in their informative value, using heterospecific social information indiscriminately can be ineffective and even detrimental. Here, we evaluated how selective use of social information might arise at a proximate level in bumblebees (Bombus terrestris) as a result of experience with demonstrators differing in their visual appearance and in their informative value as reward predictors. Bumblebees were first trained to discriminate rewarding from unrewarding flowers based on which type of "heterospecific" (one of two differently painted model bees) was next to each flower. Subsequently, these bumblebees were exposed to a novel foraging context with two live painted bees. In this novel context, observer bumblebees showed significantly more social information-seeking behavior towards the type of bees that had predicted reward during training. Bumblebees were not attracted by paint-marked small wooden balls (moved via magnets) or paint-marked non-pollinating heterospecifics (woodlice; Porcellio laevis) in the novel context, indicating that bees did not simply respond to conditioned color cues nor to irrelevant social cues, but rather had a "search image" of what previously constituted a valuable, versus invaluable, information provider. The behavior of our bumblebees suggests that their use of social information is governed by learning, is selective, and extends beyond conspecifics.

Bumblebees Use Sequential Scanning of Countable Items in Visual Patterns to Solve Numerosity Tasks.

Most research in comparative cognition focuses on measuring if animals manage certain tasks; fewer studies explore how animals might solve them. We investigated bumblebees' scanning strategies in a numerosity task, distinguishing patterns with two items from four and one from three, and subsequently transferring numerical information to novel numbers, shapes, and colors. Video analyses of flight paths indicate that bees do not determine the number of items by using a rapid assessment of number (as mammals do in "subitizing"); instead, they rely on sequential enumeration even when items are presented simultaneously and in small quantities. This process, equivalent to the motor tagging ("pointing") found for large number tasks in some primates, results in longer scanning times for patterns containing larger numbers of items. Bees used a highly accurate working memory, remembering which items have already been scanned, resulting in fewer than 1% of re-inspections of items before making a decision. Our results indicate that the small brain of bees, with less parallel processing capacity than mammals, might constrain them to use sequential pattern evaluation even for low quantities.

Randomly weighted receptor inputs can explain the large diversity of colour-coding neurons in the bee visual system.

True colour vision requires comparing the responses of different spectral classes of photoreceptors. In insects, there is a wealth of data available on the physiology of photoreceptors and on colour-dependent behaviour, but less is known about the neural mechanisms that link the two. The available information in bees indicates a diversity of colour opponent neurons in the visual optic ganglia that significantly exceeds that known in humans and other primates. Here, we present a simple mathematical model for colour processing in the optic lobes of bees to explore how this diversity might arise. We found that the model can reproduce the physiological spectral tuning curves of the 22 neurons that have been described so far. Moreover, the distribution of the presynaptic weights in the model suggests that colour-coding neurons are likely to be wired up to the receptor inputs randomly. The perceptual distances in our random synaptic weight model are in agreement with behavioural observations. Our results support the idea that the insect nervous system might adopt partially random wiring of neurons for colour processing.

Counting insects.

When counting-like abilities were first described in the honeybee in the mid-1990s, many scholars were sceptical, but such capacities have since been confirmed in a number of paradigms and also in other insect species. Counter to the intuitive notion that counting is a cognitively advanced ability, neural network analyses indicate that it can be mediated by very small neural circuits, and we should therefore perhaps not be surprised that insects and other small-brained animals such as some small fish exhibit such abilities. One outstanding question is how bees actually acquire numerical information. For perception of small numerosities, working-memory capacity may limit the number of items that can be enumerated, but within these limits, numerosity can be evaluated accurately and (at least in primates) in parallel. However, presentation of visual stimuli in parallel does not automatically ensure parallel processing. Recent work on the question of whether bees can see 'at a glance' indicates that bees must acquire spatial detail by sequential scanning rather than parallel processing. We explore how this might be tested for a numerosity task in bees and other animals.This article is part of a discussion meeting issue 'The origins of numerical abilities'.

Representation of higher-order statistical structures in natural scenes via spatial phase distributions.

Natural scenes contain richer perceptual information in their spatial phase structure than their amplitudes. Modeling phase structure of natural scenes may explain higher-order structure inherent to the natural scenes, which is neglected in most classical models of redundancy reduction. Only recently, a few models have represented images using a complex form of receptive fields (RFs) and analyze their complex responses in terms of amplitude and phase. However, these complex representation models often tacitly assume a uniform phase distribution without empirical support. The structure of spatial phase distributions of natural scenes in the form of relative contributions of paired responses of RFs in quadrature has not been explored statistically until now. Here, we investigate the spatial phase structure of natural scenes using complex forms of various Gabor-like RFs. To analyze distributions of the spatial phase responses, we constructed a mixture model that accounts for multi-modal circular distributions, and the EM algorithm for estimation of the model parameters. Based on the likelihood, we report presence of both uniform and structured bimodal phase distributions in natural scenes. The latter bimodal distributions were symmetric with two peaks separated by about 180°. Thus, the redundancy in the natural scenes can be further removed by using the bimodal phase distributions obtained from these RFs in the complex representation models. These results predict that both phase invariant and phase sensitive complex cells are required to represent the regularities of natural scenes in visual systems.