Вы находитесь на странице: 1из 10

Review

Cognition with few neurons:


higher-order learning in insects
Martin Giurfa1,2
1
Université de Toulouse (UPS), Centre de Recherches sur la Cognition Animale, 31062 Toulouse cedex 9, France
2
Centre National de la Recherche Scientifique (CNRS), Centre de Recherches sur la Cognition Animale, 31062 Toulouse cedex 9,
France

Insects possess miniature brains but exhibit a sophisti- learning. Yet, at the same time, they may tend to create
cated behavioral repertoire. Recent studies have a feeling of wonder in non-scientific members of the public,
reported the existence of unsuspected cognitive capa- usually promoted by biased media reports or by scientists
bilities in various insect species that go beyond the themselves, which revolves around a growing tendency of
traditionally studied framework of simple associative attributing too easily humanlike qualities to animals [34].
learning. Here, I focus on capabilities such as attentional In the light of such an undesired effect, the critical question
modulation and concept learning and discuss their is not, therefore, whether insects achieve ‘marvelous feats’,
mechanistic bases. I analyze whether these behaviors, but how they achieve them. Here, I focus on a selection of
which appear particularly complex, can be explained on recent findings in insect cognitive behavior and analyze the
the basis of elemental associative learning and specific neural mechanisms known to mediate such behavior. I
neural circuitries or, by contrast, require an explanatory discuss whether behaviors that appear particularly com-
level that goes beyond simple associative links. In doing plex can be explained on the basis of elemental associative
this, I highlight experimental challenges and suggest learning and specific neural circuits or, by contrast, require
future directions for investigating the neurobiology an explanatory level that goes beyond this level.
of higher-order learning in insects, with the goal of
uncovering the basic neural architectures underlying Protocols for the study of elemental associative learning
cognitive processing. in insects
Insects have been traditional models for the study of
Introduction elemental associative learning (see Glossary). Multiple
Insects have historically fascinated biologists for the access cases of Pavlovian (association between an originally
they allow to the mechanisms and organization of behav-
ior. The fact that insects possess miniature nervous sys- Glossary
tems, with a low number of neurons (e.g., 100 000 neurons
Absolute classical conditioning (A+): a case of classical conditioning in which a
in the fruit fly brain [1] or one million in the bee brain [2]
single stimulus A is learned through its association with reinforcement (+).
versus 85 billion in the human brain [3]), does not consti- Concept learning: learning about relations between objects rather than about
tute a limitation for the production of sophisticated and absolute physical features of objects (e.g., color, shape, size). Extracting such
relations allows transferring a choice to unknown objects that may differ
complex behaviors [4–6]. Numerous insect species such as dramatically in terms of their physical features but that may fulfill the learned
bees, ants, parasitoid wasps, fruit flies, moths, butterflies, relation. Learning about concepts therefore comprises extracting from a series of
hissing bugs, crickets, grasshopper, and locusts, learn and problems the rule or relation that grants the common solution to these problems.
Delayed matching to sample: a training protocol used to determine whether a
memorize different sorts of sensory cues as predictors of subject can learn the concept of sameness. The subject must match its choice
reward [7–19] or punishment [20–26] and form memories to a stimulus that corresponds to a sample previously presented. Because the
of such experiences that can be retrieved at different times sample is regularly changed during the training, animals must learn the
concept of sameness; that is ‘always choose what is shown (the sample),
after learning, from the short- long-term range. The neural independently of what else is shown’.
circuits underlying such capabilities are only simple in Delayed non-matching to sample: a training protocol used to determine
appearance but exhibit an exquisite architecture [27]. whether a subject can learn the concept of difference. The subject must choose
a stimulus that is explicitly different from a sample previously presented.
The past decade has generated a wealth of novel re- Because the sample is regularly changed during the training, animals must
search on insect learning and memory that has overcome learn the concept of difference; that is, ‘always choose the opposite of what is
the classical framework of simple forms of associative shown (the sample), independently of what else is shown’.
Differential classical conditioning (A+ BS): a case of classical conditioning in
learning to focus on more elaborate cognitive capabilities which, in its simplest form, a stimulus A is learned through its association with
[4,5]. This includes work on attention-like processes in reinforcement (+) and a stimulus B through its association with absence of
reinforcement ().
fruit flies and honey bees [28,29], pessimistic biases [30],
Elemental associative learning: learning that is mediated by simple, univocal
and concept learning in bees [31–33]. These reports shed links between two (or more) events. For instance, an animal may learn an
new light on the cognitive richness of insect behavior, elemental link between a tone and a food reward and another elemental link
and between a light and an electric shock as a punishment. Associations
which seems to transcend basic Pavlovian and operant learned are unambiguous and specific for the sensory cues learned and their
respective outcomes.
Corresponding author: Giurfa, M. (martin.giurfa@univ-tlse3.fr). Stimulus transfer: a process by which a new stimulus evokes a response instead
Keywords: associative learning; cognition; non-elemental learning; invertebrate; of a previous known stimulus that evoked originally the same response.
insect; neural circuits.

0166-2236/$ – see front matter ß 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tins.2012.12.011 Trends in Neurosciences, May 2013, Vol. 36, No. 5 285
Review Trends in Neurosciences May 2013, Vol. 36, No. 5

neutral stimulus, the conditioned stimulus [CS], and a between odor and sucrose reward such that they exhibit
biologically relevant stimulus, the unconditioned stimulus conditioned PERs to later presentations of the odor alone
[US]) and operant learning (association between a given (Figure 1a). The association built is Pavlovian and, depend-
behavior and a resulting reinforcement) have been de- ing on the number of trials, interstimulus interval, intertrial
scribed and studied in insects. For instance, in the honey interval, among other factors, can be consolidated in short-
bee Apis mellifera, olfactory conditioning of the proboscis term memory (STM), mid-term memory (MTM), or long-
extension response (PER) has been repeatedly used for the term memory (LTM) [11,36]. Classical conditioning has also
study of elemental classical conditioning and its neural been intensively studied in the fruit fly Drosophila melano-
substrates [35–37]. Individually harnessed hungry bees gaster. Flies can easily be trained to associate an odor (the
(Figure 1a) respond to stimulation of their antennae with CS) with an aversive electric shock (the US). The typical
sucrose solution (the US) with an extension of their probos- procedure consists of training groups of flies alternatively
cis (the PER). If a neutral odor (the CS) is forward-paired presented with two different odors, one paired with an
with sucrose, the bees learn an elemental association electric shock (CS+) and another non-paired with shock

(a)

Toothpick with
sucrose

Harnessed
bee CS

US

Odor delivering setup

(b)
Yaw torque signal
Laser Computer
diode Light
Torque source
meter

Color filter

Light
guides

Diffusor
Electric
motor Arena posion

TRENDS in Neurosciences

Figure 1. (a) Conditioning of the proboscis extension response (PER) in restrained honeybees. Left panel: An individual bee is immobilized in a metal tube so that only the
antennae and mouth parts (the proboscis) are free to move. The bee is set in front of an odorant stimulation setup, which is controlled by a computer and which sends a
constant flow of clean air to the bee. The air flow can be rerouted through cartridges presenting chemicals used for olfactory stimulation (conditioned stimuli [CSs]).
Sucrose solution (unconditioned stimulus [US]) is delivered by a toothpick to the antennae and to the proboscis. Right panel: After conditioning, the odor CS, which initially
did not evoke any response, triggers the PER. (b) The flight simulator used for visual conditioning of a tethered fruit fly. By courtesy of B. Brembs. Left panel: A Drosophila is
flying stationary in a cylindrical arena. The fly’s tendency to perform left or right turns (yaw torque) is measured continuously and fed into a computer, which controls arena
rotation. On the screen, four ‘landmarks’ – two Ts and two inverted Ts – are displayed to provide a referential frame for flight-direction choice. A heat beam focused on the
fly’s thorax is used as an aversive reinforcer. The reinforcer is switched on whenever the fly flies toward a prohibitive direction. The fly therefore controls reinforcer delivery
by means of its flight direction. Right panel: Detail of a tethered fly in suspended flight within the simulator.

286
Review Trends in Neurosciences May 2013, Vol. 36, No. 5

(CS–) [38]. Retention is measured in a T-maze where condi- and aversive reinforcements (mainly through dopaminer-
tioned flies must choose between the CS+ and the CS–. gic neurons for aversive reinforcement) (Figure 2b) have
Several different memories are induced in Drosophila by been relatively well characterized in bees and flies, but also
this olfactory conditioning depending on trial number, trial in other species such as ants, moths, and crickets
spacing, and retention period [21,24]. These include STM, [10,11,21,23,24,36,40,42–44]. In Drosophila, an intercon-
intermediate-term memory (ITM), and two separable forms nection between octopaminergic and dopaminergic path-
of LTM defined as anesthesia-resistant memory (ARM) and ways was recently found to mediate the appetitive
long-lasting LTM (l-LTM), of which only the latter depends reinforcement properties of sucrose [45,46].
on de novo protein synthesis.
Drosophila has also played a pivotal role in the study of Top-down modulation of elemental associative learning
operant learning. In this case, a fruit fly is suspended from in insects: the role of attentional processes
the thorax in the middle of a cylindrical arena that allows Attention describes our ability to focus our perception on
the presentation of visual landmarks (Figure 1b) [39]. The one stimulus (or group of related stimuli), while filtering
tethered fly flies stationary and, if some of these landmarks out other simultaneous stimuli that are less relevant at
are paired with the aversive reinforcement of an unpleas- any moment [47]. Several reports have recently indicated
ant heat beam pointed on the thorax, learns to fly toward a that attentional processes similar to those described in
safe direction, avoiding the dangerous-landmark direc- vertebrates can be identified in insects. Studies on bum-
tions [39,40] (Figure 1b). The fly learns to control reinforce- blebee and honey bee color learning [48,49] suggested that
ment delivery because its flight maneuvers determine the attentional processes may dramatically affect discrimina-
switching off of the heat beam if the appropriate flight tion capabilities and that the key issue is how the bees
directions are chosen [40], thus constituting a case of learn the visual task. After absolute conditioning, in which
operant learning. Classical associations can also be estab- a bee is trained with a single color rewarded with sugar
lished between the visual landmarks displayed on the water, discrimination of colors that are very similar is
cylinder wall and the reinforcer [41]. difficult, but this is facilitated after differential condition-
The neural circuits responsible for these forms of asso- ing in which the bee has to learn the distinction between a
ciative learning have been studied extensively, in particu- rewarded and a non-rewarded color, which usually takes
lar in the case of olfactory learning (Figure 2). The olfactory more trials [48]. The difference in performance suggests
circuit and the nervous circuits mediating appetitive that attentional processes are involved; in differential
(mainly through octopaminergic neurons) (Figure 2a) conditioning, the bee has to focus on the difference and

(a) (b)
Mushroom
bodies (MB) 200 μm 50 μm Mushroom
ca bodies (MB) ca

ca
PPL1 α
CN pe ml
α′
EN α PAM pe
γ
β β′
Opc lobe Opc lobe LH β PN EN
LH l-PN m-PN LH
LH Glo
DAL
SEG
Glo SEG
Antennal Antennal
lobe (AL) lobes (AL)
VUMmx1
ORN
ORN

TRENDS in Neurosciences

Figure 2. Neural pathways for conditioned stimulus (CS) and unconditioned stimulus (US) information in the honey bee and the fly brain. Note the different size scales for
each brain. (a) CS and US circuits responsible for appetitive olfactory learning (odor–sucrose association) in the honey bee. Left hemisphere – the olfactory (CS) circuit: The
antennal lobe (AL), first-order olfactory neuropil, receives input from approximately 60 000 olfactory receptor neurons (ORNs) located within sensilla on the antenna. The AL
comprises approximately 160 glomeruli (Glo), which are sites of synaptic interaction between ORNs, inhibitory local neurons (not shown), and approximately 800 projection
neurons (PNs), which convey processed information via different tracts (l-PN and m-PN) to higher brain centers, the mushroom bodies (MBs), and the lateral horn (LH). Each
MB comprises approximately 170 000 Kenyon cells. Their dendrites form the calyces (ca) and their axons form the pedunculus (Pe) and the two output lobes: the vertical (a)
lobe and the horizontal (b) lobe. Within the MBs, feedback neurons (not shown) project from the Pe and lobes back to the calyces, providing inhibitory feedback to the MB
input regions. Extrinsic neurons (ENs) take information from the Pe and the lobes and project to different parts of the protocerebrum, but most conspicuously to the LH.
Centrifugal neurons (CNs) provide feedback from the MBs to the ALs and are thought to be involved in retrograde modulation of antennal lobe circuits. Right hemisphere –
the sucrose (US) circuit: Gustatory sensory neurons on the antennae project to the dorsal lobe [95], whereas those on the mouthparts project to the subesophageal ganglion
(SEG) [96]. A single octopaminergic neuron, the ventral unpaired median neuron of the maxillary neuromere 1 (VUMmx1) (only one side is shown), represents sucrose
during appetitive conditioning. This neuron has its dendrites in the SEG and projects to the AL, the MB calyces, and the LH. (b) CS and US circuits responsible for aversive
olfactory learning (odor–electric shock association) in the fruit fly. Left hemisphere – the olfactory (CS) circuit: The AL receives input from approximately 1400 ORNs located
on the antenna and the maxillary palp. The AL comprises approximately 50 glomeruli (Glo). Approximately 150 PNs convey processed information to higher brain centers,
the MBs and the LH. Each MB comprises approximately 2500 Kenyon cells. Their dendrites form the calyces (ca); their axons form the Pe and the axon collaterals the vertical
lobes (a and a0 ) and the horizontal lobes (b, b0 , and g). ENs take information from the a and a0 lobes and project to both the ipsilateral and the contralateral LH. CNs provide
feedback from the MB to the ipsilateral AL and connect with the contralateral MB. Right hemisphere: The electric shock (US) circuit: Two clusters of dopaminergic neurons,
PP1 and PAM, whose activity is necessary and sufficient to represent the electric shock in aversive olfactory conditioning, project to different subcompartments of the MBs
and induce memories of differing stability.

287
Review Trends in Neurosciences May 2013, Vol. 36, No. 5

not the mere presence of a visual target, thus making The key mechanism for attentional processes in insects
learning slower. This hypothesis is confirmed by experi- may be found in neuromodulatory neurons that set arousal
ments on bumblebees in which incorrect choices in a color thresholds for different nervous circuits, including those of
discrimination task, which are usually followed by the mushroom bodies [54]. The fruit fly mushroom bodies are
absence of reward, are further penalized by delivery of richly innervated by different subsets of dopaminergic
quinine solution [50]; in these circumstances, the bumble- neurons[55,56] and although the classic view involves
bees’ choices become slower and more accurate, thus re- considering these neurons as the neural substrate for
vealing a trade-off between speed and accuracy in the punishment signaling in aversive olfactory learning [57],
context of foraging decisions. Also, honey bees confronted current views relate dopamine levels in the insect brain
with discrimination of colors that are perceptually similar with arousal levels [54]. Indeed, transient attenuation of
learn the task only if incorrect choices are penalized with dopamine release in fly mutants attenuates the 20–30 Hz
quinine [51], a penalization that has no effect if the colors responsiveness to the moving black bar in flies and oral
are different enough to facilitate the discrimination task. delivery of methamphetamine, which increases dopamine
These results indirectly suggest that attentional processes release, rescued this responsiveness [58]. Thus, dopami-
exist in insects and that they can be modulated by experi- nergic neurons may modulate selective attention in the
ence. insect brain, acting on a series of nervous circuits underly-
Evidence for attentional processes requires demonstra- ing different forms of sensory–motor performances.
tion of suppressed responsiveness to competing stimuli
[52]. This requirement has been met in experiments with Cognitive interpretations of elemental learning in
honey bees trained to choose a colored disc (‘target’) among insects: pessimistic biases in bees
a varying number of differently colored discs (‘distractors’) Recent work on olfactory learning in honey bees has at-
[29]. The bees’ task was to suppress responsiveness to tributed to them ‘pessimistic cognitive biases’ under nega-
distractors, and accuracy and decision time were measured tive, stressful circumstances [30]. In humans, pessimistic
as a function of distractor number and color. For all color biases are defined as the increased expectation of bad
combinations, decision time increased and accuracy de- outcomes after negative affective states such as stress or
creased with increasing distractor number, whereas per- anxiety. Is it appropriate to use this anthropocentric ter-
formance improved when more targets were present. minology to describe an insect’s behavior? Or does it exceed
These findings recall serial search strategies in primates, the specific criteria used to test the behavior of honey bees?
in which stimuli are examined sequentially. Thus, at the In this study, bees were trained using olfactory condi-
behavioral level, the strategies implemented by bees con- tioning of the PER (see above and Figure 1a) to discrimi-
verge with those of animals in which attention is commonly nate two mixtures of the same two odors A and B. As an
studied [29]. example, a mixture where the ratio A:B was 1:9 was
Although behavioral approaches only have been used to rewarded with sucrose solution (CS+), whereas another
study attention in bees, selective attention in fruit flies has mixture in which the ratio was 9:1 was punished with
been retraced to the neural level [28]. A tethered fly flying quinine solution (CS–). After training, bees were vigorous-
stationary within a cylindrical arena (similar to that in ly shaken to simulate a predatory attack and induce an
Figure 1b) and tracking a visual object (a vertical black bar ‘anxiety-like state’ that would in turn determine ‘pessimis-
displayed on the cylinder wall) moving at a constant fre- tic’ evaluation of ambiguous stimuli. Bees were then pre-
quency around it exhibits anticipatory behavior consistent sented with the trained mixtures (9:1 and 1:9), but also
with attention for the bar it tracks. The neural correlate of with novel ratios of A and B that they had not experienced
such anticipatory response is a transient increase in a during training (3:7, 1:1, and 7:3). These stimuli being
20–30 Hz local field potential recorded in a region of the novel to the bees, the question was whether shaken bees
brain called the medial protocerebrum [28]. The neural would consider, for instance, the 7:3 mixture as being more
response is not only anticipatory, but also selective and similar to the punished 9:1 mixture than would non-shak-
specific to the shape presented, increased by novelty and en bees. In this case, shaken bees would respond less to the
salience, and reduced when the fly is in a sleep-like state 7:3 mixture than would non-shaken bees, a result that
[28]. Moreover, the use of mutants showed that output of would be consistent with a pessimistic judgment bias of the
a- and b-lobe neurons of the mushroom bodies (Figure 2b), novel ambiguous mixture. This was exactly what was
which are higher-order structures of the insect brain observed [30] and it was concluded that these results
(Figure 2), is required for both the tracking response ‘add an invertebrate animal to the growing list of verteb-
and the 20–30 Hz response. In another study, competing rates. . .that when subjected to various forms of negative,
moving gratings were presented to either eye of the fly in stressful manipulations exhibit pessimistic judgment
the cylindrical arena [53]. Confronted with these percepts, biases’ [30].
flies alternated their flight direction between the right and Although this interpretation is appealing, an alterna-
left moving grating, in an attention-like process that sup- tive explanation is possible. The task inculcated by the
pressed responses to the non-followed grating. Brain ac- training was discrimination between two odor mixtures,
tivity recorded in parallel at the two optic lobes of the fly which non-shaken bees achieved at a level of approximate-
increased mostly in the 20–50 Hz range and only on the ly 35% (the difference between the percentage of responses
side associated with a behavioral choice [53], consistent to the CS+ and to the CS–). Given this small difference, the
with previous reports on the involvement of this frequency original discrimination was obviously not easy for the bees.
domain in attention-like processes [28]. After shaking, differentiation increased to 65%, because
288
Review Trends in Neurosciences May 2013, Vol. 36, No. 5

responses to the CS– decreased whereas responses to the Concept learning is particularly interesting for the
CS+ remained constant, probably due to a ceiling effect. study of non-elemental learning, because it requires trans-
One can thus argue that shaking induced better discrimi- fer that occurs independently of the physical nature of the
nation between the two original mixtures 1:9 (CS+) and 9:1 stimuli considered (e.g., colors, shape, size) [63,64] and
(CS–) and that, as a consequence of this improved discrim- relies on relations between objects [65,66]. Examples of
ination, bees responded less to the novel 7:3 mixture. In such relations are ‘same as’, ‘different of’, ‘above/below of’,
other words, shaking enhanced attentional processes, and ‘on the left/right of’. Extracting such relations allows
resulting in better discrimination performances and less transferring a choice to unknown objects that may differ
generalization between the CS– and a mixture that was dramatically in terms of their physical features but that
close to it. No terminology centered on pessimism would be may fulfill the learned relation.
required in this scenario. Rather, these findings could be Various recent reports have indicated that honey bees
depicted as a case of attentional modulation of elemental learn relational rules of different sorts. These include
olfactory learning after a negative experience. ‘sameness/difference’ [31], ‘above/below’ [32], and the
A critical question is therefore whether basic levels of mastering of two rules simultaneously, ‘above/below’ (or
interpretation are possible when studying experience- left/right) and ‘different of’ [33]. Similarly, recent reports
dependent changes in insect behavior. This argument on ‘numerosity’ in bees suggest a capacity to extract infor-
should also be applied to vertebrate studies. An interesting mation about number irrespective of the physical features
example is provided by two independent sets of experi- of the objects counted [67,68].
ments in which an entrapped rat is set free by other rats
[59] and an entrapped ant is set free by other ants of its Sameness/difference concepts
colony [60]. Whereas the former results have been pre- Learning of the concepts of sameness and difference was
sented as a straightforward case of empathy, thus assign- demonstrated through the protocols of delayed matching to
ing to rats the capacity to recognize and share sensations sample (DMTS) and delayed non-matching to sample
experienced by another individual, the ant case has been (DNMTS), respectively [31]. Honey bees foraging in a Y-
interpreted more cautiously without reference to sharing of maze (Figure 3a) were trained in a DMTS experiment in
sensations; more importantly, this difference in interpre- which they were presented with a changing non-rewarded
tation between the two studies has led to the proposition of sample (i.e., one of two differently colored disks – ‘color
an experimental agenda that should be respected before group’ – or one of two different black-and-white gratings,
speaking about empathy in animals [61]. vertical or horizontal – ‘pattern group’) at the entrance of a
maze (Figure 3b). The bees were rewarded only if they
Non-elemental learning in insects: learning about chose the stimulus identical to the sample once within the
concepts maze. Bees trained with colors and presented in transfer
A higher level of complexity is reached when animals tests with black-and-white gratings that they had not
respond in an adaptive manner to novel stimuli that they experienced before solved the problem and chose the grat-
have never encountered before and that do not predict a ing identical to the sample at the entrance of the maze.
specific outcome based on the animals’ past experience. Similarly, bees trained with the gratings and tested with
Such positive transfer of learning (also called stimulus colors in transfer tests also solved the problem and chose
transfer) brings us, therefore, to a domain that differs from the novel color corresponding to that of the sample grating
that of elemental forms of learning [62]. at the maze entrance (Figure 3c). Transfer was not limited

(a) (b) Transfer (c) Transfer tests with paerns Transfer tests with colors
test (training with colors) (training with paerns)
100
Key: Preference for vercal Key: Preference for blue
Preference for horizontal Preference for yellow

80 * * *
*
% Correct choices

60

Training 40
(60 trials)

20

0
Vercal Horizontal Blue Yellow
Paern Color
group group Sample
TRENDS in Neurosciences

Figure 3. Concept learning in honey bees. Sameness learning [31]. (a) Y-maze used to train bees in a delayed matching-to-sample task. Bees had to enter into the maze to
collect sugar solution on one of the back walls of the maze. A sample was shown at the maze entrance before bees accessed the arms of the maze. (b) Training protocol. A
group of bees was trained during 60 trials with black-and-white, vertical, and horizontal gratings (Pattern Group); another group was trained with colors, blue and yellow
(Color Group). After training, both groups were subjected to a transfer test with novel stimuli (patterns for bees trained with colors, colors for bees trained with patterns). (c)
Performance of the Pattern and Color groups in the transfer tests with novel stimuli. Both groups chose the novel stimulus corresponding to the sample, although they had
no experience with such test stimuli. *p < 0.05. Adapted from [31].

289
Review Trends in Neurosciences May 2013, Vol. 36, No. 5

to different types of visual stimulus (pattern versus color), These results showed that the honey bee possesses the
but could also operate between drastically different senso- faculty to extract a conceptual above/below relationship
ry modalities such as olfaction and vision [31]. Bees also from a set of training stimuli and to transfer this concept to
mastered a DNMTS task and learned a rule of difference newly encountered stimuli. This faculty can be useful in a
that was transferred between different sensory stimuli navigational context in which bees may decide routes
such as colors and achromatic patterns [31]. These results based on spatial relationships between landmarks, which
were the first to document that bees learn rules relating could be generalized to novel situations [72].
stimuli in their environment. They were later verified in
experiments showing that bees categorize visual images Mastering two concepts simultaneously
based on general features common to these images [69] and Processing several concepts simultaneously presupposes
in a study showing that the working memory underlying an even higher level of cognitive sophistication than deal-
the solving of the DMTS task lasts for approximately 5 s ing with one concept at a time. In a recent study [33], honey
[70], a period that coincides with the duration of other bees were shown to rapidly master two abstract concepts
visual and olfactory STMs characterized in simpler forms simultaneously, one based on spatial relationships (above/
of associative learning in honey bees [11]. The capacity to below and right/left), and another based on the perception
learn a rule of sameness has been also used to study the of difference.
presence of a sense of numerosity in bees (see [68] and Bees that learned to classify visual targets using this
below). dual concept transferred their choices to unknown stimuli
that offered a best match in terms of dual-concept avail-
Above/below concepts ability: their components presented the appropriate spa-
For many animals that must operate in complex natural tial relationship and differed from one another. As in the
environments, spatial concepts such as right, left, above, previous study [32], a series of internal within-subject
and below are of crucial importance to generate appropri- controls and simulation algorithms allowed researchers
ate relational displacements and orientation. A recent to exclude confounding low-level cues such as the global
study investigated whether honey bees learn an above/ center of gravity, the global orientation of the stimuli, and
below relationship between visual stimuli and transfer it to the retinotopic similarity between the rewarded stimuli.
novel stimuli that are perceptually different from those These results thus demonstrated that the miniature brains
used during the training [32]. Bees were trained to fly into of bees can extract abstract at least two different concepts
a Y-maze and choose visual stimuli presented above or from a set of complex pictures and combine them in a rule
below a horizontal bar. Training followed a differential for subsequent choices [33].
conditioning procedure in which one spatial relation (e.g.,
‘target above bar’) was associated with sucrose solution Neural bases of concept learning in insects
whereas the other relation (e.g., ‘target below bar’) was Neural correlates of rules have been reported in the pre-
associated with quinine solution. One group of bees was frontal cortex of both monkeys and rodents [73–75]. In the
rewarded on the ‘target above bar’ relation and another dorsolateral, ventrolateral, and orbitofrontal prefrontal
group was rewarded on the ‘target below bar’ relation. cortex of monkeys, electrophysiological recordings showed
After completing the training, bees were subjected to a the presence of neurons exhibiting greater activity during
non-rewarded transfer test in which a novel target stimu- sameness trials (DMTS) as well as neurons showing great-
lus (not used during the training) was presented above or er activity during difference trials (DNMTS), regardless of
below the bar. Despite the novelty of the test situation, which visual sample was used to build the sameness/
which preserved the spatial relationship to the bar as the difference relationship [75]. In the case of insects, with
single criterion predicting or not predicting the presence of no prefrontal cortex, the critical question is which neural
sucrose reward, bees responded appropriately: if trained architecture could mediate this performance. Given that
for the above relationship they chose the novel stimulus stimulus transfer occurs between distinct sensory modali-
above the bar and if trained for the below relationship they ties such as olfaction and vision [31], a neural candidate
chose the novel stimulus below the bar [32]. should, like the prefrontal cortex of primates, receive
Because during the training all stimuli always appeared segregated multisensory input, be associated with rein-
in the same region of the visual field (e.g., the upper field forcement systems, and display an intrinsic relationship
for bees trained to ‘above’ and the lower field for bees with memory systems [76]. Additionally, it should provide
trained to ‘below’), an alternative interpretation could posit multimodal output consistent with response transfer to
that, instead of learning a conceptual relationship, bees stimuli belonging to different domains. These require-
simply relied on the statistic distributions of image differ- ments are met by the mushroom bodies of honey bees.
ences, which are different for the two problems. A series of These central structures of the bee brain exhibit multiple
snapshots acquired during training would determine an parallel neuron architecture indicative of higher-order
average stimulus that would be spatially distinct between integration (Figure 4). In the bee, each mushroom body
the above and the below training. Furthermore, if bees comprises approximately 170 000 tightly packed, parallel
relied on a simple cue like the center of gravity of the neurons called Kenyon cells. The bee mushroom bodies
patterns [71], which would be associated with reward, the receive compartmentalized multisensory input (olfactory,
problem becomes elemental. These alternative interpreta- visual, mechanosensory, gustatory) [77–80] and their ex-
tions were excluded one by one by further experiments trinsic neurons, such as the Pe1 neuron, respond to various
specifically conceived for this aim [32]. stimuli including sucrose, odors, mechanosensory, and
290
Review Trends in Neurosciences May 2013, Vol. 36, No. 5

Kenyon cell
somata

Lip
Olf.
Mushroom
Calyx

Vis.
ring Collar Mech.gust.
body
Basal

Mech.gust.
Olf.

(α) lobe
Vercal
Lateral
Pe1 neuron horn
(extrinsic neuron)

Medial (β) lobe


TRENDS in Neurosciences

Figure 4. Scheme of a mushroom body (delineated by dashed straight lines) showing segregated multisensory input at the level of the calyx and integrated multimodal
output at the level of the vertical (a) lobe. The somata of the Kenyon cells (KCs), which integrate the mushroom body, are located in the calyx bowl. The dendrites of the KCs
form the calyx, which is subdivided into three main regions: the lip, receiving olfactory afferences; the collar, receiving mainly visual but also mechanosensory and
gustatory afferences; and the basal ring, receiving olfactory, mechanosensory, and gustatory afferences [77,79,80]. The axons of the KCs subdivide and form the vertical (a)
and medial (b) lobes. An extrinsic, multimodal pedunculus neuron, the Pe1 neuron [81,97,98], is shown, whose dendrites arborize across the vertical lobe; its axon projects
to the lateral horn (delineated by a dashed circle).

visual [81–83] (Figure 4). This multimodal convergence vice versa. The critical questions are whether and how
is consistent with a capacity for integrating sensory infor- mushroom body architecture is crucial in mediating con-
mation across various modalities and mushroom body cept learning and which neural requisites are important
subcompartments and equips the mushroom bodies for for the extraction of relational rules. From this perspective,
higher-order multimodal computations; in particular, for experiments addressing, through specific mushroom body
relational associations. Furthermore, the mushroom bodies blocking and activation, whether these structures are nec-
are tightly related to reinforcement systems (Figure 2). essary and sufficient for concept learning are of fundamen-
In the bee, octopaminergic neurons such as the ventral tal importance.
unpaired median neuron of the maxillary neuromere 1
(VUMmx1), which serves as an appetitive reinforcement Concluding remarks
system, converge with the regions of olfactory input of the The present review highlights novel studies of insect asso-
mushroom bodies. Activity in the VUMmx1 neuron substi- ciative learning that in most cases had the intention of
tutes for sucrose in the olfactory conditioning of the PER [84] transcending the traditional framework of research on
and thus may be the specialized reward system for olfactory simple stimulus–stimulus (or behavior–stimulus) associa-
cues. Also, dopaminergic neurons, which act as a punish- tions. They all represent a relatively new tendency of
ment system [20,55,57,85], converge with specific regions of appreciating the cognitive sophistication of the miniature
the mushroom bodies [55,56], thereby mediating aversive brains of insects [4,90–94]. Such a tendency is welcome in a
memories of an odor-shock association of differing temporal field where the focus on simple learning forms may have
stability [86]. Finally, mushroom bodies have been histori- sometimes overlooked the enormous richness of inverte-
cally characterized as a substrate for associative memories, brate behavior. The rigidity of some experimental designs
particularly LTMs [87–89]. frequently used in the laboratory to study insect behavior
In summary, the question of whether the mushroom leads to a danger that the restricted animal can only do
bodies are the insect equivalent of the primate prefrontal what the experimenter allows it to. Concluding from such
cortex is irrelevant, even if studies of the prefrontal cortex experiments that this is the only behavior the animal
might inspire research on mushroom body function and possesses may be fundamentally wrong.
291
Review Trends in Neurosciences May 2013, Vol. 36, No. 5

Box 1. Outstanding questions thus allowing crosstalking and information transfer. These
characteristics may allow positive transfer from a set of
 Experiments on elemental and non-elemental learning have
concentrated on few insect species (mainly bees, fruit flies, stimuli to novel ones, even if these belong to different
moths, cockroaches, parasitoid wasps, crickets, and ants). In- sensory modalities. This principle seems crucial for certain
creasing the number of species studied should allow the tasks such as rule learning. With this in mind, a qualitative
determination of species-specific differences and common me- step forward in the study of these phenomena would be to
chanisms of learning and memory formation. The huge variety of
integrate this knowledge in a characterization of the neu-
species and ecological niches occupied by insects should
stimulate an ecological and evolutionary analysis of the neuro- ral architectures that make non-elemental learning forms
biology of neural circuits underlying learning and memory. possible in insects.
 Studies revealing higher-order, non-elemental learning in insects
have been conducted almost exclusively in a single species, the Acknowledgments
honey bee A. mellifera. Are honey bees special with respect to The author thanks all members of his research team at the University of
other insects in cognitive terms? Which role do sociality and Toulouse for providing a stimulating and productive environment. He
central-place navigation (the necessity of always returning to a also thanks the French Research Council (CNRS), the University of
constant point in space, the hive) play in the development of Toulouse and the Institut Universitaire de France (IUF) for support.
cognitive skills in bees?
 To what extent are higher-order learning capacities uncovered by
experiments in insects applied in the day-to-day activities References
performed by these insects? Do they all have an adaptive value? 1 Chiang, A.S. et al. (2011) Three-dimensional reconstruction of brain-
Or are they the result of neural wiring whose primary function is wide wiring networks in Drosophila at single-cell resolution. Curr. Biol.
different but which nevertheless allows the emergence of these 21, 1–11
capacities (exaptation)? 2 Witthöft, W. (1967) Absolute Anzahl und Verteilung der Zellen im Hirn
 So far, studies of attention in fruit flies have suggested a role of der Honigbiene. Z. Morphol. Tiere 61, 160–184
dopaminergic signaling in attentional modulation. Do other 3 Herculano-Houzel, S. (2009) The human brain in numbers: a linearly
biogenic amines/neurotransmitters modulate attention in insects? scaled-up primate brain. Front. Hum. Neurosci. 3, 31
How do stress, experience, and reinforcements affect attention in 4 Chittka, L. and Niven, J. (2009) Are bigger brains better? Curr. Biol. 19,
the insect brain? R995–R1008
 Why do mushroom bodies have two calyces and a fused 5 Avarguès-Weber, A. et al. (2011) Visual cognition in social insects.
pedunculus in some species and a single calyx and pedunculus Annu. Rev. Entomol. 56, 423–443
in other species? Does this difference relate to relational (includ- 6 Greenspan, R.J. and Swinderen, B. (2004) Cognitive consonance:
ing spatial) learning capacities and demands? complex brain functions in the fruit fly and its relatives. Trends
 Mushroom bodies in the fruit fly D. melanogaster are unimodal Neurosci. 27, 707–711
and devoted to olfactory processing, whereas they are multi- 7 Hoedjes, K.M. et al. (2011) Natural variation in learning rate and
modal and process visual, olfactory, gustatory, and mechanosen- memory dynamics in parasitoid wasps: opportunities for converging
sory information in honey bees. Does this difference account for ecology and neuroscience. Proc. Biol. Sci. 278, 889–897
differences in behavioral sophistication between these two 8 Fan, R.J. et al. (1997) Behavioural analysis of olfactory conditioning in
species? the moth Spodoptera littoralis (Boisd.) (Lepidoptera: Noctuidae). J.
 Are there neurons in the insect brain that are directly related to Exp. Biol. 200, 2969–2976
concept and context-dependent learning? 9 Matsumoto, Y. and Mizunami, M. (2000) Olfactory learning in the
 Can insects reveal the minimal neural architecture mediating cricket Gryllus bimaculatus. J. Exp. Biol. 203, 2581–2588
concept learning and other forms of higher-order learning? 10 Giurfa, M. (2007) Behavioral and neural analysis of associative
learning in the honeybee: a taste from the magic well. J. Comp.
Physiol. A 193, 801–824
Yet, as underlined in this review, not all feats presented 11 Menzel, R. (1999) Memory dynamics in the honeybee. J. Comp. Physiol.
as highly cognitive are in fact distinct from non-elemental A 185, 323–340
12 Dupuy, F. et al. (2006) Individual olfactory learning in Camponotus
learning forms. A fundamental goal in conceiving and ants. Anim. Behav. 72, 1081–1091
interpreting complex insect behavior is, therefore, critical- 13 Daly, K.C. and Smith, B.H. (2000) Associative olfactory learning in the
ly to determine whether basic levels of interpretation are moth Manduca sexta. J. Exp. Biol. 203, 2025–2038
possible and to what extent they account for plastic insect 14 Raubenheimer, D. and Tucker, D. (1997) Associative learning by
locusts: pairing of visual cues with the separate consumption of
behavior. Additionally, it is important to implement a
protein and carbohydrate. Anim. Behav. 54, 1449–1459
strict agenda in which the use of anthropocentric terms 15 Bernays, E.A. and Wrubel, R.P. (1985) Learning by grasshoppers:
should be avoided when interpreting insect behavior. association of colour/light intensity with food. Physiol. Entomol. 10,
Focusing on the neural bases of insect higher-order learn- 359–369
ing is a way to avoid this, because the characterization of 16 Weiss, M.R. (1995) Associative colour learning in a nymphalid
butterfly. Ecol. Entomol. 20, 298–301
neural architectures should be a dispassionate endeavor.
17 Vinauger, C. et al. (2011) Learning the way to blood: first evidence of
The study of simple learning forms in insects, both dual olfactory conditioning in a blood-sucking insect, Rhodnius
appetitive and aversive, has shed light on how neural prolixus. I. Appetitive learning. J. Exp. Biol. 214, 3032–3038
circuits interact to produce plastic behavior and which 18 Guerrieri, F.J. and d’Ettorre, P. (2010) Associative learning in ants:
functional and structural changes occur at different stages conditioning of the maxilla-labium extension response in Camponotus
aethiops. J. Insect Physiol. 56, 88–92
of these circuits to support elemental learning and memory 19 Lucchetta, P. et al. (2008) Foraging and associative learning of visual
formation. Although many outstanding questions await an signals in a parasitic wasp. Anim. Cogn. 11, 525–533
answer (Box 1), two basic neural architectural principles 20 Vergoz, V. et al. (2007) Aversive learning in honeybees revealed by
that many invertebrate brains share with larger brains are the olfactory conditioning of the sting extension reflex. PLoS ONE 2,
the existence of specialized brain structures and circuits, e288
21 Davis, R.L. (2005) Olfactory memory formation in Drosophila: from
which refer to specific sensory domains, and of higher- molecular to systems neuroscience. Annu. Rev. Neurosci. 28, 275–302
order integration centers, in which information pertaining 22 Fiala, A. (2007) Olfaction and olfactory learning in Drosophila: recent
to these different domains converges and is integrated, progress. Curr. Opin. Neurobiol. 17, 720–726

292
Review Trends in Neurosciences May 2013, Vol. 36, No. 5

23 Keene, A.C. and Waddell, S. (2007) Drosophila olfactory memory: 53 Tang, S. and Juusola, M. (2010) Intrinsic activity in the fly brain
single genes to complex neural circuits. Nat. Rev. Neurosci. 8, gates visual information during behavioral choices. PLoS ONE 5,
341–354 e14455
24 Busto, G.U. et al. (2010) Olfactory learning in Drosophila. Physiology 54 Van Swinderen, B. and Andretic, R. (2011) Dopamine in Drosophila:
25, 338–346 setting arousal thresholds in a miniature brain. Proc. Biol. Sci. 278,
25 Simoes, P.M.V. et al. (2012) A long-latency aversive learning 906–913
mechanism enables locusts to avoid odours associated with the 55 Aso, Y. et al. (2010) Specific dopaminergic neurons for the formation of
consequences of ingesting toxic food. J. Exp. Biol. 215, 1711–1719 labile aversive memory. Curr. Biol. 20, 1445–1451
26 Vinauger, C. et al. (2011) Learning the way to blood: first evidence of 56 Mao, Z.M. and Davis, R.L. (2009) Eight different types of dopaminergic
dual olfactory conditioning in a blood-sucking insect, Rhodnius neurons innervate the Drosophila mushroom body neuropil:
prolixus. II. Aversive learning. J. Exp. Biol. 214, 3039–3045 anatomical and physiological heterogeneity. Front. Neural Circuits
27 Strausfeld, N.J. (2012) Arthropod Brains: Evolution, Functional 3, 5
Elegance, and Historical Significance, Belknap Press 57 Schwaerzel, M. et al. (2003) Dopamine and octopamine differentiate
28 van Swinderen, B. and Greenspan, R.J. (2003) Salience modulates 20– between aversive and appetitive olfactory memories in Drosophila. J.
30 Hz brain activity in Drosophila. Nat. Neurosci. 6, 579–586 Neurosci. 23, 10495–10502
29 Spaethe, J. et al. (2006) Do honeybees detect colour targets using serial 58 Andretic, R. et al. (2005) Dopaminergic modulation of arousal in
or parallel visual search? J. Exp. Biol. 209, 987–993 Drosophila. Curr. Biol. 15, 1165–1175
30 Bateson, M. et al. (2011) Agitated honeybees exhibit pessimistic 59 Bartal, I.B-A. et al. (2011) Empathy and pro-social behavior in rats.
cognitive biases. Curr. Biol. 21, 1070–1073 Science 334, 1427–1430
31 Giurfa, M. et al. (2001) The concepts of ‘sameness’ and ‘difference’ in an 60 Nowbahari, E. et al. (2009) Ants, Cataglyphis cursor, use precisely
insect. Nature 410, 930–933 directed rescue behavior to free entrapped relatives. PLoS ONE 4,
32 Avarguès-Weber, A. et al. (2010) Conceptualization of above and below e6573
relationships by an insect. Proc. Biol. Sci. 278, 898–905 61 Vasconcelos, M. et al. (2012) Pro-sociality without empathy. Biol. Lett.
33 Avarguès-Weber, A. et al. (2012) Simultaneous mastering of two 8, 910–912
abstract concepts by the miniature brain of bees. Proc. Natl. Acad. 62 Robertson, S.I. (2001) Problem Solving, Psychology Press
Sci. U.S.A. 109, 7481–7486 63 Murphy, G.L. (2002) The Big Book of Concepts, MIT Press
34 Balter, M. (2012) Animal cognition: ‘killjoys’ challenge claims of clever 64 Murphy, G.L. (2010) What are categories and concepts? In The Making
animals. Science 335, 1036–1037 of Human Concepts (Mareschal, D. et al., eds), pp. 11–28, Oxford
35 Bitterman, M.E. et al. (1983) Classical conditioning of proboscis University Press
extension in honeybees (Apis mellifera). J. Comp. Psychol. 97, 107–119 65 Zentall, T.R. et al. (2002) Categorization, concept learning, and
36 Giurfa, M. and Sandoz, J.C. (2012) Invertebrate learning and memory: behavior analysis: an introduction. J. Exp. Anal. Behav. 78, 237–248
fifty years of olfactory conditioning of the proboscis extension response 66 Zentall, T.R. et al. (2008) Concept learning in animals. Comp. Cogn.
in honeybees. Learn. Mem. 19, 54–66 Behav. Rev. 3, 13–45
37 Takeda, K. (1961) Classical conditioned response in the honey bee. 67 Dacke, M. and Srinivasan, M.V. (2008) Evidence for counting in insects.
J. Insect Physiol. 6, 168–179 Anim. Cogn. 11, 683–689
38 Tully, T. and Quinn, W.G. (1985) Classical conditioning and retention 68 Gross, H.J. et al. (2009) Number-based visual generalisation in the
in normal and mutant Drosophila melanogaster. J. Comp. Physiol. honeybee. PLoS ONE 4, e4263
Psychol. 156, 263–277 69 Zhang, S.W. et al. (2004) Grouping of visual objects by honeybees. J.
39 Götz, K.G. (1964) Optomotorische Untersuchung des visuellen Exp. Biol. 207, 3289–3298
Systems einiger Augenmutanten der Fruchtfliege Drosophila. 70 Zhang, S.W. et al. (2005) Visual working memory in decision making by
Kybernetik 2, 77–92 honey bees. Proc. Natl. Acad. Sci. U.S.A. 102, 5250–5255
40 Heisenberg, M. et al. (2001) Flexibility in a single behavioral variable of 71 Ernst, R. and Heisenberg, M. (1999) The memory template in
Drosophila. Learn. Mem. 8, 1–10 Drosophila pattern vision at th flight simulator. Vision Res. 39,
41 Brembs, B. and Heisenberg, M. (2000) The operant and the classical in 3920–3933
conditioned orientation of Drosophila melanogaster at the flight 72 Chittka, L. and Jensen, K. (2011) Animal cognition: concepts from apes
simulator. Learn. Mem. 7, 104–115 to bees. Curr. Biol. 21, R116–R119
42 Mizunami, M. et al. (2009) Roles of octopaminergic and dopaminergic 73 Miller, E.K. et al. (2003) Neural correlates of categories and concepts.
neurons in appetitive and aversive memory recall in an insect. BMC Curr. Opin. Neurobiol. 13, 198–203
Biol. 7, 46 74 White, I.M. and Wise, S.P. (1999) Rule-dependent neuronal activity in
43 Dupuy, F. et al. (2010) Calcium imaging in the ant Camponotus fellah the prefrontal cortex. Exp. Brain Res. 126, 315–335
reveals a conserved odour-similarity space in insects and mammals. 75 Wallis, J.D. et al. (2001) Single neurons in prefrontal cortex encode
BMC Neurosci. 11, 28 abstract rules. Nature 411, 953–956
44 Daly, K.C. et al. (2004) Learning modulates the ensemble 76 Barbey, A.K. and Patterson, R. (2011) Architecture of explanatory
representations for odors in primary olfactory networks. Proc. Natl. inference in the human prefrontal cortex. Front. Psychol. 2, 162
Acad. Sci. U.S.A. 101, 10476–10481 77 Mobbs, P.G. (1982) The brain of the honeybee Apis mellifera. I. The
45 Liu, C. et al. (2012) A subset of dopamine neurons signals reward for connections and spatial organization of the mushroom bodies. Philos.
odour memory in Drosophila. Nature 488, 512–516 Trans. R. Soc. Lond. B: Biol. Sci. 298, 309–354
46 Burke, C.J. et al. (2012) Layered reward signalling through octopamine 78 Ehmer, B. and Gronenberg, W. (2002) Segregation of visual input to the
and dopamine in Drosophila. Nature 492, 433–437 mushroom bodies in the honeybee (Apis mellifera). J. Comp. Neurol.
47 Posner, M.I. et al. (1980) Attention and the detection of signals. J. Exp. 451, 362–373
Psychol. 109, 160–174 79 Strausfeld, N.J. (2002) Organization of the honey bee mushroom body:
48 Giurfa, M. (2004) Conditioning procedure and color discrimination in representation of the calyx within the vertical and gamma lobes. J.
the honeybee Apis mellifera. Naturwissenschaften 91, 228–231 Comp. Neurol. 450, 4–33
49 Dyer, A.G. and Chittka, L. (2004) Fine colour discrimination requires 80 Schröter, U. and Menzel, R. (2003) A new ascending sensory tract to the
differential conditioning in bumblebees. Naturwissenschaften 91, 224– calyces of the honeybee mushroom body, the subesophageal-calycal
227 tract. J. Comp. Neurol. 465, 168–178
50 Chittka, L. et al. (2003) Psychophysics: bees trade off foraging speed for 81 Rybak, J. and Menzel, R. (1998) Integrative properties of the Pe1-
accuracy. Nature 424, 388 neuron, a unique mushroom body output neuron. Learn. Mem. 5, 133–
51 Avarguès-Weber, A. et al. (2010) Aversive reinforcement improves 145
visual discrimination learning in free-flying honeybees. PLoS ONE 82 Grünewald, B. (1999) Physiological properties and response
5, e15370 modulations of mushroom body feedback neurons during olfactory
52 Miller, S.M. et al. (2011) Attentional switching in humans and flies: learning in the honeybee Apis mellifera. J. Comp. Physiol. A 185,
rivalry in large and miniature brains. Front. Hum. Neurosci. 5, 188 565–576

293
Review Trends in Neurosciences May 2013, Vol. 36, No. 5

83 Homberg, U. and Erber, J. (1979) Response characteristics and individual differences in animal performance. J. Exp. Biol. 215,
identification of extrinsic mushroom body neurons of the bee. Z. 387–395
Naturforsch. 34, 612–615 92 Giurfa, M. (2003) Cognitive neuroethology: dissecting non-elemental
84 Hammer, M. (1993) An identified neuron mediates the unconditioned learning in a honeybee brain. Curr. Opin. Neurobiol. 13, 726–735
stimulus in associative olfactory learning in honeybees. Nature 366, 93 Zhang, S.W. et al. (2012) Visually guided decision making in foraging
59–63 honeybees. Front. Neurosci. 6, 88
85 Claridge-Chang, A. et al. (2009) Writing memories with light- 94 Srinivasan, M.V. (2010) Honey bees as a model for vision, perception,
addressable reinforcement circuitry. Cell 139, 405–415 and cognition. Annu. Rev. Entomol. 55, 267–284
86 Aso, Y. et al. (2012) Three dopamine pathways induce aversive odor 95 Haupt, S.S. (2007) Central gustatory projections and. side-specificity of
memories with different stability. PLoS Genet. 8, e1002768 operant antennal muscle conditioning in the honeybee. J. Comp.
87 Menzel, R. et al. (1974) Learning and memory in the honeybee. Physiol. A 193, 523–535
In Experimental Analysis of Insect Behaviour (Barton-Browne, L., 96 Rehder, V. (1989) Sensory pathways and motoneurons of the proboscis
ed.), pp. 195–217, Springer reflex in the suboesophageal ganglion of the honey bee. J. Comp.
88 Erber, J. et al. (1980) Localization of short-term memory in the brain of Neurol. 279, 499–513
the bee, Apis mellifera. Physiol. Entomol. 5, 343–358 97 Mauelshagen, J. (1993) Neural correlates of olfactory learning
89 Menzel, R. and Müller, U. (1996) Learning and memory in honeybees: paradigms in an identified neuron in the honeybee brain. J.
from behavior to neural substrates. Annu. Rev. Neurosci. 19, 379–404 Neurophysiol. 69, 609–625
90 Menzel, R. and Giurfa, M. (2001) Cognitive architecture of a mini- 98 Okada, R. et al. (2007) Learning-related plasticity in PE1 and other
brain: the honeybee. Trends Cogn. Sci. 5, 62–71 mushroom body-extrinsic neurons in the honeybee brain. J. Neurosci.
91 Dyer, A.G. (2012) The mysterious cognitive abilities of bees: why 27, 11736–11747
models of visual processing need to consider experience and

294

Вам также может понравиться