Download Olfaction in Invertebrates: Manduca. In: Squire LR (ed). Encyclopedia of Neuroscience, vol 7, pp 49-57. Oxford: Academic Press.

This article was originally published in the Encyclopedia of Neuroscience
published by Elsevier, and the attached copy is provided by Elsevier for the
author's benefit and for the benefit of the author's institution, for noncommercial research and educational use including without limitation use in
instruction at your institution, sending it to specific colleagues who you know,
and providing a copy to your institution’s administrator.
All other uses, reproduction and distribution, including without limitation
commercial reprints, selling or licensing copies or access, or posting on open
internet sites, your personal or institution’s website or repository, are
prohibited. For exceptions, permission may be sought for such use through
Elsevier's permissions site at:
http://www.elsevier.com/locate/permissionusematerial
Menzel R (2009) Olfaction in Invertebrates: Honeybee. In: Squire LR (ed.)
Encyclopedia of Neuroscience, volume 7, pp. 43-48. Oxford: Academic Press.
Author's personal copy
Olfaction in Invertebrates: Honeybee 43
Olfaction in Invertebrates: Honeybee
R Menzel, Freie Universität Berlin, Berlin, Germany
ã 2009 Elsevier Ltd. All rights reserved.
Olfactory Receptor Neurons and
Glomeruli
The honeybee’s strong points as a model system for
olfaction research are the facts that odor processing
can be studied through learning tests, with odor as an
appetitive stimulus, and that odor-induced neural excitation can be monitored both at the single-neuron level
and at the neural network level. A total of 60 000
olfactory receptor neurons (ORNs) on each antenna
project their axons to the antennal lobes (ALs), which
are subdivided into approximately 160 identified glomeruli. Most of the ORNs are in groups of 25–20 in
pore plates (sensilla placodea). The number and distribution of receptor molecules are unknown, but in analogy
with the findings in Drosophila and the mouse – for
which a correspondence between the number of olfactory receptor genes and the number of glomeruli has
been found – it is assumed that approximately 160
different ORNs may exist on the bee antennae.
Extracellular recordings from ORNs in pore plates
showed that the respective chemoprofiles are broad
and overlapping. Furthermore, the temporal response
characteristic of single ORNs can be complex, including both excitatory and inhibitory components to
particular odors, indicating antagonistic processing
either at the level of intracellular cellular pathways
initiated by more than one receptor protein in each
ORN, or by neural processing between ORNs within
one pore plate.
Each ORN innervates only one glomerulus, and
axons from ORNs within one pore plate reach different glomeruli. The glomeruli are interconnected by
about 4000 local interneurons, and about 800 projection neurons (PNs) lead from the glomeruli to higher
order brain centers, such as the mushroom bodies and
the lateral protocerebrum. Using calcium imaging, it
is possible to measure odor-evoked glomerular activity patterns in about 40 of the 160 glomeruli. Using a
computerized morphological atlas of AL glomeruli, it
was possible to map the identity of glomerular units
onto the physiological recordings done with calcium
imaging. With this procedure, it was possible to compare the response profiles of individual glomeruli
among specimens (Figure 1).
Glomerular Activity Pattern and Odor
Identity
Two main issues were addressed with the imaging
technique: first, whether odor representation is conserved within the species, and second, whether the
activity pattern elicited by an odor is sufficient to
predict the odor stimulus. The combinatorial pattern
of glomerular activity is indeed highly conserved
between individuals, indicating a tight developmental
control of ORN input processing in the AL. Furthermore, statistical analysis carried out on 18 of the 160
glomeruli of the bee (11%) using a discriminant analysis showed that odor representations of each given
odor form a coherent ‘cloud’ in the multidimensional
space where axes are defined by each of the identified
glomeruli. In 86% of these cases the odor could be
identified from the glomerular activity.
In a more extensive analysis the generalization profiles of 16 16 pairs of odors were determined using
the odor conditioning paradigm and compared with
the results from optophysiological recordings. The
perceptual distances correlate well with physiological
distances. It was concluded that the functional groups
of the primary and secondary aliphatic alcohols, aldehydes, and ketones, and carbon-chain lengths, are inner
dimensions of the honeybee olfactory space and that
neural activity in the AL reflects the perceptual quality
of odors. In both of these studies the perceptual similarity measure was the generalization profile, the response probability to an odor that was not learned after
the animal had been trained to a particular odor.
Another measure of stimulus similarity can be response time. In most examples studied so far, response
time and accuracy are inversely related to each other
and this appears to apply also for color discrimination
of honeybees. Under given conditions of reward and
motivation, the brain appears to accumulate the evidence against or in favor of a certain choice until a
determined threshold is reached. Free-flying bees can
be tested regarding how long they sniff an odor probe,
and whether sniffing time differs for the discrimination
of more similar or less similar odors. In this study, a
behavioral assay was used in which bees had to discriminate odors consisting of mixtures of two components, thus odors that varied greatly between very
similar and very different odors. Bees learned to discriminate all of the mixtures, including very similar
ones, and even different concentrations of the same
odor. Even though discriminating two very similar
Encyclopedia of Neuroscience (2009), vol. 7, pp. 43-48
Author's personal copy
44 Olfaction in Invertebrates: Honeybee
T1-17
T1-33
Relative response (%)
T1-28
100
C10
a
c
C5
C6
C7
C8
C9
C10
Figure 1 Representation of aliphatic alcohols. (a) Schematic view of the honeybee antennal lobe (AL), with the three glomeruli T1-28,
T1-17, and T1-33 as indicated. These three glomeruli are direct neighbors. (b) Responses of the T1-28, T1-17, and T1-33 glomeruli to a
series of alcohols, varying in carbon-chain length from C5 (1-pentanol) to C10 (1-decanol). Note that T1-28 responds most strongly to
short-chain alcohols, T1-17 to intermediate-chain lengths, and T1-33 to longer chain lengths. Each point represents the average of 14-21
individuals (error bars shown). Responses are shown relative to the response to hexanol in glomerulus T1-28 (asterisk). The shaded area
indicates noise levels. (c) Spatial activity patterns elicited by the same six alcohols as those in (b). Red indicates strong activity and dark
blue indicates low activity. Glomeruli that could not be measured are shown in gray. Note the continuous shift in activity between T1-28,
T1-17, and T1-33 as the carbon-chain length increases. Also note that the response patterns are not limited to these neighboring
glomeruli. Adapted from Sachse S, Rappert A, and Galizia CG (1999) The spatial representation of chemical structures in the AL of
honeybees: Steps towards the olfactory code. European Journal of Neuroscience 11: 3970–3982.
odors appears to be a more difficult task than discriminating two very distinct substances, it was found that
the time needed to make a choice (690 ms) for or
against an odor was independent of odor similarity.
These data suggest that, irrespective of the nature
of the olfactory code, the bee olfactory system evaluates odor quality after a constant interval. This
may ensure that odors are only assessed after the
olfactory network has optimized its representation,
an aspect which will be addressed in the following
sections.
Neural Processing in the AL
Glomeruli are connected by about 4000 local interneurons, many of which are immunoreactive to
g-aminobutyric acid (GABA). Comparing the activity
patterns of glomeruli as measured predominantly
from the input and the output sites of the glomeruli
indicated a sharpening of the odor-induced patterns,
decorrelating the activity patterns for different odors.
Frequently inhibitory responses and ‘off’ responses
were found in the output neurons of the glomeruli,
the PNs. Interglomerular inhibition is most likely
to occur between glomeruli with similar response
profiles in order to sharpen their somewhat fuzzy
response profiles.
Glomeruli with similar responses were often found
to be direct neighbors, and such neighbors may
mutually inhibit each other demonstrating lateral
inhibitory mechanisms. However, interglomerular
inhibition is not limited to direct neighbors, for the
following reasons. First, all glomeruli are approximately equidistant from the center of the antennal
lobe: the AL is spherical, with all glomeruli covering
the outside of the sphere. Local interneurons innervate
several glomeruli; however, their neurites do not travel
from one glomerulus to a neighbor, but rather from
one glomerulus to the central neuropil and from there
to other glomeruli. Second, not all glomeruli with
similar response profiles are direct neighbors. To investigate the underlying mechanisms, the GABA-receptor
antagonist picrotoxin (PTX) was applied and two separate inhibitory networks were found: one that is
GABAergic and modulates overall AL activity, another
that is PTX-insensitive and glomerulus-specific.
The net result of the two inhibitory networks
together is a globally modulated, contrast-enhanced,
and predictable representation of odors in the olfactory output neurons.
Encyclopedia of Neuroscience (2009), vol. 7, pp. 43-48
Author's personal copy
Olfaction in Invertebrates: Honeybee 45
Connecting the ALs with the Mushroom
Bodies
The second-order neuropils of olfactory processing
in the insect brain are the mushroom bodies and
the lateral horns (LHs) (Figure 2 (a)). Four different
morphological types of PNs leave the AL, two of
Mushroom
body
d
Clawed
KC
l-ACT
PN
v
Antennal
a
Clawed KCs
(∼20 000)
KC
+
−
+
PN
GABA
−
l-ACT PNs
(∼500)
b
c
Figure 2 The olfactory pathway in the bee brain and olfactory
processing in the MB. (a) Scheme of the bee brain. Lateral antennocerebral tract (l-ACT) projection neurons (PN, green) were
optically recorded at their dendrites in the antennal lobe (AL) and
at their presynaptic boutons in the lip of the MB calyx. Dendrites
and somata of clawed Kenyon cells (clawed KC, red) were
recorded in the MB calyx. Yellow arrows indicate sites of dye
injection: 1 for recording PN dendrites in the AL, 2 for recording
PN boutons in the MB calyces, and 3 for recording KCs in the MB
calyces. Squares represent the two imaged areas, AL (for measuring PN dendrites) and MB calyx (for measuring PN boutons
and/or clawed KCs). (b) The wiring diagram illustrates the divergent and convergent connectivity between PNs and clawed KCs.
About 400 cholinergic l-ACT PNs synapse onto roughly 20 000
clawed KCs. The dendrites of clawed KCs are small, arranged in
columns, and feature few clawlike synaptic specializations (black
circles). (c) Within the lip region, PNs synapse onto GABAergic
neurons which, in turn, make inhibitory synapses with PNs and
KCs. Moreover, GABAergic feedback neurons receive input in the
MB lobes and send their axons to calyx lip region. Since they leave
out the ventral vertical lobe, they presumably do not receive input
from clawed KCs. Thus, GABAergic neurons may provide local,
PN-driven inhibitory microcircuits within the MB calyx lip, and a
more global inhibitory feedback loop between the MB output and
input region. Adapted from Ganeshina O and Menzel R (2001)
GABA-immunoreactive neurons in the mushroom bodies of the
honeybee: An electron microscopic study. Journal of Comparative
Neurology 437: 335–349.
which project to the lip region of the mushroom
body (MB) and to the lateral protocerebrum, the
LH. PNs in these two pathways have been analyzed
more closely by intracellular recordings. PNs in the
median antennocerebral tract (m-ACT) code odors by
latency differences or specific inhibitory phases in
combination with excitatory phases, have more specific activity profiles for different odors, and thus
convey the information with odor-specific delay.
The PNs of the lateral antennocerebral tract (l-ACT)
code odors by spike rate differences, have broader
activity profiles for different odors, and convey the
information more quickly. Thus, less precise information about the olfactory stimulus appears to reach the
MB and the LH via neurons of the l-ACT, and odor
information subsequently becomes more specified by
activities of neurons of the m-ACT.
It was concluded that the separation into two neural pathways from the primary to the secondary centers of olfactory processing is not related to the
distinction between different odors, but rather relates
to a dual coding of the same odor by two different
neuronal strategies in the time domain. Part of the
olfactory information may also lie in the timing of
action potentials. For example, when olfactory processing is disturbed by applying the GABA-antagonist
PTX (which also affects the temporal pattern of AL
neuron firing, but may also affect the spatial representation of odors), similar odors are no longer
distinguished by honeybees.
Many more experiments are needed to understand
the relationship between temporal and spatial odor
representation in both the insect and the vertebrate
brain. One difficulty lies in the paucity of electrophysiologically recorded cells whose innervated glomeruli
have been identified, and which can therefore be used
to correlate the observed spatial and temporal activity
patterns. The atlas of the AL in the bee brain is an
excellent tool to establish, for each glomerulus, an olfactory response profile and temporal response patterns,
both for the ORN input and the output via the PNs.
Learning-Related Plasticity in the AL
The plasticity of the neural network of the AL has
been studied from two perspectives, that of neural
correlates of sensory memory and that of associative
memory after olfactory reward conditioning.
Sensory memory is a short-lived persistence of a
sensory stimulus in the nervous system, such as iconic
memory in the visual system. The effect of an
odor stimulus on the postsynaptic responses in PNs
was measured within the glomeruli. A single-odor
presentation changed the timing of spontaneous
activity across glomeruli, enhancing the probability
Encyclopedia of Neuroscience (2009), vol. 7, pp. 43-48
Author's personal copy
46 Olfaction in Invertebrates: Honeybee
of coactivity of glomeruli that had been active during
odor stimulation shortly before. Moreover, during the
first few minutes after odor presentation, correlations
between the spontaneous activity fluctuations suffice
to reconstruct the stimulus. These results were interpreted to reflect modifiable fluctuations as substrates
for Hebbian reverberations and sensory memory, a
mechanism that might well be generalized to other
neural systems.
Imaging glomerular activity during appetitive
learning of an odor leads to an increased response in
those glomeruli that are activated by the learned
odor, but not in those that respond to a specifically
untrained odor. The odor-specific activity pattern was
not changed qualitatively, but only quantitatively.
Since glomerular activity can also be enhanced by
higher concentration of the odor, it is not yet clear
how the learning-induced enhancement can be separated from the odor-concentration effect. Since these
experiments were carried out under conditions in
which the animal was trained to an odor, they not
only documented a learning-specific plasticity in the
AL, but also proved that the staining with the
Ca-fluorescence dye and the imaging procedure did
not interfere with the neural processing in the bee brain.
Recently similar associative conditioning experiments combined with Ca imaging were performed
with specific staining of a subset of the postsynaptic
elements in the glomeruli, the uniglomerular PNs of
the lateral antennoglomerular tract. It was found that
their responses to odors were remarkably resistant to
plasticity following a variety of appetitive olfactory
learning paradigms. There was no significant difference in the changes of odor-evoked activity between
single- and multiple-trial forward or backward conditioning, differential conditioning, or unrewarded
successive odor stimulation. PNs of the lateral antennoglomerular tract may thus be more involved in reliable odor coding rather than modified by learning.
The role in olfactory learning and memory retrieval of
PNs other than those of the imaged lateral antennoglomerular tract remains to be investigated.
Odor Processing in the MB
PNs convey the output of glomeruli to the MB and the
LH (Figure 2 (a)). Whereas nothing is known so far
about odor processing in the LH, the MBs have been
studied recently with Ca-imaging techniques. The MBs
are multisensory integration centers that play a dominant role in odor learning. They are densely packed
with 170 000 Kenyon cells (KCs), which receive
second-order sensory input in the MB calyces, with
different modalities innervating spatially distinct areas.
Olfactory input is confined to the lip region. KC axons
target output neurons in the vertical and medial lobes
of the MB. In these areas, GABAergic feedback
neurons receive input and project back to the MB lip
region, forming an inhibitory loop (Figure 2 (b)).
Electrophysiological recordings in locusts and imaging experiments in Drosophila indicated that odor
representations do indeed differ remarkably in the AL
and the MB. Unlike PNs, KCs respond to odors in
a sparse way. However, it is unclear whether this
transformation of odor representation is a result of
the KCs’ integration properties (as suggested for locust), or pre- and postsynaptic processing within the
MB lip region. Since the bouton-like PN terminals in
the MB lip are involved in reciprocal and feedback
and forward microcircuits between GABAergic neurons and PNs, information flow from PNs onto KCs
may not only be shaped by feed-forward processes,
but also may include interactions between PNs and
GABAergic neurons within the MB microcircuits.
It was, therefore, asked whether output activity of
PN boutons is modified by presynaptic processing
within the MB, by comparing odor-evoked responses
in PN dendrites within the AL glomeruli and their
boutons at the KC synapse. To reveal transformations
taking place in the postsynaptic KCs, odor-evoked
responses in KCs were recorded and compared to
their presynaptic input from PNs (Figure 3). At all
three processing stages (AL glomeruli, presynaptic
boutons of PNs in the MB lip, and postsynaptic
spine activity of KCs) odors reliably evoke combinatorial activity patterns. However, in contrast to PNs,
KCs code odors in a sparse way, and generate only
brief responses at stimulus onset. The KCs’ high degree of odor specificity originates at two steps: first, in
a presynaptic sharpening of PN synaptic output, and
second, in a diminishing KC response. Interestingly,
the temporal sharpening of KCs’ responses is established only at the postsynaptic side. These results also
show that PN activity generated within the first
200 ms determines whether a KC will respond. Thus,
the complex temporal response patterns of PNs are
transformed into brief phasic responses in KCs.
Thus, two types of transformations occur within
the MB: diminishment of a combinatorial code,
mediated by pre- and postsynaptic processing within
the MB microcircuits, and temporal sharpening of
postsynaptic KC responses, probably involving a
broader loop of inhibitory recurrent neurons.
Sniffing and Odor Discrimination
As mentioned previously, bees need a constant sniffing time to discriminate similar and different odors.
A task-independent choice time implies that quality is
assessed after a predetermined time span, and not as
soon as the neural codes for distinct odors differ to
a sufficient degree in order to justify a decision. As
Encyclopedia of Neuroscience (2009), vol. 7, pp. 43-48
Author's personal copy
Olfaction in Invertebrates: Honeybee 47
Excitation
Somata
5%
Inhibition
1%
a
b
a
hx1
lim
lio
oc2
b
c
2s
1%
1%
a
b
mint
d
Max
∆F/F
0
Merged, hx1-lio
e
hx1
lim
lio
oc2
Min
Merged, hx1-oc2
f
Odor
Figure 3 Response properties of l-ACT PNs. (a) Dendritic responses to odor stimulation in uniglomerular l-ACT PNs imaged in the
frontal AL glomeruli. Traces represent the time courses of Ca2þ transients evoked by 1-hexanol in two glomeruli, (a, b). d, dorsal; m,
medial; l, lateral; v, ventral. (b) Color-coded Ca2þ signals superimposed on the raw fluorescence image show that different odors activated
overlapping combinatorial sets of PN dendrites. (c) Traces show representative excitatory and inhibitory dendritic odor responses in the
AL (n ¼ 10 bees). Ca2þ transients typically followed complex temporal dynamics. Odor stimulus is highlighted. (d) Odor responses in axon
terminals (boutons) of PNs in the MB calyx. Boutons are visible in the raw fluorescence image. Stimulation with peppermint induced Ca2þ
increase in many of the boutons (single measurement). (e) Different odors evoked excitatory and inhibitory responses in distinct boutons.
The merged images of excitatory 1-hexanol (hx1, red) and linalool (lio, green) responses visualize distinct sets of activated boutons. In
contrast, 1-hexanol (red) and 2-octanol (oc2, green) activated overlapping sets of boutons. All traces represent single measurements;
images represent the averages of three stimulations. (f) Odor-evoked Ca2þ transients in PN boutons (n ¼ 8 bees) show complex
dynamics, as is the case in the dendrites. In the entire figure, traces ¼ DF/F; scale bars ¼ 50 mm. Adapted from Szyszka P, Ditzen M,
Galkin A, et al. (2005) Sparsening and temporal sharpening of olfactory representations in the honeybee mushroom bodies. Journal of
Neurophysiology. 94: 3303–3313.
Encyclopedia of Neuroscience (2009), vol. 7, pp. 43-48
Author's personal copy
48 Olfaction in Invertebrates: Honeybee
pointed out by Ditzen et al. in 2003, this finding may be
important in the context of olfactory coding, since it
implies that information about the odor is only available at a predefined time point after stimulus onset. In
the case of a ‘combinatorial code,’ one would have to
postulate that the across-glomeruli activity patterns are
only read out after a given delay. An internal ‘clock’
may provide such a signal in the form of odor-evoked
oscillations. If these oscillations have a frequency of
about 30 Hz in the bee brain – as in the locust brain –
a time span of 690 ms gives a high-end estimate of
21 cycles, from which the time not involved in odor
detection, such as further processing in the brain, and
motor commands, will have to be subtracted.
Both the ‘synchrony code’ (as suggested for the locust) and a combinatorial code that develops its optimal spatial coding over time (as shown for the
honeybee AL) rely on the sequence of activity patterns.
The sequences of two very different odors may differ
earlier than the sequences of two very similar odors.
Under these conditions, one would expect that the
brain would continuously monitor the evidence in
favor of each alternative, and make a choice as soon
as the evidence suffices for one, as has been shown in
other experimental paradigms. However, the results
reported in the foregoing text suggest that odor evaluation is not incremental, but rather occurs in single
information units, consisting of the entire sequence of
neural activity following an odor sample. It has been
suggested by Ditzen et al. that an interim storage of
the sequence would be needed in order to explain the
time constancy. Obviously, in the odor coding models
provided so far, there is a missing component: a mechanism that would allow for a constant readout time
of olfactory quality. This constant time requirement
for odor choice may guarantee that odor quality is
assessed when its coding is optimal.
See also: Olfaction in Invertebrates: Manduca; Olfaction in
Invertebrates: Drosophila; Olfactory Receptors; Olfactory
Cortex Physiology; Olfactory System Theory; Olfactory
System: Circuit Dynamics and Neural Coding in the Locust.
Further Reading
Chittka L, Dyer AG, Bock F, et al. (2003) Psychophysics: Bees trade
off foraging speed for accuracy. Nature 424: 388.
Ditzen M, Evers JF, and Galizia CG (2003) Odor similarity does
not influence the time needed for odor processing. Chemical
Senses 28: 781–789.
Faber T, Joerges J, and Menzel R (1999) Associative learning
modifies neural representations of odors in the insect brain.
Nature Neuroscience 2: 74–78.
Galán RF, Weidert M, Menzel R, et al. (2006) Sensory memory for
odors is encoded in spontaneous correlated activity between
olfactory glomeruli. Neural Computation 18: 10–25.
Galizia CG, Küttner A, Joerges J, et al. (2000) Odour representation in honeybee olfactory glomeruli shows slow temporal
dynamics: An optical recording study using a voltage-sensitive
dye. Journal of Insect Physiology 46: 877–886.
Galizia CG and Menzel R (2000) Odour perception in honeybees:
Coding information in glomerular patterns. Current Opinion in
Neurobiology 10: 504–510.
Galizia CG, Sachse S, Rappert A, et al. (1999) The glomerular code
for odor representation is species specific in the honeybee Apis
mellifera. Nature Neuroscience 2: 473–478.
Ganeshina O and Menzel R (2001) GABA-immunoreactive neurons in the mushroom bodies of the honeybee: An electron
microscopic study. Journal of Comparative Neurology 437:
335–349.
Getz WM and Akers RP (1994) Honeybee olfactory sensilla behave
as integrated processing units. Behavioral and Neural Biology
61: 191–195.
Guerrieri F, Schubert M, Sandoz JC, et al. (2005) Perceptual and
neural olfactory similarity in honeybees. PLoS Biology 3: e60.
Joerges J, Küttner A, Galizia CG, et al. (1997) Representation of
odours and odour mixtures visualized in the honeybee brain.
Nature 387: 285–288.
Laurent GJ (2003) Olfactory network dynamics and the coding
of multidimensional signals (review). Nature Reviews Neuroscience 3: 884–895.
Müller D, Abel R, Brandt R, et al. (2002) Differential parallel
processing of olfactory information in the honeybee, Apis
mellifera L. Journal of Comparative Physiology A 188:
359–370.
Peele P, Ditzen M, Menzel R, et al. (2006) Appetitive odor learning
does not change olfactory coding in a subpopulation of honeybee
antennal lobe neurons. Journal of Comparative Physiology A
192: 1083–1103.
Sachse S and Galizia CG (2002) The role of inhibition for temporal
and spatial odor representation in olfactory output neurons:
A calcium imaging study. Journal of Neurophysiology 87:
1106–1117.
Sachse S and Galizia CG (2003) The coding of odour-intensity in
the honeybee antennal lobe: Local computation optimizes
odour representation. European Journal Neuroscience. 18:
2119–2132.
Stopfer M and Laurent GJ (1999) Short-term memory in olfactory
network dynamics. Nature. 402: 664–668.
Stopfer M, Bhagavan S, Smith BH, et al. (1997) Impaired odour
discrimination on desynchronization of odour-encoding neural
assemblies. Nature. 390: 70–74.
Stopfer M, Jayaraman V, and Laurent G (2003) Intensity
versus identity coding in an olfactory system. Neuron. 39:
991–1004.
Szyszka P, Ditzen M, Galkin A, et al. (2005) Sparsening and
temporal sharpening of olfactory representations in the
honeybee mushroom bodies. Journal of Neurophysiology. 94:
3303–3313.
Encyclopedia of Neuroscience (2009), vol. 7, pp. 43-48