INS IGHTS | P E R S P E C T I V E S

NEUROBIOLOGY

Cross-modal sensory transfer: Bumble bees do it

Stored sensory input permits two sensory channels to exchange and compare information

By Gerhard von der Emde1 and content. This means that both senses provide or only vision (in light, but where they could
Theresa Burt de Perera2 information about the same characteristic of not touch the objects). After training, the
an object, such as its shape or surface struc- bees discriminated between the same objects

A
nimal sensory systems acquire in-
formation about the physical world
by transforming external stimuli
into signals that can be read and
interpreted by the nervous system.
Animals possess several senses that
provide separate streams of information
based on different physical stimuli. How-
ever, objects and environments contain
ture. In addition, the sensory inputs must
be encoded in a way that allows temporally
disjointed information from two senses to
be identified as identical, even though these
senses rely on different physical stimuli.
Thus, characteristic object features are stored
in a neuronal representation that can be ac-
cessed by multiple senses. Up to now, spon-
taneous cross-modal object recognition has
using only the information from the other
sensory modality. This observation suggests
that the brain encoded the sensory input in
a way that allowed the two sensory channels
to exchange information and to compare and
match object-related inputs.
The mechanisms by which bees accomplish
this task have yet to be elucidated. Perhaps
the bee’s brain stores a representation of the

Downloaded from http://science.sciencemag.org/ on February 21, 2020

inherently multimodal information. Thus,
neurobiologists have sought to define the
mechanisms by which information is pro-
cessed when it is received by different senses
(1). Cross-modal recognition—the ability to
transfer information across senses, irre-
spective of the sense that first accessed that
information—is a highly complex
cognitive capacity that was thought
to be limited to vertebrates. Now, on
page 910 of this issue, Solvi et al. (2)
show that bumble bees are capable
of performing the same task.
Humans execute spontaneous
cross-modal recognition fairly eas-
ily (3). For example, people often
rummage blindly through pockets
to find, by touch, a set of keys that
they first encountered visually. Cross-
modal recognition is also useful for
increasing the flexibility of object-
recognition systems. The finding by
Solvi et al. illustrates that tiny inver-
tebrates, with brain structures that
differ greatly from those of verte-
brates, also can experience an object
with one sensory modality and later
recognize that same object with a dif-
ferent sensory modality.
Simple forms of cross-modal infor-
mation transfer, which are based on
direct associations between two spe-
cific stimuli, have been observed in
various animal groups, including in-
sects (4, 5). However, cross-modal ob-
ject recognition requires additional
and more complex conditions—in
been described only in mammalian species
(humans, apes, monkeys, dolphins, and rats)
(6–8) and in one aquatic creature, the weakly
electric elephantnose fish (9).
Solvi et al. trained bumble bees to discrim-
inate two differently shaped objects (cubes
and spheres) using only touch (in darkness)
General scheme of cross-modal transfer
In both vertebrates and invertebrates, sense organs respond to physical
stimuli and transfer information to specific monomodal sensory nuclei
in the brain (dark yellow, vision; light yellow, touch). These nuclei then
communicate, in a reciprocal manner, with multimodal sensory nuclei
in higher brain centers. In principle, cross-modal sensory transfer can
occur at any stage after the first monomodal sensory nuclei.
Higher centers
Multimodal
Secondary
sensory nuclei
Multimodal
Primary
sensory nuclei
Monomodal
object in a way that allows it to be accessed
and understood across different sensory sys-
tems. The sensory information would need to
be integrated to form a representation that is
independent of the sensory modality through
which the information was introduced (that
is, both senses use a matched encoding for-
mat). This would enable bees to rec-
ognize objects cross-modally without
any previous experience, and cross-
modal object recognition would not
require training. Alternatively, infor-
mation that originates from multiple
senses might initially be unmatched
in format, and cross-modal object
recognition could depend on sensory
experience and learning. In this case,
bees might have learned to associate
visual and tactile inputs of basic fea-
tures common in other environmen-
tal objects when exposed to these
features in the past. For example, a
bee might have learned to associate a
visual and a tactile image of a curved
edge or a corner. Subsequently, these
associations would generalize to new
objects and new situations.
Whatever the mechanism, cross-
modal object recognition requires
storage (in the brain) of a representa-
tion of the object’s features that can
be accessed by different senses. This
implies that an object can be recog-
nized with a sense through which it
had never before been experienced.
Solvi et al. show that this capability
exists in an insect brain, which con-
GRAPHIC: V. ALTOUNIAN/SCIENCE

particular, that the two senses pro-
vide information that is matched in
(e.g., vision)
1
Institute of Zoology, University of Bonn,
Bonn, Germany. 2Department of Zoology,
University of Oxford, Oxford, UK.
Email: vonderemde@uni-bonn.de;
theresa.burt@sjc.ox.ac.uk
tains a small fraction of the numbers
Sense 1 Sense 2
of neurons in vertebrate brains (10).
(e.g., touch) How the organization of sensory
systems achieves cross-modal object
recognition remains unanswered.
Sense organs respond to the physi-
cal stimuli of the outside world and

850 21 FEBRUARY 2020 • VOL 367 ISSUE 6480 sciencemag.org SCIENCE

Published by AAAS
 transduce them into action potentials en-
coding the specific stimulus parameters.
Connections are then made with nerve cells
in dedicated sensory nuclei in the brain,
which in turn project to sensory nuclei in
higher brain centers for further processing.
These nuclei also project back to lower pro-
cessing centers, forming reciprocal circuits
and adding to the complexity (see the figure).
For multisensory integration and cross-
modal transfer to occur, information from
the different sensory modalities that encode
characteristics of the same object must come
together, eventually forming a multisensory
representation of the object. In most cases,
scientists still do not know exactly where in
the sensory pathway this occurs, particularly
in bees and fish, which do not possess spe-
cialized cortical structures. One possibility is
AGING
that higher brain structures, such as the cor-

Downloaded from http://science.sciencemag.org/ on February 21, 2020

tex in mammals or the mushroom bodies and
central complex in bees, form the anatomical
substrate for multisensory representations A time to grow
and a time to pause
(6, 11). Alternatively, because cross-talk be-
tween the senses also occurs at lower levels
(e.g., in the midbrain and diencephalon) (1,
12, 13) and because sensory nuclei on differ- Mechanisms of programmed arrest
ent levels are connected in a reciprocal man-
ner, a distributed representation of object protect animals from the passage of time
properties might occur in several intercon-
nected multisensory nuclei.
The fact that bees can achieve cross-modal
object recognition might have implications By Marc Van Gilst An African turquoise killifish embryo developmentally
for how we think about cognition in general. suspended in the state of diapause. This vertebrate

In humans, scientists assume that this ability
involves mental imagery (14) based on inter-
nal representations in higher brain centers.
Thus, some researchers have argued that
this task relies on awareness. Whether that
is also the case in bumble bees is a matter
of debate, as simpler explanations are pos-
sible. Whatever the underlying mechanism,
the newly found ability of bumble bees to
perform cross-modal recognition shows that,
like humans, they possess a sensory integra-
tion system that allows them to form a com-
plex representation of their world. j
REFERENCES AND NOTES
1. B. E. Stein et al., J. Neurosci. 40, 3 (2020).
2. C. Solvi, S. Gutierrez Al-Khudhairy, L. Chittka, Science
367, 910 (2020).
3. M. O. Ernst, M. S. Banks, Nature 415, 429 (2002).
4. J. Guo, A. Guo, Science 309, 307 (2005).
5. L. Proops et al., Proc. Natl. Acad. Sci. U.S.A. 106, 947
(2009).
6. B. D. Winters, J. M. Reid, J. Neurosci. 30, 6253 (2010).
7. L. M. Herman et al., J. Comp. Psychol. 112, 292 (1998).
8. A. Cowey, L. Weiskrantz, Neuropsychologia 13, 117
(1975).
9. S. Schumacher et al., Proc. Natl. Acad. Sci. U.S.A. 113,
7638 (2016).
10. M. Giurfa, Trends Neurosci. 36, 285 (2013).
PHOTO: HU ET AL. (2)
S
easonal or environmental pressures
are sometimes best dealt with by
putting growth and reproduction on
hold. Many animals have evolved
mechanisms for reversibly arresting
development at discrete developmen-
tal stages, so that the arrested embryo or
larva can wait for more favorable conditions
in which to resume development, grow, and
reproduce. One common form of develop-
mental arrest is diapause, which generally
constitutes a genetically programmed ar-
rest at a discrete point in development,
most often in an embryonic, larval, or pupal
stage (1). Diapause also involves physiologi-
cal changes that protect the arrested animal
from aging. On page 870 of this issue, Hu et
al. (2) identify some of the mechanisms that
maintain embryonic diapause in an emerg-
ing model of vertebrate aging, the African
turquoise killifish. These findings might
provide insight into the mysteries of aging
and longevity.
Examples of diapause are found through-
out the animal kingdom and have been well
characterized in insects, fish, and mam-
model could improve the understanding of aging.
facultative diapause, which is induced in
response to environmental challenges such
as starvation or dehydration, or an obligate
diapause, which is generally brought about
by seasonal pressures such as a harsh win-
ter (3). Although diapause involves an ar-
rest of development, the arrested state is
not static. Indeed, diapause usually involves
considerable physiological changes that
help to prepare the animal to survive the
forthcoming environmental challenges (4).
These physiological changes also protect
arrested animals, allowing them to be pre-
served for long periods of time until condi-
tions are more favorable for development
and reproduction.
In the simplest sense, aging is considered
the inevitable wear and tear brought on by
the passage of time. The basic idea is that
the more time passes, the more an animal
ages and the more it progresses toward its
ultimate demise. This simplistic perspective
is somewhat fatalistic and defines time as
the ultimate enemy of youth. However, it has

11. F. Peng, L. Chittka, Curr. Biol. 27, 224 (2017).

12. M. Zeymer et al., Front. Neuroanat. 12, 79 (2018).
13. H. F. Sperdin et al., Front. Neurosci. 4, 9 (2010).
14. B. Nanay, Cortex 105, 125 (2018).
10.1126/science.aba8519
mals (1). Diapause may take the form of a been established in many animals that ag-
ing is also heavily influenced by genetic and
Department of Anesthesiology and Pain Medicine, University physiological programs, such that aging may
of Washington, Seattle, WA, USA. Email: vangilst@uw.edu not necessarily be an inevitable consequence

SCIENCE sciencemag.org 21 FEBRUARY 2020 • VOL 367 ISSUE 6480 851

Published by AAAS
Cross-modal sensory transfer: Bumble bees do it
Gerhard von der Emde and Theresa Burt de Perera

Science 367 (6480), 850-851.

DOI: 10.1126/science.aba8519

Downloaded from http://science.sciencemag.org/ on February 21, 2020

ARTICLE TOOLS http://science.sciencemag.org/content/367/6480/850

RELATED http://science.sciencemag.org/content/sci/367/6480/910.full
CONTENT

REFERENCES This article cites 14 articles, 6 of which you can access for free
http://science.sciencemag.org/content/367/6480/850#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement of
Science, 1200 New York Avenue NW, Washington, DC 20005. The title Science is a registered trademark of AAAS.
Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of
Science. No claim to original U.S. Government Works