See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/9035679

Understanding the mind's eye...and nose

Article in Nature Neuroscience · December 2003

DOI: 10.1038/nn1103-1124 · Source: PubMed

CITATIONS READS
24 468

1 author:

Stephen M Kosslyn
Harvard University
266 PUBLICATIONS 37,706 CITATIONS

SEE PROFILE

All content following this page was uploaded by Stephen M Kosslyn on 20 May 2014.

The user has requested enhancement of the downloaded file.

 NEWS AND VIEWS

Understanding the mind’s eye…and nose

Stephen M Kosslyn

New findings reveal that people sniff when imagining odors and that sniffing can modify the pleasantness or unpleasantness of
the mental image. The work parallels findings in other systems and suggests common neural principles for mental imagery.
© 2003 Nature Publishing Group http://www.nature.com/natureneuroscience

Mental imagery is useful when recalling of the mental images and made the imagined time we move our eyes, the input replaces
information that may have been stored only quality of pleasant odors less distinct from what we previously saw, requiring our brains
incidentally. For example, try to imagine unpleasant odors. Therefore, sniffing was to store a series of images. But it is even more
the shape of a cat’s ears. You have probably not simply a byproduct of the mental image, complicated than that. In addition, we must
never thought about it before, and may but rather modulated aspects of the image store information about where we are look-
have simply registered their shape in pass- itself. The findings parallel the work on ing at each fixation point, so that we later
ing. Nevertheless, it was stored as a percep- visual mental imagery and provide new know how the separate ‘snapshots’ relate to
tual memory. When recalling the shape insight into the processes that give rise to all each other. In short, visual perception occurs
later, most people will visualize the animal’s forms of mental imagery. over time and requires us to store informa-
head and then ‘look’ at the ears1. This To understand how sniffing might func- tion not only about what we see, but also
process of creating a mental image involves tion in olfactory imagery, it is perhaps worth- where we see it.
components of the visual perceptual appa- while to examine what is known about visual Thus, because perceived events are stored
ratus itself. In fact, when visualizing an imagery. One possible reason we move our over time as a series of separate internal rep-
object, people move their eyes in very much eyes during visual mental imagery is rooted resentations, we need to retrieve a sequence
the same way as when actually seeing the in the anatomy of the eye. The high-resolu- of representations when later forming men-
object—and if these eye movements are tion central portion of the retina (the fovea) tal images. Experimental evidence supports
prevented, creating the mental image is subtends only about two degrees of visual this interference: for example, the more
more difficult2. angle, or about the same space as your thumb parts an object has, the more time people
In this issue, Bensafi et al.3 show that when held out at arm’s length. Thus, to view require to visualize it2. Eye movements may
mental imagining of odors works in a simi- all of the parts and characteristics of an help us recall the sequence of parts and
lar fashion. This may seem odd at first object with high resolution, we typically need characteristics by indexing their relative
glance. Vision works when radiation of spe- to make multiple fixations. However, every locations. If so, then eye movements func-
cific ranges of wavelengths stimulate recep-
tors, whereas olfaction works by fitting
molecular ‘keys’ into receptor ‘locks’. We
move our eyes rapidly, but the nose is fixed
to a relatively slow-moving head. Vision
must be able to localize a stimulus source
precisely in space, but olfaction need not
specify as many dimensions as vision.
Although the findings about mental
imagery in the visual system may seem
unlikely to bear on the nature of olfactory
imagery, the current study suggests that
mental images of scents and sights have
much in common.
Bensafi et al.3 measured nasal airflow in
people while they imagined an odor and
found that people sniff when performing
this task (Fig. 1). Furthermore, they sniff
more deeply when imagining a pleasant odor
(e.g., chocolate) than an unpleasant one
(urine). Blocking participants’ noses to pre-
vent sniffing reduced the emotional impact

Stephen Kosslyn is in the Department of

Psychology, 832 William James Hall, Harvard
University, 33 Kirkland Street, Cambridge Figure 1 Bensafi et al.3 measured inhalation volumes (‘sniffs’) and found that people sniffed when
Massachusetts 02138, USA. asked to imagine the smell of an odor. Furthermore, sniffing could modulate the quality (pleasantness
e-mail: smk@wjh.harvard.edu or unpleasantness) of the odor image. Image courtesy of Noam Sobel.

1124 VOLUME 6 | NUMBER 11 | NOVEMBER 2003 NATURE NEUROSCIENCE

 NEWS AND VIEWS

tion as part of a feedback loop, providing
cues that trigger subsequent memories.
Sniffing during olfaction is also a sequential
activity, and it may have exactly the same
function as eye movements during visual
imagery—the activity itself may provide
cues that trigger specific memories.
This line of reasoning may shed light on a
fundamental question about imagery: why
does the brain create mental images at all?
© 2003 Nature Publishing Group http://www.nature.com/natureneuroscience
that occur in olfactory imagery trigger an
internal representation that is then used to
create the image. This image makes explicit
and accessible the information that is only
implicit and inaccessible in long-term mem-
ory. For example, imagine the smell of hot
coffee or tea at the instant when you lift the
mug to your lips and are just about to take a
sip. Did you sniff? The perceptual informa-
tion probably was stored in long-term mem-
receive input from the eyes during perception
(Brodmann areas 17 and 18) also are acti-
vated—even with eyes closed—when one
visualizes objects with high resolution7. If
these high-resolution details are not included
in the image, only ‘higher’ visual cortex is acti-
vated. It would be fascinating to discover
whether the same distinction exists in other
sensory modalities. If true, this would suggest
a general principle: when fine details of an

Mental images are present for only a matter
of seconds and must be created on the basis
of information stored in long-term memory.
Why should the brain go to the trouble of
forming a mental image if the necessary
information is already stored in long-term
memory? Why not just access the stored
information directly? One answer is that the
long-term memory representations are not
simply a copy of the representations that
underlie short-term images, just as the code
on a DVD is not simply a copy of the picture
created on our computer or television screen.
Representations differ in what information
they make explicit and accessible4. The sniffs
ory in a concise format, as appears to be the
case in visual memory5, and your sniff may
have triggered a process that constructed an
image on the basis of this stored information.
You are then able to savor the scent once
again. The fact that odors can be mentally
mixed via olfactory imagery6 is consistent
with this idea of forming ‘odor images’ by
making information that is only implicit in
long-term memory explicit and accessible.
Indeed, Bensafi et al.3 suggest that sniffs may
drive a pattern of neural activity in the cortex
that underlies the odor image itself.
Neuroimaging studies of visual imagery
have found that the first cortical areas to
object or event must be included in a mental
image in order to perform a task, the brain can
recruit the ‘lowest’ levels of neural process-
ing—in both perceptual and motor systems.
1. Kosslyn, S. M. Image and Mind (Harvard Univ. Press,
Cambridge, Massachusetts, 1980).
2. Laeng, B. & Teodorescu, D.-S. Cognit. Sci. 26,
207–231 (2002).
3. Bensafi, M. et al. Nat. Neurosci. 6, 1142–1144
(2003).
4. Marr, D. Vision (MIT Press, Cambridge,
Massachusetts, 1982).
5. Tanaka, K. Annu. Rev. Neurosci. 19, 109–139
(1996).
6. Elmes, D.G. Chem. Senses 23, 443–445 (1998).
7. Kosslyn, S. M. & Thompson, W. L. Psychol. Bull.
129, 723–746 (2003).

Pulling the Soxs off proliferation

A paper in this issue discloses the secret of eternal
youth—at least for neural progenitors. During neuro-
genesis, self-renewing progenitor cells that reside in
the ventricular zone of the developing CNS exit the
cell cycle, migrate toward the marginal zone, and dif-
ferentiate into neurons. Previous studies have identi-
fied transcription factors that promote the formation
of new neurons by directing stem cells to stop divid-
ing. However, the mechanisms responsible for coun-
tering differentiation to maintain a pool of
undifferentiated neural progenitor cells have been less
clear.
On pages 1162–1168, Jonas Muhr and colleagues
report that the transcription factors Sox1, Sox2 and
Sox3 are important for maintaining the progenitor cell
pool in the chick spinal cord. They find that Sox-3
maintains neural progenitor cells in an undifferenti-
ated state and suppresses neuronal differentiation.
Sox1–3 proteins countered neurogenesis by repressing
downstream differentiation signals from bHLH proteins, a family of proneural transcription factors. Moreover, when Sox1-3 target
genes were actively repressed, independently of proneural activity, progenitor cells differentiated prematurely. The figure shows a
coronal section of an embryonic chick spinal cord stained with Sox3 (green), NeuroM (red) and Tuj1 (blue). NeuroM is a bHLH
protein, expressed during neurogenesis, whereas Tuj1 is a neuronal marker.
The Sox family of transcription factors are expressed by most stem cells in the developing CNS and are typically downregulated
during differentiation. Sox genes are important in the generation of neuroblasts in Drosophila, and in maintaining precursor cells of
the mouse blastocyst in a multipotent state. A Sox gene family member, Sox 10, expressed in migrating neural crest cells, maintains
the multipotency of neural crest cells in culture. Moreover, Sox proteins are expressed in self-renewing stem cells in the adult CNS.
Together with the work of Muhr and colleagues, these findings suggest that several Sox genes may keep cells from different origins in
an undifferentiated state.

Kalyani Narasimhan

NATURE NEUROSCIENCE VOLUME 6 | NUMBER 11 | NOVEMBER 2003 1125

View publication stats