Mechanisms of scent-tracking in humans

Abstract

Whether mammalian scent-tracking is aided by inter-nostril comparisons is unknown. We assessed this in humans and found that (i) humans can scent-track, (ii) they improve with practice, (iii) the human nostrils sample spatially distinct regions separated by approximately 3.5 cm and, critically, (iv) scent-tracking is aided by inter-nostril comparisons. These findings reveal fundamental mechanisms of scent-tracking and suggest that the poor reputation of human olfaction may reflect, in part, behavioral demands rather than ultimate abilities.

Full text

Mechanisms of scent-tracking
in humans
Jess Porter
1
, Brent Craven
2
, Rehan M Khan
3,4
, Shao-Ju Chang
5
,
Irene Kang
5
, Benjamin Judkewitz
3
, Jason Volpe
2
, Gary Settles
2
&
Noam Sobel
1,3–6
Whether mammalian scent-tracking is aided by inter-nostril
comparisons is unknown. We assessed this in humans and
found that (i) humans can scent-track, (ii) they improve with
practice, (iii) the human nostrils sample spatially distinct
regions separated by ~3.5 cm and, critically, (iv) scent-tracking
is aided by inter-nostril comparisons. These findings reveal
fundamental mechanisms of scent-tracking and suggest that
the poor reputation of human olfaction may reflect, in part,
behavioral demands rather than ultimate abilities.
Two major roles of olfaction are identifying odorants and spatially
localizing their sources. Whereas odor plume navigation in air (for
example, in moths
1
) and water (for example, in lobsters
2
)hasreceived
some attention, the mechanisms of scent-trail tracking, a critical ability
for macrosmatic mammals ranging from rats to dogs
3
(see Fig. 1a),
remain unknown, and key questions, such as whether mammals use
inter-nostril comparisons to aid scent-tracking, remain unanswered.
Humans are an appealing animal model for addressing such ques-
tions because they can follow task instructions and accurately report
behavioral strategies. Humans also tolerate manipulations, such as
nostril occlusion, that may aggravate even well-trained dogs. However,
whether humans are a valid model for this task is unknown. Therefore,
we first set out to ask whether humans can scent-track.
In Experiment 1 we asked whether 32 naive human subjects were
capable of using only their noses (all other sensory input being
blocked) to follow a B10-m-long scent trail in an open grass field
(Supplementary Methods online). All subjects gave informed consent
to procedures approved by the University of California Berkeley
Committee for the Protection of Human Subjects. Two-thirds of the
subjects were capable of following the scent trail (21 of 32 subjects,
9 women, 12 men). Figure 1b shows a time-lapsed image of one trial,
and Supplementary Video 1 online contains a movie of one trial. To
ask whether subjects were aided by any unintended non-olfactory cues,
we repeated the task with nostril occlusion. None of the subjects were
able to follow the scent trail under these conditions, assuring the
olfactory nature of the task.
We next asked whether subjects could improve with practice. In
Experiment 2, four subjects (two men, two women) trained on this
same task, three times a day, for 3 d within a 2-week period. With
training, subjects decreased their deviation from the scent track (decay-
ing exponential fit: R
2
¼ 0.2862, F
1,28
¼ 10.82, asymptote 0.1 m,
P ¼ 0.0028, Fig. 2a) and increased their velocity (0.026 ms
–1
±0.003on
day 1, 0.057 ms
–1
± 0.01 on the last day; local linear fit, R
2
¼ 0.94,
P ¼ 0.0006, Fig. 2b). Considering that tracking velocity more than
doubled within a few days, we suggest that longer-term training would
lead to further increases in tracking velocity. The plateau we observed in
lateral deviation reflects the zigzagging nature of the tracking path
(Fig. 1b and S upplementary Fig. 1 online), a characteristic also
observed in macrosmatic animals during scent-tracking
3
.
An important factor in mammalian olfactory behavior is active
sniffing
4,5
. We therefore next asked whether sniffing behavior was
related to humans’ ability to follow a scent trail. We calculated mean
sniffing frequencies for each trial (Supplementary Methods). Whereas
performance on the initial day of testing (mean velocity or deviation)
did not correlate with sniffing frequency (R
2
¼ 0.0135, t
1,9
¼ 0.1094,
P ¼ 0.75), sniffing frequency increased with tracking velocity over the
three subsequent days of training (frequency versus day: R
2
¼ 0.2932,
t
1,28
¼ 11.20, P ¼ 0.0024; velocity versus day: R
2
¼ 0.3608,
F
1,28
¼ 15.23, P ¼ 0.0006; Fig. 2c). We interpret these results to suggest
that, as subjects increased their speed, it was necessary for them to sniff
more quickly to get the same quality of information. One notable
difference between these results and those from dogs is that dogs
© Louie Psihoyos/Science Faction Images
Figure 1 Human subject’s path following a scent trail, as compared to a dog’s
path. (a) Path of a dog following the scent trail of a pheasant dragged through
a field (scent trail in yellow, dog’s path in red; from ref. 15). (b)Pathofa
human following a scent trail of chocolate essential oil through a field (scent
trail in yellow, human’s path in red).
Received 7 September 2006; accepted 22 November 2006; published online 17 December 2006; corrected online 4 January 2007; doi:10.1038/nn1819
1
299 Life Science Addition, MC 3200, Program in Biophysics, University of California Berkeley, Berkeley, California 94720, USA.
2
Department of Mechanical and Nuclear
Engineering, Penn State University, University Park, Pennsylvania 16802, USA.
3
Helen Wills Neuroscience Institute,
4
Department of Psychology and
5
Program in
Bioengineering, University of California Berkeley, Berkeley, California 94720, USA.
6
Department of Neurobiology, The Weizmann Institute of Science, Rehovot, Israel.
Correspondence should be addressed to J.P. (jessiep@berkeley.edu) or N.S. (noam.sobel@weizmann.ac.il).
NATURE NEUROSCIENCE VOLUME 10
[
NUMBER 1
[
JANUARY 2007 27
BRIEF COMMUNICATIONS
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience

sniff much faster (B6Hz)
3
, which may partially account for their
greater scent-tracking proficiency.
Mammals localize auditory sources by comparing simultaneous
inputs across two ears, converting differences in sound timing and
intensity to spatial coordinates. It has been asserted that mammals
cannot similarly exploit their two-nostril geometry to localize and track
scent trails, because the nostrils are too closely spaced to provide
spatially distinct information
6
.
In Experiment 3, we tested this assertion using particle image
velocimetry (PIV) to measure the velocity of neutrally buoyant
particles in a coronal plane intersecting the human nose during sniffing
(Fig. 3a; Supplementary Methods). Figures 3b and c shows sample
PIV images of the nasal inspiratory airstreams and Figure 3d a contour
plot of the magnitude and direction of inspired air. In contrast to the
common notion, each nostril clearly inspired air from distinct, non-
overlapping regions in space. In addition, the natural asymmetry in
airflow across nostrils shaped this reach pattern. A maximum velocity
of 0.45 ms
–1
at the right nostril and 0.30 ms
–1
at the left (see Fig. 3e)led
to a right nostril reach of B1.5–2.0 cm to the right and a left nostril
reach of B1.0–1.5 cm to the left. In other words, the two nostrils
sampled information from centroids laterally separated by B3.5 cm.
Considering that the boundary of a scent plume can be B10 mm
7
,this
result demonstrated that one nostril can be within a plume while the
other is out of the plume. Having found that the nostrils provide
spatially distinct information that could in principle
8
be exploited to
scent-track, we next asked whether this information is exploited.
In Experiment 4, 14 subjects performed the scent-tracking task, once
with one nostril taped closed, and once with both nostrils open (order
counterbalanced). Compared to dual-nostril tracking, single-nostril
tracking was less accurate (36% versus 66% accuracy, binomial
P o 0.003) and slower (26% reduction in speed, binomial P o 0.02).
To control for some limitations in the interpretation of the critical
Experiment 4 (see Supplementary Methods),inExperiment5we
re-tested the four trained subjects from Experiment 2 using a nasal
‘prism’ device that maintained input into two nostrils, while within the
prism both flow paths were conjoined to form a single virtual nostril
located in the middle of the nose (Fig. 3f). Now, the external environ-
ment was sampled by two nostrils, but without spatial separation. To
control for the effect of the prism device, we used a control prism with
two straight flow paths that maintained natural spatial separation
(schematic in Fig. 3f). After getting used to the prisms, each subject
performed three tracking runs with the control prisms and three runs
with the non-spatial prisms on two more days in counterbalanced order.
Subjects were significantly less accurate (9/12 successful with control,
5/12 with non-spatial prism; binomial, P o 0.015) (Fig. 3g)and
0.08
0.06
0.04
0.02
0
012345
Day of training
y = 0.0107x + 0.0123
Mean velocity (m s
–1
)
012345
Day of training
Mean velocit
y
(
m s
–1
)
Mean deviation (m)
Mean sniff frequency (Hz)
y = 0.01 + 0.2722e
–0.6685x
0.4
0.3
0.2
0.1
0
0.8
0.6
0.4
0.2
0
0 0.02 0.04 0.06 0.0
8
Day 1
Day 2
Day 3
Day 4
abc
Figure 2 Training increased tracking velocity, decreased deviation from track, and increased sniffing frequency. (a) The mean deviation from the scent trail is
plotted for all subjects for each day of training. Dashed line, decaying exponential fit; solid gray line, asymptote. (b) The mean tracking velocity is plotted for each
day of training. Dashed line, linear fit. (c) The mean tracking velocity is plotted against the mean sniffing frequency for each day of training. Error bars, s.e.m.
–5
0
5
10
15
y (mm)
–20 –10 0 10 20
x (mm)
V
mag
(ms
–1
)
0.450
0.350
0.250
0.150
0.050
V
mag
(ms
–1
)
0.4
0.3
0.2
0.1
024681012
Distance (mm)
Control
2 to 1
1.2
1
0.8
0.6
0.4
0.2
0
Normalized velocity (ms
–1
)
12
10
8
6
4
2
0
No. of runs completed
Control
2 to 1
*
*
Light sheet
Right naris
Left naris
a
f
gh
bc
de
Figure 3 The two-nostril advantage in sampling and tracking. (a)ThePIV
laser light sheet was oriented in a coronal plane intersecting the nostrils at
their midpoint. (b,c) PIV images of particle-laden inspired air stream for two
example sniffs. (d) A contour plot of velocity magnitude of the inspired air
stream into the nose of the subject sniffing at 0.2 Hz. (e) Velocity profiles of
the right and left naris; abcissa indicates distance from the tip of the nose
to the lateral extent of the naris. (f) Schematic diagrams of control prism
(left) and non-spatial prism (right). Arrows show the direction of sniff airflow.
Center, prism worn by subject. Inlet ports are located on bottom surface of
prism; screws are located on front surface. (g) Subjects completed fewer
trials using non-spatial prisms as compared to control prisms. (h) Subjects
were significantly slower tracking with non-spatial prisms than with control
prisms. Error bars, s.e.m.
28 VOLUME 10
[
NUMBER 1
[
JANUARY 2007 NATURE NEUROSCIENCE
BRIEF COMMUNICATIONS
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience

significantly slower (24% reduction in speed, t
1,4
¼ 4.9967, P ¼ 0.0075,
Fig. 3h) using the non-spatial prism than with the control prism. In
Experiments 4 and 5 taken together, 18 subjects who completed a total
of 52 tracks were both faster and more accurate when they were able to
make comparisons across spatially offset nostrils.
Poor olfactory abilities in humans have been attributed to the
reduction in olfactory receptor repertoire apparent in primate
evolution
9
. However, demonstrations of keen primate olfaction
10
have challenged the causal relationship between receptor repertoire
size and olfactory abilities
11
and the classical definitions of microsmat
and macrosmat
12
. Here we found that not only are humans capable of
the demanding macrosmatic behavior of scent-tracking, but they
spontaneously mimic the tracking patterns of macrosmatic mammals.
Using this model enabled us to address the key question of whether
mammals use inter-nostril comparisons to aid scent-tracking.
Our results suggest that, although comparison of sequential samples
alone can subserve tracking, there was an added benefit to simultaneous
sampling at the spatially offset locations provided by the two nostrils.
Neural and behavioral mechanisms that may subserve this behavior have
been revealed in recent studies indicating that immobile rats
13
and
humans
14
require bilateral input to localize odor sources within a single
sniff. However, these past results—obtained in highly artificial settings
with immobilized animals—did not show whether such differences were
relevant to natural spatial behavior. Here we find that mammals
performing a scent-tracking task, freely able to move their nose and
sample the olfactory environment in real time, reap added benefit from
sampling via their two spatially offset nostrils.
Note: Supplementary information is available on the Nature Neuroscience website.
ACKNOWLEDGMENTS
Studies were funded by Army Research Office grant #46666-LS and
by US National Institutes of Health, National Institute on Deafness
and other Communication Disorders grants DC006915 and DC005958.
The authors wish to thank B. Willmore for help with the video tracking
algorithm and K. Scott for comments on the manuscript, as well as
Arak Elite.
COMPETING INTERESTS STATEMENT
The authors declare that they have no competing financial interests.
Published online at http://www.nature. com/natureneuroscience
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Vickers, N.J. Chem. Senses 31, 155–166 (2006).
2. Koehl, M.A. et al. Science 294, 1948–1951 (2001).
3. Thesen, A., Steen, J.B. & Doving, K.B. J. Exp. Biol. 180, 247–251 (1993).
4. Settles, G.S. J. Fluids Eng. 127, 189–218 (2005).
5. Mainland, J. & Sobel, N. Chem. Senses 31, 181–196 (2006).
6. Moulton, D.G. Am. Zool. 7, 421–428 (1967).
7. Crimaldi, J.P. & Koseff, J.R. Exp. Fluids 31, 90–102 (2001).
8. Wilson, D.A. & Sullivan, R.M. Physiol. Behav. 66, 41–44 (1999).
9. Gilad, Y., Wiebel, V., Przeworski, M., Lancet, D. & Paabo, S. PLoS Biol. 2, 120–125
(2004).
10. Laska, M., Seibt, A. & Weber, A. Chem. Senses 25, 47–53 (2000).
11. Shepherd, G.M. PLoS Biol. 2, 572–575 (2004).
12. Smith, T.D. & Bhatnagar, K.P. Anat. Rec. B New Anat. 279, 24–31 (2004).
13. Rajan, R., Clement, J.P. & Bhalla, U.S. Science 311, 666–670 (2006).
14. Porter, J., Anand, T., Johnson, B., Khan, R.M. & Sobel, N. Neuron 47, 581–592
(2005).
15. Gibbons, B. Nat. Geogr. Mag. 170, 324–361 (1986).
NATURE NEUROSCIENCE VOLUME 10
[
NUMBER 1
[
JANUARY 2007 29
BRIEF COMMUNICATIONS
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience

CORRIGENDA
NATURE NEUROSCIENCE VOLUME 10
|
NUMBER 2
|
FEBRUARY 2007 1
Corrigendum: Mechanisms of scent-tracking in humans
Jess Porter, Brent Craven, Rehan M Khan, Shao-Ju Chang, Irene Kang, Benjamin Judkewicz, Jason Volpe, Gary Settles & Noam Sobel
Nature Neuroscience 10, 27–29 (2007); corrected after print 4 January 2007
In the version of this article initially published, the name of the sixth author was misspelled. It should be Benjamin Judkewitz.
This error has been corrected in the PDF and HTML versions online.
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience