This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Animal Behaviour 79 (2010) 1101e1107
Contents lists available at ScienceDirect
Animal Behaviour
journal homepage: www.elsevier.com/locate/anbehav
Risk-sensitive information gathering by cyprinids following release of chemical
alarm cues
Brian D. Wisenden a, *, Char L. Binstock a, Kristine E. Knoll a, Adam J. Linke a, Brandon S. Demuth b
a
b
Biosciences Department, Minnesota State University Moorhead
Department of Ecology, Evolution, and Behavior, University of Minnesota
a r t i c l e i n f o
Article history:
Received 27 March 2009
Initial acceptance 27 July 2009
Final acceptance 26 January 2010
Available online 11 March 2010
MS. number: A09-00197R
Keywords:
alarm cue
antipredator behaviour
chemical cue
field study
information gathering
risk-sensitive behaviour
zebrafish
In aquatic environments, chemical cues released during a predator attack reliably inform prey about the
presence of predation risk. Prey with information about predation risk are more successful in surviving
encounters with predators than are unwary prey. To remain prepared for attack, prey should continue to
monitor the status of predation risk, presenting a behavioural trade-off for prey: increased distance from
areas labelled with alarm cues reduces exposure to predation risk but also reduces access to information
about predation risk. In two laboratory experiments we used the presence and absence of water flow in
a laboratory fluvarium to test alarm response and subsequent risk-sensitive information gathering by
zebrafish (Danio rerio). In response to chemical alarm cues, fish significantly reduced activity and
increased use of shelters. In the absence of flow, fish sought out the shelter nearest the cue source. In the
presence of flow, fish preferred to seek shelter downstream, but not upstream, of the cue source. This
allowed fish to gather information about predation risk from a relatively safe distance. In a field
experiment on natural populations of stream fishes, fish avoided areas where chemical alarm cues were
released (versus blank water control) but primarily because they avoided the region immediately
upstream of the cue source. Fish use of the area immediately downstream of cue release did not decrease.
Taken together, these laboratory and field data are consistent with a trade-off between risk avoidance
and information gathering.
Ó 2010 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Much of behavioural decision making is guided by public
information released as a natural consequence of ecological interactions. Public information is valuable to receivers in several
important contexts ranging from habitat selection to cultural
evolution (Danchin et al. 2004). In aquatic environments, public
information about predation risk takes the form of various chemical cues released during successive stages of the predation
sequence (sensu Lima & Dill 1990; Smith 1992; Wisenden & Chivers
2006).
Animals place a high priority on gathering information about
predation risk. Predator inspection behaviour, where prey approach
a predator directly, has stimulated a large literature (e.g. Dugatkin &
Godin 1992). However, little work has been done on the inspection
of indirect indicators of predation risk, such as sources of chemical
cues. Information is valuable only to the degree to which it is
accurate, and, because of the temporally dynamic nature of
predation risk, accuracy requires frequent updating. Gathering
* Correspondence: B. D. Wisenden, Biosciences Department, Minnesota State
University Moorhead, 1104 7th Avenue S, Moorhead, MN 56563, U.S.A.
E-mail address: [email protected] (B.D. Wisenden).
information about predation risk presents a trade-off in that the
most accurate information is obtained where risk is greatest. Here,
we test for a trade-off between information gathering and risk
avoidance using zebrafish (Danio rerio) in 1.8 m long fluvaria in
which water flow could be turned on or off. Chemical alarm cues
derived from conspecific skin extract either diffused slowly from
the point of release (no flow) or was carried the length of the fluvarium by water current (flow). When water flow was turned off,
chemical alarm cues were detectable only at the shelter nearest the
site of cue release. If the function of an alarm reaction is only to
minimize predation risk, then zebrafish in both flow treatments
should seek refuge in distant shelters. If alarm reactions include
overt risk avoidance traded off against the benefits of information
gathering, then fish in the no-flow treatment should tolerate risk to
access information by seeking refuge in the shelter nearest the
location of cue release.
Laboratory experiments afford control and power to detect
biological effects, but their contrived nature may result in unnatural or spurious behavioural responses (Irving & Magurran 1997).
Therefore, we repeated our laboratory experiments on field populations of minnows occupying natural river systems. Together,
these data combine the experimental power of the laboratory
setting with the ecological realism of the field setting.
0003-3472/$38.00 Ó 2010 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.anbehav.2010.02.004
Author's personal copy
1102
B.D. Wisenden et al. / Animal Behaviour 79 (2010) 1101e1107
stimulus (10 ml of skin extract). Rigid and flexible tubing used for
stimulus injection were replaced for every trial.
METHODS
Laboratory Fluvaria
Experimental Protocol
The zebrafish, Danio rerio, is a lotic cyprinid native to India and
a model organism for molecular genetics (Engeszer et al. 2007).
Zebrafish have a well-documented antipredator response to
chemical alarm cues from conspecifics (e.g. Waldman 1982;
Suboski et al. 1990; Hall & Suboski 1995; Korpi & Wisenden
2001). Study animals were acquired from commercial suppliers
and housed in 190-litre holding tanks at the aquatic research
facility at Minnesota State University Moorhead (MSUM).
Individual fish were placed into one of two identical fluvaria.
The fluvaria were rectangular troughs, 30 cm wide and 2.44 m in
length with a viewing pane along one side (Fig. 1). Water depth was
maintained at 7.5 cm. A stack of drinking straws (diameter ¼ 6.0 mm) immediately downstream of the point of water entry
stabilized turbulence. A second identical stack of straws at the
downstream end immediately before the drain left a 180 cm
section of open stream between two visually identical ends. In the
second iteration of this experiment the downstream stack of straws
was not used, creating 210 cm of usable stream. Lines drawn on the
front viewing pane divided the tank into 18 (experiment 1) or 21
(experiment 2) 10 cm sections. The fluvarium bottom was covered
with a thin layer of silica sand that provided a smooth bottom to
minimize retention of chemical cues as they passed through the
stream system (Ferner et al. 2009).
A short length (ca. 20 cm) of rigid plastic tubing for injecting test
stimuli descended vertically from a special holder designed for this
purpose midway across the width of the fluvarium, ending midcolumn between the water surface and substrate. A 2 m length of
flexible plastic airline hosing attached to the rigid tubing extended
to the floor in front of the tank where experimenters could
surreptitiously inject test stimuli without disturbing the test
subject. Before pre-stimulus observations began, a 60 ml syringe
was used to withdraw and discard two draws of 60 ml of tank water
through the stimulus injection tube to rinse it. A third draw of 60 ml
of tank water was taken and retained to flush the control test
stimulus (10 ml of dechlorinated water). A fourth draw of 60 ml of
tank water was retained in another syringe to flush the alarm cue
The experimental protocol was identical for both iterations of
the laboratory experiment. A single zebrafish was placed into
a fluvarium and allowed at least 24 h to acclimate. Flow was on and
recirculating through the reservoir during this time. Each fish was
observed over three consecutive observation periods. Pre-stimulus
data were collected for 3 min. Activity, horizontal position and
shelter occupancy were recorded. Activity was tallied as the sum of
the number of times the fish passed one of the lines drawn on the
front viewing pane (spaced 10 cm apart). Horizontal position was
recorded as a scan sample of the 10 cm areas occupied by the fish at
15 s intervals. At the time of the scan sample, we also recorded
whether the fish occupied a shelter. When the pre-stimulus period
was complete, we redirected the outlet of the stream tank from the
reservoir to the floor drain. Thus, introduced chemical stimuli now
passed only once through the fluvarium before being flushed
permanently from the system. We then immediately began
injecting 10 ml of dechlorinated tap water (control) through the
stimulus injection tube, followed by the flush of previously retained
60 ml of tank water. Stimulus injection required about 1 min to
complete. Activity, horizontal position and shelter use were then
recorded for 3 min. This observation period was called the postwater period. When that observation was complete, we injected
10 ml of skin extract solution containing alarm cues followed by
60 ml of previously retained tank water to flush alarm cues into the
tank. Stimulus injection required about 1.5 min to complete. We
recorded activity, horizontal position and shelter use again for
3 min (post-alarm period).
Description of Flow Parameters and Fate of Odour Plumes
In the absence of flow, injected stimuli diffused a mean SE
distance of 34.9 0.98 cm (N ¼ 10 dye tests) from the point of
release (i.e. to include the nearest shelter, but not the distant one
(s)) within the 3 min observation period. Food colouring dye
released within the 0e10 cm section reached the first shelter
Side view
2
1
*
E
D
M
180 cm
To drain
U
= Straws
= Shelter (side view)
= Shelter (top view)
Reservoir
624
Pump
625
626
Heater
E
D
2
M
Top view
1
U
Figure 1. Design of the artificial stream systems used in the present study showing a 580-litre reservoir from which water was pumped to one end of a rectangular trough with one
side made of glass. The 180 cm open stream section in the centre was divided into 18 zones, 10 cm each, by a grid (not shown) drawn on the front viewing pane. A stack of drinking
straws at the upstream end served as a collimater to stabilize turbulence and create uniform current velocity. *A second stack of straws at the downstream end was used in
experiment 1 only. The drain returned water to the reservoir during acclimation periods. During data collection water was directed to the floor drain so that test stimuli were not
recirculated. In experiment 1, shelters occupied the zone between the 30 and 40 cm sections, and 150e160 cm downstream of the straws. In experiment 2, shelters were placed at
30, 100 and 170 cm. Side view and top view are shown. E ¼ end pipe; D ¼ downstream shelter; M ¼ middle shelter; U ¼ upstream shelter.
Author's personal copy
B.D. Wisenden et al. / Animal Behaviour 79 (2010) 1101e1107
(30e40 cm) by 163.4 9.17 s from the beginning of cue release, or
about 1.72 min into the 3 min post-injection behavioural observation period. When the pump was turned on for the flow treatment,
the mean SE current was approximately 4.61 0.70 cm/s (N ¼ 10
dye tests). Stimuli injected at the upstream end (0e10 cm section)
reached the first shelter in 2.4 0.16 s, the downstream shelter in
30.2 0.92 s, and the end of the tank (180 cm) by 37.0 0.70 s.
After the end of stimulus release, dye was no longer visible at the
upstream edge of the first shelter by 31.1 4.68 s or at the
upstream edge of the downstream shelter by 87.6 9.30 s, and had
been flushed from the entire stream channel by 112.5 6.65 s.
Because cue injection took about 60e90 s, by the time poststimulus observations recommenced test cues were already
flushed clear of the shelter adjacent to the site of cue injection and
were flushed clear of the stream channel within the first minute.
Experiment 1: Two-shelter Experimental Design
Two shelters were added to each stream in sections 30e40 cm
and 150e160 cm, measured from the downstream edge of the
upstream stack of straws. Shelters were rectangular ceramic tiles
(11.5 22.5 cm) resting on four 3 cm tall legs made of 2 cm diameter PVC pipe. The injected stimuli entered the tank in the 0e10 cm
interval immediately downstream from the upstream stack of
straws (Fig. 1). We conducted 15 trials with the pump on to create
flow in the stream, and 15 trials with the pump off to create a static
water body with no flow. The height of the standpipe at the drain
was lowered slightly during flow conditions to maintain a constant
depth in the stream channel. All test subjects were tested only once.
Skin extract of zebrafish was prepared in two batches. For the
first batch, 12 zebrafish (mean SE total length (TL) ¼
42.9 0.63 mm) were killed by cervical dislocation and filleted.
Sheets of dermal and epidermal tissue from each fish were weighed
and placed in a beaker of dechlorinated tap water resting on a bed
of crushed ice. A total of 0.927 g of skin tissue was collected,
equivalent to 17.9 cm2 of skin area. Skin fillets were homogenized
using a hand blender to rupture epidermal cells and release
chemical alarm cue. The resulting solution was filtered through
a loose wad of polyester fibre to remove connective tissue. The
filtrate was diluted to a final volume of 116.5 ml with dechlorinated
tap water, aliquotted into 10 ml doses and frozen at 20 C until
needed. Each 10 ml dose contained 0.080 g of skin. The second
batch of skin extract used nine zebrafish (TL ¼ 37.9 1.12 mm) to
produce 0.795 g of skin that was blended, filtered, diluted to a total
volume of 105 ml, and aliquotted into 10 ml doses. Each 10 ml dose
contained 0.078 g of skin tissue. On each occasion, 10 ml aliquots of
blank dechlorinated water were frozen at 20 C to serve as control
stimuli. The protocol for the preparation of skin extract was
approved by the MSUM Institutional Animal Care and Use
Committee (protocol number 05-R-Biol-015-N-R-1).
Experiment 2: Three-shelter Experimental Design
Results from experiment 1 (see below) showed that fish
engaged in risk-sensitive information gathering from a safe
distance when possible, but approached danger to obtain information when information was localized. To increase our confidence
in this conclusion we repeated the experiment using three shelters
in each fluvarium instead of two and injecting test stimuli immediately upstream of the middle shelter. Shelters were centred at 40,
90 and 140 cm from the upstream straw stack. Shelters were square
ceramic tiles (15 15 cm) supported by 3 cm legs of PVC pipe. All
test stimuli were released 80 cm from the upstream straw stack.
We predicted that if zebrafish engage in risk-sensitive information
gathering, then in the absence of flow they should preferentially
1103
use the middle shelter. In the presence of flow, we predicted that
fish would use downstream shelters but not upstream shelters. We
conducted 14 trials of each flow treatment. Each test subject was
used only once.
Skin extract was prepared using the same protocol as for experiment 1. We filleted 29 zebrafish (mean SE TL ¼ 36.7 0.7 mm) to
harvest 1.963 g of skin. This was diluted to 300 ml with dechlorinated tap water and aliquotted as 10 ml doses and frozen at 20 C
until needed. Each dose contained cue from approximately 65 mg of
skin. Blank dechlorinated tap water controls were also prepared and
frozen as in experiment 1.
Experiment 3: Field Experiment
Field data were collected from sites along the upper headwater
reaches of the Mississippi River (N ¼ 8) starting 1.6 km from the
river's origin in Itasca State Park, MN, U.S.A. (47 150 , 12.2800 N, 95
130 , 31.0400 W) and Coffee Pot Landing (N ¼ 6) on La Salle River (47
200 , 5700 N, 95 100 , 58.3000 W), a tributary of the Mississippi. Physical
descriptors of the sites are provided in Table 1. There were no
differences between the sites in depth, current speed or water
temperature (P > 0.1 for all), but the La Salle River was only
4.0 0.35 m wide at test sites, while the Mississippi River was
6.3 0.61 m wide at test sites (t12 ¼ 3.03, P ¼ 0.011). Airline hose
was affixed to a stick wedged vertically into the substratum. The
airline hose extended to shore (hose length ca. 3 m), whence test
stimuli could be introduced surreptitiously. An underwater camera
(Aqua-VuÒ Explorer 5, Crosslake, MN, U.S.A., with proprietary
Digital Video Recorder) was placed on the bottom of the stream
about 1 m from the stick and with the stick centred in the camera's
field of view. Airline tubing was rinsed by withdrawing and discarding 120 ml of river water.
After 10 min of acclimatization, we recorded the number of fish
in the field of view upstream and downstream of the stick for
5 min. Then 60 ml of previously retained river water (blank water
control stimulus) was injected into the airline hose followed by
120 ml of previously retained stream water to flush the test cue
out of the injection hose. Fish behaviour was then recorded for
10 min. These data served to control for the effect of stimulus
injection. We then prepared chemical alarm cues by killing
a minnow by cervical dislocation with a razor blade, then making
10 superficial cuts along each side of the fish, and rinsing the fish
with 60 ml of river water in a beaker (University of Minnesota
IACUC protocol number 0902A59884). We then recorded another
5 min of pre-stimulus area use by fish upstream and downstream
of the injection hose, then injected 60 ml of chemical alarm cues
into the hose followed by 120 ml of previously retained stream
water to flush the cue into the stream. Behaviour was then
recorded for another 10 min. Two minutes of each observation
period (pre- and post-water; pre- and post-alarm cue) was scored
and analysed from recorded video. We recorded the number of
fish in view of the camera, at 10 s intervals, for the areas upstream
and downstream of the stick holding the injection hose, for the
four observation periods pre-stimulus, post-water, pre-alarm cue
and post-alarm cue.
Table 1
Descriptors of 14 field sites where data were collected along the Mississippi (N ¼ 8)
and La Salle (N ¼ 6) rivers
Mean
SE
Min
Max
Depth (cm)
Width (m)
Current (cm/s)
Temperature ( C)
47.5
3.28
32
68
5.3
0.49
2.8
8.8
29.1
2.44
10
45
20.8
0.59
14.4
23.3
Author's personal copy
B.D. Wisenden et al. / Animal Behaviour 79 (2010) 1101e1107
Fish to be used for generating alarm cues were collected with
a seine net near the study area before trials began each day. One of
two species of fish was used to make alarm cue, depending on the
fish in that section of stream at each site. Of 14 trials, we extracted
alarm chemicals from the skin of common shiners, Notropis cornutus, for 10 trials (7 in the Mississippi River, 3 in the La Salle River),
and alarm cues from hornyhead chub, Nocomis biguttatus, for the
remaining four trials (1 in the Mississippi River, 3 in the La Salle
River). Alarm cues were used within 15 min of preparation. Stream
water was collected locally at each site.
Statistical Analysis
Activity and horizontal position from laboratory data were
analysed using repeated measures ANOVA with three time periods
(pre-stimulus, water, alarm) and two levels of flow treatment (flow,
no flow) as a categorical predictor. The interaction between time
(effect of test cue) and flow for horizontal position in experiment 1
was significant, so we used one-way repeated measures ANOVA
across times within each flow treatment and paired t tests to
examine the effect of flow for each time period. Shelter use was
scored as the total frequency of shelter occupancy over all trials for
each flowetime combination. Chi-square tests were used to
compare observed frequency of shelter use against the expectation
of equality among or between treatment groups. Field data were
normally distributed (KolmogoroveSmirnov test: P > 0.05) and
analysed using paired t tests to compare changes (post-stimulus
minus pre-stimulus) in shelter use following the release of water
versus chemical alarm cues, and changes in area use upstream
versus downstream of cue release. All statistical inferences were
interpreted from two-tailed probability distributions.
RESULTS
Change in Activity
In the first fluvarium experiment, zebrafish significantly
reduced activity in response to alarm cue and responded with equal
intensity in flow and no-flow environments; that is, there was no
significant interaction between flow and time (flow: F1,56 ¼ 1.64,
P ¼ 0.205; time: F2,56 ¼ 13.15, P < 0.001; flow*time: F2,56 ¼ 0.59,
P ¼ 0.560; Fig. 2a). In the second iteration of this experiment, fish
were more active when flow was present than when it was absent
for all three time periods (Fig. 2b). However, fish in both flow
treatments reduced activity in response to alarm cues (flow:
F1,52 ¼ 18.58, P < 0.001; time: F2,52 ¼ 10.12, P < 0.001; flow*time:
F2,52 ¼ 0.65, P ¼ 0.529).
Change in Horizontal Position
In the pre-stimulus and post-water observation periods, fish
roamed the full length of the stream tank, resulting in horizontal
positions averaging in the mid-sections of the stream (Fig. 3a, b). In
the first experiment there was a significant interaction between the
overall effect of flow and the effect of cue treatment (flow:
F1,56 ¼ 12.37, P ¼ 0.001; time: F2,56 ¼ 1.31, P ¼ 0.277; flow*time:
F2,56 ¼ 4.58, P ¼ 0.014). One-way repeated measures ANOVAs
revealed that in the absence of flow, horizontal position was not
affected by cue treatment (F2,42 ¼ 1.02, P ¼ 0.371), whereas in the
presence of flow, fish significantly shifted horizontal position
downstream in response to alarm cue (F2,42 ¼ 4.37, P ¼ 0.019). Horizontal position did not differ between flow treatments during the
pre-stimulus period (t28 ¼ 0.99, P ¼ 0.331) or the post-water period
(t28 ¼ 0.66, P ¼ 0.512), but did differ significantly between flow
treatments during the post-alarm cue period (t28 ¼ 3.55, P ¼ 0.001).
200
(a)
150
*
100
Activity (lines crossed)
1104
50
0
120
(b)
100
80
*
60
40
20
0
Pre−stimulus
Water
Alarm
Figure 2. Mean SE activity (number of grid lines crossed) for 3 min before stimuli
were introduced, 3 min after 10 ml of dechlorinated water was introduced, and 3 min
after 10 ml of skin extract containing alarm cues were introduced. Stream flow was
either on (,) or off (-). (a) Experiment 1 (N ¼ 15 per treatment). (b) Experiment 2
(N ¼ 14 per treatment).
In the second experiment, fish occupied positions further
downstream when flow was on than when it was off. There was
a significant effect of time, but no significant interaction (flow:
F1,52 ¼ 14.11, P < 0.001; time: F2,52 ¼ 3.21, P ¼ 0.048; flow*time:
F2,52 ¼ 1.64, P ¼ 0.204). As before, horizontal position did not differ
between flow treatments during the pre-stimulus period
(t26 ¼ 1.60, P ¼ 0.122) or the post-water period (t26 ¼ 0.55,
P ¼ 0.590), but did differ between flow treatments during the postalarm cue period (t26 ¼ 2.41, P ¼ 0.023).
The difference in horizontal position between flow treatments
was due to differential use of shelters (Fig. 4a, b). Overall, fish
significantly increased shelter use after introduction of alarm cues.
In the first experiment, fish occupied shelters 34 times during the
pre-stimulus period, 40 times in the post-water period and 176
times in the post-alarm period (c22 ¼ 154.78, P < 0.001). In the postalarm period, the presence or absence of flow did not affect overall
shelter use (upstream and downstream shelter use combined: fish
occupied shelters 81 times in the presence of flow and 95 times in
the absence of flow; c21 ¼ 1.11, P > 0.05); however, fish avoided the
upstream shelter in the presence of flow but not in the absence of
flow (c21 ¼ 50.03, P < 0.001; Fig. 4a). In the second experiment, fish
sought shelter in three shelters and in the standpipe at the
downstream end of the fluvarium. Fish occupied shelters 83 times
in the pre-stimulus period, 81 times in the post-water period and
257 times in the post-alarm cue period (c22 ¼ 145.5, P < 0.001). As
in the first experiment, when flow was off, fish preferentially
sought shelter in the shelter adjacent to the site where alarm cue
was released. When flow was on, fish sought shelter downstream
(but not upstream), either in the downstream shelter or next to the
standpipe at the downstream end of the fluvarium (Fig. 4b). After
the introduction of alarm cues, fish in the absence of flow occupied
Author's personal copy
B.D. Wisenden et al. / Animal Behaviour 79 (2010) 1101e1107
180
80
(a)
*
1105
(a)
60
150
40
Frequency of shelter use
120
Distance from upstream end (cm)
90
60
30
0
Flow
No flow
Pre−stimulus
80
Flow
No flow
Flow
Water
No flow
Alarm cues
(b)
60
40
0
180
20
20
(b)
*
160
0
Flow
140
No flow
Pre−stimulus
120
Flow
No flow
Flow
Water
No flow
Alarm cues
Figure 4. Frequency of shelter use during the three observation periods when flow
was off or on. (a) Experiment 1 shelters: upstream (,); downstream (-). (b) Experiment 2 shelters: upstream (,); middle ( ); downstream ( ); end pipe (-).
100
80
60
40
20
0
Pre−stimulus
Water
Alarm
Figure 3. Mean SE horizontal position (distance from upstream end) for 3 min
before stimuli were introduced, 3 min after 10 ml of dechlorinated water was introduced, and 3 min after 10 ml of skin extract containing alarm cues were introduced.
Stream flow was either on (,) or off (-). (a) Experiment 1 (N ¼ 15 per treatment).
(b) Experiment 2 (N ¼ 14 per treatment).
the middle shelter 77 times and the downstream and end pipe
locations 47 times. In the presence of flow, fish occupied the middle
shelter only 12 times and occupied the downstream shelter and
end pipe 73 times (c22 ¼ 47.5, P < 0.001).
point of alarm cue release because information about the nature of
predation risk was not available at other shelters. Only in the presence
of flow could zebrafish monitor chemical information from a relatively safe distance. Fish strongly preferred downstream locations
where information carried by water currents could be monitored.
This behavioural change is consistent with a trade-off between
information gathering and risk avoidance. In the field context, fish
primarily avoided the area upstream of the cue source where they
could not monitor chemical information about the status of risk.
Acquiring chemical information about predation risk in advance
of an attack confers a significant survival advantage to prey when
they encounter a predator. The survival value of this information
has been demonstrated for larval anurans (Hews 1988), fathead
minnows (Mathis & Smith 1993; Chivers et al. 2002), Gammarus
amphipods (Wisenden et al. 1999) and salmonids (Mirza & Chivers
2001, 2002, 2003).
Field Experiment
DISCUSSION
Fish in the no-flow treatment tolerated risk to increase access to
indirect information about predation risk. In the absence of flow,
zebrafish preferentially used shelters immediately adjacent to the
Change in number of fish in view
In the first minute, the change (post-stimulus minus prestimulus) in the number of fish in view of the camera was greater
following the release of alarm cues (9.0 4.5, N ¼ 14) than after
the release of water (þ2.5 3.5, N ¼ 14) (paired t test: t28 ¼ 2.05,
P ¼ 0.050; Fig. 5). When area use was broken down into areas
upstream and downstream of the point of cue release, the change in
the number of fish upstream of the stick decreased in response to
alarm cues (17.5 6.8) but not in response to water (þ2.8 4.3)
(paired t test: t13 ¼ 2.20, P ¼ 0.047). Downstream of the stick there
was no difference between cue treatments in change in area use
(paired t test: t13 ¼ 0.48, P ¼ 0.640; Fig. 5). By the end of the second
minute, changes in upstream and downstream area use in response
to cue type did not differ (P > 0.05).
15
NS
10
*
5
0
−5
−10
−15
−20
−25
−30
W−Up
W−Down
A−Up
A−Down
Figure 5. Change (3 min post-stimulus minus 3 min pre-stimulus) in the first minute
in the number of fish in view upstream (Up) and downstream (Down) of the point
source of the test cue. Test cues were either blank water (W) or chemical alarm cues in
skin extract (A).
Author's personal copy
1106
B.D. Wisenden et al. / Animal Behaviour 79 (2010) 1101e1107
There are three lines of research established in the literature
that support our conclusion that fish actively trade off information
gathering with risk avoidance. First, fish accept risk in exchange for
information when engaging in predator inspection behaviour.
Inspection behaviour is undertaken, in part, to gather chemical
information about the identity of the predator, the predator's diet
and the predator's likelihood of launching an attack (Dugatkin &
Godin 1992; Brown & Godin 1999; Brown et al. 2001). Note that
predator inspections occur in response to the physical presence of
a predator. In the current experiment we presented only indirect
indicators of predation risk. Second, fish engage in at least two
forms of ‘stealthy sniffing’ that allow them to remain motionless
while accessing chemical information (Wisenden & Chivers 2006).
Motion is conspicuous to predators, which is why prey reduce
activity when risk is high (Lawrence & Smith 1989). During fin
flicking, fish remain motionless in the water column and flick their
pectoral fins forward and upward while simultaneously counterthrusting forward with the caudal, anal and dorsal fins. This
action causes a waft of water past the fish's external nares without
generating net body movement. Fin flicking is commonly observed
in fish responding to chemical alarm cues (Keenleyside 1955;
Lawrence & Smith 1989; Brown et al. 1999). Fin flicking itself can
cue conspecifics to the presence of predation risk by serving as
a visual form of public information about the presence of risk
(Brown et al. 1999). Another example of surreptitious chemical
sampling is head-up behaviour by darters (Percidae). Darters are
cylindrical benthic fish that often inhabit the current-free boundary
layer next to the substrate in rivers or streams. To sample chemical
information passing overhead, darters extend their pectoral fins
downward to raise their head above the substrate, and arch their
body dorsally to extend the snout vertically upward into the water
column (Smith 1979; Wisenden et al. 1995a; Wisenden & Chivers
2006). A final example of surreptitious sampling comes from
another percid darter in which opercular rate increases in response
to exposure to alarm cues. This response potentially serves to
increase the sampling rate of chemical information about predation
risk (Gibson & Mathis 2006). A third line of evidence comes from
separate field experiments on fathead minnows, Pimephales
promelas, and brook stickleback, Culaea inconstans, where fish were
captured at multiple locations, each given a site-specific fin clip,
then released into a plume of chemical alarm cue from conspecific
skin extract (Wisenden et al. 1995b). After the source of alarm cue
was removed, all fish avoided the areas where alarm cue was
released for the first 2 h. By the end of 4 h, the number of fish using
alarm-labelled areas did not differ from that using water-labelled
areas (controls), but fish using alarm-labelled areas were unclipped individuals that had moved into the area from elsewhere.
Clipped fish (fathead minnows in one experiment, brook stickleback in the other) returned to water-labelled sites within 2 h, but
did not return to alarm-labelled sites for 6e8 h. This suggests that
knowledgeable (clipped) individuals gathered information about
risk from afar and returned to alarm-labelled areas only when it
was truly safe to do so (Wisenden et al. 1995b). The duration of
active time in these pond studies concurs with formal estimates of
active time in small lakes (Wisenden et al. 2009) but contrasts
sharply with the very short active time of 1e2 min observed in
streams in the current study. Clearly, more work needs to be done
on alarm responses in flowing water.
Except for differential shelter use, behavioural responses of fish
in the flow and no-flow treatments were quite similar. The presence or absence of flow did not affect overall shelter use or the
intensity of activity reduction, although flow stimulated more
overall activity in the second experiment. Activity reduction and
shelter use are two well-documented behavioural responses to
chemical alarm cues (Lawrence & Smith 1989; Chivers & Smith
1998). Behavioural responses observed in the flow treatment
occurred after most or all of the chemical information had been
flushed from the system. Similar heightened alertness following
detection of alarm cue occurs in poeciliids (Garcia et al. 1992) and
free-swimming natural populations of blacknose shiners
(Wisenden et al. 2004). Vigilance is indeed a form of heightened
information gathering. However, in the no-flow treatment, zebrafish were not merely vigilant; they engaged in active information
gathering by seeking out the source of the cue and remaining in the
nearest shelter.
With respect to risk-sensitive information gathering, our laboratory results were corroborated by the behavioural response of
natural populations of stream cyprinids. The avoidance of areas
where alarm cues were released was markedly pronounced
immediately upstream of the source of the cue and not detectable
downstream of the point source of the cue. The field data lead to
three conclusions. First, when chemical information about predation risk is distributed asymmetrically in space, fish vacate risky
areas that have limited access to information and instead occupy
risky areas immediately downstream where they can continue to
monitor information, rather than flee the area completely or in
random directions. The second conclusion is the counterintuitive
result that fish upstream of the source respond to the release of
chemical alarm cues. None of the sites was near conspicuous back
eddies that may transport cue upstream (Dahl et al. 1998). It seems
reasonable to conjecture that upstream fish probably responded to
the visual cue of the behavioural response of downstream fish
when alarm cue was first released, as has been shown in other
studies (Verheijen 1956; Magurran & Higham 1988). The third
conclusion, as mentioned above, is that the intensity of the
behavioural response in flowing water is transient compared with
the sustained and intense behavioural response under laboratory
conditions or lakes in the field. The current speed at the field sites
was about 10 times as fast as the current speed in laboratory fluvaria. Fish in the field tests were in shoals whereas fish in laboratory fluvaria were solitary. However, field estimates of active space
(Wisenden 2008) and active time (Wisenden et al. 2009) based in
lentic systems were conducted on minnows travelling in shoals.
Clearly, estimates of active space and time based on lentic systems
do not necessarily hold for lotic systems. Duration and intensity of
behavioural responses to chemical indicators of risk are strongly
affected by social interactions with shoalmates, enhanced by
learning and dulled by habituation. Shoals represent a form of
shelter from predatory attack (Pitcher & Parrish 1993), meaning
that shoaling fishes in nature live permanently in a form of shelter.
Fleeing the area means leaving the safety of the shoal. Thus, antipredator responses to risk may sometimes be difficult to detect in
flowing water under natural conditions (Magurran et al. 1996;
Irving & Magurran 1997).
In most ecosystems predation is the dominant arbiter of fitness.
Gathering accurate public information about predation risk is
therefore an important determinant of reproductive success,
placing steep selection on animals to include information gathering
as a component of their antipredator response. What we have
demonstrated in this study is risk-sensitive information gathering
in the absence of a predator. Benefits of monitoring indirect
chemical information about predation risk is balanced against the
cost of accessing it.
Acknowledgments
Funding was provided by Faculty Research Grants to B.D.W. and
a student research award to C.L.B. from the MSUM College of Social
and Natural Sciences. We thank Matt Cole for suggesting the threeshelter design for the second laboratory experiment. Jon Ross of the
Author's personal copy
B.D. Wisenden et al. / Animal Behaviour 79 (2010) 1101e1107
Itasca Biological Field Station of the University of Minnesota and
a UROP (undergraduate research opportunity program) grant to B.
Demuth were instrumental in facilitating the field experiment. We
also thank several iterations of referees for constructive comments
that have ultimately increased the impact this study. This work was
made possible by funding from a National Science Foundation CCLI
DUE Award (No. 0736872) to L. Fuselier, B. Wisenden and M. Malott.
References
Brown, G. E. & Godin, J.-G. J. 1999. Who dares, learns: chemical inspection behaviour
and acquired predator recognition in a characin fish. Animal Behaviour, 57, 475e481.
Brown, G. E., Godin, J.-G. J. & Pederson, J. 1999. Fin flicking: a visual antipredator
alarm signal in a characin fish (Hemigrammus erythrozonus). Animal Behaviour,
59, 469e476.
Brown, G. E., Golub, J. L. & Plata, D. 2001. Attack cone avoidance during predator
inspection visits by wild finescale dace (Phoxinus neogaeus): the effects of
predator diet. Journal of Chemical Ecology, 27, 1657e1666.
Chivers, D. P. & Smith, R. J. F. 1998. Chemical alarm signalling in aquatic predatoreprey systems: a review and prospectus. Écoscience, 5, 338e352.
Chivers, D. P., Mirza, R. S. & Johnston, J. 2002. Learned recognition of heterospecific alarm cues enhances survival during encounters with predators.
Behaviour, 139, 929e938.
Dahl, J., Nilsson, P. A. & Petterson, L. B. 1998. Against the flow: chemical detection
of downstream predators in running waters. Proceedings of the Royal Society B,
265, 1339e1344.
Danchin, É., Giraldeau, L.-A., Valone, T. J. & Wagner, R. H. 2004. Public information: from nosy neighbors to cultural evolution. Science, 305, 487e491.
Dugatkin, L. A. & Godin, J.-G. J. 1992. Prey approaching their predators: a cost
benefit perspective. Annales Zoologici Fennici, 29, 233e252.
Engeszer, R. E., Patterson, L. B., Rao, A. A. & Parichy, D. M. 2007. Zebrafish in the wild:
a review of natural history and new notes from the field. Zebrafish, 4, 21e40.
Ferner, M. C., Smee, D. L. & Weissburg, M. J. 2009. Habitat complexity alters lethal
and non-lethal olfactory interactions between predators and prey. Marine
Ecology Progress Series, 374, 13e22.
Garcia, C., Rolan-Alvarez, E. & Sanchez, L. 1992. Alarm reaction and alert state in
Gambusia affinis (Pisces, Poeciliidae) in response to chemical stimuli from
injured conspecifics. Journal of Ethology, 10, 41e46.
Gibson, A. K. & Mathis, A. 2006. Opercular beat rate for rainbow darters Etheostoma caeruleum exposed to chemical stimuli conspecific and heterospecific
fishes. Journal of Fish Biology, 69, 224e232.
Hall, D. & Suboski, M. D. 1995. Visual and olfactory stimuli in learned release of
alarm reactions by zebra danio fish (Brachydanio rerio). Neurobiology of Learning
and Memory, 63, 229e240.
Hews, D. K. 1988. Alarm response in larval western toads, Bufo boreas: release of
larval chemicals by a natural predator and its effect on predator capture efficiency. Animal Behaviour, 36, 125e133.
Irving, P. W. & Magurran, A. E. 1997. Context-dependent fright reactions in captive
European minnows: the importance of naturalness in laboratory experiments.
Animal Behaviour, 53, 1193e1201.
Keenleyside, M. H. A. 1955. Some Aspects of the Schooling Behaviour of Fish. Leiden,
Holland: E.J. Brill.
Korpi, N. L. & Wisenden, B. D. 2001. Learned recognition of novel predator odour
by zebra danios, Danio rerio, following time-shifted presentation of alarm cue
and predator odour. Environmental Biology of Fishes, 61, 205e211.
1107
Lawrence, B. J. & Smith, R. J. F. 1989. Behavioral response of solitary fathead
minnows, Pimephales promelas, to alarm substance. Journal of Chemical Ecology,
15, 209e219.
Lima, S. L. & Dill, L. M. 1990. Behavioral decisions made under the risk of predation:
a review and prospectus. Canadian Journal of Zoology, 68, 619e640.
Magurran, A. E. & Higham, A. 1988. Information transfer across fish shoals under
predation threat. Ethology, 78, 153e158.
Magurran, A. E., Irving, P. W. & Henderson, P. A. 1996. Is there a fish alarm
pheromone? A wild study and critique. Proceedings of the Royal Society B, 263,
1551e1556.
Mathis, A. & Smith, R. J. F. 1993. Chemical alarm signals increase the survival time
of fathead minnows (Pimephales promelas) during encounters with northern
pike (Esox lucius). Behavioral Ecology, 4, 260e265.
Mirza, R. S. & Chivers, D. P. 2001. Chemical alarm signals enhance survival of brook
charr (Salvelinus fontinalis) during encounters with chain pickerel (Esox niger).
Ethology, 107, 989e1006.
Mirza, R. S. & Chivers, D. P. 2002. Behavioural responses to conspecific disturbance
chemicals enhance survival of juvenile brook charr, Salvelinus fontinalis, during
encounters with predators. Behaviour, 139, 1099e1110.
Mirza, R. S. & Chivers, D. P. 2003. Response of juvenile rainbow trout to varying
concentrations of chemical alarm cues: response thresholds and survival during
encounters with predators. Canadian Journal of Zoology, 81, 88e95.
Pitcher, T. J. & Parrish, J. K. 1993. Functions of shoaling behaviour in teleosts. In:
Behaviour of Teleost Fishes (Ed. by T. J. Pitcher), pp. 363e439. London: Chapman
& Hall.
Smith, R. F. J. 1979. Alarm reaction of Iowa and Johnny darters (Etheostoma, Percidae, Pisces) to chemicals from injured conspecifics. Canadian Journal of
Zoology, 57, 1278e1282.
Smith, R. J. F. 1992. Alarm signals in fishes. Reviews in Fish Biology and Fisheries, 2,
33e63.
Suboski, M. D., Bain, S., Carty, A. E., McQuoid, L. M., Seelen, M. I. & Seifert, H.
1990. Alarm reaction in acquisition and social transmission of simulatedpredator recognition by zebra danio fish (Brachydanio rerio). Journal of
Comparative Psychology, 104, 101e112.
Verheijen, F. J. 1956. Transmission of a flight reaction amongst a school of fish and
the underlying sensory mechanisms. Experientia, 12, 202e204.
Waldman, B. 1982. Quantitative and developmental analysis of the alarm reaction
in the zebra danio, Brachydanio rerio. Copeia, 1982, 1e9.
Wisenden, B. D. 2008. Active space of chemical alarm cue in natural minnow
populations. Behaviour, 145, 391e407.
Wisenden, B. D. & Chivers, D. P. 2006. The role of public chemical information in
antipredator behaviour. In: Fish Chemoreception (Ed. by F. Ladich, S. P. Collins,
P. Moller & B. G. Kapoor), pp. 259e278. Delhi: Science.
Wisenden, B. D., Chivers, D. P. & Smith, R. J. F. 1995a. Early warning of risk in the
predation sequence: a disturbance pheromone in Iowa darters (Etheostoma
exile). Journal of Chemical Ecology, 21, 1469e1480.
Wisenden, B. D., Chivers, D. P., Brown, G. E. & Smith, R. J. F. 1995b. The role of
experience in risk assessment: avoidance of areas chemically labelled with
fathead minnow alarm pheromone by conspecifics and heterospecifics.
Écoscience, 2, 116e122.
Wisenden, B. D., Cline, A. & Sparkes, T. C. 1999. Survival benefit to antipredator
behavior in the amphipod Gammarus minus in response to injury-released
chemical cues from conspecifics and heterospecifics. Ethology, 105, 407e414.
Wisenden, B. D., Vollbrecht, K. A. & Brown, J. L. 2004. Is there a fish alarm cue?
Affirming evidence from a wild study. Animal Behaviour, 67, 59e67.
Wisenden, B. D., Rugg, M. L., Korpi, N. L. & Fuselier, L. C. 2009. Estimates of active
time of chemical alarm cues in a cyprinid fish and an amphipod crustacean.
Behaviour, 146, 1423e1442.