Content uploaded by Amiyaal Ilany
Author content
All content in this area was uploaded by Amiyaal Ilany on Apr 28, 2016
Content may be subject to copyright.
ORIGINAL PAPER
Wait before running for your life: defensive tactics
of spiny mice (Acomys cahirinus) in evading barn owl
(Tyto alba) attack
Amiyaal Ilany &David Eilam
Received: 27 February 2007 /Revised: 22 October 2007 / Accepted: 24 October 2007 / Published online: 19 December 2007
#Springer-Verlag 2007
Abstract Raptor–prey encounters were studied to evaluate
the strategies and success rate of both predator attack and
prey defense. We compared the success of barn owls in
catching stationary simulated prey (food item) with that of
moving prey (food item that was pulled in various
directions). We also tracked real encounters between barn
owls and spiny mice in a captive environment. It was found
that owls had higher success in attacking stationary prey
and that they seemed to attack the prey as soon as it became
motionless. When attacked, only a few spiny mice
remained immobile (freeze response) whereas most fled
and usually avoided capture by the owls. It was also found
that spiny mice displayed a preference to escape in those
directions in which owls had demonstrated a lower success
in catching the simulated prey. Escape initiation dichoto-
mized to a short or long (but rarely intermediate) distance
between the spiny mouse and the owl with more successful
avoidance at short-distance (last-moment) escapes. The best
predictor of escape success was the velocity of the spiny
mouse, and the second best predictor was its flight initiation
distance (FID). We present an update for Ydenberg and
Dill’s model for optimal FID in close encounters, suggest-
ing that fleeing at the last moment is advantageous.
However, a last-moment attempt to escape is also more
risky with a split second differing between life and death,
and is therefore appropriate mainly for agile prey under
close-distance attack.
Keywords Flight initiation distance (FID) .Anti-predator
behavior .Predator–prey interactions .Predation risk
Introduction
On an evolutionary time scale, predator–prey encounters
are an adaptive arms-race (Lima and Dill 1990) with each
opponent attempting to optimize its skills, whether for
catching the prey or for evading the predator. On a real-
time scale, predator–prey encounter is a dynamic conflict
in which a split-second decision can make a life or death
difference. Aerial raptors utilize various hunting strategies
(Cresswell 1996), ranging from continuous active pursuit
to perch and pounce ambush. Nocturnal owls perch
motionless, merging into the background with the camou-
flage colors of their feathers, waiting to pounce; while in
active hunt on the wing, they fly silently by virtue of the
unique structure of their wings and feathers, which
suppress the sound of airflow over the wings (Graham
1934;ThorpeandGriffin1962). These offensive strate-
gies are considered to be extremely efficient; for example,
the tawny owl (Strix aluco) is a primary predator, taking
8–10 woodland mice per night (King 1985). Indeed, owl
predation poses a significant threat for small mammals,
especially rodents that constitute a major food source for
many owl species (Mikkola 1983;MartinandBusby
1990;Selaas1993;Tome1994; Jedrzejewski et al. 1993;
Jedrzejewski et al. 1996). Despite the numerous studies on
owl predation, there is ambiguity regarding their rate of
success and their offensive strategies. For example, it was
suggested that once owls have initiated the final attack
sequence, their success rate rarely falls below 90% (Curio
1976). This raises the question of what is the ‘final attack
sequence’. Furthermore, what were the preceding phases
Behav Ecol Sociobiol (2008) 62:923–933
DOI 10.1007/s00265-007-0516-x
DO00516; No of Pages
Communicated by E. Korpimäkki
A. Ilany :D. Eilam (*)
Department of Zoology, Tel-Aviv University,
Ramat-Aviv 69978, Israel
e-mail: eilam@post.tau.ac.il
in the attack sequence, was the attack on moving or
stationary prey, etc.?
Prey species are capable of recognizing potential preda-
tors and of producing a defensive strategy that is appropriate
to the hunting strategy of that particular predator. Modulated
by the strong selective pressure of predation, defensive
behaviors converge to three categories that cross the entire
animal kingdom: freezing, fleeing, and fighting (defensive
attack). In freezing, the prey remains immobile to evade the
attention of the predator (Desy et al. 1990; Hendrie and
Weiss 1994; Hendrie et al. 1998; Ronkainen and Ylonen
1994). In fleeing, the prey escapes to remove itself from the
vicinity of the predator (Bolles 1970; Driver and Humphries
1988; King 1985). In fighting (or defensive attack), the prey
heads toward the predator to dissuade it and prevent
predation. Defensive fighting occurs only when the prey
has no possibility of freezing or fleeing and must face the
predator. Freezing and fleeing have been described in a
variety of prey species, from hermit crab (Scarratt and Godin
1992)todeer(Smith1991). For example, the white-tailed
deer (Odocoileus virginianus) typically flees flagging its tail
(Smith 1991)whereasthejerboa(Jaculus jaculus, a small
bipedal rodent) squats motionless, hiding its white ventral fur
and exposing its yellowish dorsal fur to match the desert
sand (Hendrie et al. 1998). Freezing or fleeing can be also
observed in the same individual animal under different
circumstances. For example, wood mice (Apodemous mys-
tacinus) either freeze or leap when exposed to stoats
(Mustela ermina)(Erlingeetal.1974) but scamper away
when exposed to other predators (Bolles 1970;King1985).
The response also varies with age because young white-
tailed deer tend to freeze when exposed to a predatory risk,
whereas adults typically flee (Smith 1991). Finally, freezing
and fleeing may occur in succession in the same individual.
For example, robins (Erithacus rubecula) reacted to an
approaching predator model by taking off toward the
opposite side of the cage where they hovered against the
wall for a short while before flying down to the floor and
remaining motionless (Lind et al. 1999). These examples
demonstrate that freezing and fleeing are general forms of
defensive behavior across species and individuals. In discus-
sing the question of when to freeze or flee (Eilam 2005), it
was suggested that freezing is efficient only if employed
before the prey is spotted by the predator, otherwise the prey
becomes a stationary target, which is easy to catch. Thus, a
clumsy prey that does not have access to a shelter would do
better to alternate between freezing and fleeing rather than
just freezing. In fleeing, the prey can move either directly
away and maximize its distance from the predator, move
toward the predator to confine it to a single clash point, or
dodge sideways to evade the attack (Eilam 2005). For
example, about 50% of sedge warblers (Acrocephalus
schoenobaenus) responded to an “attacking”cardboard
model of a merlin by darting sideways at an angle of almost
90° from the model (Kullberg et al. 2000). Prey can also run
in a straight path that is efficient against slow or distant
predators, or in a zigzag path that is efficient when a raptor is
close or fast (Furuichi 2002). Each one of these tactics
challenges the attacking raptor and its maneuvering abilities.
The distance between predator and prey at the moment the
prey initiates flight (flight initiation distance—FID) has
attracted attention in many studies. Ydenberg and Dill
(1986) presented a model based on economic considerations
in which FID for the prey should be at the point where the
cost of staying becomes higher than the cost of fleeing. For
example, prey should flee when the risk from an approaching
predator is higher than the revenue from a food patch. This
hypothetical model has been supported by numerous studies.
For example, woodchucks (Marmota monax) fled late
(shorter FID) when attacked closer to their burrows, reflecting
the lower risk (Bonenfant and Kramer 1996). Similarly, FID
of Bonaire whiptail lizards (Cnemidophorus murinus)was
longer when predators approached faster, heading directly
towards them; FID was shorter when food was nearby,
reflecting a higher cost of fleeing (Cooper et al. 2003). Male
broad-headed skinks (Eumeces laticeps) ran away later
(shorter FID) while guarding their females compared with
males without females (Cooper 1999). Other studies have
linked FID with different parameters, not necessarily related
to an approaching predator: 64 out of 68 species of Australian
birds fled earlier when the initial distance to the predator was
longer (Blumstein 2003). Other Australian birds have
species-typical FID (Blumstein et al. 2003). Finally, marsu-
pials in islands with no predators had shorter FIDs than their
conspecifics in the continent (Blumstein 2002).
In the present study of raptor–prey encounters, we compared
the success of barn owls in catching stationary simulated prey
with that of catching simulated prey that moved in various
directions, and we tracked real encounters between barn owls
and spiny mice in a captive environment. Specifically, we
tested: (1) if owls have a preference to attack stationary or
moving prey, and which gives them a higher success rate; (2)
what are the characteristics of owl attack in terms of timing,
velocity, and maneuverability; (3) what is the most effective
FID in spiny mice, and how is it correlated with successful
escape; and (4) which parameters of the encounters between
barn owls and spiny mice best predict a successful escape.
Materials and methods
Study animals
Barn owl (Tyto alba) An efficient raptor that feeds mainly
on rodents. It uses sounds generated by prey movement to
locate its prey, aided by its sharp night-vision (Payne 1971;
924 Behav Ecol Sociobiol (2008) 62:923–933
Ilany and Eilam, personal observation). Barn owls then
swoop down on the prey from a perch or on the wing,
catching it with their spiked talons, and killing it immedi-
ately. Eight barn owls that hatched at the research zoo of
Tel-Aviv University were hand-reared to adulthood (6–18
months old) and used in the present study. Each barn owl
was kept individually in a 80 × 60 × 60 cm cage. These cages
were placed in a large aviary (6 × 6 × 4 m) where each owl
was released separately once a week for 24 h. Like the
other zoo predators, owls were fed with freshly killed (and
from time to time also live) chicks and mice (one mouse per
night per owl) obtained from surplus stock from the animal
quarters and from chicken incubators. Thus, these captive
barn owls were accustomed to preying on live rodents. The
owls were habituated to the presence of humans by being
held in hand for 15 min daily.
Common spiny mouse (Acomys cahirinus) Weighs 38–44 g
and is 11 cm long, plus a 10-cm tail. It is an agile rodent,
common on rocky mountains where it shelters in crevices
(Shkolnik 1971; Shargal et al. 2000). The high fecundity of
spiny mice (early maturation, short weaning period,
frequent all-year-round breeding in captivity) and the
extensive use of this species in other research projects had
resulted in a surplus of these rodents at the Tel-Aviv
University research zoo, which we used in the present
study. Several weeks before testing, spiny mice were
housed in groups of 5–10 in metal cages measuring 40 ×
70×25 cm located outdoors in the zoo yard under natural
(uncontrolled) temperature and light conditions. Overturned
ceramic pots and wooden boxes were placed in each cage to
provide shelter. Seeds and diced fresh vegetables were
provided ad lib.
Apparatus
Observations took place in a 6×4.5 × 3-m aviary. A wooden
box (0.3×0.4× 0.5 m) was hung on one of the walls. The
box had two openings: one to the other side of the aviary
wall, through which an owl could be placed in the box with
no need for the experimenter to enter the aviary area; and
another opening into the aviary. The door to the aviary
could be opened silently from the outside by means of a
rope. A wooden pole was installed inside the wooden box,
10 cm above the floor where the owl could perch. From this
point, the owl could see most of the aviary area, except for
the zone just below the wooden box. For experiment 1 (see
below), a pulley block was mounted outside the aviary, to
which a 1.5-kg weight was connected. The weight was tied
to a transparent fishing line hung from a hook above the
aviary center. From this hook, the string could be tied to a
prey item; and by letting the weight fall to the ground, it
would pull the prey in the desired direction. An infrared-
sensitive video camera (Ikegami ICD47E) that allows vivid
shots in complete darkness was installed in the ceiling of
the aviary, visually encompassing the whole area of the
aviary floor. An IR light (Tracksys, IR LED Illuminator,
UK) that emits light in a range invisible to owls and spiny
mice followed the direction of the camera. The camera was
connected to another video camera (Sony DCR-HC85E)
used for recording on DV cassettes.
Procedures
Experiment 1: timing, velocity, and maneuverability in owl
attack
Before testing, the eight owls were trained for several
weeks to receive their food in the center of the aviary, and
only then underwent the present study. Each of the owls
was placed in the wooden box for 2 min of habituation
before the experimenter pulled the string that opened the
box to uncover the food item. As soon as the owl took-off
from the perch to swoop down on the food item, which was
either left stationary or pulled away via the transparent line
(by letting the weight fall to the ground) in a direction that
followed a random sequence of numbers between 0 and 4
as generated by Microsoft Excel with each number in the
sequence assigned to one pulling direction (four pulling
directions and one for stationary food item). Behavior was
videotaped, commencing at exposure of the food item and
continuing until a successful catch was made. All owls
were tested 2–3 days/week with all testing sessions taking
place in the afternoon, using the owls in a random order in
each session. There was only one trial per day for each owl.
If an owl failed to swoop within 15 min, it was tested again
only on the next test day. Accordingly, we obtained 59
trials, 16 for stationary food item, 19 where the food item
was dragged toward the owl, 13 where the food item was
dragged sideways, and 11 where it was dragged forward
(directly away from the owl).
Experiment 2: do the owls attack stationary or moving prey;
which gives them a higher success rate; what is the most
effective FID in spiny mice, and how is it correlated
with successful escape; and which parameters best predict
asuccessfulescape
Tests (one per owl per night) took place between 6.00 P.M.
and 9 P.M., at a light level of 0.1 lux, coming from street
lights near the research zoo—similar to light level on full
moon nights. At the beginning of each test period, an
experimenter switched on the equipment, placed an owl in
the box while it was closed to the aviary, and released a
spiny mouse into the enclosure. After 2 min of acclimation,
Behav Ecol Sociobiol (2008) 62:923–933 925
the box door was opened, allowing the owl to see into the
aviary area, and to launch an attack. The experimenter
waited either until attack was launched or until 15 min had
passed if the owl did not launch an attack, and then stopped
the experiment and returned the animals to their cages. In
total, we ran 54 trials.
Data acquisition
Tapes were digitized by connecting the Sony camera to a
computer using a firewire cable. Microsoft Windows Movie
Maker 2.0 created a .wmv file for each trial, saved in a 576×
720 resolution, 25 frames per second. This software was also
used for editing the films, cutting unneeded parts. We then
processed the video clips using Adobe After Effects 6.0,using
the Vector Paint function to mark the owl and the spiny
mouse in each frame. The owl was marked by a 10-pixel
diameter circle and the spiny mouse by a 5-pixel diameter
circle. These markings created a contrast against the white
background, needed by the tracking software. Markings were
added to each video clip in the part containing the attack. The
.avi files created by this software were converted to .mpg
files needed by the tracking software using Mainconcept
MPEG Encoder.Noldus Ethovision 3.1 was used to get the
location of the owl and spiny mouse in every frame of the
movie. This software produced the coordinates of the
animals, calculated the movement velocities and distances,
and created a graphic presentation of the routes of each
animal (Fig. 1). To measure the angles of escape relative to
owl direction, we used the ImageJ software.
Behavioral analysis
A moment-by-moment analysis was carried out, starting
with the initiation of the attack by the owl and ending either
with the successful capture of the spiny mouse, with the
owl landing on the aviary floor without catching the prey,
or with the return of the owl to its perch without landing. In
this analysis, twelve parameters were scored:
1. Location of spiny mouse at attack initiation:The
coordinates of the spiny mouse’s location at the
moment when the owl initiated its attack, leaving its
perch.
2. Attack timing: Whether the owl attacked the spiny
mouse when it was mobile or stationary.
Fig. 1 Data acquisition in Ethovision 3.1 from a video clip with the
owl and the spiny mouse marked with a large and a small black circle,
respectively. The bottom left panel provides the Xand Ycoordinates of
the animals for every frame. The top left panel provides the details of
the measured clip section, and the right bottom panel controls the
video display
926 Behav Ecol Sociobiol (2008) 62:923–933
3. Response of the rodent to capture attempt: For each
capture attempt, the response of the rodent was scored
as ‘move’or ‘no move’.
4. Owl speed: The barn owl average speed for the whole
attack, from leaving the perch until landing.
5. Attack duration: The duration of the barn owl attack,
from leaving the perch until landing.
6. Location of owl landing: The coordinates of the barn
owl’s location at the moment of landing on the aviary
floor, whether it caught the spiny mouse or not.
7. Distance between owl landing location and initial
spiny mouse position: The distance between the landing
position of the owl and the position of the spiny mouse
when the attack began. Because the owl was initially
aiming at the initial spiny mouse position but had to
make adjustments to its path if the spiny mouse ran
away, this parameter measures the magnitude of this
adjustment.
8. Distance between owl landing location and spiny
mouse location: The distance between the landing
position of the owl and the position of the spiny mouse
when the owl landed. If the owl failed to catch the
spiny mouse, this distance depicts the magnitude of the
failure.
9. Catch: This parameter describes whether the barn owl
caught the spiny mouse or failed in its attack.
10. FID: The distance between the owl and the spiny
mouse at the moment the spiny mouse began to flee.
11. Direction of flee: The angle between the owl’s
trajectory to the spiny mouse and the direction in
which the spiny mouse ran, as measured when the
flight began. If the spiny mouse changed its direction
substantially during the flight (>20°), this angle was
measured again.
12. Speed of flee: The average speed of the spiny mouse
during its flight, from its beginning to its end when
the spiny mouse stopped moving.
Statistics
Differences between different groups of spiny mice (i.e., those
that escaped early and those that escaped late) were tested
using t-test, unless noted otherwise. Transition of spiny mice
between mobility and immobility when attacked was tested
using McNemar’s test. Differences between expected vs
measured results were examined using χ
2
tests. Differences
between individual barn owl speeds were tested using one-
way ANOVA. A logistic regression was used to examine
which variables may best predict the success of an owl
attack. The compared variables were the speed of the fleeing
spiny mouse, owl speed, FID, escape direction, distance
between initial spiny mouse position and location of owl
landing, locomoting or stationary spiny mouse, and identity
of the individual owl.
Results
Experiment 1: owl success according to movement
direction of a simulated prey
Success rate in owl attack The owls had full success in
catching stationary food items (simulated prey) that were
not pulled away (16 successful catches in 16 trails; 2 per
owl). The owls’success in catching food items (simulated
prey) that were pulled by a string varied with the direction
of pulling. In 19 trials with a simulated prey dragged
toward the owls, there were 7 catches (37% success); in 13
trials with sideways drag direction, there were 3 catches
(23% success); and in 11 trials with drag direction forward
(directly away from the owl), there were 6 successful
catches (55% success). Owl speed did not differ between
success and failure (M=4.71, SD = 0.66 and M= 4.57, SD =
1.13 ms/s, respectively; t
41
=0.51; p=0.3); nor did the speed
of the pulled food (M=1.53, SD=0.43 and M= 1.69, SD =
0.35 m/s; t
41
=1.3; p=0.11). This latter speed matched that
of live rodents escaping from an owl, as detailed below in
Experiment 2. In all, owl success was highest when
simulated prey was dragged straight ahead of it and lowest
when it was dragged sideways.
Adjustments in flight path When attacking the simulated
prey, the owls displayed three patterns of response: (1)
giving up and turning on wing to return to the roost (23%
of trials; all counted as failures); (2) landing on the spot
where the food item had been located before pulling (7% of
trials; all counted as failures); and (3) adjusting their flight
path in an attempt to catch the moving simulated prey (70%
of trials; 53% of which were successful). When the food
item was pulled forward away from the owl, the owl
typically continued to fly forward in chase (Fig. 2a). When
the food item was pulled toward the owl, the adjustments
required in the flight path were relatively minor, generally
taking the form of a shortcut in swooping down toward the
approaching simulated prey (Fig. 2b). In sideways pulling,
the owl made a diagonal correction (Fig. 2c and d). The
following results are shown only for sideways and forward
trials where major adjustments were required because the
flight paths had to be longer (no shortcut possible). Path
adjustment significantly differed (t
22
=4.8; p<0.0001)
between successful attacks (n=9; forward and sideways
trials) where correction was of M=1.82, SD=0.41 m and
failed attacks (n=15) where correction was of M=0.90,
SD=0.48 m. Thus, a larger adjustment in flight path
Behav Ecol Sociobiol (2008) 62:923–933 927
characterized successful attacks, whereas failed attacks
involved a smaller adjustment.
Experiment 2: encounters between barn owl and spiny
mouse
Attack on moving vs stationary spiny mouse
When deciding to launch an attack, the owls could choose
whether to attack a stationary or a moving spiny mouse. In
35 out of 54 observed attacks (64.8%), the spiny mouse
was stationary at the moment of the owl’s take-off for
attack, while the remaining 19 attacks (35.2%) were on a
moving spiny mouse. This preference to attack stationary
prey was significant (χ2
1¼4:74; p=0.029). Notably, 12 out
of the 35 attacks on stationary spiny mice occurred
immediately (<1 s) after the prey had ceased to locomote.
Of the 19 attacks on locomoting spiny mice, 7 (37%) were
heading away from the owl, 7 (37%) were locomoting
sideways in relation to the owl, and 5 (26%) were
locomoting toward the owl; all directions were measured
at the moment the owl launched its attack (see the
“Materials and methods”section).
Duration and flight speed in attacking spiny mice
Duration between an owl’s take-off for attack and its
landing either on the prey or on the cage floor (when
missing the prey) was M=1.13, SD = 0.60 s. Flight speed
during attacks was M=4.49, SD = 1.03 m/s. There was no
significant difference between attack speeds of the eight
owls (one-way ANOVA; F
7,45
=1.2; p=0.31).
Adjustments in attacking path
The distance between the owl’s landing point and the position
of the spiny mouse at the moment of owl take-off describes
the actual adjustment in attack path. This distance was 0.57±
0.61 m for successful attacks and 0.71± 0.95 m for failed
attacks. The distance between the owl’s landing point and the
position of the spiny mouse at the moment of landing
assesses the magnitude of failure: in successful attacks it was
zero, while in failed attacks it was 0.58±0.47 m.
Rate of success in owl attack on fleeing vs stationary spiny
mice
Of the 35 attacks that the owls launched on stationary spiny
mice, 19 spiny mice responded by fleeing, whereas the
other 16 remained stationary. In addition to the 35 attacks
on stationary spiny mice, the owls also launched 19 attacks
on moving spiny mice. Of these, one responded by
freezing. Thus, after the initiation of the attack, 17 spiny
mice were stationary whereas 37 were moving. Altogether,
spiny mice fled from an attacking owl and did not freeze
(McNemar test; χ
2
=24.1; p<0.001). As shown in Table 1,
owls had 100% success in catching stationary spiny mice,
but their success rate substantially declined to 22% in
attacks on fleeing spiny mice. Altogether, their rate of
success was 46% of their attacks on fleeing and stationary
spiny mice. The rodents’decision whether to flee or not
together with the owls’absolute success in catching non-
fleeing spiny mice indicates that the owls’total rate of
success was primarily affected by the defensive response of
Table 1 Number of attacks and rate of success in attacking fleeing
and motionless spiny mice for each of the tested owls (rows 1–8) and
for all owls (bottom row)
Owl Total attacks
(% of success)
Attacks on
motionless prey
(100% success)
Attacks on
fleeing prey
(% of success)
1 3 (0) 0 3 (0)
2 4 (25) 1 3 (0)
3 7 (29) 2 5 (0)
4 10 (40) 3 7 (14)
5 8 (50) 3 5 20)
6 7 (57) 2 5 (40)
7 9 (67) 4 5 (40)
8 6 (67) 2 4 (50)
All owls 54 (46) 17 (100) 37 (22)
1
1
1
1
2
2
2
2
a
c
b
d
Fig. 2 Exemplary flight paths of owls in missed attacks in the
simulated prey experiment. In all four panels, the direction of owl
flight is from right to left (←). Point 1 indicates the location of the
food item when the owl launched its attack; point 2 indicates the
location of the food item at owl’s touch-down. When food item was
pulled forward (a) the owl continued flying toward the prey, but
missed and landed before reaching it. In backward pulling (b), the owl
landed while the prey was behind it. In pulling to the left (c) and right
of the owl (d), the owl took a curved path toward the prey
928 Behav Ecol Sociobiol (2008) 62:923–933
the spiny mice. Indeed, owl attack speed in successful
attacks (4.72±1.13 m/s) was not significantly higher than in
failed attacks (4.30± 0.90 m/s; t
44
=1.46; p=0.15).
Owl attack: timing of escape response in spiny mice
Figure 3a depicts the distribution of FID, which is the
distance between the attacking owl and the spiny mouse at
the moment that the spiny mouse begins to flee. As shown,
initiating escape occurred mainly at short or long FID, and
only 4 out of 37 attempted escapes occurred at the mid-
range FID. Twelve spiny mice fled when the owl was still
relatively distant; in fact, they seemed to flee as soon as the
owl launched its attack. In contrast, 21 spiny mice waited
for the owl and fled at the last moment when the owl was in
their immediate vicinity (less than 2 m). FID was not an
arbitrary outcome of the distance between the owl and
spiny mouse at the beginning of the attack. This is shown in
Fig. 3b in which the FID is shown in relation to the initial
distance between owl and spiny mouse. As shown, FIDs
dichotomized into two types: longer than 3 m and shorter
than 3 m. Correlating each group with the initial distance
between the owl and spiny mouse revealed that FIDs longer
than 3 m were in significant correlation with the initial
distance (r
2
=0.864; p=0.0001; n=10). In contrast, FIDs
shorter than 3 m were independent of the initial distance, as
illustrated by the horizontal correlation line (r
2
<0.001; p=
0.96; n=20). The latter result indicates that fleeing in the
last moment, when the owl was nearby, was not merely the
result of being attacked from a short initial distance. In
other words, the spiny mouse was waiting for the owl to get
closer and fled only at the last moment. In contrast, spiny
mice that were attacked from a relatively long initial
distance did not wait but fled immediately after launch of
attack.
Escape direction in relation to the approaching owl
Spinymiceweremorelikelytofleetoasideways
direction. Although four fled toward the attacking owl
at a range of ±45° to the approaching owl and another
four fled at a range of ±45° away from the approaching
owl, 20 fled sideways at an angle of 45° to 135° in
relation to the direction of the approaching owl. This
preference to flee sideways was significant (χ2
2¼7:07;
p=0.03). Another group of eight fleeing spiny mice were
not included in this count because they altered direction
during fleeing.
What are the parameters that may predict a successful owl
attack?
Logistic regression was used to examine which variables
may best predict the success of an owl attack revealing that
spiny mouse speed was the best predictor of owl success
(95% fit): the lower the mouse’s speed, the higher the owl’s
success (odds=0.978; 95%CI between 0.967 and 0.99; p<
0.001). Specifically, these results imply that with each spiny
mouse speed increase of 1 cm/s, the probability of being
caught declines to 0.978 of probability at the former speed.
The second best predictor was FID (odds=1.006; 95%CI
between 1.001 and 1.012; p=0.018): the lower the FID, the
lower the owl’s success. The other parameters were not
significant in predicting owl success.
0
5
10
15
20
0-2 2-4 4-6
Distance from the owl (m.)
Incidence
Predator-prey distance at the onset of attack (m.)
FID: Predator-prey distance
at the onset of fleeing (m.)
r2 = 0.864
r2 < 0.001
0
1
2
3
4
5
6
0246
a
b
Fig. 3 a Frequency distribution of FID (distance between the owl and
spiny mice at the onset of escape). bFID as a function of the initial
distance (distance between the owl and spiny mouse at the onset of
attack). Closed symbols represent successful escape; open symbols
represent successful attack. As shown, FID was usually short or long,
but not intermediate (a). In accordance, FID was plotted separately for
more than 3 m (squares) and for less than 3 m (diamonds). Trend-lines
for each of these data groups reveal that FIDs longer than 3 m were
correlated with the initial distance, whereas trend-line of FIDs shorter
than 3 m was horizontal, indicating independence between FID and
the initial distance
Behav Ecol Sociobiol (2008) 62:923–933 929
Discussion
Surprise attack: the owls’strategy
The results of the logistic regression indicate that once an
owl had launched its attack by swooping on the prey, it was
primarily the behavior of the spiny mouse that dictated the
success of the owl. In that sense, the owl’s offensive strategy
was based on surprising the prey before it could escape.
Accordingly, owl velocity and the adjustments it made in
flight path did not differ between failed and successful
attacks. However, the owl (as a predator) was the initiator of
the attack, and by timing the attack could surprise the prey,
preventing the exploitation of any defensive means. Indeed,
surprise attack is the most frequent type of attack in many
predators (Kenward 1978;Cresswell1993;Cresswell1996).
Nevertheless, in taking the decision to attack, owls have to
consider two cardinal parameters: attack on a stationary or a
moving prey, and the timing of the attack.
Attack on moving vs stationary prey
Barn owls in the present experiment showed an apparent
preference for attacking stationary prey. Moreover, tests
with simulated prey revealed 100% owl success in catching
stationary food items compared with catching food items
that were being dragged away (“simulated moving prey”).
Thus, it seems that a prey can dramatically decrease
predation risk by continuously locomoting when encoun-
tering a barn owl. Freezing, on the other hand, eliminates
the auditory and visual cues that owls use in pinpointing
prey (Mikkola 1983), and if a prey freezes before being
spotted, the owl may not be able to locate it (Kaufman
1974). However, upon noticing an owl, a prey might not
know whether it has already been spotted; and by freezing,
it may turn into a stationary, easy to catch target (Edut and
Eilam 2004; Eilam 2005). Accordingly, it may be advan-
tageous for a prey to alternate between freezing and fleeing,
combining disappearance through freezing with not being a
stationary target if freezing fails (Edut and Eilam 2004).
The present results elucidate the dynamic nature of the
encounter, revealing that barn owls optimize their offensive
behavior by attacking locomoting mice the moment these
mice cease their locomotion. The owls thereby benefit from
the easier tracking of a moving prey combined with the
easy catch of a stationary prey. However, the prey may cope
with this owl strategy by fleeing at the last moment, as
detailed below.
FID, instant escape, and last-moment escape
The timing of attempted escape was previously studied in
the context of FID, which is the distance between predator
and prey at which the prey starts to flee (Ydenberg and Dill
1986; Dill and Houtman 1989; Bonenfant and Kramer
1996; Kramer and Bonenfant 1997). FID was found to be
both species- and context-specific, ranging between two
extremes: (1) fleeing as soon as a predator is noticed and (2)
fleeing in the last moment (Ydenberg and Dill 1986; Dill and
Houtman 1989;Dill1990; Bonenfant and Kramer 1996;
Kramer and Bonenfant 1997; Cooper 1999; Blumstein et al.
2003; Cooper et al. 2003), which established the notion that
one cannot assume that a prey flees immediately upon
noticing a predator, but that it may notice the predator but
delay fleeing for some reason (economic gain). The present
results demonstrate that such a delayed attempt to escape
may also have a significant defensive value, limiting the
approaching owl’s ability to maneuver and adjust to the
route of the escaping prey. Moreover, the present results,
which were extracted from real encounters between barn
owls and spiny mice, revealed a dichotomy of escape
strategies: (1) immediate flight—a distant-dependent FID in
which flight occurred when the predator was relatively
distant (more than 3 m in the present study), in correlation
with the initial distance between them; and (2) last-moment
flight—a distance-independent FID in which flight occurred
when the predator was very close to the prey (less than 3 m
in the present study), independently of the initial distance
between them. Of these two strategies, in immediate flight,
the prey flees as soon as detecting the predator, presuming
that early flight will suffice for reaching a shelter or remove
of the prey from its immediate vicinity. In a last-moment
escape attempt, the prey flees only when the predator is
very close, as if waiting for the very final stage of the attack
(typically a distance of less than 1 m in the present study),
and then executes an agile escape. At that stage, the
predator is at maximum swooping velocity with limited
maneuverability, generally requiring it to land and then
launch a second attack. The delay to the second attack
enables the prey either to reach shelter or to move further
away from the predator.
Spiny mice are nimble (Eilam 1997; Oron et al. 1998),
living and foraging in crevices and spaces between and
under boulders, spending only little of their time in the
open (Kronfeld-Schor and Dayan 1999). Thus, an escaping
spiny mouse may quickly reach a shelter under rocks or
boulders, out of the reach of owls, and this makes both
immediate flight and last-moment flight efficient. Indeed, a
previous study found that when attacked by a barn owl, the
spiny mice fled; whereas a clumsier rodent species, the
social vole (Microtus socialis), favored alternating between
freezing and fleeing (Edut and Eilam 2004). A similar
dependence of defensive response on motor ability and
habitat structure was described in two deer species (Lingle
and Pellis 2002). White-tailed deer (Odocoileus virginianus)
that inhabit forests and are fast runners tend to flee when
930 Behav Ecol Sociobiol (2008) 62:923–933
encountering coyotes (Canis latrans), whereas mule deer
(Odocoileus hemionus) that live in relatively open spaces
and are moderate runners tend to freeze or hold their ground
and fight the predator (Lingle and Pellis 2002). Their agility,
speed, and the availability of nearby shelters thus make the
two escape strategies of spiny mice the more appropriate
response to owl attack.
The present results elaborate on Ydenberg and Dill’s
model, which presumes that early flight is always safer than
a later one (Ydenberg and Dill 1986). Indeed, this is the
case in any attempt to escape when a relatively long
distance exists between predator and prey. Moreover, this is
the range where the prey may still trade between predation
risk and other benefits, such as foraging. The possibility of
the prey to continue to forage or perform any behavior
beneficial to it other than flight is virtually nil at closer
distances, and this shorter range probably includes the
entire distance available in the present study (6 m). At the
longer distances, FID and the initial distance between
the predator and prey upon launch of attack are still
relevant. In this situation, the prey will do better by fleeing
as soon as it detects the predator. However, in attacks from a
close distance, the prey may do better by attempting a last-
moment escape. This model of FID is illustrated in Fig. 4.
Notably, most studies dealing with FID could not test its
effect on attack success; mostly because in those studies,
the simulated predator was a human approaching the prey
(Kramer and Bonenfant 1997). Comparing the two different
tactics of immediate flight and last-moment flight in the
present study revealed that spiny mice that fled later
(shorter FID) were more successful in evading barn owls
(85% vs 50% success in escaping). Thus, a shorter FID may
be advantageous in evading predators, in contrast with the
intuitive prediction that fleeing earlier is better. This finding
supports our suggestion that an agile prey may be better off
delaying its attempt to escape until the last moment.
Nevertheless, considering the velocity of the attacking
owl, a last-moment escape is risky with a split second
differing between life and death. Therefore, this defense is
probably only appropriate for an agile prey responding to a
nearby predator, especially when the prey does not have
access to a refuge.
Is there a preferred direction of escape?
The present study replicated a previous study (Shifferman
and Eilam 2004) on barn owls attacking a simulated prey—
a food item that was tied to a transparent string and pulled
away once the owl had launched its attack. As found in the
earlier study, in this study too, owls had a higher rate of
success in catching a simulated prey that was dragged
directly ahead of the direction of the attacking owl; lesser
success when the prey item was dragged toward the owl;
and least success when it was dragged sideways. In addition
to replicating the previous study (Shifferman and Eilam
2004), the present study also challenged the results
obtained with simulated prey by means of real encounters
between barn owls and spiny mice. Specifically, we tested
whether escaping spiny mice favor fleeing sideways in
relation to the approaching owl, a direction from which the
owls had had lowest success in catching the simulated prey.
Indeed, the present results unequivocally demonstrate that
spiny mice favor to flee sideways. Moreover, dodging
sideways is a common defense in numerous prey species.
For example, this is a common defense in song birds (Lima
1993). Blue tits (Parus caeruleus) jump to the side when
attacked by a swooping raptor (Lind et al. 2002; Lind et al.
2003). Ostriches (Struthio camelus) and rheas (Rhea
americana) move sideways (Farina et al. 2005), as also
do Thompson’s gazelles (Gazella thompsoni; Farina et al.
2005). In escape maneuvering, eluding the predator
successfully depends on the velocity and the turning radius
of the prey (Howland 1974).
In the present study, we reconfirmed that prey speed was
the main predictor of successful escape (the results from
logistic regression). While we did not measure the turning
radius of escaping spiny mice, this was a minor factor
considering that a spiny mouse is an 11-cm long rodent
whereas the owl is about 30 cm long with a wing span of
more than 70 cm. Thus, by moving quickly sideways, an
escaping spiny mouse can impose on the owl an impossible
maneuver. Indeed, in many of the failed attacks, the owls
touched down and immediately took off again and turned
toward the escaping mouse (Shifferman and Eilam 2004).
Howland (1974) suggested that sideways escape would be
highly efficient if made at the last moment when the
FID (flight initiation distance)
Chances for successful escape
1 – FID independent of the initial distance to predator (last moment escape)
2 – FID dependent of the initial distance to predator and alternative benefits (inst ant escape)
12
Fig. 4 A model for successful escape as a function of FID which
combines the last-moment escape (1) and instant escape (2). The
vertical lines distinguish between the escapes. In instant escape, the
chance of successful escape is lower when FID is shorter. In contrast,
in last-moment escape, the chance for a successful escape is higher
when FID is shorter, until a reversal point at which the predator is very
close and the chance of escape drops to zero
Behav Ecol Sociobiol (2008) 62:923–933 931
predator might not be able to adjust its path. This is exactly
what was found in spiny mice in their preference for a last-
moment escape attempt and in their preference to sideways
escape. In the logistic regression, however, escape direction
could not significantly predict the success of the owl, unlike
the velocity of the spiny mouse and the FID, which were
found to be reliable predictors.
Conclusions
Spiny mice under owl attack favor fleeing and by this, as
revealed in the present study, usually manage to evade the
owl. Escape initiation dichotomize to a short or long (but
rarely intermediate) distance between the spiny mouse and
the owl with higher success for the mouse attempting short-
distance (last-moment) escape. The best predictor of escape
success is the velocity of the spiny mouse and the second
best predictor is FID. While for the agile spiny mice, a high
velocity and a last-moment escape are the best strategies for
a successful escape, other strategies may better fit clumsier
prey species.
Acknowledgments We are grateful to Maoz Perlman and Yelena
Golan for their help in the experimentation and analyses, to the
zookeepers of the I. Meier Segals Garden for Zoological Research for
their help in rearing the owls and rodents, and to Naomi Paz for
editing the manuscript. This study was carried out under the
regulations and approval of the Institutional Committee for Animal
Experimentation (permit #L-05-059). The design carefully followed
the Guidelines for the Treatment of Animals in Behavioural Research
and Teaching (Animal Behaviour 2001) with special emphasis on the
Guidelines for Staged Encounters as outlined by Huntingford (1984).
References
Anonymous (2001) Guidelines for the treatment of animals in
behavioural research and teaching. Anim Behav 61:271–275
Blumstein DT (2002) Moving to suburbia: ontogenetic and evolutionary
consequences of life on predator-free islands. J Biogeogr 29:685–
692
Blumstein DT (2003) Flight-initiation distance in birds is dependent
on intruder starting distance. J Wildl Manage 67:852–857
Blumstein DT, Anthony LL, Harcourt R, Ross G (2003) Testing a key
assumption of wildlife buffer zones: is flight initiation distance a
species-specific trait? Biol Conserv 110:97–100
Bolles RC (1970) Species-specific defense reaction and avoidance
learning. Psychol Rev 77:32–48
Bonenfant M, Kramer DL (1996) The influence of distance to burrow
on flight initiation distance in the woodchuck, Marmota monax.
Behav Ecol 7:299–303
Cooper WE Jr (1999) Tradeoffs between courtship, fighting, and
antipredatory behavior by a lizard, Eumeces laticeps. Behav Ecol
Sociobiol 47:54–59
Cooper WE Jr, Perez-Mellado V, Baird T, Baird TA, Caldwell JP, Vitt
LJ (2003) Effects of risk, cost, and their interaction on optimal
escape by nonrefuging Bonaire whiptail lizards, Cnemidophorus
murinus. Behav Ecol 14:288–293
Cresswell W (1993) Escape responses by redshanks, Tringa tetanus,
on attack by avian predators. Anim Behav 46:609–611
Cresswell W (1996) Surprise as a winter hunting strategy in
sparrowhawks Accipiter nisus, peregrines Falco peregrinus, and
merlins F. columbarius. Ibis 138:684–692
Curio E (1976) The ethology of predation. Springer, Berlin
Desy EA, Batzli GO, Liu J (1990) Effects of food and predation on
behavior of prairie voles—a field experiment. Oikos 58:159–
168
Dill LM (1990) Distance-to-cover and the escape decisions of an
African cichlid fish, Melanochromis chipokae. Environ Biol
Fishes 27:147–152
Dill LM, Houtman R (1989) The influence of distance of refuge on
flight initiation distance in the gray squirrel (Sciurus carolinen-
sis). Can J Zool 67:233–235
Driver PM, Humphries DA (1988) Protean behavior: the biology of
unpredictability. Oxford University Press, London
Edut E, Eilam D (2004) Protean behavior under barn-owl attack: voles
alternate between freezing and fleeing and spiny mice flee in
alternating patterns. Behav Brain Res 155:207–216
Eilam D (1997) Postnatal development of body architecture and gait
in several rodent species. J Exp Biol 200:1339–1350
Eilam D (2005) Die hard: a blend of freezing and fleeing as a dynamic
defense—implications for the control of defensive behavior.
Neurosci Biobehav Rev 29:1181–1191
Erlinge S, Bergsten B, Kristiansson H (1974) The stoat and its prey-
hunting behaviour and fugitive reactions. Fauna Flora 69:203–
211
Farina RA, Blanco RE, Christiansen P (2005) Swerving as the escape
strategy of Macrauchenia patachonica Owen (Mammalia;
Litopterna). Ameghiniana 42:751–760
Furuichi N (2002) Dynamics between a predator and a prey switching
two kinds of escape motions. J Theor Biol 217:159–166
Graham RR (1934) The silent flight of owls. J R Aeronaut Soc
38:837–843
Hendrie CA, Weiss SM (1994) The development of an animal model
of panic with predictive and face validity. In: Cooper SJ, Hendrie
CA (eds) Ethology and pharmacology. Wiley, Chichester,
England, pp 111–132
Hendrie CA, Weiss SM, Eilam D (1998) Behavioural response of wild
rodents to the calls of an owl: a comparative study. J Zool
245:439–446
Howland HC (1974) Optimal strategies for predator avoidance: The
relative importance of speed and manoeuvrability. J Theor Biol
47:333–350
Huntingford FA (1984) Some ethical issues raised by studies of
predation and aggression. Anim Behav 32:210–215
Jedrzejewski W, Jedrzejewska B, Szymura A, Zub K (1996) Tawny
owl (Strix aluco) predation in a pristine deciduous forest
(Bialowieza National Park, Poland). J Anim Ecol 65:105–120
Jedrzejewski W, Rychlik L, Jedrzejewska B (1993) Responses of bank
voles to odours of seven species of predators: experimental data
and their relevance to natural predator–vole relationship. Oikos
68:251–257
Kaufman DW (1974) Differential predation on active and inactive
prey by barn owls. Auk 91:172–173
Kenward RE (1978) Hawks and doves: factors affecting success and
selection in goshawk attacks on woodpigeons. J Anim Ecol
47:449–460
King CM (1985) Interactions between woodland rodents and their
predators. Symp Zool Soc Lond 55:219–247
Kramer DL, Bonenfant M (1997) Direction of predator approach and
the decision to flee to a refuge. Anim Behav 54:289–295
Kronfeld-Schor N, Dayan T (1999) The dietary basis for temporal
partitioning: food habits of coexisting Acomys species. Oecologia
121:123–128
932 Behav Ecol Sociobiol (2008) 62:923–933
Kullberg C, Jackobsson S, Fransso T (2000) High migratory fuel
impair predator evasion in sedge warblers. Auk 117:1034–
1038
Lima SL (1993) Ecological and evolutionary perspectives on escape
from predator attacks: a survey of North American birds. Wilson
Bull 105:1–215
Lima SL, Dill LM (1990) Behavioral decisions made under the risk of
predation—a review and prospectus. Can J Zool 68:619–640
Lind J, Fransson T, Jakobsson S, Kullberg C (1999) Reduced take-off
ability in robins (Erithacus rubecula) due to migratory fuel load.
Behav Ecol Sociobiol 46:65–70
Lind J, Hollen L, Smedberg E, Svensson U, Vallin A, Jakobsson S
(2003) Detection distance influences escape behaviour in two
parids, Parus major and P. caeruleus. J Avian Biol 34:233–236
Lind J, Kaby U, Jakobsson S (2002) Split-second escape decisions in
blue tits (Parus caeruleus). Naturwissenschaften 89:420–423
Lingle S, Pellis SM (2002) Fight or flight? Antipredator behavior and
the escalation of coyote encounters with deer. Oecologia
131:154–164
Martin G, Busby J (1990) Birds by night. Poyser, London
Mikkola H (1983) Owls of Europe. Poyser, England
Oron U, Maltz L, Shefer G, Eilam D (1998) Histology and enzymatic
activity in the postnatal development of limb muscles in rodents.
Physiol Behav 63:651–657
Payne RS (1971) Acoustic location of prey by barn owls (Tyto alba). J
Exp Biol 54:535–573
Ronkainen H, Ylonen H (1994) Behavior of cyclic bank voles under
risk of mustelid predation—do females avoid copulations?
Oecologia 97:377–381
Scarratt AM, Godin JGJ (1992) Foraging and antipredator decisions in
the hermit crab Pagurus acadianus (Benedict). J Exp Mar Biol
Ecol 156:225–238
Selaas V (1993) A comparison of the diet of sympatric owls in Aust-
Agder country, Southern Norway. Fauna Norv 26:7–13
Shargal E, Kronfeld-Schor N, Dayan T (2000) Population biology and
spatial relationships of coexisting spiny mice (Acomys) in Israel.
J Mammal 81:1046–1052
Shifferman E, Eilam D (2004) Movement and direction of movement
of a simulated prey affect the success rate in barn owl Tyto alba
attack. J Avian Biol 35:111–116
Shkolnik A (1971) Diurnal activity in a small desert rodent. Int J
Biometeorol 15:115–120
Smith WP (1991) Ontogeny and adaptiveness of tail flagging behavior
in white-tailed deer. Am Nat 138:190–200
Thorpe WH, Griffin DR (1962) The lack of ultrasonic component in
the flight noise of owls compared with other birds. Ibis 104:256–
257
Tome D (1994) Diet composition of the long-eared owl in center
Slovenia: seasonal variations in prey use. J Raptor Res 28:253–
258
Ydenberg RC, Dill LM (1986) The economics of fleeing from
predators. Adv Study Behav 16:229–249
Behav Ecol Sociobiol (2008) 62:923–933 933
