Content uploaded by Daniel R Stahler
Author content
All content in this area was uploaded by Daniel R Stahler
Content may be subject to copyright.
ANIMAL BEHAVIOUR, 2002, 64, 283–290
doi:10.1006/anbe.2002.3047, available online at http://www.idealibrary.com on
Common ravens, Corvus corax, preferentially associate with grey
wolves, Canis lupus, as a foraging strategy in winter
DANIEL STAHLER*†, BERND HEINRICH* & DOUGLAS SMITH†
*Department of Biology, University of Vermont
†Yellowstone Center for Resources, Yellowstone National Park
(Received 28 February 2001; initial acceptance 16 May 2001;
final acceptance 22 January 2002; MS. number: A9007R)
One foraging strategy that scavengers can employ to discover unpredictable food sources is to associate
directly with predators who inadvertently provide food. The common raven, a well known feeding
generalist, is also a prominent scavenger of wolves’ kills and is found to be in close association with this
predator. We tested the hypothesis that ravens preferentially associate with wolves in winter as a
kleptoparasitic foraging strategy. The presence, absence and behaviour of ravens was documented during
winter observations of wolves, coyotes, Canis latrans, and elk, Cervus elaphus, as well as the landscape in
the absence of these three species. Ravens were found to be in close association with wolves when they
were travelling, resting and hunting prey. In comparison, ravens showed no significant association with
coyotes, elk or areas on the landscape in the absence of wolves. We also compared ravens’ discovery
success of wolf-killed and nonwolf-killed carcasses and their behavioural response upon discovery. Ravens
found all wolf kills almost immediately and remained at the carcass to feed alongside wolves after the
death of the prey. In contrast, ravens were less successful discovering experimentally placed carcasses in
the same study region, and did not land or feed despite the availability of fresh, exposed meat. Our results
show that ravens’ association with wolves is not just an incidental and proximate by-product of the
presence of fresh meat. Instead, we show that ravens preferentially associate with wolves in both the
presence and absence of food, resulting in the discovery of carcasses and suppression of ravens’ innate
fear of novel food sources. Through this mode of social foraging, ravens may experience increased
foraging efficiency in the use of an otherwise spatially and temporally unpredictable food source.
2002 The Association for the Study of Animal Behaviour. Published by Elsevier Science Ltd. All rights reserved.
The ability to gather information from either conspecifics
or nonconspecifics on the location and quality of
resources has important implications for species depen-
dent upon spatially and temporally unpredictable food
sources (Clark & Mangel 1984,1986;Pulliam & Caraco
1984). Predator-killed ungulate carcasses represent such a
food source for scavenger species, and it is presumed that
behavioural mechanisms for locating and utilizing
carcasses have evolved (Houston 1979;Heinrich 1988a;
Marzluff & Heinrich 1991). Although associations
between sympatric scavengers and carnivores have been
documented (Mech 1966;Kruuk 1972;Schaller 1972;
Cooper 1991;Gasaway et al. 1991;Caro 1994), few
studies have examined predator–scavenger relationships
with respect to the foraging behaviour of scavengers
(Kruuk 1972;Houston 1979;Paquet 1991), with the focus
instead on the behavioural ecology of the carnivore that
is being kleptoparasitized (Cooper 1991;Fanshawe
& FitzGibbon 1993;Caro 1994;Creel & Creel 1996;
Carbone et al. 1997).
Interspecific kleptoparasitism, or food stealing, is
inherent to predator–scavenger relationships. By employ-
ing kleptoparasitism as a foraging strategy, scavengers can
improve foraging success by reducing search time, energy
expenditure and risks associated with procuring the food
themselves, as well as by gaining access to large, high-
quality food items (Brockmann & Barnard 1979;Houston
1979;Heinrich 1988b). Behavioural strategies that allow
scavengers to gain information on foraging opportunities
and kleptoparasitize more efficiently, particularly in eco-
systems with high inter- and intraspecific competition,
are expected. Examples of such foraging strategies
include: travelling the landscape and visually discovering
a food source by chance (Houston 1979;Heinrich 1988b),
foraging where conspecifics or nonconspecifics are
seen to be foraging (Thorpe 1956;Kruuk 1972;Houston
Correspondence: D. R. Stahler, Yellowstone Gray Wolf Restoration
Program, Yellowstone Center for Resources, P.O. Box 168, Yellowstone
National Park, WY 82190, U.S.A. (email: dan_stahler@nps.gov). B.
Heinrich is at the Department of Biology, University of Vermont,
Burlington, VT 05405, U.S.A.
0003–3472/02/$35.00/0 2002 The Association for the Study of Animal Behaviour. Published by Elsevier Science Ltd. All rights reserved.
283
1979;Po¨ysa¨ 1992), responding to vocalizations from
other scavengers (Heinrich 1988b;Heinrich & Marzluff
1991), following conspecifics that have previously dis-
covered a food source from nocturnal roost sites
(Heinrich 1988b,1989;Marzluff & Heinrich 1991;
Marzluff et al. 1996), and associating with the predators
that make carcasses available (Mech 1966;Kruuk 1972;
Houston 1979;Cooper 1991;Paquet 1991).
The foraging behaviour of ravens living in highly
seasonal northern climates is of particular interest because
they feed in large groups and are dependent upon carrion,
an unpredictable food source (Bent 1946;Heinrich
1988a,b). In winter, ravens frequently scavenge the kills
of large mammalian carnivores (Mech 1966;Peterson
1977;Heinrich 1989) and in some northern areas, ravens
are dependent upon wolves to kill and open carcasses
(Mech 1970;Peterson 1977;Allen 1979), as they are un-
able to tear the hide of large mammals (Heinrich 1988b).
Wherever wolves persist today, ravens are found feeding
at their kills (Mech 1966,1970;Peterson 1977;Allen
1979;Bjarvall & Isakson 1982;Promberger 1992;Carbyn
et al. 1993) and researchers have hypothesized that a
unique relationship may exist between these two species
(Mech 1966,1970;Peterson 1977;Allen 1979;Carbyn
et al. 1993). Anecdotal evidence suggests that ravens
monitor wolf pack activities by either following them
directly (Mech 1966,1970;Peterson 1977;Allen 1979;
Bjarvall & Isakson 1982;Carbyn et al. 1993), following
their tracks in the snow (Mech 1966), or responding to
vocalizations to determine the location of the wolves
(Harrington 1978). Although documented, no studies
have quantified this predator–scavenger association in
detail to sort out possible proximate attraction to prey and
carcasses versus attraction to the predator itself.
In this paper, we report the degree of association
between ravens and wolves at and away from a food
source, making comparisons to raven associations with
sympatric coyotes, Canis latrans, elk, Cervus elaphus, and
the landscape in the absence of wolves. We hypothesize
that ravens preferentially associate with wolves as a
kleptoparasitic foraging strategy in winter and predict
that ravens follow wolves in the concurrent absence of a
food source, resulting in almost immediate discovery of
carcasses after a wolf predation event. Furthermore, the
degree of raven association is expected to differ depend-
ing on wolf activity, as well as between wolf and coyote
activity, due to the differential effects these variables have
on foraging opportunities for ravens. In addition, we
investigated another aspect of social foraging, the reduc-
tion of neophobia (Heinrich 1988c) through social facili-
tation. Ravens are initially fearful of large carcasses when
encountered, gradually losing their caution through
experience gained either individually or by observing
conspecific (Heinrich 1988c;Marzluff & Heinrich 1991;
Heinrich et al. 1995) and/or nonconspecific carcass feed-
ers (Heinrich 1999a). By comparing raven discovery suc-
cess and behavioural response to wolf-killed carcasses
versus carcasses not attended by wolves, we tested the
hypothesis that ravens experience reduced neophobia
through their association with this predator at carcasses,
presumably leading to greater foraging success.
STUDY AREA AND METHODS
This study was conducted primarily in the northern
portion of Yellowstone National Park (YNP), Wyoming,
in an approximate 800-km
2
area. The elevation of the
study site ranges from 1500 to 3400 m and the climate
is characterized by short, cool summers and long,
cold winters, with most of the annual precipitation
falling as snow (Houston 1982). Mean annual tempera-
ture is 1.8 C, and mean annual precipitation is 31.7 cm
(Houston 1982). The habitat in the study area included
forest, mesic meadow, mesic shrub-meadow, riparian
grassland, sage grassland and road (Gese et al. 1996).
During winter, elk are the most abundant ungulate
species, with the northern range herd estimated at
15 500–17 400 elk (Lemke et al. 1998). Elk carrion avail-
able as a result of predation or winter kill provides an
important food source for scavengers, including common
raven, black-billed magpie, Pica pica, coyote, red fox,
Vulpes vulpes, golden eagle, Aquila chrysaetos, bald eagle,
Haliaeetus leucocephalus, grizzly bear, Ursus arctos and
black bear, U. americanus.
Wolves were reintroduced into YNP in 1995–1996 after
an approximate 70-year absence (Bangs & Fritts 1996).
The Rose Creek Pack, Druid Peak Pack and Leopold Pack
were the three established northern-range wolf packs
monitored in this study. Pack sizes ranged between
13–22, 7–11 and 8–14 for Rose Creek, Druid Peak and
Leopold packs, respectively, from 1997 to 1999. Some
wolves from each pack had previously been radiocollared
by the National Park Service and the U. S. Fish and
Wildlife Department as part of long-term monitoring and
management in compliance with the Endangered Species
Act (Bangs & Fritts 1996) and we were granted permission
to use radiotelemetry to monitor these individuals. Raven
demography is largely unknown in YNP. The winter
population estimate for the study area was 60–120 indi-
viduals. However, 300–500 ravens live 20 km northwest
of the study area near Gardiner, Wyoming, an area
outside of YNP’s north entrance (T. McEneaney, personal
communication, YNP).
Raven Association with Wolves, Coyotes,
Landscape, Elk and Carcasses
We collected data on ravens associating with wolves,
coyotes, elk and the landscape from October through to
April during the winters of 1997–1998, 1998–1999, and
the first half of winter in 1999–2000. Direct open-field
observations were made of both radiocollared and
unmarked wolves that were typically associated as
members of resident packs during observations. Visual
observations were made from distances of 0.15–4.0 km.
Upon visual location of wolves, facilitated by radiotelem-
etry, known pack movement patterns, and 25–60spot-
ting scopes, an all-occurrence sampling period (Altmann
1974) began and we recorded (1) the number of wolves
present, (2) the behavioural activity, (3) the duration
(recorded within 1 min) of that behaviour bout and
(4) raven association.
284 ANIMAL BEHAVIOUR, 64, 2
Wolf activity categories were: (1) travelling: walking,
trotting or running, stopping occasionally to scan sur-
roundings; (2) resting: lying down either with head up
and alert or with head down and assumed to be resting,
and no food source was present; (3) chasing prey: active
running pursuit or interacting with ungulate prey species
in an attempt to kill; (4) mousing: stalking, searching,
orienting, chasing or pouncing on small mammals;
and (5) at-kill: feeding on an ungulate carcass or scraps
from a carcass within 100 m of that carcass. Because of
temporal behaviour patterns and pack cohesion, the
above behaviour categories were assigned without ambi-
guity and when more than half of the individuals in an
observed wolf group were performing a particular
activity. Activity bouts less than 10 min were not used,
except with chasing prey bouts. Because wolf and elk
chases frequently ended in a successful kill and were
sometimes less than 5 min in duration, the signifi-
cance in terms of raven association and foraging
success warranted that all chasing prey bouts be used for
analysis.
To determine the association between wolf activity and
ravens, continuous sampling during wolf activity bouts
recorded (1) the presence or absence of ravens, (2) the
number of ravens, (3) the duration (within 1 min) of
raven presence per wolf activity bout and (4) the proxim-
ity of ravens to wolves (within 1 m). A raven was con-
sidered ‘present’ when within 250 m of wolves and there-
fore assumed to be monitoring wolf activity. This value
was thought to be a conservative distance for association
due to the open landscape and the acuity of ravens’ visual
perception. Ravens were considered ‘at-kill’ when feeding
on an ungulate carcass or meat scraps within 100 m of
that carcass. There were no marked or known individual
ravens, and because they frequently flew in and out of the
our field of view, counts of individuals at and away from
carcasses were probably underestimates of the true
number of individuals present.
We collected data on ravens associating with coyotes
following the same protocol previously described with
wolves. Coyote behaviour categories were (1) travelling,
(2) resting and (3) mousing, under the same definitions
given for wolf behaviour. During this study, hunting
behaviour observed in coyotes was limited to mousing, so
prey-chasing bouts were not relevant. At-kill bouts were
not recorded because raven and coyote association at
carcasses was the result of two scavengers utilizing a wolf
kill.
To control for the possibility that the raven and wolf
association was simply the result of ubiquitous raven
presence on the landscape instead of with wolves per se,
we collected data on raven association with fixed points
on the landscape. Landscape observation sites were
chosen using the following criteria: (1) sites were in the
same areas that ravens were observed in association with
wolves and (2) sites had distinct landscape features such
as boulders, trees, or other topographical distinctions to
be used as reference points. Raven presence (<250 m)
with respect to fixed landscape points was recorded,
along with the number of ravens and duration of their
presence.
The fourth sampling method involved observing elk
groups in wolf territory and recording raven association.
Elk groups (XSE= 80.814.1 elk, range 4–600, N= 46)
visible from common wolf-viewing areas were selected for
observation. Raven presence (<250 m), number of indi-
viduals, duration of presence and behaviour were
recorded. These data tested the hypothesis that ravens
may associate with animals in general, or more specifi-
cally, with wolf prey species on the chance that predators
will eventually kill an individual prey animal that ravens
are monitoring.
Observations of coyotes, elk and fixed points on the
landscape were made within a wolf territory when wolves
were not present (>2 km away) to assure that raven
presence was not due to wolf presence. Assumed indepen-
dence of wolf, coyote, elk and landscape bouts was
maintained by using bouts separated by a night-time
period (10–14 h of dark) when ravens are not active and
sleeping.
Raven Discovery Time and Subsequent Behaviour
at Carcasses
We compared raven discovery of wolf-killed carcasses
versus nonwolf-killed carcasses. Discovery times of
wolf-killed carcasses were collected by observing
predation events and recording the arrival time (within
1 min) of the first raven (<250 m) to a wolf kill after the
assumed time of death of the prey species. Data on wolf
kills were the result of prey-chasing bouts that ended in
the capture of prey. Discovery time of a wolf-killed
carcass by ravens was recorded as 0 min if ravens were
present during a chasing prey bout that ended in a kill.
Following the time of death, maximum raven numbers
were recorded at 15-min intervals for the first 60 min of
the at-kill bouts and included ravens within 100 m of
the carcass. The recorded behaviour of ravens upon
discovery was either: (1) land and feed, or (2) leave
without feeding. Wolves were either feeding at the
carcass or within 20 m of the carcass in all recorded
bouts.
To test carcass discovery and behavioural response of
ravens to nonwolf-killed carcasses, we conducted 25
experimental trials by placing road-killed ungulate car-
casses in wolf territories during the same winter seasons
and conditions that wolves were killing prey. We selected
carcass placement sites that were located in the same
areas of previously observed wolf-killed carcass sites, and
conducted a 10-min observation period prior to carcass
placement to determine whether ravens were in the
vicinity (the trial was aborted if ravens were seen). We
placed cut-open carcasses (with unfrozen, exposed red
meat) on open, snow-covered areas conspicuous to avian
foragers. The snow surrounding the carcass was packed
and sprinkled with blood patches to simulate a typical
wolf-kill site. Following carcass placement, we conducted
a 60-min observation period from a hidden position and
recorded the same data as described with wolf-killed
carcasses to determine discovery success, discovery time
and behavioural response by ravens.
285STAHLER ET AL.: RAVENS FOLLOWING WOLVES TO FORAGE
Data Analysis
Our sampling unit used to determine raven association
was an activity or observation bout. We quantified the
proportion of bouts in which at least one raven was
present, as well as the proportion of time (in min) during
which ravens were present for any given bout type.
Behaviour category bouts were combined for wolves
(travelling, resting, mousing and chasing prey) and
coyotes (travelling, resting and mousing) for pairwise
comparisons of raven association with wolves versus
coyotes, elk and landscape bouts. At-kill data were
excluded from these pairwise comparisons because data
on coyotes at kills were not collected. We also made
pairwise comparisons between the two canids’ travel,
rest and mouse bouts, as well as among the wolf
behaviour categories (travel, rest, chase, mouse) to deter-
mine whether differences existed in the degree of raven
attraction.
All pairwise comparisons were performed using the
ratio estimation method (Scheaffer et al. 1996), which
provided more accurate estimates of proportions based on
varying bout lengths and gave more weight to longer
bouts. The ratio estimation value was the ratio of the total
number of minutes ravens were present for any given
bout category to the total number of minutes recorded for
that particular bout category. A ztest indicated significant
differences between the proportions being compared (Ott
1993). We adjusted level to correct for experimentwise
error rate of pairwise comparisons using Bonferroni
correction (Sokal & Rohlf 1995) for raven association
with wolves compared to coyotes, elk and landscape
(=0.017), and for comparison of raven association
among various wolf behaviour types (=0.005). The com-
parison between discovery success of wolf-killed carcasses
and nonwolf-killed carcasses was made using a ztest
comparing two binomial proportions (Ott 1993).
RESULTS
From November 1997 to December 1999, 53 resident
wolves were observed from three packs (Leopold, Druid
Peak and Rose Creek Packs) for 226.7 h. Wolf group size
for activity bouts varied depending on the pack observed
and degree of pack cohesion at a given time
(XSE= 8.60.3 wolves, range 1–22, N= 234). Of the 209
wolf activity bouts with ravens present, 56 (26.8%) were
collected on at least two of the three packs simul-
taneously in different parts of the study area by different
observers. Individual coyotes and their pack affiliations
were not identified, but a minimum of 36 coyotes
inhabited a 70-km
2
portion of the Druid Peak Pack’s
territory (Crabtree & Sheldon 1999) and because observa-
tions were made in two other wolf pack territories, at least
twice as many individual coyotes could have potentially
been observed. Coyote group size varied little (XSE
coyotes= 1.60.1, range 1–3, N= 101), and coyotes were
typically solitary or associated in pairs when observed. A
total of 58.8, 28.0 and 58.0 h were observed for coyote,
elk and landscape bouts, respectively. Activity bout
lengths varied widely for each of the four bout categories
because changes in behaviour or visibility could not be
anticipated and all-occurrence sampling was employed
(Table 1).
Preferential Association with Wolves
Ravens were present in the majority of wolf activity
bouts observed away from a known carcass, whereas
ravens were absent in the majority of coyote, elk and
landscape bouts (Table 1). The proportion of time that
ravens were observed in association with the four bout
categories also varied significantly (Table 1). Ravens spent
more time in the presence of wolves than they did with
landscape points in the absence of wolves (ztest: z=10.4,
Table 1. Comparison of the bouts recorded for wolf, coyote, elk and landscape showing the percentage of bouts and minutes with ravens
present, mean bout length and mean raven presence (min)
Bout type
% Bouts with ravens
(number of bouts with ravens/N)
% Time (min) ravens were
observed in association*
X±SE
bout length (min)
X±SE
raven presence (min)
per bout type
Wolf
Travelling 89.6 (69/77) 37.7 (1003/2660) 34.6±2.9 14.5 ±1.4
Resting 81.0 (51/63) 27.5 (1362/4959) 78.7±10.6 26.7 ±4.2
Chasing prey 87.2 (34/39) 62.9 (393/625) 16.0±2.7 11.6 ±2.2
Mousing 100 (6/6) 84.7 (166/196) 32.7±9.1 15.8 ±1.5
Activities combined† 86.5 (160/185) 34.6 (2924/8440) 45.6±4.3 18.3 ±1.6
At-kill 100 (49/49) 99.7 (5148/5162) 105.4±16.3 105.1 ±16.3
Coyote
Travelling 1.7 (1/60) 0.1 (1/1827) 30.5±2.2 N/A
Resting 6.5 (2/31) 0.2 (3/1242) 40.1±3.6 1.5 ±0.5
Mousing 0 (0/13) 0 (0/456) 35.1±7.8 N/A
Activities combined† 2.9 (3/104) 0.1 (4/3525) 33.9±2.0 1.3 ±0.3
Landscape 25.7 (19/74) 1.4 (49/3479) 47.0 ±5.4 2.6±0.8
Elk 6.5 (3/46) 0.1 (2/1680) 36.5±1.0 0.7 ±0.2
*Produced by the sum of the total number of minutes ravens were present divided by the total number of minutes the category bout was
observed.
†All behaviour types excluding at-kill.
286 ANIMAL BEHAVIOUR, 64, 2
P<0.0001). Ravens spent significantly less time associat-
ing with elk, the main prey of the wolf, than with the
predators themselves (z=10.8, P<0.0001). Ravens also
showed preferential association with wolves over coyotes
away from a carcass (z=10.3, P<0.0001), even when
comparing wolf bouts of comparable group size (X
SE= 1.70.2 wolves, range 1–3, N= 23) to coyote bouts
(z=3.4, P<0.001). Wolves travelling, resting and mousing
all attracted ravens more than coyotes engaged in these
same activities (travelling: z=26.9, P<0.0001; resting:
z=19.5, P<0.0001; mousing: z= 10.1, P<0.0001).
The number of ravens in association with wolves away
from a carcass (XSE= 2.70.2 ravens, range 1–16,
N=160) was greater than the number associating with
coyotes away from a carcass (XSE= 1.30.3 ravens,
range 1–2, N=3; Student’s ttest: t
3
=3.7, P<0.05), elk
groups (XSE= 1.30.3 ravens, range 1–2, N=3; t
3
=3.7,
P<0.05), and points on the landscape (XSE= 1.60.2
ravens, range 1–4, N=19; t
59
=4.4, P<0.0001). When
present, ravens were in closer proximity to wolves
(XSE= 25.02.2 m, N=160) than to coyotes (XSE =
56.723.3 m, N=3), elk groups (XSE =116.7 44.1 m,
N=3), or points on the landscape (XSE =113.7 14.0 m,
N=19).
Association between wolves and ravens was greatest
when wolves were at a kill, with 100% (N=49) of these
bouts having ravens present nearly the entire time, but
the proportion of time that ravens were observed in
association with each of the behaviour categories varied
(Table 1). Ravens spent more time associating with
wolves when wolves were chasing prey or mousing com-
pared with when they were travelling (ztest: z=2.7,
P<0.004; z=5.0, P<0.0001) or resting (z= 3.7, P<0.0001;
z=6.0, P<0.0001). Ravens appeared to spend more time
associating with wolves when wolves were travelling than
when wolves were resting, although the difference did
not reach significance following Bonferroni adjustment
(z=1.6, P= 0.05).
Discovery Success and Behavioural Response to
Wolf-killed and Nonwolf-killed Carcasses
Twenty-nine prey-chasing bouts that ended in the
successful predation of an ungulate species (28 elk, one
bison, Bison bison) were recorded. Discovery success of
kills by ravens within the 60-min observation period
following the time of death of the prey animal was 100%
(N=29). Wolf-killed carcasses were discovered almost
immediately because some ravens were typically follow-
ing the wolves prior to the kill. In 24 of the 29 kills
(82.8%), ravens (1–13) were present during the chase,
within 5–100 m of the wolves, hovering above or perched
on boulders nearby, whereas the other five kills were
discovered within 4 min of the time of death (Fig. 1;
overall XSE= 0.50.2 min, median=0 min, range 0–4,
N=29).
In all wolf kills, arriving ravens responded by landing
on the ground or perching on nearby boulders in close
proximity to the kill and feeding wolves. In the majority
of cases, ravens would approach to within 1 m of the
feeding wolves to take advantage of meat scraps made
available during the evisceration process or peck at blood
in the snow. In one kill, one raven landed on the prey
animal as it was dying and remained while a wolf began
evisceration. The number of ravens increased steadily
throughout the first 60 min after the time of death
(Fig. 2).
Ravens discovered wolf-killed carcasses with greater
success than they did nonwolf-killed carcasses within
60 min of the time of initial carcass presence on the
landscape (z=5.2, P<0.0001). With the experimentally
placed carcasses, only nine out of 25 (36%) were dis-
covered within the 60-min observation period following
carcass placement. Ravens did not immediately discover
these nine experimental carcasses after their initial pres-
ence (Fig. 1;XSE discovery time=36.33.7 min, range
17–52, N=9). The initial number of ravens discovering
experimental carcasses was either one or two, which was
less than initial group size discovering wolf kills (XSE
experimental carcasses= 1.40.2 ravens, N=9; XSE
0
45
Discovery time (min)
40
35
30
25
20
15
10
5
Wolf-killed carcasses Experimental carcasses
Figure 1. Discovery time of wolf-killed (N=29) and experimental
(N=9) carcasses by ravens. These values represent the arrival time
(X±SE) in min of the first raven(s) to approach to within 250 m of
the carcasses after the time of prey death (wolf-killed carcass) or
placement of carcass (experimental carcass).
0
18
Number of ravens at carcass
16
14
12
10
8
6
4
2
Time of
death/carcass
placement
15 min 30 min 45 min 60 min
Wolf kill
Experimental carcass
Figure 2. Number of ravens (X±SE) at wolf-killed (N=29) and
experimental (N=9) carcasses at 15-min intervals after initial
presence on the landscape for the first 60 min of carcasses’ presence.
No ravens stayed to feed at experimental carcasses.
287STAHLER ET AL.: RAVENS FOLLOWING WOLVES TO FORAGE
wolf kills= 3.00.5 ravens, N= 29; t
34
=3.03, P<0.01). In
all of the discovered experimental carcasses, arriving
ravens circled above the carcass once or twice before
leaving the area for the remainder of the observation
period. In contrast to the response towards wolf kills,
none of the ravens landed to feed at experimental
carcasses (Fig. 2). Magpies discovering these carcasses (in
four trials), however, did immediately feed from them.
In determining the number of ravens attending wolf-
killed carcasses, we combined data from the 29 observed
wolf predation events with data on other wolf-killed
carcasses that were discovered after the kill was made. The
number of individual ravens attending wolf-killed
carcasses at one time ranged between three and 135
(XSE= 28.62.1 ravens, N= 98), but because ravens
were unmarked and frequently coming and going to
cache food, these values probably underestimate the
actual number of ravens attending wolf kills during our
study.
DISCUSSION
In YNP during winter, ungulate carcasses represent a
spatially and temporally unpredictable and ephemeral
food source for ravens, whether they are the result of
predator-caused mortality or other causes. Ravens are the
most common and numerous vertebrate scavengers using
these carcasses in the winter (Yellowstone Wolf Project,
National Park Service, unpublished data), indicating
that this resource is an important food source for them.
Inter- and intraspecific competition at ungulate carcasses
in YNP is high, so foraging strategies employed by
scavengers that facilitate discovery of a carcass soon
after its initial presence on the landscape presumably
lead to greater foraging efficiency and resource intake.
Social interaction between ravens and wolves in YNP
demonstrates an important scavenger foraging strategy.
Ravens in winter in YNP showed routine daily associ-
ation with wolves. We observed ravens in close proximity
to wolves in the majority of wolf bouts and this presence
ranged from 27.5% to 99.7% of the observed minutes for
any given behaviour type, serving as estimates for the
degree of association in the winter. Ravens preferentially
associated with wolves compared to coyotes and elk, and
this association could not be explained by ubiquitous
raven presence on the landscape.
The infrequent association with coyotes suggests that
ravens discriminate between wolves and coyotes based on
the differential abilities of these two canids to kill larger
prey and thus provide scavenging opportunities for them.
Predation on ungulates by coyotes does not occur fre-
quently enough in YNP to provide the scavenging com-
munity with carcasses on a regular basis (Gese et al.
1996), although coyotes may benefit ravens by opening
nonwolf-killed carcasses. Wolf group size was larger on
average than coyote group size, and the activity of a larger
canid group size could, arguably, have been more attrac-
tive to ravens in the area. Our comparison of similar wolf
and coyote group sizes (1–3 individuals), however, sug-
gests that ravens are able to distinguish between these
two canids. When ravens were present during coyote
bouts, they flew overhead and then left, suggesting that
ravens can make quick assessments on canid identity
and/or the potential for food.
The lack of significant raven association with elk
groups suggested that ravens are not following elk on the
northern range of YNP. Thus, it is probably not advan-
tageous for ravens to follow elk around waiting for one to
die. We have observed on several occasions, however,
ravens being quick to locate and harass injured elk,
apparently drawing the attention of wolves and coyotes
through local enhancement.
The almost immediate presence of ravens at wolf-killed
carcasses does not itself distinguish whether ravens follow
wolves or are attracted to a kill by the activity of the chase
when they see it from a distance. However, the frequent
presence of ravens with wolves in the absence of a carcass
supported the hypothesis that some ravens were follow-
ing wolves throughout the day. On days where wolves
were in continuous visible range, we frequently observed
ravens following wolves throughout continuous activity
that changed from resting to travelling to chasing prey,
which sometimes led to the wolves making a kill.
The differences in the degree of raven association when
wolves were resting, travelling, chasing prey, mousing or
at a kill coincided with the differential potential for each
of these behaviours to provide food for foraging ravens.
The trend for ravens to spend more time associating with
travelling wolves than resting wolves probably reflects
the fact that resting wolves are less likely to kill a prey
animal, whereas travelling wolves could potentially
encounter prey and have the opportunity to kill. Ravens
were present for the majority of minutes we observed
wolves chasing prey, presumably because this activity
held the greatest potential for foraging success. When
wolves were mousing, ravens were on the ground
hopping behind the wolves, sometimes less than 1 m
away, whereas no ravens were ever seen in the presence of
a mousing coyote. Small mammals are an important part
of a coyote’s diet during winter in YNP (Gese et al. 1996)
and any rodent captured is immediately consumed. In
contrast, mousing wolves appeared less intent on con-
suming captured rodents than playing with them, giving
associating ravens a better opportunity to steal from a
wolf (as occurred in one occasion). Our results suggest
that ravens may be making decisions on how much time
to spend following wolves based on their activity type
and the potential for food acquisition.
The discovery of wolf-killed carcasses by ravens led to
greater opportunity for them to obtain meat compared
with nonwolf-killed carcasses. Not only did it take longer
for ravens to locate nonwolf-killed carcasses, but they also
did not utilize the carcasses upon discovery or within the
60-min observation time, despite the availability of red
meat. In contrast, ravens discovering wolf kills attempted
to obtain meat soon after carcass discovery, often moving
to within 1 m of feeding wolves. Apparent fear response
to large carcasses by ravens (neophobia) has been shown
elsewhere and is characterized by a cautious approach to
a carcass and retreating without feeding (Heinrich 1988c).
Fear of large carcasses upon first encounter is an innate
response (Heinrich 1988c;Heinrich et al. 1995) that is
288 ANIMAL BEHAVIOUR, 64, 2
reduced through experience (Heinrich et al. 1995)or
social facilitation (Marzluff & Heinrich 1991). Our results
suggest neophobia in ravens upon discovering carcasses
unattended by wolves, whereas discovery of wolf-killed
carcasses attended by wolves resulted in an apparent
suppression of this fear response. Thus, wolves appeared
to be a primary stimulus for ravens to start feeding.
Heinrich (1999b) found similar feeding responses by
ravens in Nova Scotia, Canada, at carcasses in the pres-
ence and absence of captive wolves in a fenced enclosure.
Wolves defend carcasses against ravens by chasing
them, and dead ravens are occasionally found near
wolves and their kills (Peterson 1977;Allen 1979; J. Ryan,
personal communication; T. Brooks, personal communi-
cation), suggesting that feeding next to wolves is not risk
free. Therefore, if foraging next to wolves is strictly a
reflection of ravens’ opportunism to feed on fresh
meat, then they should prefer to feed on meat that is
unattended by wolves, or feed equally at meat attended
and unattended by wolves. Our findings of ravens prefer-
entially feeding with wolves at carcasses suggest that the
ravens received socially facilitated benefits, such as
reduced neophobia, which outweighed potential risks
associated with feeding next to wolves. Frequent behav-
ioural interactions between these two species were
observed at and away from kill sites, such as ravens
pulling wolves’ tails, ravens interacting with wolf pups at
den sites, and playful chasing between them (Stahler
2000). Such interactions may serve to ‘educate’ ravens on
the responses and intent of potentially dangerous, large
carnivores, as has been suggested (Heinrich 1989,1999a),
ultimately benefiting ravens when feeding among wolves
at carcasses.
We could not determine whether the same ravens were
following wolves on different days, or the age and social
status of those ravens. Both adult territorial ravens with
known nests and nonterritorial foraging groups were
found in wolf pack territories. We also found ravens
travelling between wolf kills belonging to two different
packs, showing that some ravens did not just associate
with one pack, and thus probably monitor the activities
of several packs when foraging. The fact that we observed
different ravens in a variety of group sizes (1–16) associ-
ating with different wolf packs simultaneously suggests
that more than just a few members of the raven popu-
lation have adopted this foraging strategy. Our results
may also apply to the association between ravens and
wolves seen in other ecosystems (Mech 1966;Peterson
1977;Bjarvall & Isakson 1982;Promberger 1992;Carbyn
et al. 1993).
Considerably more ravens showed up to feed at wolf
kills compared with the number of ravens associating
with wolves in the absence of food. We believe this rapid
arrival of additional ravens was due to the enhancement
of kill-site locations from the activity of birds already
present. Local enhancement, or the attraction to individ-
uals already feeding (Thorpe 1956;Turner 1964), is an
important component of the foraging strategy of species
that use spatially and temporally unpredictable foods
in open habitats (Schaller 1972;Houston 1979;Po¨ysa¨
1992). In YNP, local enhancement may be influential in
directing the attention of other ravens to a kill site on the
landscape, and subsequently serving to attract other
kleptoparasites such as coyotes, magpies and eagles. Non-
breeding juvenile ravens may also be actively recruiting
other vagrant nonbreeders and overpowering territorial
adults at carcasses, as has been seen elsewhere (Heinrich
1988b;Marzluff & Heinrich 1991), although we did
not record recruitment vocalizations (Heinrich 1988b)
because our distance to carcasses was frequently out of
audible range. Larger raven aggregations at carcasses may
also distract wolves and other scavenger species, which
may increase food intake for individual ravens present
(Marzluff & Heinrich 1991).
Our results showing ravens following a nonconspecific
forager to gain access to food is similar to the symbiotic
relationship of the African ratel, Mellivora capensis,
and human honey hunters following the honeyguide,
Indicator indicator (Isack & Reyer 1989), coyotes following
wolves (Paquet 1991), hyaenas, Crocuta crocuta, following
wild dogs, Lycaon pictus, lions, Panthera leo and cheetahs,
Acinonyx jubatus (Kruuk 1972;Cooper 1991;Caro 1994),
and birds following monkeys (Terborgh 1983;Zhang &
Wang 2000). As with all scavenger–predator relation-
ships, however, the raven–wolf association reported here
demonstrates a kleptoparasitic form of social symbiosis,
in that ravens are the primary beneficiaries by stealing
food that would otherwise benefit the wolves.
It is significant that ravens associated strongly with
wolves in YNP immediately following the 1995–1996
reintroduction of this predator after a 70-year absence in
the ecosystem (Bangs & Fritts 1996). Although our data
did not distinguish the behavioural mechanisms under-
lying ravens’ preferential association with wolves, we
believe that both innate and learned behavioural
responses towards wolves are involved; suggesting that
the raven–wolf relationship is an ancient evolved one.
Regardless of the mechanism, our results are the first to
show that ravens actively seek wolves’ company to find
and gain access to large carcasses and to overcome their
apparent fear of them. In addition, we show that some
ravens follow wolves despite the concurrent absence
of food, demonstrating a kleptoparasitic foraging
strategy for scavenger species dependent on spatially and
temporally unpredictable food sources.
Acknowledgments
This study was supported by Yellowstone National Park,
Canon U.S.A. and the University of Vermont. We thank
the many Yellowstone Wolf Project volunteers and tech-
nicians who assisted with this study. We especially
thank K. Murphy, T. Drummer, S. Evans, D. Guernsey, S.
Hudman, C. Wilmers and J. Varley for field and logistical
support on this study and input on the manuscript. The
research presented here was evaluated and approved by
the Animal Behavior Society’s Animal Care Committee
on 21 February 2002.
References
Allen, D. 1979. Wolves of Minong: Their Vital Role in a Wild Commu-
nity. Boston: Houghton Mifflin.
289STAHLER ET AL.: RAVENS FOLLOWING WOLVES TO FORAGE
Altmann, J. 1974. Observational study of behaviour: sampling
methods. Behaviour,49, 227–267.
Bangs, E. & Fritts, S. 1996. Reintroduction of the gray wolf to
central Idaho and Yellowstone National Park. Wildlife Society
Bulletin,24, 402–413.
Bent, A. C. 1946. Life histories of North American Jays, Crows and
Titmice, Part 1. U.S. Natural Museum, Bulletin no. 191. New York:
Dover.
Bjarvall, A. & Isakson, E. 1982. Winter ecology of a pack of three
wolves in northern Sweden. In: Wolves of the World: Perspectives of
Behavior, Ecology, and Conservation (Ed. by F. Harrington & P.
Paquet), pp. 146–157. Park Ridge, New Jersey: Noyes.
Brockmann, H. J. & Barnard, C. J. 1979. Kleptoparasitism in birds.
Animal Behaviour,27, 487–514.
Carbone, C., Du Toit, J. T. & Gordon, I. J. 1997. Feeding success
in African wild dogs: does kleptoparasitism by spotted hyenas
influence hunting group size? Journal of Animal Ecology,66,
318–326.
Carbyn, L. N., Oosenbrug, S. M. & Anions, D. W. 1993. Wolves,
Bison, and the Dynamics Related to the Peace-Athabascan Delta in
Canada’s Wood Buffalo National Park. Edmonton: University of
Alberta.
Caro, T. M. 1994. Cheetahs of the Serengeti Plains: Group Living in an
Asocial Species. Chicago: University of Chicago Press.
Clark, C. W. & Mangel, M. 1984. Foraging and flocking strategies:
information in an uncertain environment. American Naturalist,
123, 626–641.
Clark, C. W. & Mangel, M. 1986. The evolutionary advantages of
group foraging. Theoretical Population Biology,30, 45–75.
Cooper, S. M. 1991. Optimal group size: the need for lions to
defend their kills against loss to spotted hyaena. African Journal of
Ecology,29, 130–136.
Crabtree, R. L. & Sheldon, J. W. 1999. The ecological role of
coyotes on Yellowstone’s northern range. Yellowstone Science,7,
15–23.
Creel, S. & Creel, N. M. 1996. Limitation of African wild dogs
by competition with larger carnivores. Conservation Biology,10,
526–538.
Fanshawe, J. H. & FitzGibbon, C. D. 1993. Factors influencing the
hunting success of an African wild dog hunting group. Animal
Behaviour,45, 479–490.
Gasaway, W. C., Mossestad, K. T. & Stander, P. E. 1991. Food
acquisition by spotted hyenas in Etosha National Park, Namibia:
predation versus scavenging. African Journal of Ecology,29, 64–75.
Gese, E. M., Ruff, R. L. & Crabtree, R. L. 1996. Foraging ecology of
coyotes (Canis latrans): the influence of extrinsic factors and a
dominance hierarchy. Canadian Journal of Zoology,74, 769–783.
Harrington, F. H. 1978. Ravens attracted to wolf howling. Condor,
80, 236–237.
Heinrich, B. 1988a. Foodsharing in the raven, Corvus corax.In:
The Ecology of Social Behavior (Ed. by C. N. Slobodchickoff),
pp. 285–311. New York: Academic Press.
Heinrich, B. 1988b. Winter foraging at carcasses by three sympatric
corvids with emphasis on recruitment by the raven, Corvus corax.
Behavioral Ecology and Sociobiology,23, 141–156.
Heinrich, B. 1988c. Why do ravens fear their food? Condor,90,
950–952.
Heinrich, B. 1989. Ravens in Winter. New York: Simon & Schuster.
Heinrich, B. 1999a. Mind of the Raven. New York: Cliff Street Books,
Harper Collins.
Heinrich, B. 1999b. Sociobiology of ravens: conflict and co-
operation. Sitzungsberichte der Gesellschaft Naturforschender
Freunde zu Berlin,37, 13–22.
Heinrich, B. & Marzluff, J. 1991. Do common ravens yell because
they want to attract others? Behavioral Ecology and Sociobiology,
28, 13–21.
Heinrich, B., Marzluff, J. M. & Adams, W. 1995. Fear and food
recognition in nai¨ve common ravens. Auk,112, 499–503.
Houston, D. B. 1982. The Northern Yellowstone Elk. New York:
Macmillan.
Houston, D. C. 1979. The adaptations of scavengers. In:
Serengeti: Dynamics of an Ecosystem (Ed. by A. R. E. Sinclair & M.
Norton-Griffiths), pp. 263–286. Chicago: University of Chicago
Press.
Isack, H. A. & Reyer, H. U. 1989. Honeyguides and honeygatherers:
interspecific communication in a symbiotic relationship. Science,
243, 1343–1346.
Kruuk, H. 1972. The Spotted Hyaena: a Study of Predation and Social
Behavior. Chicago: University of Chicago Press.
Lemke, T. O., Mack, J. A. & Houston, D. B. 1998. Winter range
expansion by the northern Yellowstone elk herd. Intermountain
Journal of Sciences,4, 1–9.
Marzluff, J. M. & Heinrich, B. 1991. Foraging by common ravens in
the presence and absence of territory holders: an experimental
analysis of social foraging. Animal Behaviour,42, 755–770.
Marzluff, J. M., Heinrich, B. & Marzluff, C. S. 1996. Raven roosts
are mobile information centres. Animal Behaviour,51, 89–103.
Mech, L. D. 1966. The Wolves of Isle Royale. Washington, DC: U.S.
National Park Service, U.S. Government Printing Office.
Mech, L. D. 1970. The Wolf: the Ecology and Behavior of an
Endangered Species. Minneapolis: University of Minnesota Press.
Ott, L. R. 1993. An Introduction to Statistical Methods and Data
Analysis. Pacific Grove, California: Duxbury Press.
Paquet, P. C. 1991. Winter spatial relationships of wolves and
coyotes in Riding Mountain National Park, Manitoba. Journal of
Mammology,72, 397–401.
Peterson, R. O. 1977. Wolf Ecology and Prey Relationships on Isle
Royale. U.S. National Park Service Scientific Monograph Series.
Washington, D.C.: U.S. Government Printing Office.
Po¨ysa¨, H. 1992. Group foraging in patchy environments: the
importance of coarse-level local enhancement. Ornis Scandinavica,
23, 159–166.
Promberger, C. 1992. Woelfe und scavenger (wolves and
scavengers). Diplomarbeit, Ludwig Maximillians Universitaet,
Muenchen.
Pulliam, H. R. & Caraco, T. 1984. Living in groups: is there an
optimal group size? In: Behavioural Ecology: an Evolutionary
Approach (Ed. by J. R. Krebs & N. B. Davies), pp. 122–147. Oxford:
Blackwell Scientific.
Schaller, G. B. 1972. The Serengeti Lion. Chicago: University of
Chicago Press.
Scheaffer, R. L., Mendenhall, III, W. M. & Ott, R. L. 1996.
Elementary Survey Sampling. Pacific Grove, California: Duxbury
Press.
Stahler, D. R. 2000. Interspecific interactions between the common
raven (Corvus corax) and the gray wolf (Canis lupus) in Yellowstone
National Park, Wyoming: investigations of a predator and
scavenger relationship. M.S. thesis, University of Vermont.
Sokal, R. R. & Rohlf, F. J. 1995. Biometry. New York: W. H. Freeman.
Terborgh, J. W. 1983. Five New World Primates. Princeton, New
Jersey: Princeton University Press.
Thorpe, W. H. 1956. Learning and Instinct in Animals. London:
Methuen.
Turner, E. R. A. 1964. Social feeding in birds. Behaviour,24, 1–46.
Zhang, S. & Wang, L. 2000. Following of brown capuchin monkeys
by white hawks in French Guiana. Condor,102, 198–201.
290 ANIMAL BEHAVIOUR, 64, 2
