Vigilance and fright behaviour in the insular Svalbard reindeer ( Rangifer tarandus platyrhynchus )

Vigilance and fright behaviour in the insularSvalbard reindeer (Rangifer tarandusplatyrhynchus)

Eigil Reimers, Steinar Lund, and Torbjørn Ergon

Abstract: The insular Svalbard reindeer (Rangifer tarandus platyrhynchus Vrolik, 1829) provide an opportunity to studyvigilance behaviour in the absence of predators and parasitizing insects. We measured vigilance and fright and flight re-sponse during summer 2006 in the Svalbard archipelago; in four areas on Nordenskiöld Land on the island Spitzbergen andin one area on the island Edgeøya. Vigilance was higher in reindeer on Edgeøya than in the four Spitzbergen areas. Maleswere less vigilant than lactating and barren females and vigilance decreased with increasing group size. The relaxed vigi-lance behaviour in Svalbard reindeer compared with wild reindeer in southern Norway demonstrates a vigilance threshold inthe absence of traditional predators of Rangifer Hamilton Smith, 1827. Alert, flight initiation, and escape distances were allshorter in Adventdalen, with Longyearbyen and its considerably higher amounts of human activities and infrastructure thanin the other study areas, supporting evidence of habituation towards humans. There were no systematic vigilance or differen-ces in fright and flight responses between reindeer in Colesdalen, Reindalen, and Sassendalen, indicating that a combinationof low level of human activities including hunting, recreation, and scientific activities affected the animals differently. Lowerprobability of assessing before fleeing in Edgeøya (63% vs. 94% in the Nordenskiöld Land areas), along with their highervigilance, may indicate more frequent interactions with polar bears (Ursus maritimus Phipps, 1774) in Edgeøya.

Résumé : Les rennes insulaires de Svalbard (Rangifer tarandus platyrhynchus Vrolik, 1829) représentent une occasiond’étudier le comportement de vigilance en l’absence de prédateurs et d’insectes parasites. Nous avons mesuré la vigilance etles réactions de peur et de fuite à l’été 2006 dans l’archipel de Svalbard, soit quatre sites sur la terre de Nordenskiöld surl’île de Spitzbergen et un site sur l’île d’Edgeøya. La vigilance est plus forte sur l’île d’Edgeøya que dans les quatre sites deSpitzbergen. Les mâles sont moins vigilants que les femelles nourricières et les femelles sans petit et la vigilance diminue àmesure que la taille du groupe augmente. Le comportement de vigilance plus relâché chez les rennes de Svalbard par com-paraison à celui des rennes sauvages du sud de la Norvège démontre l’existence d’un seuil de vigilance en l’absence desprédateurs traditionnels des Rangifer Hamilton Smith, 1827. L’alerte et l’initiation de la fuite se produisent à des distancesplus courtes et les distances de dérobade sont moins grandes à Adventdalen; à Longyearbyen où il y a considérablementplus d’activités et d’infrastructures humaines que dans les autres sites d’étude, il existe des indications d’une habituationaux humains. Il n’y a pas de différence systématique dans les réactions de peur et de fuite entre les rennes de Colesdalen,de Reindalen et de Sassendalen, ce qui indique qu’une combinaison d’activités humaines de basse intensité, incluant de lachasse et des activités récréatives et scientifiques, affectent les animaux différemment. Une probabilité plus faible d’évalua-tion avant la fuite sur Edgeøya (63 % par rapport à 94 % dans les sites de la terre de Nordenskiöld) et une plus grande vigi-lance peuvent refléter des interactions plus fréquentes avec les ours polaires (Ursus maritimus Phipps, 1774) à Edgeøya.

[Traduit par la Rédaction]

Introduction

Predation is recognized as a major selective force in theevolution of behavioural characteristics such as alertness andvigilance in mammals (Elgar 1989). Foraging theory predictsthat animals may sacrifice feeding time to reduce the risk ofpredation (Lima 1995; Brown 1999). Conversely, isolationfrom predators should result in a selection against costly anti-predator behaviour (Magurran 1999; Blumstein and Daniel2005). Vigilance is a good indication of an animal's risk ofpredation and varies accordingly with body size, reproductive

status, group size, and habitat characteristics (Mooring et al.2004). Vigilance is generally defined as time spent with thehead raised during feeding periods (Whittingham et al. 2004;Lung and Childress 2007) and plays a major role in the de-tection of predators (Hopewell et al. 2005). This antipredatorbehaviour is widely used (Elgar 1989), potentially increasingfitness by decreasing the risk of mortality (Lima 1998; Wat-son et al. 2007). Although head-up periods may serve addi-tional functions, such as scanning for conspecifics (Lung andChildress 2007) and handling food items (Illius and Fitzgib-bon 1994), all models of antipredator vigilance (Pulliam et al.

Received 2 November 2010. Accepted 11 March 2011. Published at www.nrcresearchpress.com/cjz on 5 August 2011.

E. Reimers and T. Ergon. Department of Biology, University of Oslo, P.O. Box 1066 Blindern, 0316 Oslo, Norway.S. Lund. Department of Ecology and Natural Resources Management, Norwegian University of Life Sciences, 1432 Ås, Norway.

Corresponding author: Eigil Reimers (e-mail: [email protected]).

753

Can. J. Zool. 89: 753–764 (2011) doi:10.1139/Z11-040 Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.

1982; Lima 1987; Packer and Abrams 1990) assume that dur-ing these head-up periods, predators will be detected morequickly than when the head is lowered. This has been sup-ported by empirical evidence (Lima and Bednekoff 1999).When isolated from predators, costly and no longer func-tional antipredator behaviour should be eliminated by selec-tion (Kavaliers 1990; Magurran 1999). A consequence ofisolation from predators is that prey lose or change their anti-predator behaviour (Byers 1997). Two studies (Beauchamp2004; Blumstein and Daniel 2005) demonstrated that isola-tion on islands with negligible predation risk was responsiblefor the loss of antipredator behaviour.The insular Svalbard reindeer (Rangifer tarandus platy-

rhynchus Vrolik, 1829) is the northernmost population ofRangifer Hamilton Smith, 1827 inhabiting an environmentwithout parasitizing insects, and except for a few observa-tions of predation by polar bears (Ursus maritimus Phipps,1774) (Derocher et al. 2000), no predators (other than man)that are elsewhere are part of their natural habitat. This situa-tion has prevailed for at least 4000 years (Van der Knaap1986; Tyler and Øritsland 1989). From where the Svalbardreindeer originate or for how long they have been isolatedfrom predators or other Rangifer populations is unknown,but is likely several thousands of years. Contrary to Rangifersubspecies elsewhere, Svalbard reindeer live individually orin small groups (Alendal and Byrkjedal 1976; Alendal et al.1979), are seasonally sedentary (Tyler and Øritsland 1989),and do not have the nomadic behaviour known from otherRangifer subspecies. Survival of Svalbard reindeer is basedupon a delicate balance between maximum energy intake interms of optimal grazing and minimal energy output (Reim-ers 1980). This results in large pre-winter fat reserves (Reim-ers et al. 1982) that enable females and young animals tocounteract low-quality winter forage (Staaland and White 1991).The objective of our study in 1996 (Colman et al. 2001)

was to examine response distances of Svalbard reindeer to di-rect provocation by humans on foot during summer in areassubject to combinations of high or low activity and huntingor no hunting. The objective of the present study was to ex-amine the vigilance behaviour of Svalbard reindeer relatingto the absence of predators, and to determine if their frightand flight distances have changed in response to increasedhuman activities during the last 12 years. We made the fol-lowing predictions:

1. Number of vigilance bouts per minute and time spent vig-ilant while grazing should be low compared with otherRangifer subspecies and (or) other ungulates subject topredation.

2. If there were area differences in vigilance, we expectedhigher vigilance in areas where hunting is allowed andin areas with a high number of human–reindeer interac-tions, and low vigilance in the remote and isolated is-land Edgeøya.

3. Based on a previous study (Colman et al. 2001), we ex-pected fright and flight response distances to be shortestin Adventdalen with extensive human activities relatedto the main city Longyearbyen, greater in the areasopen for hunting and scientific activities involving livecapturing of reindeer, and greatest in the remote islandEdgeøya.

Materials and methods

Study areasSvalbard (63 000 km2), a group of islands with Spitsbergen

as the largest, is located in the western Barents Sea, between74°N and 81°N (Fig. 1). Despite the high latitude, the cli-mate is relatively mild owing to the North Atlantic Current(mean yearly temperature: –6.7 °C; mean precipitation:190 mm; from Longyearbyen Airport between 1975 and1990: http://www.yr.no). The landscape is mountainous withpeaks up to 1700 m. Large areas are covered by glaciers, andsummer pastures for reindeer are restricted to valleys, coastalplains, and plateaus. The flora in Svalbard is classified asMid and High Arctic (Rønning 1985). Summer pasture inmost places is limited to below 100–150 m above sea leveland to the plateaus at 400 m above sea level.Four of the five study areas are located on Nordenskiöld

Land on Spitsbergen (39 000 km2). Adventdalen (150 km2),Colesdalen (94 km2), Reindalen (361 km2), and Sassendalen(193 km2) are wide U-shaped valleys surrounded by steepmountains of about 300–800 m above sea level (Fig. 1). Thefifth area, Plurdalen-Grunnlinjesletta (120 km2), is located onthe west coast of Edgeøya (ca. 5 150 km2) in the eastern partof the archipelago (Fig. 1).Helicopter surveys conducted by the Governor of Svalbard

yearly from 1997 to 2005 show that the summer subpopula-tions of Svalbard reindeer are approximately 380 in Coles-dalen, 510 in Reindalen, and 830 in Sassendalen. Thepopulation in Adventdalen summer 2006 was about 800 ani-mals (Tyler et al. 2008). The total summer population ofreindeer in Plurdalen-Grunnlinjesletta including the plateauswas estimated at 227 animals in 1977 (Reimers 1982) and181 animals in 2006 (E. Reimers, unpublished data).

HuntingHunting of reindeer has occurred in Svalbard since the

17th century, but the extent is presumed to be low until1860, when human presence on Svalbard increased owing totrapping and mining activities. Between 1865 and 1925, thedocumented harvest is 20 000 animals, but as much as twicethis number might have been killed (see references in Tyler1987). As a result of reports of rapidly decreasing densitiesof Svalbard reindeer, and locally extinct populations on Nor-denskiöld Land, hunting of Svalbard reindeer was banned in1925 when Norway was assigned sovereignty of Svalbard.Since then, reindeer densities have increased, which led tothe reopening of limited hunting in Colesdalen, Reindalen,and Sassendalen in 1983 (Table 1). Hunting intensity hasbeen low, with a yearly take of between 117 and 231 reindeer(hunting statistics from the Governor of Svalbard).

TourismTourism is the single largest human activity in the Arctic

and it is fast growing (Snyder et al. 2007). On Svalbard, tour-ism is one of the leading industries, together with coal min-ing and scientific research. Longyearbyen (2000 inhabitants)is the starting point for most of the tourist activity for all ofSvalbard, and levels of human activities are thus highest inAdventdalen. Numbers of registered guest days in hotels andguest houses in Longyearbyen increased from 30 000 in 1994to almost 78 000 in 2004 (Sysselmannen 2006). Presently,

754 Can. J. Zool. Vol. 89, 2011

Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.

there are several tourist companies offering a variety of activ-ities. Hiking trips from Longyearbyen and within Advent-dalen, as well as into tributary valleys, are important touristactivities during summer (Gill et al. 2001; Sysselmannen2006). Neither locals nor nonresident visitors are instructedto notify the Governor of Svalbard when travelling in Nor-denskiöld Land. Colman et al. (2001) used data of tourist ac-tivity from Kaltenborn (1991) to classify the study areasaccording to human activity (Table 1). Adventdalen is by farthe easiest accessible area. Because there is no network oftourist cabins in Nordenskiöld Land, hiking activities inColesdalen, and specifically in Reindalen and Sassendalen,

are limited. Edgeøya is a nature reserve and public visitingis generally prohibited.

Scientific activityIn a long-term research program in two of the study areas,

Colesdalen and Reindalen (Table 1), reindeer have been cap-tured and collared from snowmobiles in March–April eachyear since 1995 (Omsjoe et al. 2009).

Sampling procedureWe collected data on vigilance and flight responses during

3 weeks in July and August 2006. When spotting an animal

Fig. 1. The Svalbard study areas Adventdalen, Colesdalen, Reindalen, and Sassendalen in Nordenskiöld Land on Spitzbergen and on theisland Edgeøya in July and August 2006.

Table 1. Ranking of human activities, scientific activities, and hunting pressure on Svalbard reindeer (Rangifer tarandus platyrhynchus) inthe study areas in Svalbard in 2006.

Variable Adventdalen Colesdalen Reindalen Sassendalen EdgeøyaDistance to closest settlement by foot/boat (km)* 0/0 30/28 39/— 45/40 —/200Number of cabins† >200 12 4 3 0Human activity High Medium Medium–low Low NoneSeasons reindeer captured during scientific work since 1994‡ 0 12 12 0 0Scientific activity None High High None NoneHunting seasons since 1983§ 0 21 23 23 0Minimum/maximum animals hunted since 1983§ 0/0 17/54 10/49 33/95 0/0Hunting rank order None Low Low Medium None

*Colman et al. (2001).†The Governor of Svalbard and the mining company Store Norske, unpublished data.‡Omsjoe et al. (2009).§The Governor of Svalbard, unpublished data.

Reimers et al. 755

Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.

or group of animals, we video-recorded the animal(s) for latermeasurement of vigilance. Following the video recording, weencountered the animal(s) and measured their response dis-tances.

VigilanceWe video-recorded grazing animals from a hidden position

80–800 m away and defined grazing as the act of ingestingforage with the muzzle down. Each individual was recordedpreferably for 10 min (in some cases longer than 10 min) oruntil the reindeer laid down or moved out of sight. We de-fined a vigilance bout as the act of interrupting feeding tolift the head above the shoulders (Frid 1997) and observe thesurroundings for ≤10 s before returning to feeding (Bøvingand Post 1997). Bouts lasting >10 s were recorded but notincluded as vigilance because they frequently ended up inother behavioural activities. While filming, we registeredgroup size (number of individuals <50 m apart), group struc-ture (males, females and yearlings, mixed (all ages and bothsexes), or females with calves at heel), sex, and age of theanimal filmed (lactating female, female with calf at heel, bar-ren female (≥1 year old), male (≥1 year old), or calf), as wellas wind speed following the Beaufort wind scale(calm, <1 m/s; light or gentle breeze, 1.6–5.4 m/s; moderateor fresh breeze, 5.5–10.7 m/s; or gale, 10.8–17.1 m/s), andweather (sunny or partly sunny; cloudy; rain or snow; or foggy).We played back the videotapes on a 27-inch (1 inch =

25.4 mm) plasma TV monitor following individual reindeergrazing throughout the observation period and registered (allregistrations done by the lead author) the number and dura-tion of vigilance bouts. Slow walking between vegetation hotspots with head down was included in total grazing time.Slow walking with head up was excluded from total grazingtime.

Behavioural responsesA single person dressed in a dark hiking outfit (the ob-

server) encountered reindeer by directly approaching themand measured response distances between the reindeer andthe observer and the resultant displacement distance by thereindeer after taking flight using laser mononoculars LeicaRangemaster 1200 Scan; 1 m accuracy at 1000–1200 m. Be-fore an encounter, we registered the following four additionalparameters: activity of the animal or group prior to encounter(lying; grazing; moving; lying or grazing; or standing), ter-rain (level or rugged), wind direction relative to the observer(into the wind (including crossways to the wind) or tailwind), and terrain relative to the observer (downhill, flat, oruphill). The observer then measured the distance to the group(encounter distance). The observer used a “direct approachmethod” that had an “interrupted pattern”: advancing directlytowards the animal or estimated centre of the group at a con-stant speed (ca. 4 km/h) with ≤6 s stops to measure the threeadditional response distances (sight, alert, and flight initiationdistance) defined below. The observer continued to approachthe group on all occasions until reaching the position wherethe reindeer were located at the start of the encounter. Allmeasurements were made from the position of the directlyapproaching observer to the closest animal in the group.We used terminology and methodology associated with

wildlife response distance recommended by Taylor and

Knight (2003) and modified for our study by Reimers et al.(2009):

1. Encounter distance: the distance between the observer andthe individual reindeer (or the closest animal if we en-countered a group).

2. Sight distance: the distance between the observer and theclosest animal when animals in the group displayed analerted behaviour directed at the observer.

3. Alert distance: the distance when the reindeer exhibited anincreased alert response by grouping together or indivi-duals urinating with one hind leg extended outward atan exaggerated angle while staring at the directly ap-proaching observer.

4. Flight initiation distance: the distance from the directlyapproaching observer to the reindeer when the reindeerinitially took flight.

5. Escape distance: the shortest straight-line distance fromwhere the reindeer took flight in response to the directlyapproaching observer to where the reindeer resumedgrazing or bedded down. Greater escape distances(>1000 m) were measured in metres from maps.

Statistical analyses

VigilanceTo assess variation in the frequencies of vigilance bouts

among the five areas, we fitted the observed vigilance datato a Poisson model. Assuming that the expected number ofvigilance bouts observed for an individual was proportionalto the time it was observed, we included ln(time observed)as a fixed offset in the log-linear model for the number ofvigilance bouts. The random variation among days was in-cluded as a variance component in the model to account forweather effects and to facilitate robust inferences about thedifferences among the areas, as the areas were not surveyedon the same days. Hence, we assumed the number of vigi-lance bouts for an individual i observed over ti seconds dur-ing day d(i) to be Poisson-distributed with the log-linearexpectation:

lnli ¼ xibþ lnti þ ddðiÞ

where dd(i) is normally distributed, dd(i) ∼ N(0, s2), andxib models the difference between the areas and log-linear ef-fects of covariates (xi). This model was fitted with the func-tion lmer (Table 2) in the lme4 package of R version 2.11.0(R Development Core Team 2008).Initial analysis revealed a significant effect of group size,

and it was clear from GAM (generalized additive model)plots that a proportional change in group size led to a propor-tional change in vigilance frequency (i.e., ln(vigilancefrequency) appeared linearly related to ln(group size) but notto group size). Hence, we included ln(group size) in thelog-linear model. Besides effects of area and groups size, weevaluated the main effects of whether the observer was dis-covered or not, group structure (lactating females, barren fe-males, males, or mixed sex group), and functional categoryof the observed animal (lactating female, calf, barren female,and male). We included only the effects of other predictorsthat were statistically significant at the 5% level in the modelused for making inferences about difference in vigilance

756 Can. J. Zool. Vol. 89, 2011

Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.

among areas. The effect of group structure was not statisti-cally significant (P > 0.24), but the effect of whether the ob-server was discovered or not and the effect of functionalcategory were retained in the model.To compare the mean duration of the vigilance bouts

among areas and groups of individuals, a log-normal linearmodel with “day” as a random effect was fitted to the dura-tion of the first observed vigilance bouts per individual. In-spection of the residuals showed that the model fitted well tothe data, and we obtained identical results from the lme(nlme package) and the lmer (lme4 package) functions in Rversion 2.11.0 (R Development Core Team 2008). To facili-tate interpretation of time spent vigilant while grazing, wemultiplied estimates of mean vigilance frequencies (Table 2)with estimates of mean vigilance duration for lactating fe-males, barren females, and males in the five areas (Table 3)for presentation in Fig. 3.

Behavioural responsesInitial analysis revealed that sight, alert, and flight initia-

tion distances were strongly influenced when the encounterdistances were <250 m. Consequently, we analyzed the re-sponse distances only in situations where the encounter dis-tance were ≥250 m. We transformed the response distancesinto their natural logarithms and analyzed sight distance witha generalized linear model (GLM) to determine the probabil-ity of the observer to be sighted when encounter and sightdistance were ≥250 m. We analyzed alert and flight initiationdistances with linear models (LM) excluding encounter dis-tances <250 m, and because escape distance did not relateto the encounter distance, this response distance was ana-lyzed with the full data set. Following the backwards modelselection philosophy of Crawley (2005), we started with fullmodels containing biologically plausible main effects andtwo-way interactions. The full starting models contained asfixed effects are area, group size, group composition, winddirection, terrain type, activity type, and the interactionsgroup size × group structure and wind direction × terrain. Ina stepwise procedure, we removed the fixed effect term withthe highest P value based on F tests, never removing anymain effects that were included in an interaction term, and

repeated this until only significant variables were retained(P < 0.05). The exception was the main effect of area, whichwe retained in the models irrespective of significance, as it isour variable of primary interest. LM models were fitted usingstatistical software R version 2.11.0 (R Development CoreTeam 2008). Distributions of residuals were checked with di-agnostic plots to check for any strong deviation from normal-ity. We also present predicted values at the untransformedscale (in metres) in the text of the Results section. These val-ues are predicted values from the models (e.g., Table 4) ob-tained by varying only the factor of interest while keeping allother factors constant at a reference level.We analyzed probability of reindeer assessing an observer

before taking flight with a GLM. We classified the binomialresponse as one if assessment time was >1 s and zero if as-sessment time was ≤1 s (i.e., immediate flight after alert). Asassessment probability was similar for reindeer from the fourNordenskiöld Land areas but different from Edgeøya reindeer(Table 5), we grouped areas in the former (Table 6). In addi-tion to all variables included in full LM models, the full as-sessment model also included alert distance.To facilitate interpretation of the disturber on the four

fright and flight measures, we plotted predicted values fordisturbances in the areas (Fig. 4).

Results

VigilanceWe video-recorded 484 reindeer in the five areas (Advent-

dalen: n = 60; Colesdalen: n = 78; Reindalen: n = 64:Sassendalen: n = 47; Edgeøya: n = 235) during July and Au-gust 2006. Of these, 69 discovered us during filming and weincluded the parameter seen or unseen in the model. Meangroup size was 2.4 animals (median = 2), ranging from 1 to11. Observation periods per animal averaged 482 s andranged from 27 to 1267 s. Observation frequency of thenumber of vigilance bouts was strongly zero inflated(42.6%).Number of vigilance bouts per minute was higher in rein-

deer from Edgeøya than in animals in Colesdalen, Reindalen,and Sassendalen, and marginally higher than in Adventdalen

Table 2. Generalized linear model for predicting the number of vigilance bouts per minute inSvalbard reindeer (Rangifer tarandus platyrhynchus) in Nordenskiöld Land and Edgeøya inSvalbard in July and August 2006.

Variable Estimate SE z Pr >|z|Intercept –1.1703 0.1402 –8.345 ≤0.001AreaAdventdalen vs. Edgeøya –0.4102 0.2160 –1.899 0.058Colesdalen vs. Edgeøya –0.5224 0.2184 –2.392 0.017Reindalen vs. Edgeøya –0.8913 0.2281 –3.908 ≤0.001Sassendalen vs. Edgeøya –0.5165 0.2419 –2.135 0.033

Observer (discovered vs. undiscovered) 0.6631 0.1048 6.325 ≤0.001ln (group size) –0.3571 0.0707 –5.054 ≤0.001Sex and ageCalves vs. lactating females 0.1795 0.1908 0.941 0.347Barren vs. lactating females –0.1633 0.1115 –1.465 0.143Males vs. lactating females –0.3947 0.1049 –3.761 ≤0.001

Note: The standard deviation (SD) of ln (number of vigilance bouts/min) among days (random inter-cept) was estimated at 0.293. Reference levels for categorical variables are provided in the table (the levelafter “vs.”).

Reimers et al. 757

Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.

(Fig. 2, Table 2). Vigilance was not different between Adven-tdalen vs. Colesdalen (z = –0.077, P = 0.94), Reindalen (z =–0.707, P = 0.48), or Sassendalen (z = –0.876, P = 0.38)(Fig. 2). Males were 33% (95% confidence interval (CI):17% to 45%) less vigilant (in terms of the number of vigi-lance bouts per minute grazing time than lactating females

and vigilance decreased by 22% (95% CI: 14% to 29%) witha doubling in group size. Vigilance increased by 94% (95%CI: 57% to 139%) when the observer was discovered(Table 2). Duration of first vigilance bout was, however, notinfluenced (Table 3). Males, females, and lactating femaleswere more vigilant in Edgeøya than in the other areas with

Table 3. Log-normal linear model for predicting the mean duration (s) of thefirst observed vigilance bouts per individual during grazing in Svalbard reindeer(Rangifer tarandus platyrhynchus) in July and August 2006.

Value SE df t PIntercept 1.0484 0.0790 251 13.279 0.000Adventdalen vs. Edgeøya 0.2337 0.1224 21 1.909 0.070Colesdalen vs. Edgeøya 0.1390 0.1130 21 1.230 0.232Reindalen vs. Edgeøya 0.0463 0.1285 21 0.360 0.722Sassendalen vs. Edgeøya 0.3210 0.1398 21 2.296 0.032Weather (sun vs. cloudy) 0.1872 0.0828 251 2.262 0.025

Note: The standard deviation (SD) of ln (duration) among days (random intercept) wasestimated at 0.064. Reference levels for categorical variables are provided in the table (thelevel after “vs.”).

Table 4. Summary of the linear model for predicting escape distance (ln-transformed) of groups ofSvalbard reindeer (Rangifer tarandus platyrhynchus) disturbed by an approaching person in Svalbardin July and August 2006.

Variable Estimate SE t Pr >|t|Intercept 5.492 0.365 15.07 ≤0.001AreaColesdalen vs. Adventdalen 0.937 0.354 2.64 0.009Edgeøya vs. Adventdalen 0.565 0.270 2.10 0.038Reindalen vs. Adventdalen 1.099 0.333 3.30 0.001Sassendalen vs. Adventdalen 0.730 0.342 2.14 0.035

Group structureFemale(s) vs. female(s) with calf –1.575 0.351 –4.49 ≤0.001Male(s) vs. female(s) with calf –1.487 0.292 –5.09 ≤0.001Mixed vs. female(s) with calf –1.610 0.310 –5.20 ≤0.001

Note: The reference levels for categorical variables are provided in the table (the level after the “vs”).

Table 5. Assessment time in Svalbard reindeer (Rangifer tarandus platyrhynchus) in four areas in Nordens-kiöld Land and in Edgeøya in July and August in 2006.

Assessment time Adventdalen Colesdalen Edgeøya Reindalen Sassendalen≤1 s between alert and flight 1 1 26 0 0>1 s between alert and flight 6 8 36 9 8

Note: We classified the binomial response as one if assessment time was >1 s and zero if assessment time was ≤1 s (i.e.,immediate flight after alert).

Table 6. Generalized linear mixed-effects model for predicting probability of groupsof Svalbard reindeer (Rangifer tarandus platyrhynchus) assessing an observer beforefleeing (1 = assessment time ≥1 s) in Nordenskiöld Land (Adventdalen, Colesdalen,Reindalen, and Sassendalen) and in Edgeøya in July and August 2006.

Variable Estimate SE z Pr >|z|Intercept –6.418 2.432 –2.64 0.008Nordenskiöld Land areas vs. Edgøya 2.330 0.827 2.82 0.005ln(alert distance) 1.256 0.495 2.54 0.011Group size 0.350 0.192 1.82 0.068

758 Can. J. Zool. Vol. 89, 2011

Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.

0.21 (95% CI: 0.16 to 0.27), 0.26 (95% CI: 0.20 to 0.35) and0.31 (95% CI: 0.23 to 0.41) vigilance bouts/min, respectively(Fig. 2, Table 2).The mean duration of the first observed vigilance bouts

per individual was significantly longer in Sassendalen (3.9 s)than in Edgeøya (2.9 s) and in Reindalen (2.9 s) (Table 3).Multiplying estimates of mean vigilance frequencies (Table 2)with estimates of mean vigilance duration (Table 3) showedthat time spent vigilant while grazing was longest in Edgeøyaand significantly longer than in Reindalen (Fig. 3). Numberof seconds vigilant per minute grazing time varied between0.26 (95% CI: 0.16 to 0.41) in Reindalen to 0.89 (95% CI:0.64 to 1.22) in Edgeøya (Fig. 3), corresponding to 0.43%and 1.48% of grazing time, respectively.

Behavioural responsesWe approached groups of reindeer 127 times in the five

areas: Adventdalen (n = 17), Colesdalen (n = 13), Reindalen(n = 16), Sassendalen (n = 15), and Edgeøya (n = 66).The probability of being discovered at a greater distance

than 250 m was highest in Colesdalen and Reindalen (82%;95% CI: 44% to 96%), lowest in Sassendale (23%; 95% CI:6% to 59%) and Edgeøya (30%; 95% CI: 14% to 53%), andintermediate in Adventdalen (61%; 95% CI: 25% to 88%).In rugged terrain, the probability of being discovered at a

greater distance than 250 m is 2.3 times higher (95% CI: 0.7to 7.7). When reindeer groups were encountered downhill,the probability of being discovered at a greater distance than250 m is 1.3 times higher (95% CI: 0.4 to 4.3) than whenencountered in level terrain. Encountering with an uphill ap-proach, the odds of being discovered at a greater distancethan 250 m is reduced by 80% (95% CI: 5.5% to 96%).Reindeerwere alerted at greater distances in Reindalen

(159 m; 95% CI: 89 to 284 m) and in Sassendalen (153 m;95% CI: 84 to 279 m) than reindeer in Adventdalen (83 m;95% CI: 53 to 129 m) (Fig. 4). Alert distance was insignifi-cantly greater in Edgeøya (119 m; 95% CI: 74 to 190 m) andColesdalen (122 m; 95% CI: 67 to 222 m) than in Advent-dalen.Flight initiation distances was greatest in Reindalen

(109 m; 95% CI: 62 to 190 m) and shortest in Adventdalen(60 m; 95% CI: 40 to 90 m). Flight initiation distances in

Colesdalen, Sassendalen, and Edgeøya were 80–84 m with95% CI between 45 and 150 m (Fig. 4).Barren females 1+ year old fled shorter distance in Adven-

tdalen (50 m; 95% CI: 25 to 101 m) than in Colesdalen(128 m; 95% CI: 64 to 259 m), Reindalen (151 m; 95% CI:75 to 304 m), Sassendalen (104 m; 95% CI: 52 to 210 m),and Edgeøya (88 m; 95% CI: 44 to 178 m) (Table 4, Fig. 4).Lactating females escaped over greater distances than males,barren females, or mixed groups (Table 4). Escape distancein lactating females was 243 m (95% CI: 117 to 504 m) inAdventdalen, 620 m (95% CI: 305 to 1258 m) in Colesdalen,729 m (95% CI: 374 to 1418 m) in Reindalen, 504 m(95% CI: 254 to 998 m) in Sassendalen, and 427 m(95% CI: 249 to 733 m) in Edgeøya (Table 4).Probability of assessing the observer before fleeing was

lower in Edgeøya (61%; 95% CI: 47% to 75%) than in theNordenskiöld Land areas (94%; 95% CI: 85% to 103%) (Ta-bles 5 and 6, Fig. 4). Probability of assessing increased withincreasing alert distance and tended to increase with increas-ing group size (Table 6). A doubling of group size predicts a27% increase in the probability to assess the observer beforefleeing (95% CI: 10% decrease and 66% increase).

Discussion

Vigilance responsesThe way in which antipredator behaviour is modified de-

pends on both heritable predisposition (Riechert and Hedrick1990), as well as experience (Berger et al. 2001; Blumstein etal. 2004). Experience-dependant behaviour may be lost afterthe first generation in the absence of predators, while more“hard-wired” antipredator behaviour acquired on an evolu-tionary time scale may persist for thousands of years follow-ing isolation from predators (Diamond 1990; Byers 1997;Coss 1999; Blumstein et al. 2000). Given an island as amodel, terrestrial prey become naïve to predation risk whenpredators have been absent for long periods (Byers 1997;Blumstein and Daniel 2005). In accordance with expecta-tions, Svalbard reindeer show a relaxed vigilance behaviourwith a low frequency of vigilance bouts (between 0.09 and0.31 bouts/min in the five areas) and little time spent vigilant

Fig. 2. Predicted values (±2SE; ca. 95% CI) of the number of vigi-lance bouts per minute grazing in Svalbard reindeer (Rangifer tar-andus platyrhynchus) in different areas in July and August 2006.Reference level of categorical variable observer is undiscovered.

Fig. 3. Predicted total time spent vigilant (s) with head aboveshoulder height ≤10 s while grazing in Svalbard reindeer (Rangifertarandus platyrhynchus) in different areas in July and August 2006.Error bars are ±2SE (ca. 95% CI). Reference level of categoricalvariable observer is undiscovered.

Reimers et al. 759

Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.

while grazing (between 0.43% and 1.48%). However, wildreindeer in southern Norway display a comparable vigilancerate (0.11–0.38 bouts/min), but more time vigilant whilegrazing (0.74%–2.49% of grazing time) (E. Reimers, unpub-lished data). As reindeer in southern Norway either coexistor share a recent past with predators, their moderately highervigilance reflects a behavioural adaptation towards predators,whereas Svalbard reindeer in the four Nordenskiöld Landareas presumably reflect baseline vigilance. Although thereare differences in vigilance between reindeer on Svalbard,mainland Norway, and caribou in West Greenland andAlaska (Bøving and Post 1997), vigilance behaviour in theseRangifer subspecies is surprisingly low regardless of absenceor presence of predators. An explanation most probably com-bines two factors; the generally open, easily surveyable treelessalpine–arctic environment inhabited by Rangifer, and the anat-omy and physiology of the ungulate eye (Sjaastad et al. 2003).In Rangifer as in many other herbivores, the combined visualfield of both eyes covers 360° even when grazing, allowingthe animals some environmental control also when subgrazing.A key prediction of the multipredator hypothesis (Blum-

stein 2006) is that isolation from all predators may lead to arapid loss of antipredator behaviour including loss of the ef-fect of group size and breakdown of abilities to recognizepredators. For example, both reindeer and caribou frompredator-free regions (Svalbard and West Greenland) wereabout 3.5 times less vigilant to playback of howls from wolves

(Canis lupus L., 1758) at control sites (Denali National Park,Alaska, and Tetlin National Wildlife Refuge, Alaska)(Berger 2007). A commonly reported benefit of aggregationis seen when individuals decrease the time devoted to vigi-lance and increase the time to foraging as group size in-creases (Lima and Dill 1990). Costly antipredator behaviourshould not persist once there is no net benefit. And indeed,the traditional grouping behaviour that characterize Rangifersubspecies elsewhere is absent in Svalbard, where animalslive individually or in small groups (Alendal and Byrkjedal1976; Alendal et al. 1979). The lack of group behaviour inSvalbard reindeer as a response to the absence of predatorsand parasitizing insects facilitates energy optimization andenables these animals to cope with this harsh environment(Reimers 1980; Reimers et al. 1982).There was, however, an effect of group size, as vigilance

in Svalbard reindeer decreased with increasing group size.Surprisingly, there was also a higher vigilance rate displayedby reindeer in Edgeøya compared with reindeer in the fourNordenskiöld Land areas. Besides rare incidences, polarbears are generally not regarded as predators on reindeer(Derocher et al. 2000). However, after the international har-vest ban in 1973, the population of polar bears in the Sval-bard archipelago has increased (Aars et al. 2009). Inaddition, the sea-ice cover in the arctic region during summerhas decreased in recent years (Singarayer et al. 2006), result-ing in more bears on land during all seasons and thus more

Ale

rtdis

tance

(m)

0

20

40

60

80

100

120

140

160

180

200

Sa Co Re Ad Ed

Flig

htdis

tance

(m)

0

20

40

60

80

100

120

140

160

180

200

Sa Co Re Ad Ed

Escap

edi

sta

nce

(m)

0

20

40

60

80

100

120

140

160

180

200

Sa Co Re Ad Ed

Pro

bofasse

ssin

gafterale

rt

0.0

0.2

0.4

0.6

0.8

1.0

Nordenskiöld Land Edgeøya

Fig. 4. Predicted values (±SE) of alert distance, flight initiation distance, escape distance, and probability of assessing the observer beforeflight in groups of Svalbard reindeer (Rangifer tarandus platyrhynchus) disturbed by a person during July and August in Sassendalen (SA),Colesdalen (Co), Reindalen (Re), and Adventdalen (Ad) in Nordenskiöld Land, and in Edgeøya (Ed), Svalbard, in 2006. Reference level ofcategorical variables included in the model (see Tables 2 and 4) is female(s) 1+ year old. For numerical variables, the predictions are basedon the mean.

760 Can. J. Zool. Vol. 89, 2011

Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.

frequent encounters between reindeer and polar bears. Edge-øya, being an important denning area for polar bears (J. Aars,Norwegian Polar Institute, personal communication), has ahigh density of polar bears during summer, greatly exceedingthe density in Nordenskiöld Land (Derocher et al. 2000). Ex-perience-dependant behaviour may be restored quickly thefirst time individuals encounter predators (Brown et al.1997; Griffin et al. 2000; Berger et al. 2001). A comparisonof the response behaviour of reindeer from encounters with apolar bear and persons disguised as polar bears during ourstudy in Edgeøya (E. Reimers and S. Eftestøl, unpublisheddata) support a predator–prey relationship between the twospecies that may explain increased vigilance among reindeerin the more densely polar bear populated Edgeøya.The number of vigilance bouts and time vigilant per mi-

nute grazing time tended to be higher or longer among fe-males with calf at foot than among females without calvesand significantly higher than among males. This is consistentwith results reported for other wild ungulates (Lipetz andBekoff 1982; Burger and Gochfeld 1994; Linnell and Ander-sen 1995; Toïgo et al. 1999), but also for domesticated ungu-lates (Kluever et al. 2008).

Fright and flight responsesThe probability of being discovered by reindeer at greater

distance than 250 m was highest in Colesdalen and Rein-dalen, possibly indicating an effect of 22 years of huntingand 12 years of live capturing of reindeer in these areas. Aperson was more easily discovered at greater distances onrugged terrain than on level terrain, and also discovered atgreater distances when the observer approached the reindeerdownhill compared with an approach on level terrain. Ap-proaching the reindeer uphill reduced the probability of beingdiscovered at greater distances than 250 m. These resultsconcur with observations in herds of wild reindeer in south-ern Norway (Reimers et al. 2009, 2010) that reindeer appa-rently feel more safe and in control on level terrain andwhen they are at an altitudinal advantage.None of the alert, flight initiation, or escape distances

measured in the Nordenskiöld Land areas in July–August2006 differed from those that we measured in the same areasin 1994 (Colman et al. 2001). Alert, flight initiation, and es-cape distances were all shortest in Adventdalen, with Long-yearbyen and an associated extensive human infrastructure,indicating habituation to humans as predicted and shown byColman et al. (2001). Greater response distances in Sassen-dalen, Colesdalen, and particularly in Reindalen (Fig. 4) mayrelate to negative interactions with humans related to hunting(Sassendalen, Colesdalen, and Reindalen) and high level ofscientific activity (live capturing of reindeer) (Colesdalenand Reindalen). Svalbard reindeer operate mostly solitary orin small groups frequently dispersed at 0.5–2 km apart (Tylerand Øritsland 1989). Hunters most often approach, shoot,and dress a reindeer without being noticed by other reindeer,facilitating a non-negative stimuli approach to hunting. Asthe response distances have remained unchanged in Sassen-dalen in 2006 compared with 1994 (Colman et al. 2001) inspite of medium hunting rank order during these years, hunt-ing probably does not evoke negative response behaviour to-wards humans on foot in this area with a low level of humanactivities. In Colesdalen and Reindalen, live capturing from

snowmobiles has occurred annually since 1994 in addition tohunting (Omsjoe et al. 2009). However, response distances inthese two areas have also remained unchanged from 1994. Aswas the case in 1994, response distances in Reindalen in2006 were greater than in Sassendalen and Colesdalen. Thislikely reflects an effect of negative human interference interms of hunting and capturing combined with less habitua-tion because of lower levels of “undisturbing” recreationalactivities. Nevertheless, it appears that a moderate level ofapparently negative human–reindeer interactions like huntingand recreational and scientific activities have a negligible ef-fect on reindeer behavioural responses on Svalbard. Our find-ings support Omsjoe et al. (2009), who report no evidencefor a lower calf success in animals that had been capturedthe previous winter. Likewise, there was no relationship be-tween the strength of the acute stress response (measured ascortisol consentrations) and the probability of females havinga calf at foot the subsequent summer (Omsjoe et al. 2009).Although reindeer sex and age did not influence alert and

flight initiation distances, females with calves at foot hadgreater escape distances (e.g., in Edgeøya, the distance was427 vs. 97 m in males and 88 m in females without calves).This corresponds with the earlier Svalbard study (Colman etal. 2001) and other studies on Rangifer (de Vos 1960; Hor-ejsi 1981). If reindeer mothers are at a behavioural disadvant-age towards possible predators, demographic consequenceswill be severe, hence their great escape distances.In our study in 1994 (Colman et al. 2001), we recorded the

greatest response distances in Reinsdyrflya; a remotely lo-cated and rarely visited peninsula 320 km2 north of Long-yearbyen. We suggested that this behaviour could be aresponse to a low level of interaction with humans. Conse-quently, we expected great response distances among reindeerin the even more isolated Edgeøya. It is conceivable thatreindeer with little or no previous experience with humansdisplay a behaviour that maximizes their escape options.Contrary to our expectations, response distances in Edgeøyawere similar to or even shorter than in the three areas outsideAdventdalen. In hindsight, it appears equally conceivable thatreindeer with no previous experience with humans allow acloser approach for examination before flight.This may explain the short alert distance (118 m) that al-

lows a closer examination between sight and alert but tradesoff lower probability of assessing the observer before fleeing(61%) in Edgeøya vs. 94% in the Nordenskiöld Land areas.Probability of assessing increased with increasing alert dis-tance and tended to increase with increasing group size. Thismight indicate the importance of greater distances and largergroup size for the animals to feel in control of their surroundings.Interestingly, sight, alert, and flight initiation distances re-

corded in wild reindeer with a semidomestic origin in south-ern Norway during summer (E. Reimers, unpublished data)did not differ from those of Svalbard reindeer. Escape dis-tance was, however greater, as were all response distances insouthern Norway reindeer with a wild origin. It appears thenthat reindeer have a hard-wired flight response threshold ac-quired on an evolutionary time scale and is reflected in Sval-bard, primarily in Edgeøya but possibly also in Sassendalen.This threshold may change in accordance with two key fac-tor: presence or absence of predators and presence or absenceof humans. The predator effect is straightforward, i.e., the

Reimers et al. 761

Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.

threshold decreases and response distances increase (Klop-pers et al. 2005; Stankowich and Blumstein 2005). The hu-man factor is more complicated. For example, when animalsover time are exposed to extensive human activities as in Ad-ventdalen (with Longyearbyen) or in herds of wild reindeerin southern Norway (Reimers et al. 2009, 2010), thresholdincreases possibly because humans are not considered to bepredators (Stankowich and Blumstein 2005). Alternatively, inareas with few interactions between reindeer and humans, theanimals may perceive humans as predators and decrease theresponse threshold. The combined effect of few humans andhunting may result in a further decrease as recorded in rein-deer with a wild origin (E. Reimers, unpublished data).There was no apparent relationship between fright and

flight responses and vigilance (Figs. 2 and 4). Reindeer inReindalen were least vigilant but showed the strongest frightand flight responses to people, whereas reindeer on Edgeøyawere most vigilant but responded moderately to human en-counters. Reindeer in the densely trafficked Adventdalen dis-played very relaxed behavioural responses to humans, butwere next to Edgeøya in vigilance. A possible explanationfollows from the preceding discussion: reindeer in areas withfrequent but unthreatening encounters with humans (Advent-dalen) maintain short response distances but a high vigilancelevel as a result of a variety of possible alerting stimuli(sound, sight, and smell from surrounding infrastructures). InReindalen with few but hostile human encounters (huntingand live capturing), vigilance is at a low, but when encoun-tered by humans the response distances are great. InEdgeøya, reindeer rarely encounter humans and they maintainshort response distances. However, because of possible en-counters with polar bears, they keep up the alert guard interms of vigilance. Reindeer in Sassendalen and Colesdalenrespond similarly. Few and mildly hostile encounters with hu-mans in Sassendalen (hunting), as well as many unthreaten-ing encounters with hikers combined with hostile encountersin Colesdalen (hunting and live capturing), result in moderatelevels of vigilance and response distances.

AcknowledgementsWe thank the Governor on Svalbard for the necessary

landing allowances on Edgeøya; The Norwegian Polar Insti-tute for rent of necessary field equipment; and S. Eftestøl, R.Steen, K.F. Øi, M. Kardel, and K. Reimers Kardel for fieldassistance; L.E. Loe for statistical advice; J.E. Colman forlinguistic support; and two anonymous reviewers for manu-script improvements. Financial support was provided by theNorwegian Science Foundation, the Norwegian Polar Insti-tute, and Framkomiteens Polarfond.

ReferencesAars, J., Marques, T.A., Buckland, S.T., Andersen, M., Belikov, S.,

Boltunov, A., and Wiig, Ø. 2009. Estimating the Barents Sea polarbear subpopulation size. Mar. Mamm. Sci. 25(1): 35–52. doi:10.1111/j.1748-7692.2008.00228.x.

Alendal, E., and Byrkjedal, I. 1976. Population size and reproductionof the reindeer (Rangifer tarandus platyrhynchus) on Nordens-kiöld Land, Svalbard. Norsk Polarinstitutt Årbok, 1974: 139–152.

Alendal, E., de Bie, S., and van Wieren, S.E. 1979. Size andcomposition of the wild reindeer (Rangifer tarandus platyr-

hynchus) population in the Southeast Svalbard Nature Reserve.Holarct. Ecol. 2: 101–107.

Beauchamp, G. 2004. Reduced flocking by birds on islands withrelaxed predation. Proc. R. Soc. Lond. B Biol. Sci. 271(1543):1039–1042. doi:10.1098/rspb.2004.2703. PMID:15293857.

Berger, J. 2007. Carnivore repatriation and holarctic prey: narrowingthe deficit in ecological effectiveness. Conserv. Biol. 21(4): 1105–1116. doi:10.1111/j.1523-1739.2007.00729.x. PMID:17650259.

Berger, J., Swenson, J.E., and Persson, I.L. 2001. Recolonizingcarnivores and naive prey: conservation lessons from Pleistoceneextinctions. Science (Washington, D.C.), 291(5506): 1036–1039.doi:10.1126/science.1056466. PMID:11161215.

Blumstein, D.T. 2006. The multipredator hypothesis and theevolutionary persistence of antipredator behavior. Ethology,112(3): 209–217. doi:10.1111/j.1439-0310.2006.01209.x.

Blumstein, D.T., and Daniel, J.C. 2005. The loss of anti-predatorbehaviour following isolation on islands. Proc. R. Soc. Lond. BBiol. Sci. 272(1573): 1663–1668. doi:10.1098/rspb.2005.3147.PMID:16087420.

Blumstein, D.T., Daniel, J.C., Griffin, A.S., and Evans, C.S. 2000.Insular tammar wallabies (Macropus eugenii) respond to visual butnot acoustic cues from predators. Behav. Ecol. 11(5): 528–535.doi:10.1093/beheco/11.5.528.

Blumstein, D.T., Daniel, J.C., and Springett, B.P. 2004. A test of themulti-predator hypothesis: rapid loss of antipredator behavior after130 years of isolation. Ethology, 110(11): 919–934. doi:10.1111/j.1439-0310.2004.01033.x.

Bøving, P.S., and Post, E. 1997. Vigilance and foraging behaviour offemale caribou in relation to predation risk. Rangifer, 17: 55–63.

Brown, J.S. 1999. Vigilance, patch use, and habitat selection:foraging under predation risk. Evol. Ecol. Res. 1: 49–71.

Brown, G.E., Chivers, D.P., and Smith, R.J.F. 1997. Differentiallearning rates of chemical versus visual cues of a northern pike byfathead minnows in a natural habitat. Environ. Biol. Fishes, 49(1):89–96. doi:10.1023/A:1007302614292.

Burger, J., and Gochfeld, M. 1994. Vigilance in African mammals:differnces among mothers, other females, and males. Behaviour,131(3): 153–169. doi:10.1163/156853994X00415.

Byers, J.A. 1997. American pronghorn: social adaptations and theghosts of predators past. University of Chicago Press, Chicago.

Colman, J.E., Jacobsen, B.W., and Reimers, E. 2001. Summer responsedistances of Svalbard reindeer Rangifer tarandus platyrhynchus toprovocation by humans on foot. Wildl. Biol. 7: 275–283.

Coss, R.G. 1999. Effects of relaxed natural selection on the evolutionof behavior. In Geographic variation in behavior: perspectives onevolutionary mechanisms. Edited by S.A. Foster and J.A. Endler.Oxford University Press, Oxford. pp. 180–208.

Crawley, M.J. 2005. Statistical computing: an introduction to dataanalysis using S-Plus. John Wiley and Sons Ltd., Chichester, U.K.

de Vos, A. 1960. Behavior of barren-ground caribou on their calvinggrounds. J. Wildl. Manage. 24(3): 250–258. doi:10.2307/3797511.

Derocher, A.E., Wiig, Ø., and Bangjord, G. 2000. Predation ofSvalbard reindeer by polar bears. Polar Biol. 23(10): 675–678.doi:10.1007/s003000000138.

Diamond, J.M. 1990. Biological effects of ghosts. Nature (London),345(6278): 769–770. doi:10.1038/345769a0.

Elgar, M.A. 1989. Predator vigilance and group size in mammals andbirds: a critical review of the empirical evidence. Biol. Rev. Camb.Philos. Soc. 64(1): 13–33. doi:10.1111/j.1469-185X.1989.tb00636.x. PMID:2655726.

Frid, A. 1997. Vigilance by female Dall’s sheep: interactions betweenpredation risk factors. Anim. Behav. 53(4): 799–808. doi:10.1006/anbe.1996.0345.

762 Can. J. Zool. Vol. 89, 2011

Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.

Gill, J.A., Norris, K., and Sutherland, W.J. 2001. Why behaviouralresponses may not reflect the population consequences of humandisturbance. Biol. Conserv. 97(2): 265–268. doi:10.1016/S0006-3207(00)00002-1.

Griffin, A.S., Blumstein, D.T., and Evans, C. 2000. Training captive-bred or translocated animals to avoid predators. Conserv. Biol.14(5): 1317–1326. doi:10.1046/j.1523-1739.2000.99326.x.

Hopewell, L., Rossiter, R., Blower, E., Leaver, L., and Goto, K. 2005.Grazing and vigilance by Soay sheep on Lundy island: influenceof group size, terrain and the distribution of vegetation. Behav.Processes, 70(2): 186–193. doi:10.1016/j.beproc.2005.04.009.PMID:15963661.

Horejsi, B.L. 1981. Behavioral response of barren ground caribou to amoving vehicle. Arctic, 34: 180–185.

Illius, A.W., and Fitzgibbon, C. 1994. Costs of vigilance in foragingungulates. Anim. Behav. 47(2): 481–484. doi:10.1006/anbe.1994.1067.

Kaltenborn, B.P. 1991. Utkast til forvaltningsplan for turisme ogfriluftsliv p Svalbard. NINA, Lillehammer, Norway.

Kavaliers, M. 1990. Responsiveness of deer mice to a predator, theshort-tailed weasel: population differences and neuromodulatorymehanisms. Physiol. Zool. 63: 388–407.

Kloppers, E.L., St. Clair, C.C., and Hurd, T.E. 2005. Predator-resembling aversive conditioning for managing habituated wildlife.Ecology and Society, 10(1): 31.

Kluever, B.M., Breck, S.W., Howery, L.D., Krausman, P.R., andBergman, D.L. 2008. Vigilance in cattle: the influence ofpredation, social interactions, and environmental factors. Range-land Ecol. Manag. 61(3): 321–328. doi:10.2111/07-087.1.

Lima, S.L. 1987. Distance to cover, visual obstructions, and vigilancein house sparrows. Behaviour, 102(3–4): 231–237. doi:10.1163/156853986X00144.

Lima, S.L. 1995. Back to the basics of anti-predatory vigilance: thegroup-size effect. Anim. Behav. 49(1): 11–20. doi:10.1016/0003-3472(95)80149-9.

Lima, S.L. 1998. Stress and decision making under the risk ofpredation: recent developments from behavioral, reproductive, andecological perspectives. Adv. Stud. Behav. 27: 215–290. doi:10.1016/S0065-3454(08)60366-6.

Lima, S.L., and Bednekoff, P.A. 1999. Back to the basics of antipredatoryvigilance: can nonvigilant animals detest attack? Anim. Behav. 58(3):537–543. doi:10.1006/anbe.1999.1182. PMID:10479369.

Lima, S.L., and Dill, L.M. 1990. Behavioral decisions made under therisk of predation: a review and prospectus. Can. J. Zool. 68(4):619–640. doi:10.1139/z90-092.

Linnell, J.D.C., and Andersen, R. 1995. Site tenacity in roe deer:short term effects of logging. Wildl. Soc. Bull. 23: 31–35.

Lipetz, V., and Bekoff, M. 1982. Group size and vigilance inpronghorns. Z. Tierpsychol. 58(3): 203–216. doi:10.1111/j.1439-0310.1982.tb00318.x.

Lung, M.A., and Childress, M.J. 2007. The influence of conspecificsand predation risk on the vigilance of elk (Cervus elaphus) inYellowstone National Park. Behav. Ecol. 18(1): 12–20. doi:10.1093/beheco/arl066.

Magurran, A.E. 1999. The causes and consequences of geographicvariation in antipredator behavior: perspectives from fish popula-tions. In Geographic variation in behavior: perspectives onevolutionary mechanisms. Edited by S.A. Foster and J.A. Endler.Oxford University Press, Oxford. pp. 139–163.

Mooring, M.S., Fitzpatrick, T.A., Nishihira, T.T., and Reisig, D.D.2004. Vigilance, predation risk, and the allee effect in desertbighorn sheep. J. Wildl. Manage. 68(3): 519–532. doi:10.2193/0022-541X(2004)068[0519:VPRATA]2.0.CO;2.

Omsjoe, E.H., Stien, A., Irvine, J., Albon, S.D., Dahl, E., Thoresen,

S.I., Rustad, E., and Ropstad, E. 2009. Evaluating capture stressand its effects on reproductive success in Svalbard reindeer. Can. J.Zool. 87(1): 73–85. doi:10.1139/Z08-139.

Packer, C., and Abrams, P. 1990. Should cooperative groups be morevigilant than selfish groups. J. Theor. Biol. 142(3): 341–357.doi:10.1016/S0022-5193(05)80557-0.

Pulliam, H.R., Pyke, G.H., and Caraco, T. 1982. The scanningbehavior of juncos: a game theoretical approach. J. Theor. Biol.95(1): 89–103. doi:10.1016/0022-5193(82)90289-2.

R Development Core Team 2008. lme4 package in R version 2.11.0[computer program]. R Foundation for Statistical Computing,Vienna, Austria. Available from http://www.r-project.org/.

Reimers, E. 1980. Activity pattern; the major determinant for growthand fattening in Rangifer? In Proceedings of the 2nd InternationalReindeer/Caribou Symposium, Røros, Norway, 1979. Edited by E.Reimers, E. Gaare, and S. Skjenneberg. Direktoratet for Vilt ogFerskvannsfisk, Trondheim, Norway. pp. 466–474.

Reimers, E. 1982. Winter mortality and population trends of reindeeron Svalbard, Norway. Arct. Alp. Res. 14(4): 295–300. doi:10.2307/1550792.

Reimers, E., Ringberg, T., and Sørumgard, R. 1982. Bodycomposition of Svalbard reindeer. Can. J. Zool. 60(8): 1812–1821. doi:10.1139/z82-235.

Reimers, E., Loe, L.E., Eftestøl, S., Colman, J.E., and Dahle, B. 2009.Effects of hunting on response behaviors of wild reindeer. J. Wildl.Manage. 73(6): 844–851. doi:10.2193/2008-133.

Reimers, E., Røed, K.H., Flaget, Ø., and Lurs, E. 2010. Habituationresponses in wild reindeer exposed to recreational activities.Rangifer, 30: 45–59.

Riechert, S.E., and Hedrick, A.V. 1990. Levels of predation andgenetically based anti-predator behaviour in the spider, Ageno-lopsis aperta. Anim. Behav. 40(4): 679–687. doi:10.1016/S0003-3472(05)80697-9.

Rønning, O.I. 1985. Vegetasjonslære, 2. Universitetsforlaget, Oslo,Norway.

Singarayer, J.S., Bamber, J.L., and Valdes, P.J. 2006. Twenty-first-century climate impacts from a declining Arctic Sea ice cover. J.Clim. 19(7): 1109–1125. doi:10.1175/JCLI3649.1.

Sjaastad, Ø.V., Hove, K., and Sand, O. 2003. Physiology of domesticanimals Scandinavian Veterinary Press, Oslo, Norway.

Snyder, J., Snyder, J.M., and Prokosch, P. 2007. Tourism in the polarregions: the sustainability challenge United Nations EnvironmentProgramme, Nairobi, Kenya.

Staaland, H., and White, R.G. 1991. Influence of foraging ecology onalimentary-tract size and function of Svalbard reindeer. Can. J.Zool. 69(5): 1326–1334. doi:10.1139/z91-187.

Stankowich, T., and Blumstein, D.T. 2005. Fear in animals: a meta-analysis and review of risk assessment. Proc. R. Soc. Lond. B Biol.Sci. 272(1581): 2627–2634. doi:10.1098/rspb.2005.3251. PMID:16321785.

Sysselmannen. 2006. Turisme og friluftsliv p Svalbard. Sysselman-nen p Svalbard, Longyearbyen, Norway.

Taylor, A.R., and Knight, R.L. 2003. Behavioral responses of wildlifeto human activity: terminology and methods. Wildl. Soc. Bull. 31:1263–1271.

Toïgo, C., Gaillard, J.M., and Michallet, J. 1999. Cohort affectsgrowth of males but not females in alpine ibex (Capra ibex ibex). J.Mammal. 80(3): 1021–1027. doi:10.2307/1383272.

Tyler, N.J.C. 1987. Natural limitation of the abundance of the High ArcticSvalbard reindeer. Ph.D. thesis, University of Cambridge, Cambridge.

Tyler, N.J.C., and Øritsland, N.A. 1989. Why don’t Svalbard reindeermigrate? Holarct. Ecol. 12: 369–376.

Tyler, N.J.C., Forchhammer, M.C., and Oritsland, N.A. 2008.Nonlinear effects of climate and density in the dynamics of a

Reimers et al. 763

Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.

fluctuating population of reindeer. Ecology, 89(6): 1675–1686.doi:10.1890/07-0416.1. PMID:18589531.

Van der Knaap, W.O. 1986. On the presence of reindeer (Rangifertarandus L.) on Edgeøya, Spitzbergen in the period 3800–5000BP. Circumpolar Journal, 2: 3–10.

Watson, M., Aebischer, N.J., and Cresswell, W. 2007. Vigilance andfitness in grey partridges Perdix perdix: the effects of group sizeand foraging-vigilance trade-offs on predation mortality. J. Anim.

Ecol. 76(2): 211–221. doi:10.1111/j.1365-2656.2006.01194.x.PMID:17302828.

Whittingham, M.J., Butler, S.J., Quinn, J.L., and Cresswell, W. 2004.The effect of limited visibility on vigilance behaviour and speed ofpredator detection: implications for the conservation of granivor-ous passerines. Oikos, 106(2): 377–385. doi:10.1111/j.0030-1299.2004.13132.x.

764 Can. J. Zool. Vol. 89, 2011

Published by NRC Research Press

Can

. J. Z

ool.

Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Gua

ngzh

ou J

inan

Uni

vers

ity o

n 06

/06/

13Fo

r pe

rson

al u

se o

nly.