Recolonizing Carnivores and Naïve Prey: Conservation Lessons from Pleistocene Extinctions

See allHide authors and affiliations

Science  09 Feb 2001:
Vol. 291, Issue 5506, pp. 1036-1039
DOI: 10.1126/science.1056466


The current extinction of many of Earth's large terrestrial carnivores has left some extant prey species lacking knowledge about contemporary predators, a situation roughly parallel to that 10,000 to 50,000 years ago, when naı̈ve animals first encountered colonizing human hunters. Along present-day carnivore recolonization fronts, brown (also called grizzly) bears killed predator-naı̈ve adult moose at disproportionately high rates in Scandinavia, and moose mothers who lost juveniles to recolonizing wolves in North America's Yellowstone region developed hypersensitivity to wolf howls. Although prey that had been unfamiliar with dangerous predators for as few as 50 to 130 years were highly vulnerable to initial encounters, behavioral adjustments to reduce predation transpired within a single generation. The fact that at least one prey species quickly learns to be wary of restored carnivores should negate fears about localized prey extinction.

The spectacular post-Pleistocene extinctions of many genera of large animals in areas ranging from Australia to North America have been attributed primarily to human overkill as hunters encountered naı̈ve prey—the “blitzkrieg hypothesis” (1)—and/or to climate change (2). An inadvertent consequence of today's extinction of many large carnivores is that prey in otherwise intact areas may lose knowledge about current predators (3, 4). These extinctions, however, offer opportunities to assess the generality of components of the blitzkrieg hypothesis and to address concerns about the ecological consequences of carnivore restoration. In Western Europe and the United States (outside of Alaska), wolves (Canis lupus) and brown bears (Ursus arctos) were eliminated within 100 years from more than 95% of their range. The cessation of predation has released mammalian prey from past selection pressures (3–5), but the current expansion of large carnivores into formerly depopulated ecosystems is bringing them into contact with naı̈ve prey, a situation roughly similar to some post-Pleistocene events. Whereas naı̈ve large animals first encountered colonizing paleolithic hunters 10,000 to 50,000 years ago, species such as bison, moose, and elk are now reexperiencing bears and wolves after they had been locally extinct.

To understand the possible susceptibility of extant prey to unfamiliar predators, we focus on two predictions: that naı̈ve prey (i) fail to display appropriate behavior to reduce predation and (ii) experience heightened mortality along colonizing fronts (6). We studied a common circumpolar ungulate, the moose (Alces alces), focusing on sites (i) with predators, (ii) lacking predators, and (iii) with recent carnivore recolonization. Two types of macrogeographic contrasts were made: one in Europe, the other in North America. Three Scandinavian sites [(i) Dalarma-Hedmark (Sweden-Norway), (ii) Harjedalen (Sweden), and (iii) Norrbotten (Sweden)] had predator-naı̈ve moose that were becoming reexposed to bears along colonizing fronts (7, 8). In the United States, six sites (four in Alaska and two in Wyoming) were studied. The Wyoming sites included naı̈ve moose in and away from the paths of colonizing wolves and grizzly bears in the southern region of the Greater Yellowstone Ecosystem. Bears, wolves, and moose have overlapped in mainland Alaska for about 9000 years (9), and in these and other systems, juvenile moose mortality is three times greater than when these predators are absent (10, 11).

Predator-naı̈ve individuals may be less sensitive to cues that signify the presence of dangerous carnivores. We examined this idea by comparing (i) vigilance, (ii) predator-directed aggression, and (iii) abandonment of feeding sites by adult female moose during and after systematic playback experiments with novel and familiar auditory and olfactory cues. In Alaska, we used three sites with intact carnivore communities (Talkeetna Mountains, Denali National Park and Preserve, and Matanuska Valley) and three predator-free systems (Kalgin Island in Alaska's Cook Inlet, Grand Teton National Park, and adjacent forest lands); in the latter, prey had not experienced grizzly bears or wolves for 40 to 75 years (12).

Responses to auditory or olfactory cues were conspicuously less among predator- naı̈ve moose (Figs. 1 and 2). Wolf calls increased vigilance by about 250% in predator-experienced Alaskan populations relative to predator-free ones (F 1, 221 = 25.771, P < 0.0001) (13). The sound of a control—running water—produced no differences (F 1,180 = 0.541, P = 0.744). Additionally, the reactions of predator-experienced moose to raven calls were about six times greater than those of moose in predator-free zones (F = 44.26; P < 0.0001) (Fig. 1), presumably because these scavengers are highly associated with grizzly bears and wolves (4, 14). Likewise, the odors of wolves and bears generally elicited parallel responses (Fig. 2); a striking absence of site abandonment occurred in naı̈ve moose from either Alaska or Wyoming when they were exposed to wolf urine (abandonment occurred in 0 out of 70 trials; Fig. 3) (15).

Figure 1

Descriptive summary of nontransformed mean (cumulative) moose vigilance per 180-s bout during and after experimental playback of sounds. Sounds, sites, and sample sizes per site (followed by SEM in parentheses) were as follows. Water: Grand Teton National Park (Tet), 196 (1.1); U.S. Forest Service (USFS), 30 (1.5); Kalgin Island (Kal), 12 (1.9); Matanuska Valley (Mat), 28 (1.0); Talkeetna Mountains (Tal), 25 (1.4); and Denali National Park (Den), 35 (1.1). Wolf: Tet, 215 (4.3); USFS, 30 (4.3); Kal, 12 (8.7); Mat, 27 (4.6); Tal, 24 (5.9); and Den, 31 (5.1). Coyote: Tet, 137 (5.3); USFS, 21 (2.9); and Den, 26 (4.7). Raven; Tet, 215 (3.2); USFS, 30 (1.6); Kal, 12 (7.3); Mat, 28 (4.8); Tal, 23 (5.4); and Den, 31 (4.9). Tiger: Tet, 187 (4.3); USFS, 20 (6.2); Kal, 12 (6.8); Mat, 22 (5.5); Tal, 18 (5.5); and Den, 27 (5.0).

Figure 2

Descriptive summary of nontransformed mean and SEM of moose responses to odor trials by site. The number of trials with the various substances per site was as follows (numbers appear in parentheses in the following order: Alaska-yes, Wyoming, and Alaska-no): control, observational only (27, 95, 1); snow (24, 78, 1); human urine (18, 59, 7); fermented potato (0, 10, 0); grizzly bear feces (19, 80, 4); black bear feces (6, 31, 6); wolf urine (22, 88, 6); coyote urine (12, 70, 4); tiger feces (12, 59, 3). Alaska-yes indicates sites with predators; Alaska-no indicates a 2500-acre predator exclosure on Kenai Peninsula.

Figure 3

Frequency of local site abandonment by female moose after odor deposition. The number of trials (N) is as indicated. Alaska-yes data are from Denali and Talkeetnas; Wyoming data are from Grand Teton National Park. Pairwise Alaska-Wyoming contrasts (excluding the Kenai site) are as follows: wolf (P = 0.016), wolf and coyote combined (P = 0.003), grizzly bear (P =0 .065), grizzly and black bear combined (P = 0.005), and tiger (P = 0.553).

One additional measure—the demeanor of moose—also supports the notion that individuals from predator-depopulated regions are more vulnerable to predation. Naı̈ve moose were one-quarter as likely to behave aggressively in response to odor cues than were predator-savvy conspecifics (11 versus 47%, respectively;N = 86 naı̈ve animals and 24 predator-savvy animals; χ2 = 22.397; P < 0.0001) (16). The former failed to drop their heads, retract their ears, and piloerect their nape fur—all patterns observed when moose fight off bears or wolves (17, 18). Because naı̈ve moose approached, rather than avoided, odors in 16% of the trials (N = 86 trials), but predator-experienced moose never did (N = 24 trials, P < 0.0001), we can only presume that approaches increase, rather than decrease, the chances of predation.

These data show clear and repeatable differences in behavior associated with predators but do not indicate whether prey lacking knowledge of either novel or recolonizing predators experience heightened mortality. We examined this possibility using moose in two discrete systems with recolonizing (i) Scandinavian brown bears and (ii) Yellowstone wolves and bears. In the Dalarna-Hedmark area, bears were functionally extinct by 1897. To evaluate the efficiency of bear predation on naı̈ve and experienced moose, we used spatial contrasts, predicting a greater proportion of kills along bear colonizing fronts than in the center of their range. Bear colonizing fronts typically involve males at low densities, with females more centrally concentrated at higher densities (7,8).

Our measures reveal that (i) both sexes of radio-marked bears fed on adult moose carcasses along dispersal fronts at more than twice the rate than away from them; (ii) male bears alone did so about two-thirds more often along the front than in the center (Table 1); and (iii), most prominent, predation was more successful along the periphery than at the center (38 versus 0%; Table 1) (19), a pattern that was also repeated in North America. In Yellowstone National Park, where grizzly bears and moose have occurred sympatrically since at least the 1880s, predation on moose was not detected between 1959 and 1992 (20). However, along the current recolonization front in Jackson Hole, at least 10 adult moose were killed between 1996 and 2000. Our findings (Table 1) indicate that naı̈ve individuals are (i) conspicuously lacking in astuteness and (ii) experiencing a blitzkrieg.

Table 1

Summary of the major effects of colonizing carnivores on predator-experienced and -naı̈ve moose in Europe and North America. Sample sizes are given in parentheses under the “Predation” heading.

View this table:

If predation has the capacity to be both intense and swift, then how do naı̈ve prey avoid extinction, either now or post-Pleistocene? One survival model posits the rapid development of predator recognition through individual experience. Predator-naı̈ve mothers whose calves were killed by wolves colonizing Jackson Hole elevated their post-playback vigilance to wolf calls by about 500%, and the consequent latency to resumption of feeding increased from a mean of less than 30 s (in years before predation) to more than 6 min after (F 1, 89 = 32.516, P < 0.0001); additionally, a fourfold increase in site abandonment accompanied these feeding shifts (χ2 = 5.831, P < 0.016) (21). That such dramatic changes were caused by wolf predation per se and not by other factors is supported by two comparative analyses: (i) Mothers whose calves died because of starvation or collision with vehicles (N = 3) had lower vigilance (x̄ = 29) levels during playbacks than did mothers with wolf experience (x̄ = 92) (F 1, 16 = 24.625, P < 0.0001), and (ii) differences between these control mothers (who had not lost offspring to wolves) and non–wolf-savvy ones (x̄ = 17) were not detectable (P = 0.772). Moreover, mothers who lost young to predators at the Talkeetna (Alaska) site relocated their subsequent birth sites more than 1.75 times the distance from the prior birth site in the next year than did mothers who reared offspring successfully (22).

Our results suggest that offspring loss to predators may cause maternal hypersensitivity, and they illustrate how differences in the detection of cues may contribute to survival. The sounds of wolves elicited much stronger responses, at least after the initial phase of the blitzkrieg south of Yellowstone Park, than did other cues, and the magnitude of reactions exceeded those of predator-experienced Alaskan moose; responses to odors or the calls of scavengers did not (Fig. 4). Although the lack of predator recognition by naı̈ve prey may have been but one of several factors that contributed to reduced post-Pleistocene population viability (6), for the survivors it remains unclear whether the acquisition of antipredatory skills is maintained horizontally or vertically, or if it is just a transitory process (23).

Figure 4

Summary of comparative mean annual post-playback responses of adult female moose to four cues: olfactory (wolf urine and grizzly bear feces) and auditory (raven calls and wolf howls). The predation blitz is shown once only, but post-blitz responses to all four cues were possible (the differences between females of Teton-South and Teton-North are F = 24.625, P < 0.0001, and are reflected by the bold dotted line). Data on Alaskan females are as follows: Olfactory, wolf and grizzly (Talkeetnas and Denali); auditory, raven and wolf (Matunsuka, Talkeetnas, and Denali).

What is evident is that naı̈ve prey have the capacity to process information about predators swiftly—in the case of these moose, in a single generation. Where both young and adults are killed rapidly, opportunities for learning will be diminished and possibilities for local extinction increased. In contrast, where young are primary targets, learning may be transmitted vertically (24, 25), possibly reducing extinction thresholds. Whether behavioral mechanisms some 10,000 to 12,000 years ago retarded or facilitated prey viability as human hunters advanced remains uncertain. Nevertheless, the evidence involving recolonizing brown bears and wolves and moose is partially consistent with a blitzkrieg model but tempered by rapid adaptive learning (Fig. 4 andTable 1). If climate change was not responsible for the Quaternary extinctions (2), then perhaps species that failed to survive post-Pleistocene hunters were simply not quick learners.

Because large native carnivores continue to be reintroduced to parts of Africa, Europe, and North America (26,27), our findings have conservation relevance. In extant prey, rapid learning may prevent a complete blitzkrieg, which is something to be welcomed not only by contemporary human hunters but by those of us who favor fewer extinctions.

  • * To whom correspondence should be addressed. E-mail: berger{at}

  • Present address: Department of Animal Ecology, Swedish University of Agricultural Sciences, S-90183 Umea, Sweden.


View Abstract

Navigate This Article