Efﬁcacy of Firearms for Bear Deterrence
TOM S. SMITH,
Wildlife Sciences Program, Faculty of Plant and Wildlife Sciences, Brigham Young University, 451 WIDB, Provo,
UT 84602, USA
STEPHEN HERRERO, Environmental Science Program, Faculty of Environmental Design, University of Calgary, Calgary, AB, Canada T2N 1N4
CALI STRONG LAYTON, Wildlife Sciences Program, Plant and Wildlife Sciences Program, Brigham Young University, 448 WIDB, Provo,
UT 84602, USA
RANDY T. LARSEN, Wildlife Sciences Program, Faculty of Plant and Wildlife Sciences and Monte L. Bean Life Sciences Museum, Brigham Young
University, 407 WIDB, Provo, UT 84602, USA
KATHRYN R. JOHNSON,
Alaska Science Center, USGS, 1011 E. Tudor Road, Anchorage, AK 99502, USA
ABSTRACT We compiled, summarized, and reviewed 269 incidents of bear–human conﬂict involving
ﬁrearms that occurred in Alaska during 1883–2009. Encounters involving brown bears (Ursus arctos;
218 incidents, 81%), black bears (Ursus americanus; 30 incidents, 11%), polar bears (Ursus maritimus;
6 incidents, 2%), and 15 (6%) unidentiﬁed species provided insight into ﬁrearms success and failure. A
total of 444 people and at least 367 bears were involved in these incidents. We found no signiﬁcant difference
in success rates (i.e., success being when the bear was stopped in its aggressive behavior) associated with long
guns (76%) and handguns (84%). Moreover, ﬁrearm bearers suffered the same injury rates in close encounters
with bears whether they used their ﬁrearms or not. Bears were killed in 61% (n¼162) of bear–ﬁrearms
incidents. Additionally, we identiﬁed multiple reasons for ﬁrearms failing to stop an aggressive bear. Using
logistic regression, the best model for predicting a successful outcome for ﬁrearm users included species and
cohort of bear, human activity at time of encounter, whether or not the bear charged, and if ﬁsh or game meat
was present. Firearm variables (e.g., type of gun, number of shots) were not useful in predicting outcomes in
bear–ﬁrearms incidents. Although ﬁrearms have failed to protect some users, they are the only deterrent that
can lethally stop an aggressive bear. Where ﬁrearms have failed to protect people, we identiﬁed contributing
causes. Our ﬁndings suggest that only those proﬁcient in ﬁrearms use should rely on them for protection in
bear country. ß2012 The Wildlife Society.
KEY WORDS Alaska, bear deterrence, bear–human interactions, black bears, brown bears, firearms, grizzly bears, polar
bears, Ursus americanus,Ursus arctos,Ursus maritimus.
People who work and recreate in North American bear
habitat often fear bear encounters. Although the vast major-
ity of bear–human interactions are benign, some yield a
variety of adverse outcomes including destruction of proper-
ty, injuries, and fatalities to both bears and humans (Herrero
2002). Bear attacks are of great interest to the media and can
result in negative consequences for bear conservation
(Craighead and Craighead 1971, Miller and Chihuly
1987, Loe and Roskaft 2004). In 1967, for example, 2 human
fatalities in Glacier National Park led some to call for the
elimination of grizzly bears from America’s national parks
(Moment 1968, 1969). Ultimately, bears were seen as an
integral part of ecosystems and remained (Herrero 1970,
Craighead and Craighead 1971), but ongoing bear–human
interactions and attendant consequences persist.
Until the advent of bear spray in the 1980s, ﬁrearms were
the primary deterrent for safety in bear country (Smith et al.
2008). Even now, private, state, and federal agencies in
North America often require employees to carry ﬁrearms
while working in bear country. Although bear safety manuals
acknowledge the value of ﬁrearms, they also caution that
users must be proﬁcient under duress (Shelton 1994, Smith
2004, Gookin and Reed 2009). Furthermore, data regarding
ﬁrearm performance in aggressive bear encounters are lack-
ing. In fact, we could ﬁnd little published information
quantitatively addressing the effectiveness of ﬁrearms as
Herrero (2002) noted that ﬁrearms have their place in
protecting people from aggressive bears, but did not present
data regarding ﬁrearm use or efﬁcacy. Similarly, Bromley
et al. (1992) discussed many aspects of ﬁrearm use as bear
deterrents, but provided no supporting data. The United
States Fish and Wildlife Service (2002) stated that people
using ﬁrearms in bear encounters were injured 50% of the
time, but no data or references were provided as support
for this ﬁgure. Similarly, Meehan and Thilenius (1983)
Received: 20 December 2010; Accepted: 22 October 2011
Additional Supporting Information may be found in the online version
of this article.
Present Address: P.O. Box 4374, Palmer, AK 99645, USA.
The Journal of Wildlife Management; DOI: 10.1002/jwmg.342
Smith et al. Bears and Firearms 1
presented data regarding bullet performance at short range
with reference to bear attacks, but did not present ballistics
information from actual bear encounters.
Hence, a study of bear–human conﬂicts involving ﬁrearms
has not been conducted. Moreover, ﬁrearm efﬁcacy in
resolving bear–human conﬂict has not been quantiﬁed and
remains speculative. Our speciﬁc objectives for this paper
were to 1) review and summarize Alaskan bear–ﬁrearm
incidents and 2) identify factors associated with successful
use of ﬁrearms in bear–human conﬂicts in order to promote
both human safety in bear country and bear conservation.
Alaska is located in the northwestern portion of North
America and occupies an area of 1,530,699 km
. The human
population in Alaska was estimated to be 698,473 in 2009.
The brown or grizzly bear (Ursus arctos) ranges throughout
the state with the most recent estimate at 31,700 (Miller
1993). Black bears (Ursus americanus) are found in most
forested areas of Alaska. Formal population estimates do
not exist for black bears, but an Alaska Department of
Fish and Game (ADFG) biologist roughly estimated
them to number more than 50,000 (Harper 2007). Polar
bears (Ursus maritimus) are marine mammals that rarely
venture onto land (Amstrup 2003). In Alaska, polar bears
from both the Chukchi and Southern Beaufort Seas sub-
populations occasionally range up to 80 km inland, primarily
for maternal denning. Recent estimates put their numbers at
about 3,800 (Amstrup 2003).
Compilation and Summary
We compiled information on bear attacks from readily ac-
cessible state and federal records, newspaper accounts, books,
and anecdotal information that spanned the years 1883–
2009. We deﬁned an incident as a single bear–ﬁrearm event
that involved 1 or more people and 1 or more ﬁrearms. For
each incident we recorded the following variables to the
extent data were available: date, time, month, year, location
of incident, number of people, sex of people, activity at time
of interaction, whether or not people were making noise
prior to the encounter, probable cause of encounter, distance
to bear at time of encounter, bear species and cohort (age–sex
class), whether or not the bear charged, minimum distance to
the bear, presence of ﬁsh and/or game meat, type of ﬁrearms
used, number of shots ﬁred, warning shots, ﬁrearm efﬁcacy,
ﬁrearm ratio (number of ﬁrearms/number of people), dis-
tance to bear when shot, visibility of habitat (subjectively
rated poor, fair, good based on terrain and vegetation),
reasons for ﬁrearm ineffectiveness, extent of human injuries,
and extent of bear injuries. To overcome problems associated
with missing or unclear information, we limited the contri-
butions of each record to what we deemed were the most
trustworthy pieces of information. The review process was
subjective, but we feel conﬁdent that we limited our infer-
ences to a minimum while gleaning useful information for
For each of these incidents, we also used the following
categories to characterize probable causes for bear–human
encounters: surprised (people startled the bear), curiosity (the
bear’s motivation appears to have been curiosity), provoked
(e.g., a photographer crowding it or a hunter pursuing it),
predatory (the bear treated the human as potential prey), and
carcass defense (the bear defending a food source). We
also subjectively evaluated injuries as follows: slight injuries
included nips, limited biting, and scratches where hospitali-
zation was not required; moderate injuries required hospi-
talization to some degree, and included punctures, bite
wounds and broken bones; and severe injuries resulted in
extended hospitalization and often permanent disability.
We deﬁned a charge as an agonistic behavior typiﬁed by a
sudden rush, or lunge, toward the perceived threat. Some
charges terminated prior to contact (i.e., bluff charges)
whereas others resulted in contact. For distance to bears,
we regrouped values into broad categories (i.e., <10 m,
10–25 m, 26–50 m, and >51 m) based on speciﬁcity in
the accounts. We used reported distances for greater accuracy
(e.g., distance to bear when shot) whenever possible.
We deemed use of a ﬁrearm successful (response coded as
1) when it stopped the offensive behavior of the bear. These
successes included incidents where bears no longer pursued a
person, broke off an attack, abandoned attempts to acquire
food or garbage, were killed, or turned and left the area as a
result of ﬁrearm use. Conversely, ﬁrearm failures (response
coded as 0) occurred when the bear continued its pursuit,
persisted in attempts to acquire food or garbage, or showed
no change in behavior after ﬁrearm use. We excluded inci-
dents from our analysis where ﬁrearms were available but no
attempt to use them was made.
We used the G-test of independence (Dytham 2003) when
we had 2 nominal variables, each with 2 or more possible
values, and we wanted to compare frequencies of one to the
other. We also tested the equality of sample means with a
2-sample t-test. We used the Z-test to compare the propor-
tions from 2 independent groups to determine if they were
different. We set signiﬁcance at P<0.05.
To understand the relationship between variables and
incident outcomes, we used logistic regression where the
response variable was success (1) or failure (0) of ﬁrearms.
We identiﬁed candidate models representing different
hypotheses related to ﬁrearm success as a function of bear,
human, ﬁrearm, and spatio-temporal factors (Table 1). Our
analysis followed several steps. First, we evaluated the
number of records and odds associated with different
categories of our explanatory variables to determine which
should be collapsed or combined. Second, we used Akaike’s
Information Criterion adjusted for small sample sizes (AIC
to rank models (Akaike 1973, Burnham and Anderson 2002)
for each variable type (i.e., bear, ﬁrearm, human, spatio-
temporal). We then used the top model and competing
<2.0) within each type in the third stage
of analysis similar to Doherty et al. (2008). Here we com-
bined variables and ranked models based on smallest AIC
2 The Journal of Wildlife Management
identify a best approximating model. We then evaluated
these models and their associated variables to identify any
uninformative parameters that did not improve AIC
discarded them (Arnold 2009).
Because the best approximating model had high AIC
¼0.96), we used it to evaluate the direction
and strength (odds ratios) of associations between explana-
tory variables and ﬁrearm efﬁcacy. To test for lack of model
ﬁt, we calculated Hosmer and Lemeshow’s (2000) goodness
of ﬁt statistic. We also used the top model in a 5-fold cross
validation exercise where we withheld 1/5th of the data and
estimated model coefﬁcients. We then used the estimated
coefﬁcients to predict incident outcomes for the withheld
data. We calculated the proportion of outcomes accurately
predicted (estimated probability of success >0.50 for suc-
cessful outcomes and 0.50 for unsuccessful outcomes) for
the withheld data and repeated this process until we obtained
a prediction and accuracy for each observation.
A total of 444 people were involved in 269 incidents. At least
357 bears, including dependent offspring, were involved in
269 incidents, including 300 brown bears (84%), 36 black
bears (10%), 6 polar bears (2%), and 15 of unknown species
(4%). Bear-inﬂicted injuries occurred in 151 of 269 (56%)
incidents (see Supplemental Information available online at
www.onlinelibrary.wiley.com for additional details regarding
characteristics of bear incidents).
Success rates by ﬁrearm type were similar with 84% of
handgun users (31 of 37) and 76% of long gun users (134
of 176) successfully defending themselves from aggressive
bears (Z¼1.0664, P¼0.2862). When we compared out-
comes for people who used their ﬁrearm in an aggressive bear
encounter (n¼229) to those who had ﬁrearms but did not
use them (n¼40), we found no difference in the outcome
¼0.691, P¼0.708), whether the outcome was no
injury, injury, or fatality. However, we found a difference
in the outcome for bears with regard to ﬁrearm use: 172 bears
died when people used their ﬁrearms, whereas no bears were
killed when ﬁrearms were not used.
Firearms failed to protect people for a variety of reasons
including lack of time to respond to the bear (27%), did not
use the ﬁrearm (21%), mechanical issues (i.e., jamming;
14%), the proximity to bear was too close for deployment
(9%), the shooter missed the bear (9%), the gun was emptied
and could not be reloaded (8%), the safety mechanism was
engaged and the person was unable to unlock it in time to use
the gun (8%), people tripped and fell while trying to shoot
the bear (3%), and the ﬁrearm’s discharge reportedly trig-
gered the bear to charge that ended further use of the gun
With respect to efforts to model ﬁrearm efﬁcacy, we
classiﬁed 156 incidents as successful. Our initial evaluation
of the number of incidents assigned to each category and
associated odds (Appendix 1) suggested (95% CI on odds
ratio included 1) collapsing the species category to black
bears and other (brown, polar, and unknown bears). This
same evaluation suggested our initial breakdown of cohorts
and seasons could be collapsed to female and other (pairs,
males, unknown). Similarly, we combined seasons into 2
categories for summer and other (spring, fall, winter,
The best model for ﬁrearm success relative to bear variables
included species, cohort, and whether the bear charged
(Table 2). For ﬁrearm variables, 2 models received enough
<2.0) to be advanced to the third stage of
analysis. These models included ﬁrearm ratio, number of
warning shots, and success of warning shots. We found little
support for models that included ﬁrearm type (Table 2).
Analysis of human and spatio-temporal variables indicated
activity, distance, group size, noise, presence of ﬁsh or game,
season, and visibility inﬂuenced ﬁrearm success (Table 2).
After combining variables across categories, a model in-
cluding species, cohort, whether the bear charged, group size,
human activity, noise, and presence of ﬁsh or game had the
weight. Because separate removal of group size
and noise improved (reduced) the AIC
values compared to
removal of other variables, we considered them uninforma-
tive and eliminated them from the top model. These same 2
variables also had 85% conﬁdence intervals that spanned
zero—further suggesting they were uninformative (Arnold
2009). The resulting top model accounted for 96% of the
weight and was more than 6 DAIC
better than the next
competing models (Table 3). Hosemer–Lemeshow’s good-
Table 1. Description of variables used in models of firearm success in bear incidents in Alaska, USA during 1883–2009.
Variable Category Description
Species Bear Species of bear involved in event (black, brown, polar)
Cohort Bear Cohort of bear involved in event (unknown, pairs, female with young, female, and male)
Charge Bear Bear charged (yes, no, unknown)
Group size Human Group size (count of people present)
Activity Human Activity code of humans (active, intermediate, sedentary, unknown)
Fish/game Human Presence of fish or game (no, game, fish, unknown)
Noise Human Noise associated with activity (no, yes, unknown)
Firearm type Firearm Firearm type (handgun, long gun, both, unknown)
No. of shots Firearm Number of shots fired
Warning shots Firearm Number of warning shots fired
Firearm ratio Firearm Number of guns divided by number of individuals in group
Distance Spatio-temporal Distance from bear when firearm discharged (<10 m, 10–20 m, 21–30 m, 31–40 m, >40 m)
Visibility Spatio-temporal Visibility at site (poor, good, unknown)
Season Spatio-temporal Season of incident (spring, summer, fall, winter, unknown)
Smith et al. Bears and Firearms 3
ness of ﬁt statistics (P¼0.26) provided no evidence of lack
of ﬁt. Similarly, 5-fold cross validation using the top model
produced an accurate prediction of incident outcomes on
withheld data for 71.9% of incidents—further suggesting
When the animal involved in the incident was a black bear,
odds of ﬁrearm success were more than 38 times greater than
when the bear was a brown, polar, or unknown bear
(Table 4). Similarly, females without young were associated
with a nearly 7-fold increase in odds of ﬁrearm success
(Table 4). Conversely, odds of ﬁrearm success were nega-
tively associated with human activity level and charging
behavior by involved bears. Odds of ﬁrearm success were
12 and 24 times greater for intermediate and sedentary
activity levels, respectively, compared to people considered
active (Table 4). Once a bear charged, odds of ﬁrearm success
decreased nearly 7-fold (Table 4). Interestingly, the presence
of ﬁsh or game meat was associated with increases of 4 and 8,
respectively, in odds of ﬁrearm success.
Brown bears were disproportionately involved (81%) in these
encounters, a ﬁnding that is consistent with the widely held
perception that brown bears are considerably more aggressive
and hence, more likely to be involved in bear–human conﬂict
leading to injury, than the other 2 species (Herrero and
Table 3. Ranking of supported models (DAIC
<10.0) describing firearm success as a function of bear, firearm, human, and spatio-temporal influences in
Alaska, USA during 1883–2009.
Model structure AIC
Species þCohort þCharge þActivity þFish/Game 276.11 0.0 0.96 11 253.00
Success Species þCohort þCharge þDistance þVisibility þSeason 282.93 6.8 0.03 11 259.82
Success Group size þActivity þFish/Game þNoise þDistance þVisibility þSeason 285.28 9.2 0.01 16 250.94
Akaike’s Information Criterion adjusted for small sample sizes.
Change in AIC
from top model.
Number of estimable parameters.
Reduced model after removal of group size and noise which were uninformative.
Table 2. Ranking of supported models (DAIC
<10.0) describing firearm success as a function of bear (species, cohort, charge), firearm (firearm type, no.
shots, warning shots, firearm ratio), human (group size, activity, fish/game, noise), and spatio-temporal (distance, visibility, and season) influences in Alaska,
USA during 1883–2009.
Model structure AIC
Success Species þCohort þCharge 296.30 0.0 0.97 5 286.05
Success Species þCohort 303.82 7.5 0.02 3 297.72
Success Species þCharge 305.51 9.2 0.01 4 297.35
Success No. shots þFirearm ratio þWarning shot 321.31 0.0 0.40 7 306.85
Success No. shots þFirearm ratio 322.34 1.0 0.24 4 314.18
Success Firearm type þNo. shots þFirearm ratio þWarning shot 324.23 2.9 0.09 10 303.31
Success No. shots þWarning shot 324.45 3.1 0.08 5 314.20
Success Firearm ratio þWarning shot 325.80 4.5 0.04 6 313.45
Success Firearm type þNo. shots þFirearm ratio 325.98 4.7 0.04 7 311.52
Success No. shots 326.28 5.0 0.03 2 322.23
Success Firearm type þNo. shots þWarning shot 327.12 5.8 0.02 8 310.52
Success Firearm ratio 327.22 5.9 0.02 3 321.11
Success Warning shot 328.83 7.5 0.01 4 320.67
Success Firearm type þFirearm ratio þWarning shot 328.93 7.6 0.01 9 310.18
Success Firearm type þNo. shots 329.78 8.5 0.01 5 319.53
Success Firearm type þFirearm ratio 331.32 10.0 0.00 6 318.97
Success Group size þActivity þFish/Game þNoise 305.27 0.0 0.59 10 284.35
Success Group size þFish/Game þNoise 307.91 2.6 0.16 7 293.45
Success Activity þFish/Game þNoise 308.16 2.9 0.14 9 289.41
Success Group size þActivity þFish/Game 310.27 5.0 0.05 8 293.67
Success Fish/Game þNoise 311.11 5.8 0.03 6 298.76
Success Group size þActivity þNoise 313.03 7.8 0.01 7 298.57
Success Activity þNoise 314.12 8.8 0.01 6 301.77
Success Distance þVisibility þSeason 300.26 0.0 0.74 7 285.79
Success Distance þSeason 302.36 2.1 0.26 6 290.01
Akaike’s Information Criterion adjusted for small sample sizes.
Change in AIC
from top model.
Number of estimable parameters.
4 The Journal of Wildlife Management
Higgins 1999, 2003; Herrero 2002). Female bears with
dependent young comprised the second-most common co-
hort involved in ﬁrearm incidents. Surprise encounter was
the reason most often given for conﬂict with this cohort.
Brown bear family groups suddenly confronted by people
were commonly aggressive-defensive, as they protected their
cubs from a perceived threat. Too few incidents involving
black and polar bear family groups (n¼4 and 1, respective-
ly) occurred to support meaningful conclusions, but Herrero
(2002) reported that black bears rarely attack people in
response to sudden encounters. Single bears comprised the
most common cohort involved in ﬁrearms incidents, a
reﬂection of the relative frequency of that cohort in North
American bear populations (Jonkel and Cowan 1971,
Schwartz et al. 2003) and the fact that single bears are the
most hunted cohort.
Although ﬁrearms were successful (84% handgun; 76%
long gun) in deterring aggressive bears in the records we
studied, we do not claim that these rates represent the
outcome for all bear–ﬁrearm incidents throughout Alaska.
When we initiated this study in the late 1990s, we had access
to the Alaska Department of Fish and Game’s defense of life
or property (DLP) records. However, privacy laws restricted
our access to records from 2001 to present. This incomplete
record potentially affects 3 ﬁndings: the number and type of
human injuries, the number and type of bear injuries, and
ﬁrearm success rates. First, because bear-inﬂicted injuries are
closely covered by the media, we likely did not miss many
records where people were injured. Therefore, even if more
incidents had been made available through the Alaska DLP
database, we anticipate that these would have contributed
few, if any, additional human injuries. Second, including
more DLP records would have increased the number of
bears killed by ﬁrearms. Finally, additional records would
have likely improved ﬁrearm success rates from those
reported here, but to what extent is unknown.
Our modeling results indicated that models with ﬁrearm
variables had very limited support (Tables 1 and 3). The type
of ﬁrearm, number of shots taken, whether or not the people
ﬁred warning shots, and how many ﬁrearms were present in
the group had minimal inﬂuence on the outcome. Success
was best predicted by a model that included species and
cohort of bear, whether or not the bear charged, human
activity level, and if ﬁsh or game meat were available. These
ﬁndings coupled with odds ratios from univariate analyses
(Appendix 1) afﬁrm some of the conventional advice for
avoiding bear encounters: hike in a group, avoid areas of poor
visibility, be more cautious when in brown bear country, and
make noise to avoid startling females with dependent young
(Herrero 2002, Smith 2004).
Although bear spray, pyrotechnics, noise makers, and other
deterrents may alter a bear’s behavior, only a ﬁrearm provides
a lethal force option. However, interviews revealed that some
people were hesitant to use lethal force for fear of shooting
the person being attacked, or because they did not want to
have to skin the bear and pack out its hide, skull, and claws as
required by law. Additionally, some people admitted that
they were reluctant to shoot a protected species. In some
cases, this reluctance proved detrimental when split second
decisions were required for the person to defend themselves
from an aggressive, attacking bear. The decision regarding
which deterrent to use is a personal one, but the consequen-
ces of attempting to use lethal force should be carefully
Firearm type received very little support, suggesting that
efﬁcacy of the ﬁrearm was unrelated to whether people used a
handgun or long gun. Considering the high intensity, rapidly
unfolding, close-quartered, and chaotic nature of bear
attacks, these results are not surprising. Hence, we cannot
recommend one class of weapon over the other. We did not
have data regarding the level of expertise associated with
those who carried ﬁrearms. Regardless, a person’s skill level
plays an inﬂuential role in determining the outcome in bear–
Firearms should not be a substitute for avoiding unwanted
encounters in bear habitat. Although the shooter may be able
to kill an aggressive bear, injuries to the shooter and others
also sometimes occur. The need for split-second deployment
and deadly accuracy make using ﬁrearms difﬁcult, even for
experts. Consequently, we advise people to carefully consider
their ability to be accurate under duress before carrying a
ﬁrearm for protection from bears. No one should enter bear
country without a deterrent and these results show that
ﬁrearms are not a clear choice. We encourage all persons,
Table 4. Logistic regression coefficients, standard errors (SE), odds ratios, and 95% confidence intervals from the highest ranked model (Akaike’s Information
Criterion adjusted for small sample sizes (AIC
) weight ¼0.96) of firearm success as a function of bear, firearm, human, and spatio-temporal influences in
Alaska, USA during 1883–2009.
Coefficient Estimate SE Odds ratio Lower 95% CI Upper 95% CI
Intercept 0.984 1.017
Black bear 3.652 1.180 38.56 5.92 855.38
Female 1.916 0.718 6.79 1.93 34.80
Charge 1.921 0.783 0.15 0.02 0.58
Charge unknown 3.283 0.980 0.04 0.00 0.22
Intermediate activity 2.485 0.912 12.01 2.40 97.04
Sedentary activity 3.180 0.983 24.04 4.11 216.71
Unknown activity 2.141 1.140 8.51 1.01 97.67
Game 1.427 0.482 4.17 1.67 11.21
Fish 2.109 0.690 8.24 2.40 38.86
Unknown 0.662 0.366 1.94 0.95 4.02
Smith et al. Bears and Firearms 5
with or without a ﬁrearm, to consider carrying a non-lethal
deterrent such as bear spray because its success rate under a
variety of situations has been greater (i.e., 90% successful for
all 3 North American species of bear; Smith et al. 2008) than
those we observed for ﬁrearms.
The United States Geological Survey Alaska Science Center
and Brigham Young University provided support for this
project. We gratefully acknowledge a number of reviewers
who have provided guidance regarding this manuscript, the
Associate Editor and Editor-in-Chief for the Journal of
Wildlife Management in particular. D. Hardy (ADFG re-
tired) provided useful comments that have helped make this
more valuable. Additionally, we appreciate those who helped
us collect bear incident data.
Akaike, H. 1973. Information theory as an extension of the maximum
likelihood principle. Pages 267–281 in B. Petran and F. Csaki, editors.
International Symposium on Information Theory. Second edition.
Akademiai Kiado, Budapest, Hungary.
Amstrup, S. C. 2003. Polar bear. Pages 587–610 in G. A. Feldhamer, B. C.
Thompson, and J. A. Chapman, editors. Wild Mammals of North
America: Biology, Management, and Conservation. Second edition.
The Johns Hopkins University Press.
Arnold, T. W. 2009. Uninformative parameters and model selection using
Akaike’s information criterion. Journal of Wildlife Management 74:1175–
Bromley, M., L. H. Graf, P. L. Clarkson, and J. A. Nagy. 1992. Safety in
bear country: a reference manual. Department of Renewable Resources,
Yellowknife, Northwest Territories, Canada.
Burnham, K. P., and D. A. Anderson. 2002. Model selection and multi-
model inference: a practical information-theoretic approach. Second
edition. Springer-Verlag, New York, New York, USA.
Craighead, J. J., and F. C. Craighead. 1971. Grizzly bear–man relationships
in Yellowstone National Park. BioScience 21:845–857.
Doherty, K. E., D. E. Naugle, B. L. Walker, and J. M. Graham. 2008.
Greater sage-grouse winter habitat selection and energy development.
Journal of Wildlife Management 72:187–195.
Dytham, C. 2003. Choosing and using statistics: a biologist’s guide.
Blackwell Publishing, Boston, Massachusetts, USA.
Gookin, J., and T. Reed. 2009. NOLS bear essentials: hiking and camping in
bear country. Stackpole Books, Mechanicsburg, Pennsylvania, USA.
Harper, P. 2007. Black bear management of survey-inventory activities.
Alaska Department of Division of Wildlife Conservation, Juneau, USA.
Herrero, S. 1970. Human injury inﬂicted by grizzly bears. Science 170:593–
Herrero, S. 2002. Bear attacks: their causes and avoidance. Lyons & Burford
Publishers, New York, New York, USA.
Herrero, S., and A. Higgins. 1999. Human injuries inﬂicted by bears in
British Columbia: 1960–1997. Ursus 11:209–218.
Herrero, S., and A. Higgins. 2003. Human injuries inﬂicted by bears in
Alberta: 1960–1998. Ursus 14:44–54.
Hosmer, D., and S. Lemeshow. 2000. Applied Logistic Regression. Second
Edition. John Wiley & Sons, Inc., New York.
Jonkel, C. J., and I. McT. Cowan. 1971. The black bear in the spruce-ﬁr
forest. Wildlife Monographs 27.
Loe, J., and E. Roskaft. 2004. Large carnivores and human safety: a review.
Meehan, W. R., and J. F. Thilenius. 1983. Safety in bear country: protective
measures and bullet performance at short range. USDA Forest Service
General Technical Report PNW-152, Portland, Oregon, USA.
Miller, S. D. 1993. Brown bears in Alaska: a statewide management over-
view. Alaska Department of Fish and Game, Wildlife Technical Bulletin
Number 11, Juneau, USA.
Miller, S. D., and M. A. Chihuly. 1987. Characteristics of nonsport brown
bear deaths in Alaska. International Conference on Bear Research and
Moment, G. B. 1968. Bears: the need for a new sanity in wildlife conserva-
tion. BioScience 18:1105–1108.
Moment, G. B. 1969. Bears and conservation: realities and recommenda-
tions. BioScience 19:1019–1020.
Schwartz, C. C., S. D. Miller, and M. A. Haroldson. 2003. Grizzly bear.
Pages 556–586 in G. A. Feldhamer B. C. Thompson, and J. A. Chapman,
editors. Wild mammals of North America: biology, management, and
conservation. Second edition. Johns Hopkins University Press, Baltimore,
Shelton, J. G. 1994. Bear encounter survival guide. Shelton Publishing,
Hagensborg, British Columbia, Canada.
Smith, D. 2004. Backcountry bear basics: the deﬁnitive guide to avoiding
unpleasant encounters. The Mountaineers, Seattle, Washington, USA.
Smith, T. S., S. Herrero, T. D. DeBruyn, and J. M. Wilder. 2008. Efﬁcacy of
bear deterrent spray in Alaska. Journal of Wildlife Management 72:640–
United States Fish Wildlife Service [FWS]. 2002. Bear spray vs. bullets:
which offers better protection? Fact Sheet No. 8. <http://www.fws.gov/
Accessed 10 Dec 2010.
Associate Editor: Scott McCorquodale.
6 The Journal of Wildlife Management
Appendix 1. Univariate analysis showing log-likelihood, odds, and associated 95% confidence intervals for firearm success in relation to bear, firearm, human,
and spatio-temporal influences in Alaska, USA 1883–2009.
Category Variable Log-likelihood Df Odds Lower 95% CI Upper 95% CI
Bear Species (polar) 152.05 4 2.00 0.39 14.43
Black 13.50 1.07 334.95
Brown 0.65 0.09 3.40
Unknown 2.75 0.26 30.43
Cohort (single) 157.26 5 1.19 0.79 1.80
Female w/young 1.12 0.59 2.15
Female 7.00 2.25 30.86
Male 1.81 0.92 3.63
Pair 1.68 0.31 12.56
Charge (none) 156.77 3 7.33 2.54 30.98
Yes 0.23 0.05 0.69
Unknown 0.07 0.01 0.28
Firearm Type (handgun) 163.40 4 2.44 1.16 5.60
Long gun 0.64 0.27 1.44
Both 0.27 0.03 1.91
Unknown 0.76 0.27 2.05
No. shots 161.12 2 1.20 1.04 1.38
Warning shot (none) 160.33 4 1.63 1.22 2.19
Success 2.05 0.60 9.37
Unsuccessful 1.60 0.75 3.66
Unknown 0.23 0.05 0.82
Firearm ratio (1/group) 160.56 4 6.33 2.16 26.96
1/Person 0.22 0.05 0.67
Unknown 0.31 0.07 1.03
Human Group Size 160.99 2 1.68 1.14 2.53
Activity (active) 157.41 4 0.30 0.07 0.98
Intermediate 5.71 1.68 26.13
Sedentary 11.33 2.86 58.44
Unknown 2.86 0.55 17.52
Fish/game (none) 157.50 4 1.18 0.79 1.76
Game 2.74 1.21 6.68
Fish 5.94 1.89 26.32
Unknown 1.25 0.70 2.24
Noise (none) 155.83 3 1.25 0.87 1.81
Yes 4.16 1.99 9.40
Unknown 1.01 0.55 1.83
Spatio-temporal Distance (<10 m) 151.90 5 1.20 0.81 1.78
11–25 m 2.45 1.23 5.08
26–50 m 1.91 0.93 4.10
>51 m 13.38 2.58 246.08
Unknown 0.44 0.17 1.07
Visibility (poor) 163.69 3 1.59 1.12 2.29
Good 0.78 0.37 1.68
Unknown 1.30 0.73 2.33
Season (spring) 154.04 5 1.17 0.54 2.57
Summer 3.48 1.35 9.06
Fall 1.25 0.51 3.01
Winter 0.86 0.31 2.33
Unknown 0.38 0.08 1.49
Smith et al. Bears and Firearms 7