Oecologia (1991) 88:161 166
Oecologia
9 Springer-Verlag 1991
Pathogens as a factor limiting the spread of cannibalism
in tiger salamanders
David W . Pfennig, Michael L.G. Loeb, and J a m e s P. Collins
Department of Zoology, Arizona State University, Tempe, AZ 85287 1501, USA
Received March 4, 1991 / Accepted in revised form July 1, 1991
S u m m a r y . Intraspecific predation is taxonomically wide-
spread, but few species routinely prey on conspecifics.
This is surprising as conspecifics could be a valuable
resource for animals limited by food. A potential cost of
cannibalism that has been largely unexplored is that it
may enhance the risk of acquiring debilitating pathogens
or toxins from conspecifics. We examined how pathogens
affect variation in the incidence of cannibalism in tiger
salamander larvae (Ambystoma tigrinum nebuIosum),
which occur as two environmentally-induced morphs,
typicals and cannibals. Salamanders from one population were more likely than those in another to develop
into cannibals, even when reared under identical conditions. Variation in the propensity to become a cannibal
may be caused by variation in pathogen density. In the
population with cannibals at low frequency, bacterial
blooms in late summer correlated with massive die-offs
of salamanders. The frequency of cannibals correlated
significantly negatively with bacterial density in ten different natural lakes. In the laboratory, cannibals exposed
to a diseased conspecific always preyed on the sick animal. As a result, cannibals were more likely to acquire
and die from disease than were typicals that were similarly exposed, or cannibals that were exposed to healthy
conspecifics. Since conspecifics often share lethal pathogens, enhanced risk of disease may explain why cannibalism is generally infrequent. Pathogens may constrain not
only the tendency to be behaviorally cannibalistic, but
also the propensity to develop specialized cannibal morphologies.
Key words: Adaptive plasticity - C a n n i b a l i s m - Mass
mortality - Pathogens - Trophic polymorphism
More than any other intraspecific interaction, cannibalism, the ingestion of a living conspecific, consistently
Offprint requests to: J.P. Collins
generates the greatest fitness differential between interactants. Largely for this reason researchers have long speculated on the possible adaptive significance of this behavior. Only recently, with the greater emphasis on developing causal explanations based on selection of individuals,
has cannibalism become regarded as adaptive in certain
situations (Polis 1981). For example, organisms in
ephemeral habitats may increase development and
growth, and hence survivorship, through cannibalism
(Eickwort 1973; Crump 1983, 1990). Despite this apparent advantage, organisms that spend at least part of
their life in ephemeral habitats, e.g., larval insects and
amphibians, typically avoid cannibalism and entire cohorts may die for lack of food and time to develop
(exceptions include rare species with specialized cannibal
morphs.) Why is cannibalism infrequent, especially in
animals for which food may consistently become limiting?
There are at least three major costs to cannibalism.
First, cannibals risk injury or death from struggling conspecifics (Polis 1981) or their guardians (Sherman 1981).
Nonetheless, interspecific predation is frequent although
similar costs should accrue to any predator. Second,
cannibals may experience reduced inclusive fitness if they
ingest kin (Polis 1981 ; Crump 1983). Unless close kin are
consumed at high rates, however, a reduction in fitness
due to consuming kin may be more than offset by gains
in individual fitness (e.g., Eickwort 1983). Third, cannibals may be debilitated or killed by pathogens or toxins
acquired from victims (Polis 1981). This cost has not
been studied in detail, which is surprising given recent
theoretical and empirical studies indicating that selection
to avoid parasites may affect many behaviors (e.g., see
Read 1988). In this paper we evaluate the potential role
of pathogens in affecting the incidence of cannibalism in
a larval amphibian.
Amphibians are an excellent model to study the factors influencing cannibalism. Cannibalism has been documented in 12 of the 21 families of anurans, seven of the
nine families of salamanders, and one family of caecilians
162
( C r u m p 1992). Species t h a t c o n s u m e conspecifics o n a
r e g u l a r basis, often as larvae, o c c u r in o n l y 5 a n u r a n
families a n d one s a l a m a n d e r f a m i l y (see Polis a n d M y e r s
1985; C r u m p 1992). R e s o l v i n g the q u e s t i o n o f w h y c a n n i b a l i s m is n o t m o r e f r e q u e n t w i t h i n a m p h i b i a n s m a y be
a c c o m p l i s h e d b y e x a m i n i n g a species w i t h discrete cannibalistic a n d n o n c a n n i b a l i s t i c p h e n o t y p e s a n d a s k i n g
w h a t affects the f r e q u e n c y o f e a c h m o r p h .
C e r t a i n p o p u l a t i o n s o f at least three subspecies o f
tiger s a l a m a n d e r s , Ambystoma tigrinum G r e e n , have discrete j u v e n i l e m o r p h s (Collins et al. 1980; L a n n o o a n d
B a c h m a n n 1984): a t y p i c a l m o r p h t h a t feeds m o s t l y o n
p l a n k t o n a n d insect larvae, o r a b e h a v i o r a l l y , d e v e l o p m e n t a l l y , a n d m o r p h o l o g i c a l l y distinctive c a n n i b a l
m o r p h t h a t feeds m o s t l y o n conspecifics (Collins et al.
1980; Collins a n d H o l o m u z k i 1984; L a n n o o a n d Bachm a n n 1984). C a n n i b a l s are also larger, m e t a m o r p h o s e at
a y o u n g e r age ( L a n n o o a n d B a c h m a n n 1984; u n p u b l .
d a t a ) , a n d differ f r o m typicals in s h a p e o f h e a d a n d
t o o t h - b e a r i n g b o n e s a n d length o f teeth ( P o w e r s 1907;
P e d e r s e n 199l). C a n n i b a l s a r e i n d u c e d b y high conspecific densities (Collins a n d C h e e k 1983).
W e s u s p e c t e d t h a t p a t h o g e n s m a y affect the incidence
o f c a n n i b a l i s m in A. t. nebulosum H a l l o w e l l . C o n s i d e r a b l e v a r i a t i o n in the f r e q u e n c y o f c a n n i b a l s occurs w i t h i n
A r i z o n a (Collins 1981). W h e r e c a n n i b a l s are i n f r e q u e n t
in n o r t h e r n A r i z o n a a n d U t a h , s a l a m a n d e r p o p u l a t i o n s
t y p i c a l l y experience m a s s m o r t a l i t y in w h i c h m o s t salam a n d e r l a r v a e b e c o m e l e t h a r g i c a n d die w i t h i n a w e e k
( W o r t h y l a k e a n d H o v i n g h 1989; B e r n a 1990). M a s s
m o r t a l i t y o f t e n occurs as the s a l a m a n d e r s a p p r o a c h
m e t a m o r p h o s i s ( B e r n a 1990) a n d m a y be m e d i a t e d b y
b a c t e r i a ( W o r t h y l a k e a n d H o v i n g h 1989) o r o t h e r p a t h o gens. W e also n o t e d t h a t in l a k e s with d i s e a s e d s a l a m a n ders c a n n i b a l s o c c u r r e d d i s p r o p o r t i o n a t e l y a m o n g the
casualties. This s t u d y e x a m i n e s w h e t h e r the incidence o f
p a t h o g e n s influences the f r e q u e n c y o f c a n n i b a l s in these
animals.
Materials and methods
Study area
We studied A. t. nebulosum on two plateaus in Arizona, USA, that
separated 5 mya by incision (1500 m deep) of the Grand Canyon
(Rice 1988): Kaibab Plateau ( ~ 2312 km 2, 2000-2700 m in elevation) and Mogollon Plateau ( ~ 20000 km 2, 2000-2400 m in elevation). The Kaibab Plateau extends north of the Grand Canyon; the
Mogollon Plateau is about 110 km south. The plateaus are covered
by Petran montane conifer forest or Petran subalpine conifer forest
(Brown 1982). Both have numerous, natural, sinkhole lakes that fill
in winter, and both receive about 63 cm of precipitation annually
(Green and Sellers 1964). The lakes fill from snowmelt during
March to April and from rainfall during late summer. In the summer most lakes are small (< 50 m diameter), shallow (< 1 m depth)
and filled with emergent vegetation. In addition to salamanders, the
lakes support algae, protozoans, rotifers, cladocerans, fairy shrimp
(Anostraca), molluscs, numerous aquatic insects, and anuran larvae. None of the lakes contain fish or other species of salamanders.
A complete description of invertebrates is in Holomuzki (1986) and
Holomuzki and Collins (1987).
Comparison of frequency of cannibals across two regions
Earlier studies suggested that cannibals were less frequent on the
Kaibab Plateau than on the Mogollon Plateau (e.g., Berna 1990).
To test this, we compared frequencies of cannibals within lakes in
the two regions. We compiled a list of lakes in both regions that
supported salamanders previously. We used a random number table
to select 8 lakes in each region from this list. About 8 w after the
first eggs were laid in each lake we censused larval salamanders in
these populations by placing a drop box at 5-15 randomly-selected
positions in each lake. Drop box sampling of anuran tadpole densities is highly reliable (Pfennig 1990). For lakes that ultimately
experienced mass mortality of salamanders, we conducted censuses
well before there were any obvious signs of diseased salamanders.
Our censuses indicated that natural habitats on the Kaibab
Plateau had significantly lower frequencies of cannibals than did
those on the Mogollon Plateau (see results). To see if this difference
was due to interregional differences in density, the cue that induces
cannibals (Collins and Cheek 1983), we used two approaches. First,
we compared densities of salamanders in each region using estimates derived from the drop boxes. Second, we reared salamanders
from each habitat under common conditions in the laboratory. We
sampled eggs from each lake soon after the snow began melting (late
March). By comparing developmental stages of eggs from the lakes
with eggs laid in the laboratory, we found that most eggs collected
in the field were laid recently (< 1 d old). We kept the eggs from each
lake in a common container in the laboratory. After the eggs
hatched we placed larvae in separate 250 ml cups filled with dechlorinated tap water and exposed them to common conditions (20 ~ C,
11:13 h photoperiod). Within 5-7 days of hatching we randomly
placed the larvae in 20 1 aquaria at 4 different densities (3, 5, 7, and
15 salamanders per 20 1). These densities encompass the limits of
estimates from the field. We provided each larva with 0.15 g of live
brine shrimp (Artemia sp.) per d. At 6 w after hatching we scored
each animal's morphotype using criteria in Powers (1907). We used
an analysis of variance to examine the effects of region, initial
density, and their interaction on the number of cannibals produced.
To reduce the effects of unequal variances, we used the
square root of the sum of the number of cannibals in each treatment
plus 0.375 (Bartlett 1971). An initial one-way analysis of variance
indicated no significant differences in the number of cannibals
produced among lakes within each region (.F7,88 = 1.374, P = 0.226;
Fv,ss = 1.426, P = 0.205; Mogollon and Kaibab Plateau Lakes, respectively). We therefore pooled these sums of squares and degrees
of freedom with the error term within each region and treated the
eight lakes within a treatment as replicates. We analyzed the data
with a balanced design, two-way, fixed-effects analysis of variance.
A Games-Howell test was used to contrast means.
Causes of differences between regions in frequency of
cannibals
To see if disease influences frequency of cannibals, we noted the
number of lakes in each region with signs of mass mortality. Typically, in the late summer most larval salamanders in lakes on the
Kaibab Plateau become lethargic, develop reddening around the
venter and legs, and die (Berna 1990). The entire process may take
a week. We operationally defined a salamander population as
having experienced mass mortality if we found in the lake any dead
or diseased salamanders exhibiting the above symptoms. All such
assignments were unambiguous (i.e., in diseased populations, the
vast majority of salamanders were floating on the lake surface). We
visited lakes in each region periodically and used a seine to ascertain
the health of the salamanders.
To see if disease selects differentially against cannibals, we collected 7 cannibals and 5 typicals from a lake on the Kaibab Plateau
with no diseased salamanders. We then collected lethargic salamanders from another lake that was in the early stages of mass mortality. We randomly assigned each healthy animal to a 20 1 aquarium
163
with a diseased animal. We reared 5 healthy cannibals and 15
healthy typicals from the same lake in separate aquaria as controls.
The control cannibals were fed healthy typicals. We examined
animals daily for signs of cannibalism and disease.
To determine the proximate cause of disease, we censused bacterial populations in randomly chosen, natural lakes in both regions. We assayed aerobic, gram negative bacteria since these are
typically abundant and are generally indicative of overall pathogen
levels in polluted water (Mara 1974). We serially diluted one ml of
water to 10 .6 ml after collecting water from randomly-chosen sites
along the shore of each lake and at its center. Using a glass spreader
we plated the samples immediately after collection on to 90 mm
plates containing Emb agar. We kept the plates at 18-21 ~ C for
3 12 h during transfer to the laboratory where they were incubated
at 37 ~ C for 17 h. For analysis, we averaged the number of bacterial
colonies from the shore and center samples of equal dilution. At
highest concentrations, most plates contained too many bacterial
colonies to count. Values reported in the results are based on the
highest concentration not containing colonies too numerous to
count.
Comparison of parasites in cannibals and typicals
Cannibals were more likely to acquire pathogens from the conspecifics they consumed (see results). We therefore predicted that
compared with typicals cannibals should bear heavier loads of
macroscopic pathogens (i.e., parasites). To test this, we compared
levels of intestinal nematodes, a parasite of amphibians (Sprent
1984), in the gastrointestinal tracts of metamorphosing cannibals
and typicals. We used a microscope to count the numbers of nematodes in the stomachs and intestines of six cannibals and five typicals that we had collected from two different lakes on the Mogollon
Table 1. Frequency of cannibals, larval
salamander density, and bacterial counts
(aerobic, gram negative bacteria) for lakes
on the Mogollon and Kaibab Plateaus;
( - = not sampled); w denote salamander
populations that suffered mass mortality in
1990. Alphabetic superscripts indicate populations compared with a Mann-Whitney
U-test
Lake
Plateau. The person who counted the nematodes was unaware of
whether each gastrointestinal tract had come from a cannibal or a
typical. An initial one-way analysis of variance indicated no significant differences between lakes in numbers of parasites (F1,9= 4.558,
P = 0.062). We therefore pooled the data from the two lakes and
used a Mann Whitney U test to compare parasite loads in both
morphs.
Results
Comparison of frequency of cannibals across two regions
N a t u r a l h a b i t a t s o n the K a i b a b P l a t e a u h a d s i g n i f i c a n t l y
l o w e r f r e q u e n c i e s o f c a n n i b a l s t h a n d i d t h o s e o n the
M o g o l l o n P l a t e a u ( T a b l e 1). T h r e e lines o f e v i d e n c e
i n d i c a t e t h a t this difference w a s n o t s i m p l y d u e to i n t e r r e g i o n a l differences in d e n s i t y , t h e c u e t h a t i n d u c e s c a n n i bals. F i r s t , d e n s i t y d i d n o t differ s i g n i f i c a n t l y b e t w e e n
r e g i o n s ( T a b l e 1). S e c o n d , for a g i v e n s a l a m a n d e r d e n s i t y
c a n n i b a l s were less f r e q u e n t i n n a t u r a l p o p u l a t i o n s o n
the K a i b a b P l a t e a u t h a n o n the M o g o l l o n P l a t e a u
(Fig. 1). T h i r d , K a i b a b P l a t e a u s a l a m a n d e r s w e r e less
likely t h a n M o g o l l o n P l a t e a u s a l a m a n d e r s to b e c o m e a
cannibal even when reared under common conditions
( T a b l e 2, Fig. 2). S i g n i f i c a n t l y m o r e c a n n i b a l s were
produced overall among Mogollon Plateau animals than
K a i b a b P l a t e a u a n i m a l s ( P < 0.01, G a m e s - H o w e l l m u l t i ple c o m p a r i s o n test). F u r t h e r s t u d y is n e e d e d to d e t e r -
Cannibal frequency
(mean :k s.e.m.
% cannibals)
Mogollon Plateau:
Nelson Lake
0 +_ 0
Baker Lake
6.7-+ 6.7
Lost Lake
12.4_+ 6.2
Dude Lake
20.0_+ 10.0
Potato Lake
Alder Lake
2.9_+ 1.8
Myrtle Lake
22.8_+ 3.3
Horseshoe Lake
17.1 _+ 9.9
means (s.e.m.) = 11.7 (3.3) a
Salamander density
(mean =ks.e.m.
# salamanders/m 3)
Bacterial density
(mean :k 95 % C. I.
# bacterial colonies/ml)
1.7_+
1.6_+
16.0_+
5.1_+
385+ 86
840_+ 124
20_+ 26
0_+ 0
0_+ 0
0.9
0.9
3.2
1.4
63.1 _+ 12.0
75.3 _+ 7.9
27.8 _+ 3.9
27.2 (11.5)b
249 (165) ~
Kaibab Plateau:
Kaibab National Forest (livestock present) :
Lookout Lakew
Oquer Lakew
Dog Lakew
Snipe Lakew
Fracas Lake
Warm Springs L.
Crane Lake
J&W'sTankw
means (s.e.m.)=
0 + 0
0 _+ 0
0 • 0
0 _+ 0
0 _+ 0
0 • 0
0 _+ 0
5.6_+ 2.8
0.7 (0.7) a
2.7_+ 1.8
32.6_+ 8.4
6.5_+ 2.8
0.4_+ 0.4
1.7_+ 0.5
42.6_+ 5.3
14.4 (7.0) b
505+ 98
1,100+140
285_+ 76
1,340_+ 154
1,980_+186
1,042 (303) c
Grand Canyon National Park (livestock absent)."
Little Park Lake
Greenland Lake
means(s.e.m.)=
3.6_+ 0.3
3.6 (0)
15.2_+ 4.7
15.2 (0)
325• 80
50+ 36
188 (138)
P=0.005, b P=0.568, c P=0.046; (all two-tailed Mann-Whitney U tests)
164
30
Table 2. Two-way A N O V A of number of cannibals produced
among laboratory-reared animals from two regions at four initial
densities. Data were transformed before analysis (/[number of cannibals+ 0.375]). The untransformed means are in Fig. 2
[] M o g o l l o n P l a t e a u I
9 Kaibab Plateau
I
20
o
-(3
rt-"
O
10
[]
z
:
0 "~:~" :
0
S
60
# larvae / m
90
3
Fig. 1. Comparison of the relationship between mean frequency of
cannibals and salamander larval densities for natural lakes on the
Kaibab and Mogollon Plateaus. horizontal lines - + one standard
deviation
2.5
@ Kaibab
[] M o g o l l o n
.I3
ct-
2.0
O
1.5
"Q~
E~_
1.0
c:
0.5
0.0
v
3
5
7
15
Initial number of animals
Fig. 2. Mean n u m b e r of cannibals produced in the laboratory under
different densities among Kaibab Plateau and Mogollon Plateau
animals, vertical lines around each point
4- one s.e.m.
2000
q)
.~~~
.~
$~
A M a s s mortality
]
9 N o m a s s mortality
15oo
r =- 0.69
s
P = 0.039
1000'
.Q--
~_E
OL-
. d)
O D-
df
MS
Region
Initial density
Region x Initial density
Residual
1
3
3
56
1.866 19.063 0.0001
0 . 8 5 4 8 . 7 2 4 0.0001
0 . 0 9 7 0 . 9 8 7 0.4054
0.098
F
P
i
I
30
Source of variation
500
mine if variation between regions in the propensity to
become a cannibal is heritable.
Causes of differences between regions in frequency of
cannibals
In 1990, 5 of 10 natural lakes on the Kaibab Plateau and
0 of 7 natural lakes on the Mogollon Plateau exhibited
the characteristic signs of mass mortality [Zz=4.51,
P < 0 . 0 5 ; G-test]. Mass mortality of salamanders must
have an extrinsic cause, like disease, since in the laboratory animals from the Kaibab Plateau (where mass mortality occurred) had greater survivorship than animals from
the Mogollon Plateau (Z2= 64.89, P < 0.0001).
The presence of disease appeared to affect the distribution of the cannibalistic and typical morphs. A significantly lower percentage of cannibals occurred in lakes
on the Kaibab Plateau containing diseased animals
( m e a n i one s.e.m. = 0 + 0% cannibals; n = 5 lakes) than
in those not containing diseased animals (10.64-3.1%
cannibals; n = 8 lakes; P = 0.014; two-tailed Mann-Whitney test). In an experiment designed to measure the
effects of the presence of diseased conspecifics on the
frequency of cannibals, we found that all the cannibals
consumed the diseased animal in their tank; none of the
typicals did so. All 7 cannibals died within one week. All
but one typical m o r p h survived to metamorphosis (Zz =
5.56, P < 0.05; G-test). Only one cannibal used as a control died (Z2= 5.56, P < 0.05; G-test). O f the larvae collected because they had symptoms o f disease, the five that
were not cannibalized died within one week confirming
that all the animals were indeed ill.
The bacterial censuses, which were designed to identify the proximate cause of the disease, revealed that natural lakes with mass mortality had significantly more
bacteria than did natural lakes without mass mortality
(Table 1). Bacterial levels correlated significantly negatively with the frequency of cannibals (Spearman rank
correlation coefficient = - 0.69, n = 10, P = 0.039 ; Fig. 3).
Compar&on of parasites in cannibals and typicals
Z
0
5
10
15
20
Percent cannibals
Fig. 3. Frequency of cannibals as a function of bacterial density in
ten different natural lakes
Cannibals contained in their gastrointestinal tracts significantly more parasitic nematodes than did typicals
( m e a n + l s.e.m, number of nematodes: in cannib a l s = 2 3 . 5 + 7 . 1 , n = 6 ; in typicals --1.6 :k 0.5, n = 5 ;
P = 0.006, two-tailed Mann Whitney U test). Cannibals
did not carry more intestinal nematodes than typicals
165
simply because cannibals were larger: Nematode number
did not correlate significantly with body size (snout-tovent length) in either morph (cannibals: Spearman rank
correlation coefficient= - 0.62, n = 6, P = 0.167; typicals: Spearman rank correlation coefficient=-0.21,
n = 5, P = 0.682).
Discussion
The cannibals in our laboratory experiment likely died
of a disease transmitted from their victim. Cannibalism
per se could not have killed the cannibals as we routinely
rear salamanders on healthy conspecifics. The cannibals
died after displaying symptoms (e.g., reddening about
the venter and on the ventral surface of the limbs) like
those seen in salamanders from natural lakes undergoing
mass mortality (Worthylake and Hovingh 1989; Berna
1990).
What is the proximate cause of disease in these salamanders and how is it transmitted? Cannibals appear to
contain higher levels of both bacteria and parasites.
A comparison of parasitic nematode levels in cannibals
and typicals revealed higher levels in the gastrointestinal
tracts of the former. Pathogenic bacteria also seem to be
present at higher levels in cannibals. An analysis of
5 diseased and 5 healthy animals from lakes on the
Kaibab Plateau revealed abnormally high concentrations
of Clostridium in livers of the former (D. Nichols, pers.
comm.). Harmless in the intestine, this bacterium can be
pathogenic in other tissues (Mara 1974). Many pathogenic bacteria and parasites enter an amphibian's body
through oral exposure (Hoff and Hoff 1984; Sprent
1984), such as through ingestion of an infested conspecific. As with nematodes, an enhanced presence of bacteria
in cannibals indicates higher levels of overall pathogens
in this morph. Three lines of evidence implicate bacteria
(or some closely associated pathogen), however, as the
agent of mortality. First, in neighboring Utah mass mortality of the same subspecies of salamander is also correlated with late summer bacterial blooms (Worthylake
and Hovingh 1989). Second, in our study natural lakes
with mass mortality had significantly more bacteria than
did natural lakes without mass mortality (Table 1 ; Fig.
3). Third, the frequency of cannibals and bacterial levels
correlated significantly negatively (Fig. 3).
An alternative explanation for the correlation between bacterial density and the frequency of cannibals
warrants mention. Bacteria may be especially abundant
in organically rich lakes that contain greater densities of
insects and microorganisms on which typicals feed. Thus,
differences in resource level, not probability of acquiring
disease, may select against cannibals in lakes containing
more bacteria. This explanation seems unlikely since
lakes with high concentrations of bacteria generally harboured few aquatic macroscopic invertebrates (D. Pfennig, personal observ.).
Introduced livestock may have transmitted many of
the pathogens that caused mass mortality in the Kaibab
Plateau population. Many pathogens emanate from
faeces (Mara 1974; Hoff and Hoff 1984), and livestock
contribute considerable faeces to the lakes that we stud-
ied. Kaibab Plateau lakes that were visited by livestock
(e.g., lakes in the Kaibab National Forest) had higher
bacterial counts than did those that were not visited by
livestock (e.g., lakes in Grand Canyon National Park;
Table 1). Since livestock were not introduced on the
Kaibab Plateau until 1860 (Anderson 1990), selection
against cannibals in Ambystoma may be a recent
phenomenon.
Circumstantial evidence further implicates pathogens
as affecting the incidence of cannibalism in this subspecies. In the White Mountains of east-central AZ
where cannibals are rarer than on the Mogollon Plateau,
but more common than on the Kaibab Plateau (Collins
1981), mass mortality also occurs, but not as frequently
as on the Kaibab Plateau (J. Collins, personal obser.).
Mass mortality also occurs in populations of A. t. stebbinsi, which lack cannibals in nature (Collins et al. 1988).
Selection to escape pathogens may have affected the
evolution of other life history traits in this species. In our
common garden experiment, cannibals from the Kaibab
Plateau metamorphosed significantly faster than those
from the Mogollon Plateau (mean+ 1 s.e.m, length of
larval period = 86 • 4 d and 123 4- 27 d for cannibals from
the Kaibab Plateau and Mogollon Plateau, respectively;
P < 0.05, two-tailed t-test).
Many studies (Sherman 1981; Meffe and Crump
1987; Crump 1983, 1990, 1992) focus on cannibalism's
benefits, and show convincingly that cannibalism can be
adaptive (e.g., cannibalism can increase developmental
rate and reduce the risk of being trapped in a lethal
environment [Crump 1990]). No study has empirically
examined why cannibalism is infrequent even among
animals that occur in ephemeral habitats. Two potentially important costs of cannibalism are that a struggling
victim (or its parents [Sherman 1981]) may kill or injure
a cannibal, and a cannibal may consume kin and suffer
a decrement in inclusive fitness (Polis 1981). Our results
support an additional, underappreciated cost of cannibalism: cannibals may die from pathogens or toxins
acquired from victims. Pathogens often are host specific
(Anderson and May 1982), and cannibals may be more
likely to acquire pathogens from conspecifics than from
the environment or heterospecific prey. Thus, pathogens
may affect the evolution of foraging behavior in addition
to other traits, such as mate choice (Read 1988) or the
propensity to reproduce sexually (Lively 1987). Intraspecific predation may be sufficiently costly to constrain not
only an organism's behavior but also its tendency to
produce certain morphologies, which may explain why
polyphenic cannibal morphs are rare.
Acknowledgement. We thank H. Reeve, G. Polis, P. Sherman,
S. Stearns, J. Travis and an anonymousreviewer for commenting
on various drafts of the manuscript, D. Nichols for examining
diseased salamanders, C. Schnaitmanand M. Bloomfor advice on
conducting bacterial censuses, and the United States Forest Service
and Arizona Game and Fish Department for their cooperation.
Fundingwas providedby a MaytagPostdoctoralFellowship(Zoology Department, Arizona State University) to D. P. and the NSF
(BSR-8919901) to J. C. Additionalindirect support was provided
by the Center for Insect Science, University of Arizona (through
D. P.).
166
References
Anderson J (1990) Rangeland success on the Arizona Strip. In:
Brock JH (ed) Proceedings of the First Conservation Symposium. [A] p 37
Anderson RM, May RM (1982) Population biology of infectious
diseases. Springer, Berlin Heidelberg New York
Bartlett MS (1971) The use of transformations. In: Steger JA (ed)
Readings in statistics for the behavioral scientist. Holt, Rinehart
and Winston, New York, pp 169-178
Berna HJ (1990) Ecology and life history of the tiger salamander
Ambystoma tigrinum nebulosum Hallowell on the Kaibab Plateau. unpubl. MS thesis, Arizona State University, Tempe, AZ,
pp 1 121
Brown DE (1982) Biotic communities of the American Southwest
- United States and Mexico. Desert Plants 4:1-342
Collins JP (1981) Distribution, habitats and life history variation in
the tiger salamander, Ambystoma tigrinum, in east central and
southeast Arizona. Copeia 1981:666-675
Collins JP, Cheek JE (1983) Effect of food and density on development of typical and cannibalistic salamander larvae in Ambystoma tigrinum nebulosum. Am Zool 23:7%84
Collins JP, Holomuzki JR (1984) Intraspecific variation in diet
within and between trophic morphs in larval tiger salamanders
(Ambystoma tigrinum nebulosum). Can J Zool 62:168-174
Collins JP, Jones TR, Berna HJ (1988) Conserving genetically
distinctive populations: the case of the Huachuca tiger salamander (Ambystoma tigrinum stebbinsi Lowe). In: Szarzo RC,
Severson KC, Patton DR (eds) Management of amphibians,
reptiles, and small mammals in North America. USDA Forest
Service GTR-RM-166, Rocky Mountain Forest and Range
Experiment Station, Fort Collins, CO
Collins JP, Mitton JB, Pierce BA (1980) Ambystoma tigrinum:
a multispecies conglomerate? Copeia 1980:938-941
Crump ML (1983) Opportunistic cannibalism by amphibian larvae
in temporary aquatic environments. Am Nat 121:281-289
Crump ML (1990) Possible enhancement of growth in tadpoles
through cannibalism. Copeia 1989:794-797
Crump ML (1992) Cannibalism in amphibians. In: Elgar M, Crespi
B (eds) Cannibalism: ecology and evolution among diverse taxa.
Oxford University Press, New York (in press)
Eickwort KR (1973) Cannibalism and kin selection In Labidomera
clivicollis (Coleoptera: Chrysomelidae). Am Nat 107 :452-453
Green CR, Sellers WD (1964) Arizona climate. Univ Arizona Press,
Tucson, AZ
Hoff GL, Hoff DM (1984) Salmonella and Arizona. In: Hoff GL,
Frye FL, Jacobson ER (eds) Diseases of amphibians and reptiles. Plenum Press, New Yor~c, pp 69-82
Holomuzki JR (1986) Effect of microhabitat on fitness components
of larval tiger salamanders, Ambystoma tigrinum nebulosum.
Oecologia 71 : 142-148
Holomuzki JR, Collins JP (1987) Trophic dynamics of a top predator, Ambystoma tigrinum nebulosum (Caudata: Ambystomatidae), in a lentic community. Copeia 1987:949-957
Lannoo M J, Bachmann MB (1984) Aspects of cannibalisticmorphs
in a population of Ambystoma t. tigrinum larvae. Am Midl Nat
112:103-109
Lively CM (1987) Evidence from a New Zealand snail for the
maintenance of sex by parasitism. Nature 328:519 521
Mara BD (1974) Bacteriology for sanitary engineers. Longman,
New York
Merle GK, Crump ML (1987) Possible growth and reproductive
benefits of cannibalism in the mosquitofish. Am Nat
129:203-212
Pedersen SC (1991) Dental morphology of the cannibal morph in
the tiger salamander, Ambystoma tigrinum. Amphibia-Reptilia
12:1-14
Pfennig DW (1990) The adaptive significance of an environmentally-cued developmental switch in an anuran tadpole. Oecologia 85 101-107
Polls GA (1981) The evolution and dynamics of intraspecific predation. Ann Rev Ecol Syst 12:225-251
Polls GA, Myers CA (1985) A survey of intraspecific predation
among reptiles and amphibians. J Herp 19: 99-107
Powers JH (1907) Morphological variation and its causes in Ambystoma tigrinum. Stud Univ Nebraska 7:197-274
Read AF (1988) Sexual selection and the role of parasites. Trends
Ecol Evol 3 : 97-102
Rice RJ (1988) Fundamentals of geomorphology, 2nd ed. John
Wiley and Sons, New York
Sherman PW (1981) Reproductive competition and infanticide in
Belding's ground squirrels and other animals. In: Alexander
RD, Tinkle DW (eds) Natural selection and social behavior.
Chiron Press, New York, pp 311-331
Sprent JFA (1984) Ascaroid nematodes. In: Hoff GL, Frye FL,
Jacobson ER (eds) Diseases of amphibians and reptiles. Plenum
Press, New York, pp 219 245
Worthylake KM, Hovingh P (1989) Mass mortality of salamanders
(Ambystoma tigrinum) by bacteria (Acinetobacter) in an oligotrophic seepage mountain lake. Great Basin Nat 49:364-372