Journal of Theoretical Biology 394 (2016) 137148

Contents lists available at ScienceDirect

Journal of Theoretical Biology

journal homepage: www.elsevier.com/locate/yjtbi

How residency duration affects the outcome of a territorial contest:

Complementary game-theoretic models
Mike Mesterton-Gibbons a,n, Tom N. Sherratt b
a
Department of Mathematics Florida State University, 1017 Academic Way, Tallahassee, FL 32306-4510, USA
b
Department of Biology, Carleton University, 1125 Colonel By Drive, Ottawa, Ontario, Canada K1S 5B6

H I G H L I G H T S

First discovers often become owners, but sometimes it is unclear who got there rst.

When rival claims to property arise, contest duration increases with residency time.

Game-theoretic models based on increasing belief or motivation explain these effects.

art ic l e i nf o a b s t r a c t

Article history: While the rst individuals to discover and maintain territories are generally respected as owners, under
Received 20 November 2015 some conditions there may be ambiguity as to who got there rst. Here we attempt to understand the
Received in revised form evolutionary consequences of this ambiguity by developing a pair of game-theoretic models in which we
10 January 2016
explicitly consider rival residency-based claims to ownership. Following earlier qualitative explanations
Accepted 12 January 2016
Available online 22 January 2016
for residency effects, we assume that either the value of the territory (Model A) or an interloper's self-
belief that it is the owner (Model B) increases with duration of residency. Model A clearly demonstrates
Keywords: that if the value of a territory increases to a resident over time, so should its motivation to ght in terms
Ownership of the effort it invests in ghting. Indeed, only a small increase in territory value with residency duration
Territoriality
can be sufcient for longer established residents to win disputes, even without any arbitrary convention
Animal contests
or other form of priority effect. Likewise, Model B shows that the observed increase in ghting persis-
Game theory
Evolutionarily stable strategy tence with residency duration can be readily explained as a consequence of increasing condence on
Asymmetric war of attrition behalf of the interloper that it is the rightful owner. Collectively, the models help to explain some general
ndings long observed by empiricists, and shed light on the nature of conicts that can arise when
individuals do not have complete information about rival claims to ownership.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction both contestants consider themselves the resident (Waage, 1988).

It has long been appreciated that respect for ownership can
play a role in settling disputes over contested resources, although
the extent to which this deference arises as a convention or as a
consequence of asymmetry in resource holding potential and/or
value remains unclear (see Kokko, 2013; Kemp, 2013; Sherratt and
Mesterton-Gibbons, 2015 for reviews). Ironically, however, it
appears that this respect can also result in escalated contests if
n
Corresponding author. Tel.: +1 850 644 2580; fax: +1 850 644 4053.
E-mail addresses: mesterto@math.fsu.edu (M. Mesterton-Gibbons),
Tom.Sherratt@Carleton.ca (T.N. Sherratt).
URLS: http://www.math.fsu.edu/  mesterto (M. Mesterton-Gibbons),
https://carleton.ca/biology/people/tom-sherratt/ (T.N. Sherratt).
If a territory is large or complex, for instance, then a territory
holder may be present when another individual arrives, but the
two individuals do not detect one another until both have been
present for some time. Waage (1983, 1988) observed precisely this
set of events in territorial Calopteryx maculata damselies and
showed experimentally that the protracted contests that arose
under these conditions were best understood as a consequence of
habitat-mediated confusion over ownership. Likewise, a resident
may move away temporarily from its territory to forage or repel
intruders, only to be replaced by an interloper. Particularly long
disputes between interlopers and the returning resident have been
widely observed under natural conditions (e.g., Wickman and
Wiklund, 1983; Waage, 1988; Gribbin and Thompson, 1991;
Rutowski, 1992; Kemp, 2003) and they have also been frequently

http://dx.doi.org/10.1016/j.jtbi.2016.01.016
0022-5193/& 2016 Elsevier Ltd. All rights reserved.
138 M. Mesterton-Gibbons, T.N. Sherratt / Journal of Theoretical Biology 394 (2016) 137148
reported in manipulative experiments following the removal and
subsequent re-introduction of the former resident (e.g., Davies,
1978; Krebs, 1982; Alcock and Bailey, 1997; Kemp, 2000; Kemp
and Wiklund, 2004; Takeuchi, 2006; Peixoto and Benson, 2012).
Intriguingly, while the temporary interlopers in the above
experiments are frequently successful at repelling any non-
resident intruders, the original residents often win the escalated
contests that arise on their return (Wickman and Wiklund, 1983;
Krebs, 1982; Alcock and Bailey, 1997; Kemp, 2000; Kemp and
Wiklund, 2004; Takeuchi, 2006). This is a key observation, because
it rules out the possibility that animals with rival time-based
claims to the same territory use entirely arbitrary conventions
based on current occupancy, and instead suggests that the sense of
ownership and/or motivation to retain the property develops over
time. Indeed, there is clear evidence that the period of time an
interloper is in residence affects both the duration of the sub-
sequent contest with the former owner and the likelihood that it
will win. For example, Krebs (1982) removed resident pairs of
great tits (Parus major) from their territories and released them
back after replacement pairs had occupied the areas for a con-
trolled period of time. The duration of the contests between the
former resident great tits and their interlopers was an increasing
function of replacement time, as was the probability that the
interlopers would retain the territory. Alcock and Bailey (1997)
conducted an analogous experiment with territorial tarantula
hawk wasps Hemipepsis usulata and found that interlopers that
had been on the territory for longer tended to ght longer and
harder against the original owner, before ultimately losing.
Krebs (1982) interpreted his results to have arisen as a con-
sequence of an increase in value of the territory over time to an
interloper, and a decrease in value over time to the former owner.
Possible reasons for this gradual change in value of the territory
include the residents' increasing familiarity with their physical
surroundings, but also with their neighbors. Tobias (1997) likewise
conducted removal experiments in European robins Erithacus
rubecula and found that newcomers were more likely to resist
eviction by the original owners, the longer they were allowed to
hold the territory. Just as Krebs (1982) had proposed, Tobias (1997)
argued that newcomers had to pay non-transferable settlement
costs when negotiating boundaries with neighbors, so that the net
value of a territory to a resident increased over time. Indeed a
signicant asymmetry in net value favoring the interloper is pos-
sible if, the longer a resident is absent from its territory, the more
likely it will have to renegotiate its boundaries when it returns.
Of course, not all territories will increase in value to the resi-
dent as a consequence of its increasing familiarity with its sur-
roundings and its neighbors. Indeed, Kemp and Wiklund (2001)
felt that such effects were unlikely in buttery species, where
territories are small and individuals appear not to be able to
specically recognize their neighbors. Alcock and Bailey (1997)
proposed that the residency-duration effect they observed in hawk
wasps could also be mediated by a change in perceived value, but
suggested that it might arise indirectly through the interloper
being increasingly convinced that the former resident would not
return. Walton and Nolan (1986) made a similar case for the
importance of uncertainty over roles when describing the terri-
torial behavior of prairie warblers (Dendroica discolor) after their
spring migration. Naturally, increasing familiarity with one's sur-
roundings and confusion over ownership are not mutually exclu-
sive explanations for the time-in-residence effects. For instance,
Takeuchi (2006) found that green hairstreak (Chrysozephyrus
smaragdinus) butteries that replaced an original owner fought
longer against original occupants as their residence duration in the
territory increased, and invoked both the above arguments to
explain his results.
Given the role of time in residency in affecting both the dura-
tion and outcome of disputes involving co-owners with joint
claims for ownership, it is somewhat surprising that no model of
territorial conict has been presented to help understand these
fundamental phenomena. To elucidate the signicance of these
behaviors from an adaptive perspective, we develop two separate
but complementary game-theoretic models. The rst model
addresses how an increase of resource value with time on territory
affects the effort expended in a contest over ownership, when no
role asymmetry is perceived by the contestants. This model is new.
The second model addresses how increasing belief in ownership
on the part of an interloper affects contest duration, when the
roles of owner and intruder are perceived by two contestants, but
imperfectly. This model is an adaptation of Hammerstein and
Parker's (1982) asymmetric war of attrition. Separate mathema-
tical models allow us to focus, as far as possible, on one of these
two effects in isolation from the other. We will refer to the rst
model as Model A and to the second as Model B.
2. Model A: no role asymmetry perceived by the contestants
First we explore how an increase of resource value with time
on territory affects effort expended in a contest over ownership,
when no role asymmetry is perceived by the contestants.
Accordingly, consider two animals that have discovered an
otherwise unoccupied territory, whose value at time s since dis-
covery is V(s). That is, V(s) is the site value to an animal that has
been in residence for time s, assuming that it retains the resource
after contesting it. If one animal has been on site for time sO and
the other for time sI, then in general sO a sI , and so the respective
site values, VsO and VsI , also differ. We assume that site value is
zero initially and increases with time on territory towards a
maximum value of 1. Specically, we assume
V0 0; V 0 s 4 0; V 1 1; 1
where a prime denotes differentiation with respect to argument.
We note that some territories may have immediate value to resi-
dents on arrival (so that V0 4 0), and their ultimate value may
even decline with increasing residency duration (so that V 0 s o0).
However, the inequalities assumed in (1) are the simplest starting
point to quantitatively evaluate earlier verbal hypotheses as to
why interlopers should ght longer to retain a territory, the longer
they have been in residence.
Each of the animals is initially unaware of the other's attach-
ment to the sitethere is habitat-mediated confusion (Waage,
1983, p. 25), with two co-discoverers both continually present.
One of these individuals is the focal individual or Player 1, from
whose perspective we analyze their interaction; this individual
plays a potential mutant strategy, and the other individualPlayer
2plays the population strategy. Eventually, these two animals
will meet. Let t be the time that has then elapsed since Player
1 discovered the territory, and let T be the corresponding time for
Player 2. Although t is then known to Player 1, from Player 1's
perspective, T is a random variable, which we assume to be con-
tinuously distributed over 0; 1 with probability density function
g, and distribution function G dened by
Z s
Gs g d: 2
0
Without loss of generality, we assume that time is measured in
units of the mean of this distribution, so that its mean becomes
1 and its variance is
Z 1
2 s  12 gs ds: 3
0
 M. Mesterton-Gibbons, T.N. Sherratt / Journal of Theoretical Biology 394 (2016) 137148 139

This is also the distribution from which t has been independently Adopting an intercept (base effort) and gradient (incremental
drawnthe distribution of times to rst encounter. effort) allows us to consider the possibility that individuals do not
Because T is continuously distributed, one of the two animals invest more effort in ghting, the longer they are in residence
must have arrived before the other, i.e., ProbT t 0. We will (gradient 0); and conversely, the possibility that they do not
refer to the rst arrival as the owner, and to the other individual as invest in ghting until they have occupied the territory for a while
the intruder. We assume, however, that neither individual is aware (intercept0). Thus, if times sO and sI have elapsed since the
of which role it is in: owner and intruder are merely convenient owner and the intruder, respectively, have been on territory (so
labels by which an external observer can distinguish the animals. that sO 4 sI ), then the efforts expended by Player 1 and Player 2 are
Thus we rule out the possibility of a Bourgeois-like convention x u1 u2 sO and y v1 v2 sI , respectively, if Player 1 has been the
our interest is rather how time on territory affects contest effort. owner and Player 2 the intruder, but by y u1 u2 sI and x v1
When the animals meet, they engage in a contest, on which v2 sO if Player 1 has been the intruder and Player 2 the owner. Effort
each expends effort. Their efforts may differ or be equal. If their E expended by either animal is assumed to cost c(E), where
efforts are equal, then the probability of victory is assumed to be
determined by priority advantage , that is, the degree to which c0 0; c0 E 4 0; c E 4 0: 7
an owner's probability of winning is increased beyond 12 (where If T ot, which happens with probability G(t), then Player 1 is the
0) toward 1 (where 1) in a contest between two otherwise owner and Player 2 is the intruder, with s t  T. Because the
evenly matched opponents by virtue of its greater familiarity with
focal u-strategist is the owner, its probability of victory is
the territory. Specically, the probability that the owner wins this
pO u1 u2 t; v1 v2 T; s, with expected payoff
contest is 1=2  , where we assume for simplicity that
depends on the difference s between the owner's and the F 1 u; v; t; T VtpO u1 u2 t; v1 v2 T; t  T  cu1 u2 t: 8
intruder's time on territory according to
If, on the other hand, T 4 t, which happens with probability
s 1 e  s : 4 1  Gt, then Player 2 is the owner and Player 1 is the intruder,
with s T  t, and so Player 1's expected payoff is
We can interpret as the rate at which greater experience of the
territory translates into a tactical advantage in a contest. Note that F 2 u; v; t; T VtpI v1 v2 T; u1 u2 t; T  t  cu1 u2 t: 9
if the animals themselves perceived an asymmetry of ownership,
So the reward at time t to Player 1 from playing strategy u against
then we would refer to as owner advantage; but because own-
population strategy v is
ership is perceived only by an external observer, we use priority
Z t
advantage instead.  
Fu; v; t Gt F 1 u; v; t; sg 1 s ds 1  Gt
In the case where their efforts differ, as will happen often, let x 0
and y denote effort expended by owner and intruder, respectively. Z 1 Z 1
Vtgs
We assume that, in the absence of any priority advantage, prob- F 2 u; v; t; sg 2 s ds ds cu1 u2 t 10
t 0 1 e  Q u;v;s;t
ability of victory for the owner would increase sigmoidally with
respect to effort difference x  y according to 1=1 e  rfx  yg , where
where we interpret r as the reliability of greater effort as a pre-
Q u; v; s; t ft  sg rfu1 u2 t  v1  v2 sg; 11
dictor of contest outcome. More generally, let pO x; y; s denote
the probability of victory for the owner when it has been on ter- after simplication with (5) and (6). Before the focal animal dis-
ritory for s longer than the intruder. Then we assume that the covers the territory, however, the time t is itself a random variable,
effect of priority advantage combines with that of effort difference drawn from the same distribution as for T. Hence, integrating over
according to that distribution, the unconditional reward to Player 1 from
1 1 playing strategy u against population strategy v is
pO x; y; s ; 5 Z 1 Z 1Z 1
1 f1  sge  rx  y 1 e  fs rx  yg Vtgsgt
f u; v Fu; v; tgt dt ds dt
which is readily seen to satisfy pO x; y; 1 1 and 0 0 0 1 e  Q u;v;s;t
Z 1
pO x; y; 0 1=2  Z 12. We denote the corresponding probability
 cu1 u2 tgt dt: 12
of victory for the intruder by 0

1  s 1
pI x; y; s 1  pO x; y; s :
1  s erx  y 1 efs rx  yg
6 3. The evolutionarily stable strategy for Model A
Note that our notation suppresses a dependence of pO and pI on
and r. A population strategy v is a strong evolutionarily stable strat-
It is assumed that an animal's motivation to contest a territory egy or ESS sensu Maynard Smith (1982) when it is uniquely the
can vary with its duration of residency. Accordingly, a strategy for best reply to itself, that is when f v; v 4 f u; v for any potential
each player consists of two quantities, a minimum or base effort mutant strategy u a v, which requires
 
that even a new arrival is willing to expend on contesting the f  f 
0 13
resource and a rate of increase of latent effort with respect to u1 u v u2 u v
residency duration, which together determine the effort that the
animal will actually expend in a contest over the territory. We will with
 
refer to the rst quantity as base effort and to the second as
2 f  2 f 
incremental effort, and denote this pair of quantities by u u1 ; u2  o 0;  o0 14
u21  u22 
for the focal individual and by v v1 ; v2 for the corresponding uv uv

population strategy, where u1 ; u2 ; v1 ; v2 Z 0. Note that a sub- for an interior ESS such that v1 ; v2 4 0, but instead
scripted 1 or 2 is used for either player's base or incremental effort, 
respectively, and in no way signies Player 1 (focal individual) or f 
0 15
Player 2 (population strategy). u1 u v
140 M. Mesterton-Gibbons, T.N. Sherratt / Journal of Theoretical Biology 394 (2016) 137148

owner has been in residence for time T, then the intruder's

probability of victory is
1
pI vn1 vn2 T; vn1 vn2 t; T  t 20
1 e rv2 T  t
n

by (6). To calculate I , we must condition on T 4 t. Hence

Z 1
1 gs
I t ds: 21
1  Gt t 1 e rvn2 s  t
The probability of victory for an intruder discovered soon after
arrival is well approximated by taking the limit of (21) as t-0. Let
g max denote the maximum value of g over 0; 1. Then by standard
analysis we obtain the upper bound
Z 1 Z 1
gs ds g ln2
Fig. 1. Probability density functions of a uniform (dashed) and a Weibull (solid) I 0 dsr g max n 22
distribution with mean 1 for four different values of its shape parameter . The 0 1 e rv2 s
n max
0 1 e rv2 s
n
rv2
variances for 1 (exponential distribution), 2, 3 and 10 are 1, 0.273,
0.132 and 0.014, respectively, whereas the uniform distribution has variance 13.
which suggests that I 0, the probability of victory by a newly
arrived and instantly discovered intruder, is typically smallas
with illustrated in Fig. 3.
 
2 f  f  3.1. ESS for a uniform distribution of times to rst encounter
 o 0; o0 16
u21  u2 u v
uv
Here we assume that times to rst encounter are drawn from a
for a boundary ESS of the form v v1 ; 0 with v1 4 0 (see, e.g.,
uniform distribution with mean 1, and hence probability density
Broom and Rycht, 2013). No boundary ESS of the form v 0; v2
function
with v2 Z 0 arises. Explicit expressions for the quantities appearing (
in (13)(16) are given by Appendix A. Calculation of the ESS 1
if 0 o s o 2
gs 2 23
requires specifying the value function V, the cost function c and 0 if 2 o s o 1
the distribution of times to rst encounter. Accordingly, we satisfy
(1) and (7) with and variance 2 13, by (3); its graph is the dashed curve in Fig. 1.
Although in principle this choice of g together with our choices for
Vt 1  e  t ; 17 V and c enable the left-hand sides of (A.3) to be evaluated in terms
of hypergeometric functions, the resulting analytical expressions
cE E2 18 are far too cumbersome to be of any use. We therefore proceed
and choose either a uniform or a Weibull distribution for time to numerically instead.
rst encounter. Thus V 0 0 measures the rate at which the How vn and O depend on (the rate at which resource value
resource rises in value towards its asymptotic level, whereas rises towards its asymptotic levelsee (17)) is illustrated by the
measures (twice) the rate at which the marginal cost of ghting left-hand column of Fig. 2 for 0:01, r 2 and three different
effort c0 E 2 E increases with effort. Our chosen distributions for values of , namely, 4, 1 and 0. We see that base effort
time to rst encounter are plotted in Fig. 1. vn1 is strictly increasing with respect to whereas incremental
For any values of , , , r and any choice of g (residency effort vn2 is highest for intermediate values of , and that vn2
duration distribution density), the ESS equations(13) or (15), for increases much more rapidly with respect to than vn1 at lower
which (A.3) yields explicit expressionsare readily solved by values of . Most importantly, the expected probability O of vic-
numerical means to yield a unique pair of values of v1 and v2, with tory for the owner remains high even when priority advantage is
(14) or (16), as appropriate, invariably satised. low. Indeed the dashed curve in Fig. 2(c) corresponds to the limit
We denote this ESS by vn vn1 ; vn2 . Our focus is on how vn of zero priority advantage ( 0).
depends on , , r and especially . We denote the expected How I (i.e., the probability of winning for the individual that
probability of victory for the owner at the ESS by O ; ; ; r. We has been an occupant for a shorter time) increases with t is illu-
can calculate O either as the expected value of 1=f strated in Fig. 3(a) for the same values of , r and and for three
n n n n
1 e  Q v1 v2 T;v1 v2 S;S;T g conditional on T 4 S or as the expected different values of , namely, 0:5, 1:5 and 2:5. An
value of 1=f1 e  Q vn1 vn2 S;vn1 vn2 T;T;S
g conditional on S 4 T, where S interesting feature of these graphs is how little inuences the
and T are random draws from the distribution of times to rst intruder's probability of victory after any given time in residence,
encounter, and Q is dened by (11). Thus, by integrating over despite the increase of contest effort with . Moreover, the inu-
either the lower or the upper triangle of the joint sample space of ence of on these curves is likewise very weak, which is why we
S and T (and dividing by 12, the probability that one exceeds the have plotted them only for 0.
other, to obtain a conditional probability density function), we
obtain 3.2. ESS for a Weibull distribution of times to rst encounter
Z 1Z t
gsgt
O ; ; ; r 2  rvn2 t  s
ds dt From, e.g., Mesterton-Gibbons et al. (1996, p. 68), the Weibull
1 e
distribution with mean 1 and shape parameter ( Z1) on 0; 1
0 0
Z 1Z s
gsgt
2  rvn2 s  t
dt ds: 19 has probability density function
0 0 1 e
 
gs 1 1= s  1 e  f 1 1=sg ; 24a
Let us now consider the prospects at this ESS of an intruder that
has been on territory for time t. Its effort at the ESS is E vn1 vn2 t. distribution function
Let us denote its expected probability of victory at time t by I t,

using notation that suppresses a dependence on , , and r. If the Gs 1 f 1 1=sg 24b
 M. Mesterton-Gibbons, T.N. Sherratt / Journal of Theoretical Biology 394 (2016) 137148 141

Fig. 2. Left-hand column: How (a) base effort vn1 , (b) incremental effort vn2 and (c) expected probability O of victory for the owner vary with maximum rate of resource value
increase, , at the ESS when time to rst encounter has a uniform distribution, with 0:01 and r 2 for 4 (thick solid curves), 2 (thin solid curves) and 0 (thick
dashed curves). Right-hand column: How (d) vn1 , (e) vn2 and (f) O vary with at the ESS when time to rst encounter has a Weibull distribution with shape parameter 2
(and hence variance 4  1  0:273), for the same values of , r and . By (4), 0 corresponds to no priority advantage.

and hence, by (2) and (3), variance variance than the uniform distribution. Again we see that base
effort vn1 is strictly increasing with respect to , incremental effort
2 2=
2 1 25 vn2 is highest for intermediate values of and the expected prob-
f 1=g
2
ability O of victory for the owner remains high. A difference,
where R denotes Euler's function (dened for z 40 by however, is that whereas vn2 is always positive for a uniform dis-
z 01 wz  1 e  w dw). The graph of g is plotted in Fig. 1 for four tribution, for a Weibull distribution there is a critical value of at
different values of . which it falls to zero. This critical value decreases with as illu-
How vn and O depend on when residency durations are strated in Fig. 2(e); it is approximately 2.193 for 4, but is off the
drawn from a Weibull distribution is illustrated by the right-hand graph to the right for 0 and 1. In effect, for high values of
column of Fig. 2 for 0:01, r 2 and three different values of in (and a Weibull distribution of residency duration) there should be
the case where 2, so that the Weibull distribution has a lower no increase in ghting effort with residency duration. How I
142 M. Mesterton-Gibbons, T.N. Sherratt / Journal of Theoretical Biology 394 (2016) 137148

Fig. 3. How an intruder's expected probability of victory increases with time t in residence when time to rst encounter has (a) a uniform distribution and (b) a Weibull
distribution with 2 when 0:01, r 2 and 0 for 0:5 (dashed curves), 1:5 (thin solid curves) and 2:5 (thick solid curves). In both cases, the longer the
individual has been on the territory, the higher its chances of winning (up to a maximum of 0.5), even though it has been in residence for a shorter period of time than
its rival.

increases with t is illustrated in Fig. 3(b). Again, and have only by A, and the other contestant in the disfavored role, denoted by B
a weak inuence on the intruder's probability of victory after any (Hammerstein and Parker, 1982, pp. 671672). Favored and dis-
given time in residence. A difference this time is that whereas I is favored are dened with respect to payoffs. For I A; B, let VI
an accelerating function of t for a uniform distribution, it is denote the value of the resource in role I and let CI denote the cost
decelerating for a Weibull distribution. per unit time of contesting it, until the animal with the shorter
persistence time withdraws. Then role A is favored (and role B
3.3. Interim conclusions disfavored) if
VA VB
Our model assumes that the territory has value to both animals 4 : 26
CA CB
at the time of their encounter, and this value is non-transferable,
sensu Kokko (2013, p. 13), in that the benets of residency accrue Let us dene cost and value quotients qC, qV by
to the individual only. Moreover, we have assumed that the win- CA VA
qC ; qV : 27
ner will retain its value in the site while the loser will have to seek CB VB
it afresh elsewhere. Under these conditions, the owner is much
Then a sufcient condition for (26) to hold is qC o 1 o qV , which is
more likely to win because longer residency induces greater
most readily satised by setting
incremental effort, which reinforces its priority advantage. Base
effort also reinforces priority advantage, but effort is costly, and 1
qV ; qC 28
neither contestant knows either which animal has priority
advantage or the extent of it; and because winning is not guar-
with 4 1. The roles A and B are perceived by the contestants, but
anteed, base effort is expected to be smaller, especially when is
imperfectly. Let PI be the probability that a contestant in role I
low. In other cases, if the distribution of residency times is
identies with role I, hence 1  P I the probability that it identies
clumped (Weibull distribution) and is high, then there may be
with the alternative role. Then, provided there is at least a small
no effect of residency duration on ghting effortindividuals are
positive probability that both opponents identify with the same
highly motivated to ght from the get-go.
role in a contest, requiring
Model A shows that a high probability of victory by the owner
that is, the contestant with the longer residencydoes not require 0 o P A o 1; 0 o P B o1; 29
the animals to observe a role asymmetry. However, the model Hammerstein and Parker (1982, p. 672) have shown that at the
considers contest effort only, and so does not directly address ESS, a contestant identifying with role I should be prepared to
contest duration. For that purpose we need a different model, and persist for a time drawn from the continuous distribution with
so we turn to Model B. probability density function gI and cumulative distribution func-
tion GI, where

4. Model B: roles of owner and intruder perceived by the 0 if 0 o x o s
g A x ;
contestants rers  x if s o x o 1
Z x
GA x g A d
Here we explore how increasing belief in ownership on the part 0
of an intruder affects contest duration, when the roles of owner 
0 if x r s
and intruder are imperfectly perceived by two contestants. For this 30a
1  ers  x if x 4 s
purpose we adapt an existing game-theoretic model of a contest
over a resource, namely, Hammerstein and Parker's asymmetric and
war-of-attrition model, but we need only the special case of their (
Ke  rx if 0 ox os
model in which there exists an objective asymmetry between g B x ;
two contestants, with one contestant in the favored role, denoted 0 if s o x o 1
 M. Mesterton-Gibbons, T.N. Sherratt / Journal of Theoretical Biology 394 (2016) 137148 143

Fig. 4. Region of the unit square in which the point P A ; P B must lie for the ESS to exist (light and dark shading) and be unique (dark shading), for various values of . The
signicance of the L-shape inside the top right-hand quadrant is discussed in Appendix B.

Z x
invariably prepared to persist for longer than the animal in the
GB x g B d
0 disfavored role. Note that we have recast the expressions given by
8
<K Hammerstein and Parker (1982, p. 672) to present the results we
1  e  rx if x r s
r 30b need in a form more suited to our purpose.
:
1 if x 4 s: Nevertheless, (29) is not the only condition that PA and PB must
Here r, K and s are dened by satisfy for this ESS to exist and be unique. We show in Appendix B
that existence also requires
C A C B 1  P A C A P B C B 1  r
r ; K ; s  ln 1  31
V A V B P B 1  P A V A V B r K
PA 4 1  PB ; 32
with (29) implying K 4r. Thus whenever the two contestants
identify with different roles, the animal in the favored role is that is, the contestant in the favored role identies with it more
144 M. Mesterton-Gibbons, T.N. Sherratt / Journal of Theoretical Biology 394 (2016) 137148

Fig. 5. and 2 versus intruder's probability 1 P B of belief in ownership for

0 o o P A with PA 0.95, VA 2, V B 1 C B and C A 12 (implying r 12) to satisfy
(28) with 2. By (32)(34), the ESS is unique only for 0 o o 45 (where the curve
is shown solid) but continues to exist for 45 r o P A (where the curve is shown
dashed).

than its opponent does. When (32) holds, the point P A ; P B lies in
the (light or dark) shaded triangle in Fig. 4.
We further show in Appendix B that uniqueness additionally
requires P A ; P B to lie in the dark shaded curvilinear quadrilateral
within the upper right-hand quadrant; and that a sufcient con-
dition for P A ; P B to lie in this region is

1
PA 4 ; o 33
2 1 2 1

where

1  PB : 34

It is convenient for our purposes to assume that (33) holds.

Let us now identify ownership with role A and temporary
residency with role B. Then the temporary resident will win with
probability 12 if either player is mistaken about its role (so that both
draw either from the lower or the upper part of the distribution)
and with probability 1 if both players are mistaken, but will lose if
both players are correct. Thus the probability that the temporary
resident wins is
1
2  P B 1  P A 12  1  P B P A 1  1  P B 1  P A 0
 P B P A 12 f1  P A g 35

where ( 1  P B ) is now the probability that the intruder

believes itself to be the owner. The above expression is strictly
increasing with respect to .
The distribution over 0; 1 of contest duration is determined
in Appendix C, where we also calculate its mean and variance 2.
In Fig. 5, and 2 are plotted against the intruder's probability Fig. 6. , 2 and pI versus time on territory t for 0:025 and three rates of increase
1  P B of belief in ownership for PA 0.95 and 2, so that in belief in ownership with PA 0.95, VA 2, V B 1 C B and C A 12 (implying
(33) is satised for o 25. r 12) to satisfy (28) with 2. By (32)(34) and (36), the ESS is unique only for

Increasing 1  P B corresponds to increasing time on territory, 0o t o 1 ln P APA  where the curve is shown solid, but continues to exist for
2 1
which we denote by t. Let us make this correspondence explicit 
PA 
1
ln P A  o t o 1 where the curve is shown dashed.
2 1
 M. Mesterton-Gibbons, T.N. Sherratt / Journal of Theoretical Biology 394 (2016) 137148
with
 t
P B 1  P A P A  e 36
where is the rate at which belief in ownership increases,
(small) is initial belief in ownership and the intruder's belief in
ownership (1  P B ) approaches that of the true owner (PA) as t-1.
Then the probability that the temporary resident wins becomes
n o
pI 12 1  P A  e  t : 37
This expression increases with t at a rate determined by , from
2f1  P A g as t-0 to 2 as t-1; because is small with P A  1,
1 1
the lower limit is small (  12). In Fig. 6(a), the mean contest
duration, , is plotted against time on territory t for the same
values of PA, VA, VB, CA and CB as in Fig. 5 with 0:025 in (36) and
three different values of . Fig. 6(b) and (c) shows the corre-
sponding curves for the variance, 2, and pI.
5. Discussion
We have developed two complementary models to understand
why both the contest length and the probability of an individual
winning against a co-owner might vary with the duration of
residency. Both models assume conditions under which there is a
degree of habitat-mediated, or behaviorally mediated confusion
over ownership (Waage, 1983), and both models consider contests
which are decided on the basis of the costs individuals are pre-
pared to incur in order to win. The rst model (A) assumes no
asymmetry in roles (neither individual knows the arrival time of
the other) and simply investigates how a change in value of the
resource during occupancy affects the ESS contest effort on
encounter. This ESS contest effort takes the form of an initial
investment and a time-dependent additional effort, reecting a
potential increase in the motivation to win as the resource
becomes more valuable to the resident. By contrast, the second
model (B) considers a contest over a resource that remains xed in
value over time, but assumes an asymmetry in roles and a time-
dependent degree of confusion over which role an interloper is in
(the real owner continues to consider itself the owner with high
probability). The ESS identied in this second model is a
persistence-time bid, with the resource going to the individual
prepared to contest the resource for the longest (i.e., an asym-
metric war of attrition, Hammerstein and Parker, 1982).
Model A shows that if a territory becomes more valuable to a
resident over time, then the evolutionarily stable effort expended
in excess of base effort is in general positive and increases with
time spent in residency, since incremental effort tends to be
positive at the ESS. An exception to this general rule arises only
when the distribution of residency times is clumped and the rate
of increase of resource value with time in residency is very high; in
this case, the ESS effort is high (because base effort is high), but
may not increase with time in residency (Fig. 2, right-hand col-
umn, thick solid curves). In short, the model predicts that if
resources become more valuable to residents over time, then
whenever a dispute arises, residents should invest more to retain
their resources the longer they have been in residence. Intrigu-
ingly, however, even for a relatively modest increase in the value
of the resource to an occupant over time (i.e., a low ), the longer
residing co-owner is much more likely to win any contest that
ensues. Put simply, long-term owners tend to invest more in ghts
and thereby win, because they have already invested more non-
transferable costs into the territory and will therefore have more
to lose than a more temporary resident if they fail to keep it. This
effect of longer residency reinforces any priority advantagethat
is, any tactical advantage in contesting the territory from greater
145
experience of itbut is also a quite distinct effect, since it arises
even in the absence of priority advantage, i.e., when 0 (corre-
sponding to the dashed curves in Fig. 2), which implies 0 in
(4).
Resources such as cases for caddis ies, silk shelters for cater-
pillars and shells for hermit crabs are not breeding territoriesbut
they are objects of value that are fought over, just as territory
occupancy is frequently disputed. Given the close parallels, beha-
vioral ecologists have long tended to consider contests over non-
breeding resources under the same general umbrella as conicts
over breeding resources. We have presented our model in terms of
disputes over territory occupancy, because rival time-based claims
are much more likely to arise under these conditions. Never-
theless, when discussing how investment in ghting might change
with territory value, we feel it is appropriate to spread the net
wide and consider examples of disputes over all resources, rather
than only territorial ones.
In this regard, there is considerable evidence that the value of a
resource inuences an individual's motivation to obtain or retain
it. Caddis y larvae residents alter their aggressive response
according to the quality of the case they occupy (Englund and Otto,
1991), while spiders increase their signalling rates when they
occupy webs that are larger in area (Riechert, 1984). Bergman et al.
(2010) found that male Pararge aegeria butteries that had inter-
acted with a female had a higher probability of becoming domi-
nant and reversing the contest outcome in staged competitions. In
a recent study, Yack et al. (2014) found that the signalling rates of
resident masked birch caterpillars (Drepana arcuata) increased
with increased duration on the leaf prior to introducing an
intruder. Moreover, squatters placed on a leaf containing a full silk
shelter (made by another individual) signalled more to intruders
than did caterpillars placed on a fresh leaf a similar amount of
time, while residents whose shelters had been removed signalled
less than those occupying an intact shelter. Collectively, their
experiments indicate that the motivation of caterpillars to signal is
a function of both prior residency duration and territory quality,
although the two are clearly interlinked. Silk is a costly material to
make, and if the caterpillars have invested limited time and
resources in creating a shelter, then they will have gradually
accumulated costs that are at least in part non-transferrable. It is
this non-refundable investment that may explain the endowment
effect, with individuals valuing the objects they own more than
identical objects owned by others (Kahneman et al., 1991) simply
because the consequences of loss become higher for an investing
resident than a random intruder.
In a classic paper, Dawkins and Brockmann (1980) reported
eld observations of 23 incidences of co-occupation of nests by the
digger wasp Sphex ichneumoneus. These instances of co-occupation
are unusual in that two wasps, seemingly temporarily unaware of
each other's presence, provision the same burrow with insect prey
(paralyzed katydids). The co-occupants rarely meet because they
spend most of their time hunting, but when they do a ght gen-
erally ensues. The duration of these ghts in such instances was
found to increase with the total number of katydids in the nest,
but was most strongly correlated with the number of prey items
the loser had contributed. Although our rst model was based on
contest effort rather than ght duration, it provides quantitative
support for the intuitive proposal that individuals should be more
motivated to ght to retain a property when it is more valuable to
them. Intriguingly, the winner of the above contests tended to be
the wasp who had brought most prey to the nest (although the
effect was a marginal one). One might wonder why the greater
investor tends to win, since the katydid resources are entirely
transferable and therefore equally valuable to both parties. One
possibility, suggested by Dawkins and Brockmann (1980), is that
the wasps count only their own investment when considering
146 M. Mesterton-Gibbons, T.N. Sherratt / Journal of Theoretical Biology 394 (2016) 137148

Z 1 Z 1 Z 1
value, in which case our model readily explains why the individual f 1 tVtgsgt
r ds dt  tc0 u1 u2 tgt dt;
who places greater value on the resource tends to win it. u2 2 0 0 1 coshQ u; v; s; t 0
Our complementary model B conrms that as an interloper A:1b
grows increasingly condent that it is the rightful owner, a belief
Z 1Z 1
that we have assumed to increase with time in residency (see also 2 f 1 VtsinhQ u; v; s; tgsgt
 r2 ds dt
Alcock and Bailey, 1997), then the duration of the contest with the u12 2 0 0 f1 coshQ u; v; s; tg
2
Z 1
original resident will increase. In addition, the model predicts that
 c u1 u2 tgt dt A:2a
in the absence of any RHP asymmetry, the probability that the 0
interloper will win in a contest against the original resident rises
and
as the interloper's belief in ownership increases. Thus the model
Z 1Z 1 2
helps explain directly why ghts are longer when both contestants 2 f 1 t VtsinhQ u; v; s; tgsgt
 r2 ds dt
have held the territory in the past, yet tend to be short with ran- u22 2 0 0 f1 coshQ u; v; s; tg
2
Z 1
dom intruders. Mesterton-Gibbons et al. (2014) analysed a pair of
territorial games with discrete strategies and also found that  t 2 c u1 u2 tgt dt: A:2b
0
confusion over roles increases the frequency of ghting. Here,
Setting u1 v1 and u2 v2 in (A.1) reduces (13) to
however, we have considered a continuous game, allowing us to
Z 1Z 1 Z 1
consider contest duration directly. 1 Vtgsgt
r ds dt c0 v1 v2 tgt dt
An increase in contest duration with residency has been widely 2 0 0 1 coshf rv2 gft  sg 0
observed, as has a time-dependent increase in the probability that A:3a
the interloper will win (see Section 1). For example, in aquarium Z Z Z
1 1 1
experiments with brown trout (Salmo trutta), temporary replace- 1 tVtgsgt
r ds dt tc0 v1 v2 tgt dt
ments that were allowed to be resident for 4 days fought for 2 0 0 1 coshf rv2 gft  sg 0

longer and won more contests against the original owners than A:3b
interlopers who were resident for only 2 days (Johnsson and For- for an interior ESS, with v2 0 in (A.3a) for a boundary ESS. As
ser, 2002). Likewise, Figler and Einhorn (1983) staged contests noted in Section 3, for any values of , , , r and any choice of g,
between territorial convict cichlids, Cichlasoma nigrofasciatum and Eqs. (A.3) are readily solved by numerical means to yield a unique
found that individuals must be the sole resident for somewhere pair of values of v1 and v2, with (14) or (16), as appropriate,
between 1 and 2 days before consistently dominating intruders. invariably satised.
Moreover, they found that residents temporarily removed from
their territories after 3 days dominated their interloper after 1 h
removal, but not after 2 h removal. Collectively, this work Appendix B. Constraints on Hammerstein & Parker's ESS
demonstrates that the prior-residency effect among individuals
with competing claims to ownership is not instantaneous and Let wIJ be the probability that Player 1 identies with role I
develops over a period of time, perhaps (as the authors suggest) as while Player 2 identies with role J for I; J A; B. Then
a consequence of the decline in a fear response (another form of
wAA P A 1  P B
non-transferable cost analogous to negotiating boundaries with
wAB 12 P A P B 12 1  P B 1  P A
neighbors).
Naturally, it is quite possible for a territory to increase in net wBA 12 P B P A 12 1  P A 1 P B wAB
value with increasing familiarity, and for contestants to have role wBB P B 1  P A B:1
asymmetry with imperfect identication of roles. Such a model
with
would involve a modication of Model B to allow territory value to
vary with time in residency, as currently assumed in Model A. wAA wAB wBA wBB 1: B:2
Indeed, when this arises there may be conditions under which the Let VIJ and CIJ denote value and cost, respectively, to Player
interloper is eventually in the favored role with respect to payoffs. 1 conditional upon event (or situation) IJ, that is, Player 1 identies
Nevertheless, by treating these phenomena separately we can not with role I while Player 2 identies with role J; let ProbDj E
only elucidate the effects of motivation and confusion over roles in denote the probability of event D, conditional upon event E; and
isolation but also ensure that the models are analytically tractable. let events I, J and K all be either A or B. Then, by Bayes' Theorem,
ProbIJj KProbK
ProbK j IJ
ProbIJj AProbA ProbIJj BProbB
Acknowledgements
ProbIJj K
B:3
ProbIJj A ProbIJj B
This work was partially supported by a grant from the Simons
Foundation (#274041 to Mike Mesterton-Gibbons), a COFRS award because ProbA 12 ProbB. First suppose that I A, J B.
from Florida State University and a grant from NSERC (to Tom Then ProbABj A P A P B and ProbABj B 1  P B 1  P A ,
Sherratt). We thank our editor and an anonymous reviewer for from which V AB V A  ProbAj AB V B  ProbBj AB with ProbAj
helpful comments. AB P A P B =fP A P B 1  P B 1  P A g and ProbBj AB 1  P B 1 
P A =fP A P B 1  P B 1  P A g. Continuing in this fashion, we
readily obtain
Appendix A. The ESS equations for Model A V AA 12 V A V B
P P B V A 1  P B 1  P A V B
From (12) we obtain V AB A
P A P B 1  P B 1  P A
Z 1Z 1 Z 1 1  P A 1  P B V A P A P B V B
f 1 V tgsgt V BA
r ds dt  c0 u1 u2 tgt dt; P A P B 1  P A 1 P B
u1 2 0 0 1 coshQ u; v; s; t 0
A:1a V BB 12 V A V B
 M. Mesterton-Gibbons, T.N. Sherratt / Journal of Theoretical Biology 394 (2016) 137148 147

C AA 12 C A C B Appendix C. The contest duration distribution for Model B

P A P B C A 1  P B 1  P A C B
C AB To calculate the distribution function for contest lengths, let
P A P B 1  P B 1  P A
random variable D denote (unconditional) contest duration; and
1  P A 1  P B C A P A P B C B let random variable DIJ denote contest duration conditional upon
C BA
P A P B 1  P A 1 P B situation IJ, that is, Player 1 identies with role I while Player
2 identies with role J. Let D or DIJ have distribution function H or
C BB 12 C A C B B:4
HIJ and density h or hIJ, let Player 1's draw when identifying with
and hence role I be denoted by XI, and let Player 2's draw when identifying
with role J be denoted by YJ. Then
V AB V BA f1  P A 1 P B P A P B gP A P B  1V A C B  V B C A
 ProbDAA 4 ProbX A 4 ; Y A 4 ProbX A 4
C AB C BA fP A P B C A 1 P A 1  P B C B gf1  P A 1  P B C A P A P B C B g
B:5 ProbY A 4 f1  GA g2
ProbDAB 4 ProbY B 4 1  GB
in which the denominator and the term in curly brackets in the ProbDBA 4 ProbX B 4 1  GB
numerator are both positive. Favored is rst dened by inequality
ProbDBB 4 ProbX B 4 ; Y B 4 ProbX B 4 ProbY B 4 f1  GB g2
(5) of Hammerstein and Parker (1982, p. 653) as requiring the left-
C:1
hand side of (B.5) to be positive, which translates to (26), i.e., V A =C A
4 V B =C B (Hammerstein and Parker, 1982, p. 672) only if also implying
P A P B 4 1, that is, if (32) holds. The uniqueness of the ESS also
H AA ProbDAA r 1  f1  GA g2
requires the weak asymmetry condition, that is, both wAB V AB 4 wBB
H AB ProbDAB r GB
V BB and wBA C BA 4 wAA C AA (Hammerstein and Parker, 1982, p. 653).
H BA ProbDBA r GB
We have
  H BB ProbDBB r 1  f1  GB g2 C:2
wAB V AB  wBB V BB 12 2P A 1P B V A 1 P A 1  2P B V B
  and hence
wBA C BA  wAA C AA 12 1  2P A 1  P B C A 2P B  1P A C B B:6
H ProbD r
For both these quantities to be positive, we require ProbDAA r  wAA ProbDAB r  wAB ProbDBA r
wBA ProbDBB r  wBB
2P A 1P B qV 1  P A 1  2P B 4 0 B:7a P A 1  P B GA f2  GA g f1  P A P B gGB
and  P B 1  P A fGB g2 C:3

1  2P A 1  P B qC 2P B  1P A 4 0 B:7b after simplication with (B.1). Differentiating and using (30), we

nd that D has density
where qC and qV are dened by (27). Because (B.7a) excludes
h H 0
P A r 12; P B Z 12, (B.7b) excludes P A Z 12; P B r 12 and (32) excludes
2P A 1 P B g A f1  GA g f1  P A P B gg B
P A r 12; P B r 12, (B.7) can hold only if PA and PB both exceed 12, that is, if
 2P B 1  P A GB g B
the point P A ; P B lies in the top right-hand quadrant of the square (
dened by (29). Within this smaller square, (B.7) will actually hold Kf1 P A P B  2KP B 1  P A 1  e  r =rge  r if 0 o o s

within the region dened by the inequalities 2P A 1  P B re2rs  if s o o1;

qV 2P B  1 2P B  1
PA 4
2qV 1P B  1 2 1P B  1
mean
P 2P A  1qC 2P A 1 Z 1
PB 4 A B:8 2  P A P B 2r  KP B s K1  P A fKP B  rs  P B g
2P A 2P A  1qC 2 1P A  1 h d
0 2r r2
This region is shaded dark in Fig. 4 for various values of ; it lls the C:4
top right-hand quadrant of the unit square dened by (29) in the limit and variance
as -1. Z 1
4  P A  P B 2 P A  P B 41  P A P B Ks
A sufcient condition for (B.8) to hold is for P A ; P B to lie in the 2  2 h d
0 4r 2
still smaller square with solid lower and left-hand boundaries,
which lies wholly within the region that is shaded dark in Fig. 4. It K1  P A P B 3P A  P B  1Ks P A P B  1

is now straightforward to show from (B.8) that P A ; P B lies within r3
this square when (33) holds. KK1 P A  rKP B  r1  P A P B s2

The weak asymmetry condition is a sufcient condition for r3
uniqueness, not a necessary condition for existence. Even if it fails K 2 1  P A P B f1  P A P B 2Ks  1  K1  P A  rKP B  rs2 g
:
to hold, the ESS dened by (30) and (31) will still exist as long as r4
w2AB C BA V AB 4 wAA wBB C AA V BB (Hammerstein and Parker, 1982, p. C:5
665) in addition to (29) and (32). It is straightforward to show that
(B.1), (B.4), (27) and (28) imply
References
w2AB C BA V AB  wAA wBB C AA V BB 14 P A P B  1fP A P B V B C B

1 P A 1  P B V A C A 14 P A P B  12 V B C B ; B:9 Alcock, J., Bailey, W.J., 1997. Success in territorial defence by male tarantula hawk
wasps Hemipepsis ustulata: the role of residency. Ecol. Entomol. 22, 377383.
which is always positive. So existence requires only (29) and (32): Bergman, M., Olofsson, M., Wiklund, C., 2010. Contest outcome in a territorial
buttery: the role of motivation. Proc. R. Soc. Lond. B 277, 30273033.
the ESS exists throughout the shaded upper triangle of Fig. 4, Broom, M., Rycht, J., 2013. Game-Theoretical Models in Biology. CRC Press, Boca
although it is only in the darker region that the ESS is unique. Raton, FL.
148 M. Mesterton-Gibbons, T.N. Sherratt / Journal of Theoretical Biology 394 (2016) 137148
Davies, N.B., 1978. Territorial defence in the speckled wood buttery (Pararge
aegeria): the resident always wins. Anim. Behav. 26, 138147.
Dawkins, R., Brockmann, H., 1980. Do digger wasps commit the concorde fallacy?
Anim. Behav. 28, 892896.
Englund, G., Otto, C., 1991. Effects of ownership status, weight asymmetry, and case
t on the outcome of case contests in two populations of Agrypnia pagetana
(Trichoptera: Phryganeidae) larvae. Behav. Ecol. Sociobiol. 29, 113120.
Figler, M.H., Einhorn, D.M., 1983. The territorial prior residence effect in convict
cichlids (Cichlasoma nigrofasciatum Gunther): temporal aspects of establish-
ment and retention and proximate mechanisms. Behaviour 85, 157183.
Gribbin, S.D., Thompson, D.J., 1991. The effects of size and residency on territorial
disputes and short-term mating success in the damsely Pyrrhosoma nymphula
(Sulzer) (Zygoptera: Coenagrionidae). Anim. Behav. 41, 689695.
Hammerstein, P., Parker, G.A., 1982. The asymmetric war of attrition. J. Theoret.
Biol. 96, 647682.
Johnsson, J.I., Forser, A., 2002. Residence duration inuences the outcome of ter-
ritorial conicts in brown trout (Salmo trutta). Behav. Ecol. Sociobiol. 51,
282286.
Kahneman, D., Knetsch, J.L., Thaler, R.H., 1991. The endowment effect, loss aversion,
and status quo bias. J. Econ. Perspect. 5, 193206.
Kemp, D.J., 2000. Contest behavior in territorial male butteries: does size matter?.
Behav. Ecol. 11, 591596.
Kemp, D.J., 2003. Twilight ghting in the evening brown buttery, Melanitis leda (l.)
(Nymphalidae): age and residency effects. Behav. Ecol. Sociobiol. 54, 713.
Kemp, D.J., 2013. Contest behaviour in butteries: ghting without weapons. In:
Hardy, M., Briffa, M. (Eds.), Animal Contests. Cambridge University Press,
Cambridge, pp. 134146.
Kemp, D.J., Wiklund, C., 2001. Fighting without weaponry: a review of male-male
contest competition in butteries. Behav. Ecol. Sociobiol. 49, 429442.
Kemp, D.J., Wiklund, C., 2004. Residency effects in animal contests. Proc. R. Soc.
Lond. B 271, 17071711.
Kokko, H., 2013. Dyadic contests: modelling ghts between two individuals. In:
Hardy, M., Briffa, M. (Eds.), Animal Contests. Cambridge University Press,
Cambridge, pp. 532.
Krebs, J.R., 1982. Territorial defence in the great tit (Parus major): do residents
always win?. Behav. Ecol. Sociobiol. 11, 185194.
Maynard Smith, J., 1982. Evolution and the Theory of Games. Cambridge University
Press, Cambridge.
Mesterton-Gibbons, M., Karabiyik, K., Sherratt, T.N., 2014. The iterated Hawk-Dove
game revisited: the effect of ownership uncertainty on Bourgeois as a pure
convention. Dyn. Games Appl. 4, 407431. http://dx.doi.org/10.1007/
s13235-014-0111-5.
Mesterton-Gibbons, M., Marden, J.H., Dugatkin, L.A., 1996. On wars of attrition
without assessment. J. Theoret. Biol. 181, 6583.
Peixoto, P.E.C., Benson, W.W., 2012. Inuence of previous residency and body mass
in the territorial contests of the buttery Hermeuptychia fallax (Lepidoptera:
Satyrinae). J. Ethol. 30, 6168.
Riechert, S.E., 1984. Games spiders play. III. Cues underlying context-associated
changes in agonistic behaviour. Anim. Behav. 32, 115.
Rutowski, R.L., 1992. Male mate-locating behavior in the common eggy, Hypo-
limnas bolina (nymphalidae). J. Lepid. Soc. 46, 2438.
Sherratt, T.N., Mesterton-Gibbons, M., 2015. The evolution of respect for property. J.
Evol. Biol. 28, 11851202. http://dx.doi.org/10.1111/jeb.12648.
Takeuchi, T., 2006. Matter of size or matter of residency experience? Territorial
contest in a green hairstreak Chrysozephyrus smaragdinus (Lepidoptera:
Lycaenidae). Ethology 112, 293299.
Tobias, J., 1997. Asymmetric territorial contests in the European robin: the role of
settlement costs. Anim. Behav. 54, 921.
Waage, J.K., 1983. Sexual selection, ESS theory and insect behavior: some examples
from damselies (Odonata). Fla. Entomol. 66, 1931.
Waage, J.K., 1988. Confusion over residency and the escalation of damsely terri-
torial disputes. Anim. Behav. 36, 586595.
Walton, R., Nolan Jr., V., 1986. Imperfect information and the persistence of pre-
tenders: male prairie warblers contesting for territory. Am. Nat. 128, 427432.
Wickman, P.O., Wiklund, C., 1983. Territorial defence and its seasonal decline in the
speckled wood buttery Pararge aegeria. Anim. Behav. 31, 12061216.
Yack, J.E., Gill, S., Drummond-Main, C., Sherratt, T.N., 2014. Residency duration and
shelter quality inuence vibratory signalling displays in a territorial caterpillar.
Ethology 120, 354364.