Behavioral Ecology
doi:10.1093/beheco/arj059
Advance Access publication 8 March 2006
Signal residuals during shell fighting in hermit
crabs: can costly signals be used deceptively?
Mark Briffa
Marine Biology and Ecology Research Centre, School of Biological Sciences, University of Plymouth,
Plymouth PL4 8AA, UK
Animals advertise aspects of individual quality in a range of situations including during agonistic contests over access to resources.
The advertisement can be produced by different types of signal, which can be broadly classified into ‘‘conventional signals’’ and
‘‘costly signals.’’ In both cases, the signals are thought to be honest such that any benefits to cheating are very limited due to
prohibitive costs of advertising an inaccurately high level of quality. However, some individuals may benefit by low-level cheating
in systems that are otherwise honest. Here, I test this possibility for a costly signal by analyzing residuals from relationships
between a key parameter of signal magnitude and two measures of fighting ability from ‘‘shell fights’’ in hermit crabs. Contrary to
the expectations for significant use of bluffing, individuals that performed at a greater magnitude than expected for their ability
were more likely to win the encounter but used fewer repetitions than those that performed below the level expected for their
ability. This indicates that the level of performance is primarily driven by the cost and that the functional significance of signal
residuals in this case is that they provide information on several measures of fighting ability. Key words: contest, cost, deception,
hermit crab, honesty, signal. [Behav Ecol 17:510–514 (2006)]
nimals use signals to advertise individual quality in a range
of situations, including contests over resources. During
agonistic encounters, opponents advertise their fighting
ability or resource holding potential (RHP) by using ‘‘conventional signals’’ (Hurd, 1997) based on displaying an ornament
or with ‘‘costly’’ (Maynard Smith and Harper, 1995) signals by
performing an activity, as in hermit crabs (Briffa and Elwood,
2005), crickets (Hack, 1997), and fiddler crabs (Matsumasa
and Murai, 2005) for example. For signals to be evolutionarily
stable, most instances of signaling must be ‘‘honest’’ with performances that accurately reflect the sender’s RHP. Zahavi
(1977) argued that the honesty of biological signals could
be guaranteed if the costs are correlated with the feature
being advertised. On average, it should not be adaptive to
produce a signal that indicates an inaccurately high level of
quality because the costs of producing that signal would outweigh the potential benefits of an exaggerated advertisement.
Although this ‘‘handicap principle’’ of honest signals is often
discussed within the framework of sexually selected conventional signals (Hurd, 1997), signals that advertise other aspects
of individual quality such as RHP may also be guaranteed by
cost, and costly signals as well as conventional signals are expected to be honest (Grafen, 1990; Zahavi, 1977).
Although signals are expected to be honest on average, it is
possible that low levels of ‘‘cheating’’ are stable within an
otherwise honest population (Adams and Mesterton-Gibbons,
1995; Bond, 1989; Getty, 1997). Cheating is defined as benefiting from an exaggerated advertisement that results from producing an inaccurately high level of the signal. The presence
of cheating may be detected by the analysis of residuals from
a population-based relationship between a measure of RHP
and a measure of signal magnitude (Hughes, 2000). Performances with positive residuals are greater than expected for the
A
Address correspondence to M. Briffa. E-mail: mark.briffa@
plymouth.ac.uk.
Received 21 September 2005; revised 7 February 2006; accepted 12
February 2006.
The Author 2006. Published by Oxford University Press on behalf of
the International Society for Behavioral Ecology. All rights reserved.
For permissions, please e-mail: [email protected]
sender’s RHP, whereas those with negative residuals are lower
than expected (see figure 1 in Hughes, 2000). If individuals
that exaggerate benefit from doing so, they should perform
more repetitions of the signaling activity than those that do
not exaggerate (Hughes, 2000). This exaggeration is only possible, however, if the signal residuals do not correlate with
competitive ability. Deceptive use of the signal is not predicted
in cases where either (1) signal residuals vary between outcomes or (2) individuals with lower than expected signal magnitude perform more or the same number of repetitions than
those with positive residuals.
Hughes (2000) demonstrated ‘‘incomplete honesty’’ in the
use of chelar displays in the snapping shrimp Alpheus heterochaelis, and similar results were reported in mantis shrimp
Gonodactylus oerstedii (Adams and Caldwell, 1990) and fiddler
crabs Uca annulipes. In the case of the latter two studies, it was
suggested that cheating is possible due to a temporary disparity between the visual appearance of the chelipeds and their
actual strength or weight.
Hermit crabs, Pagurus bernhardus, use ‘‘shell-rapping’’ signals to resolve contests over ownership of empty gastropod
shells (Briffa and Elwood, 2000a; Briffa et al., 1998, 2003).
An individual in a suboptimal shell (either because it is too
small or of an unpreferred species) may attempt to evict another individual from its shell if that shell is of higher quality.
Attackers initiate a fight by grabbing the defender’s shell, and
the defender withdraws tightly into its shell. The attacker then
performs bouts of shell rapping where it repeatedly strikes the
surface of its shell against that of the defender, using its abdominal muscles to provide the power. Shell rapping is performed in a series of bouts that are separated by pauses. After
a series of bouts, the contest can end in one of two ways. The
defender may make a decision to ‘‘give up’’ and release its grip
of its shell, allowing the attacker to evict it by pulling it out
through the aperture. Alternatively, the attacker may decide to
give up without evicting the defender.
Although it involves direct contact between the two animals,
shell-rapping functions as a ‘‘repeated signal’’ advertising the
RHP of the attacker (Briffa and Elwood, 2000a,b,c; Briffa
et al., 1998). It is an energetically demanding activity (Briffa
Briffa • Signal residuals during shell fighting in hermit crabs
and Elwood, 2001), and the pattern of change in vigor during
the fight (Briffa and Elwood, 2000b; Briffa et al., 1998) shows
that it must function as a signal of stamina, a key correlate of
RHP (Payne and Pagel, 1997). Physiological data show that
the level of resistance offered by defenders is based on their
perception of the attacker’s RHP (Briffa and Elwood, 2005).
Different parameters of the pattern of signaling contribute to
the defender’s decision to give up, but the critical feature is
the duration of the pauses between the bouts. These are inversely proportional to the attacker’s stamina and thus advertise the attacker’s RHP (Briffa et al., 1998).
The aim of this study is to analyze signal residuals, where
the signal is the inverse of pause duration (1/P) in order to
determine whether shell rapping can be used deceptively.
The relationship between signal magnitude and RHP will be
examined for two different measures of RHP: (1) the relative
difference in weight (RWD) between attacker and defender
and (2) the postcontest lactic acid concentration in attackers,
which have both previously been shown to vary with outcome
and attacker performance (Briffa and Elwood, 2001; Briffa
et al., 1998).
If there is no difference in signal residuals between outcomes and if attackers with positive residuals perform a greater
number of bouts, then shell rapping may be used dishonestly.
It is important to note that contest duration is determined by
the decision of the loser to terminate the contest by giving up
rather than by any decision made by the winner. Therefore,
the difference in the number of bouts can only be meaningfully tested for in attackers that failed to evict the defender
because these attackers decided their own number of performances, whereas in successful attackers the number was
determined by the duration of resistance in the defender.
By analyzing these signal residuals, it will be possible to determine whether they vary between outcomes and whether
the number of performances in attackers that give up varies
between attackers with positive and negative signal residuals.
511
ferred to an arena consisting of a crystalizing dish filled with
a 1-cm layer of sand and aerated seawater. Observations were
conducted behind a one-way mirror such that the observer
could not be seen by the crabs. During the fight, the pattern
of shell rapping was recorded using a handheld computer
configured as a time-event recorder using The Observer 3.0
behavioral software package (1993, Noldus IT bv, Wageningen,
The Netherlands). Fights were recorded until they were
terminated either by the attacker evicting the defender or
the attacker giving up. In all four studies, crabs were only used
once before being returned to the sea or sacrificed for lactate
analysis.
The combined data sets yielded a total of 364 fights for
analysis of which the attacker gave up without evicting the
defender in 82 cases. A subsample of 105 fights came from
studies where postfight lactic acid was assayed in attackers and,
of these, the attacker failed to evict the defender in 38 cases.
The data were analyzed using Statview 3.0. Statistical tests were
two tailed in all cases. Lactate concentration was performed
using a standard enzymatic technique described elsewhere
(Briffa and Elwood, 2005). Because postfight lactate is negatively correlated with RHP, the inverse, 1/[lactate], was used
as the measure of ability.
RESULTS
Relationships between RHP and signal magnitude
Regression analyses were used to determine the relationships
between each measure of RHP (independent variable) and 1/P
(dependent variable). In both cases, there was a highly significant positive relationship (RWD: F1,358 ¼ 22.99, p , .0001,
1/[lacate]: F1,103 ¼ 13.24, p , .0005) (Figure 1a,b) but a low
R2 (RWD: R2 ¼ .06, 1/[lactate]: R2 ¼ .12) indicating high
variance about each relationship.
Differences in signal residuals between outcomes
METHODS
Data sets were pooled from four previous studies (Briffa and
Elwood, 2000c, 2001, 2002; Briffa et al., 1998) in order to
analyze signal residuals. These all involved staged encounters
between pairs of crabs, but they drew crabs from two different
size classes (small ¼ 0.16–0.48 g and large ¼ 0.9–2.7 g). Although the distribution of attacker weight was bimodal, the
mean relative difference in weight between attacker and defender (RWD), calculated by RWD ¼ 1 (weight of defender/
weight of attacker), following Briffa et al. (1998), was not
different between the groups (small: 0.316 6 SE ¼ 0.012,
large: 0.333 6 SE ¼ 0.01, t362 ¼ 1.1, ns) and could thus be
used as a continuous measure of RHP. Crabs were collected
from Ballywalter and Minerstown, Northern Ireland, and
taken to the laboratory in Belfast. Fights were staged by removing the crabs from their shells, by cracking the shell in
a bench vice, and then allocating the crabs to pairs comprising
a large individual, the potential attacker, and a smaller individual, the potential defender. Only male crabs that were free
from recent moult, missing appendages, and obvious parasites
were used in the observations. Optimal shell sizes were calculated from regression lines relating crab weight to preferred
shell weight, obtained during previous shell-choice experiments (Jackson and Elwood, 1989). In each case, the large
crab was supplied with a new shell of the preferred species
but only 50% of the preferred size, whereas the small crab was
supplied with a shell 100% adequate in size for the corresponding large crab. Thus, by evicting the small crab, attackers could make a substantial gain in shell quality. After
a settling period of not less than 16 h, the crabs were trans-
Unpaired t tests were used to determine whether there was
a difference in the two sets of residuals between winners (attackers that evicted the defender) and losers (attackers that
gave up without effecting an eviction). In both cases, the residuals of losers were significantly lower than those of winners
(RWD versus 1/P: t359 ¼ 7.44, p , .0001, 1/[lactate] versus
1/P: t102 ¼ 3.83, p , .0002) (Figure 2a,b). A series of one-group
t tests showed that all four mean residuals were significantly
different from zero, indicating that winner residuals were significantly positive (RWD versus 1/P: mean ¼ 10.003, t278 ¼
3.35, p , .001, 1/[lactate] versus 1/P: mean ¼ 10.004, t66 ¼
2.14, p , .05) and that there was a highly significantly negative
direction to loser residuals (RWD versus 1/P: mean ¼ 0.1,
t80 ¼ 8.3, p , .0001, 1/[lactate] versus 1/P: mean ¼ 0.007,
t36 ¼ 6.15, p , .0001).
Difference in number of bouts between attackers with
positive and negative residuals
The number of bouts performed by attackers with positive
and negative residuals, from fights where the attacker failed
to evict the defender, was compared with unpaired t tests.
Attackers that left shorter pauses than expected for their
RWD (n ¼ 12) performed fewer bouts than those that left
longer pauses than expected (n ¼ 80) (t79 ¼ 3.89, p ,
.0002) (Figure 3a), whereas there was no difference between
the number of bouts performed by attackers with positive
(n ¼ 8) and negative (n ¼ 30) residuals from the relationship
between 1/[lactate] and the total number of bouts (t35 ¼ 0.89,
p ¼ .4, power ¼ 0.13) (Figure 3b).
Behavioral Ecology
512
0.006
0.25
(a)
Mean residual (RWD v 1/P)
0.004
1/Pause duration
0.20
0.15
0.10
0.05
(a)
0.002
0.000
-0.002
-0.004
-0.006
-0.008
-0.010
0.00
-0.012
0.0
0.2
0.4
0.6
No eviction
0.8
0.008
0.25
Mean residual (1/[Lactate] v 1/P)
(b)
1/Pause duration
0.20
0.15
0.10
0.05
0.00
Eviction
Outcome
RWD
0.006
0.004
0.002
0.000
-0.002
-0.004
-0.006
-0.008
-0.010
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1/[Lactate]
Figure 1
Relations between (a) relative weight difference ‘‘RWD’’ (increasing
RHP) and (b) 1/Lactic acid concentration (increasing RHP) with
1/P (increasing signal magnitude).
DISCUSSION
Use of repeated signals during aggressive interactions is a
widespread feature of animal contests. Male red deer, Cervus
elephas, for example, perform bouts of vocalizations during contests over access to females (Clutton-Brock and Albon, 1979),
as do anurans (e.g., Ryan, 1988) and birds (e.g., Ballintijn and
ten Cate, 1999; ten Cate et al., 2002). In all these cases, it has
been suggested that the signaling activity is energetically demanding and, therefore, costly to perform. The aim of the
present study was to determine whether such costly signals
could be used deceptively. In the shell fights I analyzed, signal
magnitude (1/P) increased significantly with both measures
of RHP (RWD and 1/[lactate]), but in each case, there was a
large amount of variation around the relationship (Figure 1).
Nevertheless, it appears that deception, by performing an
inaccurately high level of the shell-rapping signal, is not a
feature of these contests.
Although successful attackers on average performed significantly better than would be expected according to either
measure of RHP (Figure 2), attackers that gave up performed
significantly worse than expected. Furthermore, the negative
deviation from the expected signal magnitude in this group
was far greater than the positive deviation seen in successful
attackers (Figure 2). Thus, it appears that it is only possible for
successful attackers to achieve small increases in signal mag-
(b)
No eviction
Eviction
Outcome
Figure 2
Differences in signal residuals between unsuccessful (no eviction)
and successful (eviction) attackers. Residuals differed between
outcomes and were significantly different from zero in each case
(*p , .05, **p , .001, ***p , .0001).
nitude. By contrast, the performance of those attackers that
failed to effect an eviction fell far below that expected for
their RHP. Thus, positive signal residuals are unlikely to exaggerate the perceived ability of the attacker because signal residuals correlate closely with the attacker’s competitive ability.
Furthermore, if significant levels of cheating were a feature of
this system, we would expect to see those unsuccessful attackers that performed with greater than expected vigor performing more bouts before giving up than those showing negative
signal residuals (Hughes, 2000), but this was not the case
(Figure 3).
In the analysis based on postfight lactate, there was no difference in the number of bouts, whereas in the case of residuals derived from RWD, the opposite pattern is seen with
attackers that leave shorter than expected pauses performing
fewer bouts than those with negative residuals (Figure 3). It
has been previously noted that variation in the pattern of
rapping during the fight can both escalate and de-escalate,
depending on the outcome (Briffa et al., 1998), and this pattern of variation in signal magnitude fits some of the predictions of an energetic war of attrition model of repeated signals
(E-WOA) where the signals are assumed to advertise stamina
(Payne and Pagel, 1997). Of the attackers that give up, those
that perform at a level higher than expected for the weight
difference between the two roles (RWD) do not persist for as
Briffa • Signal residuals during shell fighting in hermit crabs
1.2
(a)
Log mean number of bouts
1.0
0.8
0.6
0.4
0.2
0.0
Positive
Negative
Residual direction
1.4
Log mean number of bouts
(b)
1.2
1.0
0.8
0.6
0.4
0.2
0.0
Positive
Negative
Residual direction
Figure 3
The effect of direction of residuals from the relationships between
(a) RWD and 1/P and (b) postfight lactate and 1/P on the total
number of bouts performed by unsuccessful attackers (no eviction)
before deciding to give up. Different symbols above means denote
a significant difference.
long as those that perform at a lower than expected level
(Figure 3), the opposite of the pattern expected for cheating.
It is interesting that the effect of signal residual direction
(positive or negative) was significant in the case of RWDderived residuals but not 1/lactate-derived residuals. This suggests that the costs associated with RWD are more important
than those associated with physiological status. Given the energetic nature of shell rapping, this is surprising but may be
due to the possibility that RWD drives a variety of (presumably
physiological) costs of fighting in addition to lactate accumulation. Hughes (2000) suggested that in the absence of
cheating, there are two possible explanations for the variance
around relationships between signal magnitude and RHP
(Figure 1). These are (1) that the variation has no functional
significance because it is below the threshold of perception in
receivers and (2), as noted above, some of the variation must
be due to additional correlates of RHP. In the case of hermit
crabs, the first possibility is unlikely given the clear differences
in the vigor of rapping seen between outcomes (Briffa et al.,
1998). On the other hand, it has been demonstrated that a
range of different factors may contribute to RHP (Briffa and
Elwood, 2001, 2002, 2004; Briffa et al., 2003).
513
As noted above, in addition to agonistic encounters, costly
signals are used in other situations such as courtship, for example, abdominal drumming in wolf spiders (Kotiaho et al.,
1998) and displays in frigate birds (Dearborn et al., 2005). If
the relations between signal residuals, contest outcomes, and
the number of repetitions are similar to those seen in the
present analysis, it is unlikely that such signals are easy to bluff
or that individuals that signal at a higher than expected rate
gain any benefits from doing so. Even in the case of conventional signals, the costs associated with bluffing appear to be
prohibitive. In sparrows, for example, experimental elevation
of rank results in high costs in terms of increased risk of injury
(Moller, 1988). Perhaps then, the incomplete honesty described during contests in mantis shrimp (Adams and Caldwell,
1990) and fiddler crabs (Backwell et al., 2000) is most likely
in cases where there is a temporary disparity in relations between the visual appearance of an appendage used in displays
and true fighting ability. Previous studies on signal honesty
have focussed on conventional signals (e.g., Moller, 1988)
rather than signals given by performance of a demanding
activity, but the present data show that any benefits to exaggerating the shell-rapping signal must be extremely limited.
Indeed, individuals that signal with a greater than expected
magnitude also give up sooner than expected (Figure 3).
There are several interpretations of the handicap principle
(Grafen, 1990), and costly signals may act as ‘‘strategic choice
handicaps’’ where the cost varies between individuals and encounters and senders ‘‘choose’’ the cost they are prepared to
pay. Previous studies suggest that shell rapping could work in
this way as attackers that give up have high lactic acid levels
(Briffa and Elwood, 2001), such that the lactate accumulation
represents the ‘‘cost’’ they are prepared to pay. An alternative
explanation, however, is that high lactate causes a giving up
decision because it prevents effective rapping (Briffa and
Elwood, 2005). In order to test this, it would be necessary to
monitor energy use, for example, by respirometry, during the
contest rather than measuring accumulated costs at the end of
the contest. Further questions arise from the problem of multiple factors that may contribute to signal magnitude and from
the possibility that several signal components may contribute
to the overall signal. In costly signals, where the magnitude is
given by the rate of performance of an activity, the situation
is clearly more complex than in the case of conventional
signals given by display of an ornament or weapon.
Shell rapping, however, is evidently an energetically demanding activity (Briffa and Elwood, 2001, 2002, 2004,
2005), and pause duration has consistently been shown to
be the critical feature that determines attacker success but it
increases during the encounter. The high metabolic costs
have been shown to limit the agonistic ability of attackers
(Briffa and Elwood, 2005) and may make physical eviction
of a resisting defender more difficult. Thus, attackers could
potentially benefit from exaggerating their RHP, but the present analysis suggests that the signal is unlikely to be used in
this way. Analysis of residuals from other examples of costly
signals used during fights can potentially reveal whether an
absence of significant exaggeration is a typical feature of such
systems.
I thank Jon M. Russ and two anonymous referees for comments on an
earlier version of the manuscript.
REFERENCES
Adams ES, Caldwell RL, 1990. Deceptive communication in asymmetric fights of the stomatopod crustacean Gonodactylus-Bredini.
Anim Behav 39:706–716.
514
Adams ES, Mesterton-Gibbons M, 1995. The cost of threat displays
and the stability of deceptive communication. J Theor Biol 175:
405–421.
Backwell PRY, Christy JH, Telford SR, Jennions MD, Passmore NI,
2000. Dishonest signalling in a fiddler crab. Proc R Soc Lond B
267:719–724.
Ballintijn MR, ten Cate C, 1999. Variation in number of elements
in the perch-coo vocalization of the collared dove (Streptopelia
decaocto) and what it may tell about the sender. Behaviour
136:847–864.
Bond AB, 1989. Toward a resolution of the paradox of aggressive displays .1. Optimal deceit in the communication of fighting ability.
Ethology 81:29–46.
Briffa M, Elwood RW, 2000a. Analysis of the finescale timing of repeated signals: does shell rapping in hermit crabs signal stamina?
Anim Behav 59:159–165.
Briffa M, Elwood RW, 2000b. Cumulative or sequential assessment
during hermit crab shell fights: effects of oxygen on decision rules.
Proc R Soc Lond B 267:2445–2452.
Briffa M, Elwood RW, 2000c. The power of shell rapping influences
rates of eviction in hermit crabs. Behav Ecol 11:288–293.
Briffa M, Elwood RW, 2001. Decision rules, energy metabolism and
vigour of hermit-crab fights. Proc R Soc Lond B 268:1841–1848.
Briffa M, Elwood RW, 2002. Power of shell-rapping signals influences
physiological costs and subsequent decisions during hermit crab
fights. Proc R Soc Lond B 269:2331–2336.
Briffa M, Elwood RW, 2004. Use of energy reserves in fighting hermit
crabs. Proc R Soc Lond B 271:373–379.
Briffa M, Elwood RW, 2005. Rapid change in energetic status in fighting animals: causes and effects of strategic decisions. Anim Behav
70:119–124.
Briffa M, Elwood RW, Dick JTA, 1998. Analysis of repeated signals
during shell fights in the hermit crab Pagurus bernhardus. Proc R
Soc Lond B 265:1467–1474.
Briffa M, Elwood RW, Russ JM, 2003. Analysis of multiple aspects of
a repeated signal: power and rate of rapping during shell fights in
hermit crabs. Behav Ecol 14:74–79.
Behavioral Ecology
Clutton-Brock TH, Albon SD, 1979. The roaring of red deer and the
evolution of honest advertisement. Behaviour 69:145–170.
Dearborn DC, Anders AD, Williams JB, 2005. Courtship display by
great frigate birds, Fregata minor: an energetically costly handicap
signal? Behav Ecol Sociobiol 58:397–406.
Getty T, 1997. Deception: the correct path to enlightenment? Trends
Ecol Evol 12:159–160.
Grafen A, 1990. Biological signals as handicaps. J Theor Biol 144:
517–546.
Hack MA, 1997. Assessment strategies in the contests of male crickets,
Acheta domesticus (L.). Anim Behav 53:733–747.
Hughes M, 2000. Deception with honest signals: signal residuals and
signal function in snapping shrimp. Behav Ecol 11:614–623.
Hurd PL, 1997. Is signalling of fighting ability costlier for weaker
individuals? J Theor Biol 184:83–88.
Jackson NW, Elwood RW, 1989. How animals make assessments—
information gathering by the hermit crab Pagurus bernhardus.
Anim Behav 38:951–957.
Kotiaho JS, Alatalo RV, Mappes J, Nielsen MG, Parri S, Rivero A, 1998.
Energetic costs of size and sexual signalling in a wolf spider. Proc R
Soc Lond B 265:2203–2209.
Matsumasa M, Murai M, 2005. Changes in blood glucose and lactate
levels of male fiddler crabs: effects of aggression and claw waving.
Anim Behav 69:569–577.
Maynard Smith J, Harper DGC, 1995. Animal signals: models and
terminology. J Theor Biol 177:305–311.
Moller AP, 1988. Badge size in the house sparrow Passer domesticus—
effects of intrasexual and intersexual selection. Behav Ecol Sociobiol 22:373–378.
Payne RJH, Pagel M, 1997. Why do animals repeat displays? Anim
Behav 54:109–119.
Ryan MJ, 1988. Energy, calling, and selection. Am Zool 28:885–898.
ten Cate C, Slabbekoorn H, Ballintijn MR, 2002. Birdsong and malemale competition: causes and consequences of vocal variability in the
collared dove (Streptopelia decaocto). Adv Study Behav, 31:31–75.
Zahavi A, 1977. The cost of honesty (further remarks on the handicap
principle). J Theor Biol 67:603–605.