aB
a
e
hypothesis whereby juvenile ?sh are compelled to emerge earlier in order to resume feeding. In the
,
edownstream sites this effect was slightly offset by the relatively greater predation threat for smaller ?sh,
such that they delayed their emergence from cover. We discuss the underlying importance of variation in
boldness and its effects on other behavioural and life history traits.

 2004 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
The study of individual variation in animal behaviour has
become increasingly active over the last decade (Wilson
1998). Central to understanding this variance is the
potential effect of consistent ?personality? traits, such as
the boldness?shyness continuum (Wilson et al. 1994;
Coleman & Wilson 1998), differences in coping styles
(Benus et al. 1991), and demographic variables such as
sex, age and reproductive state.
The propensity to take risks (boldness) has implications
for survival, reproduction and many other life history and
behavioural traits (e.g. Budaev 1997a, b). Bold ?sh, for
example, are more likely to explore novel objects, discover
novel food sources or patches and stray from their normal
home range (Wilson et al. 1993; Fraser et al. 2001). Bold
minnows, Phoxinus phoxinus, inspect predators more
often, show higher rates of skittering and forage more
persistently in the face of predation threat than shy
minnows (Murphy & Pitcher 1997). Female guppies,
Poecilia reticulata, prefer to mate with bolder males (Godin
& Dugatkin 1996), and bolder males are most likely to
The propensity to take risks (boldness) has long been
associated with an individual?s nutritional state (Sih
1997), so it may also be indirectly in?uenced by de-
mography. Hungry ?sh are more likely to take risks than
satiated ?sh by emerging from shelters sooner to maxi-
mize foraging opportunities, but at the expense of in-
creasing their exposure to predation. If boldness is affected
by internal states and related demographic variables, then
it is likely that individual boldness scores will vary from
day to day or over the lifetime of an individual. As
contexts change over an individual?s lifetime so too might
its degree of boldness. One potential axis of contextual
change is variation caused by growth that results in
differences in motivational states within an individual
on an ontogenetic timescale.
In the banded killi?sh, Fundulus diaphanus, the time
spent in a refuge after exposure to a model trout predator
decreased with increasing body size suggesting that small
killi?sh are more vulnerable to predation than larger
killi?sh (Dowling & Godin 2002). The reverse was trueSize matters: a test of bold
of the poeciliid Br
CULUM BROWN & VICT
Institute of Evolutionary Biology, School of
(Received 22 July 2003; initial
?nal acceptance 28 April 2004; published on
Individual variation in behaviour within populations
term, stable individual psychological differences. We e
time to emerge from a shelter and explore a novel env
poeciliid Brachyraphis episcopi originating from sites u
that run into the Panama Canal. The relation betwe
positive, with larger ?sh taking longer to emerge. This
sites, being signi?cant in the upstream populations o
ANIMAL BEHAVIOUR
doi:10.1016/j.anbinspect approaching predators (Godin & Davis 1995). Bold
?sh may also escalate ?ghts with conspeci?cs in the
presence of a predator (Brick & Jakobsson 2002).
Correspondence: C. Brown, Institute of Evolutionary Biology, School
of Biological Sciences, Kings? Buildings, University of Edinburgh,
Edinburgh EH9 3JT, U.K. (email: culum.brown@ed.ac.uk).
132
0003?3472/04/$30.00/0 
 2004 The Association for the Sness in eight populations
chyraphis episcopi
ORIA A. BRAITHWAITE
iological Sciences, University of Edinburgh
cceptance 18 September 2003;
line 29 September 2004; MS. number: 7792)
may be explained in part by demographics and long-
xamined the relation between boldness (taken as the
ironment) and body size in eight populations of the
pstream and downstream of waterfalls in four rivers
n body size and time to emerge from a shelter was
relation differed between downstream and upstream
nly. These results are best explained by a metabolic
2004, 68, 1325?1329
hav.2004.04.004for a similar study involving sticklebacks, Gasterosteus
aculeatus, that were exposed to a model king?sher (Krause
et al. 1998) suggesting that the type of threat may be
important in determining emergence latencies. In two
further studies of the killi?sh Rivulus hartii, Fraser et al.
(2001) and Wilson et al. (1993) both found no relation
between boldness and body size. Despite these con?icting
5
tudy of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
1326results, there are compelling reasons to suspect that body
size should be correlated with boldness. First, small ?sh
are far more vulnerable to predation (reviewed in Sogard
1997) and therefore should be more wary in novel
situations and emerge from shelter later than large ?sh
(predation hypothesis). Second, juveniles have lower body
fat reserves, faster metabolic rates and higher drag coef-
?cients than adults and are therefore compelled to emerge
from shelters and begin foraging sooner than larger ?sh
(metabolic hypothesis; Wooton 1994; Krause et al. 1998;
Skalski & Gilliam 2002). These two hypotheses result in
con?icting predictions with respect to the relation be-
tween body size and boldness. The ?rst predicts a positive
relation whereas the latter predicts a negative one. Thus,
the degree of boldness displayed by an individual is likely
to change with ontogeny, re?ecting the perceived risk to
the individual and its metabolic status.
Previous studies have considered variation in the time
to emerge from a shelter in a single population of ?sh. The
decision to emerge, however, should be in?uenced by
a number of habitat variables, including predation pres-
sure, foraging competition or food availability (Godin
1997; Sih 1997; Dowling & Godin 2002). We investigated
the tendency of eight populations of the poeciliid Brachyr-
aphis episcopi, originating from sites above and below
waterfalls, to emerge from a shelter and explore a novel
environment (taken as an indicator of an individual?s
boldness, see Sneddon 2003). We predicted that ?sh from
below falls (high-predation sites) would be less bold on
average than ?sh from above falls (low-predation sites),
and therefore emerge from shelter later. Furthermore, we
predicted the relation between body size and boldness
would differ between upstream and downstream sites
owing to different selection pressures operating in these
regions.
METHODS
We collected eight populations of ?sh from four streams
along the Pipeline Road near Gamboa, Panama (Table 1)
Table 1. The relative percentage of fish making up the fauna at
each site
River
Brachyraphis
episcopi
Predatory
fish
Other
fish
Rio Agua Salud
Above falls 99 0 1
Below falls 55 45 0
Rio Macho
Above falls 79 0 21
Below falls 37 54 9
Rio Limbo
Above falls 86 2.5 11.5
Below falls 5 80 15
Quebrada Juan Grande
Above falls 99 1 0
Below falls 16 78 6
ANIMAL BEHAVIOUR, 68, 6Predators include characins, cichlids and tigerfish. Other fish largely
comprised catfish, Rivulus and other poeciliids.under a permit issued by the National Authority for the
Environment (ANAM). Since B. episcopi does not inhabit
the Panama Canal and the rivers run directly into the
canal, each represents an independent sample. In each
river, we collected one population above a waterfall that
represented a signi?cant barrier to upstream ?sh move-
ment. The distance between upstream and downstream
sites was kept to a minimum to minimize genetic
differentiation. In all cases, the distance between sites
was less than 500 m. Above the falls, the ?sh fauna is
highly depauperate and is dominated by B. episcopi and
the killi?sh R. brunneus. Below the falls, the fauna is
primarily made up of members of the characin family as
well as cichlids (Geophagus and Aequidens spp.), tiger?sh
(Haplies spp.) and other potential predators. Nevertheless,
B. episcopi is present below the falls but in much lower
abundance (Table 1). In many respects, the system and
species involved mimic that of the well-studied Trinida-
dian guppy (Reznick & Endler 1981).
Brachyraphis episcopi is a small (to 5 cm) poeciliid that is
an opportunistic omnivore feeding primarily on insects
and fruit. Its general behavioural ecology is not well
documented but it is clear that large females are territorial
whereas smaller females tend to form loose shoals. Males
roam as individuals and are highly aggressive towards one
another. When threatened, all ?sh tend to hide rather
than form schools and preliminary investigation suggests
that there is no difference in schooling responses between
upstream and downstream locations (unpublished data).
Jennions & Telford (2002) reported that reduced predation
pressure on ?sh from upstream locations has a signi?cant
?ow-on effect on their life history strategies. Upstream
(low-predation) ?sh reach a much larger size and tend to
mature later than ?sh from downstream (high-predation)
sites. These data are similar to those for other Brachyraphis
species (Reznick et al. 1993) and poeciliids in general.
Approximately 40 B. episcopi were collected from each
site using seine nets, brought into a laboratory and
transferred to eight housing aquaria (58! 54 cm).
Males and females were housed together. Each housing
aquarium contained gravel obtained from the appropriate
rivers and a standard corner ?lter. Water was maintained
at 15 cm depth. Water temperature was controlled by
ambient air temperature but it remained at approximately
25 
C. Lighting consisted of twin overhead ?uorescent
tubes on a 12:12 h light:dark cycle to mimic natural light
patterns for the time of year (late July). The ?sh were fed
standard ?ake food twice a day (morning and evening)
except on the day of testing when they were fed after the
experiment.
Fish were tested approximately 9 days after capture. A
single female was chosen at random from a population
and placed into the test arena. Only females longer than
2 cm were tested. Fish smaller than 2 cm are immature
and therefore cannot be sexed and males are relatively
rare. The test arena consisted of a tank of identical
dimensions to the housing tanks, but with shallower
water (8 cm) and no gravel. In one corner, a blue bucket
(20 cm diameter) was positioned upside down to form
a shelter. A small hole (3! 3 cm) cut in the bucket
allowed the ?sh to exit the shelter and explore the rest
327of the arena. Within the bucket, a clear plastic cylinder of
15 cm diameter allowed the experimenter to constrain the
test ?sh during a 5-min settling period. After this period,
a remote pulley lifted the cylinder clear of the water and
the ?sh was free to emerge from the shelter and explore
the test arena. The observer was able to watch the
behaviour of the test subject remotely on a computer
monitor connected by ?rewire (Belkin FireWire cable IEEE
1394) to a digital camera mounted over the test arena. We
recorded the time until the ?sh?s head emerged from the
shelter. Fish that failed to emerge within 10min of the
bottle being lifted were given the maximum score of 600 s
(only 9 of 160 ?sh tested failed to emerge: ?ve from
regions of high predation and four from low-predation
areas). After emerging from the shelter the ?sh were
allowed to swim freely around the test arena and explore
it before being netted and measured (standard
lengthG 1 mm). No ?sh was tested more than once. We
tested 20 ?sh from each of the eight locations. After the
experiment, the ?sh were transported to the University of
Edinburgh where they were incorporated into a breeding
programme for ongoing studies.
Data on the time to emerge from shelter were log
transformed and examined for a relation with standard
length using regression analysis. A test of parallelism as
a post hoc analysis under an analysis of covariance
(ANCOVA) was then used to determine whether the
gradient of the two regression slopes differed signi?cantly.
An ANCOVA was further used to examine the effects of
location (upstream and downstream) and stream identity
as independent ?xed variables, with log (time to emerge)
as the dependent variable and standard length as the
covariate.
RESULTS
Regression analysis revealed a signi?cant positive relation
between standard length and the time the ?sh took to
emerge from the shelter (F1,158Z 15.09, PZ 0.0002).
Smaller ?sh emerged from the shelter sooner than larger
?sh, which supports the metabolic hypothesis. We further
examined the relation by investigating the data from
high-predation and low-predation sites independently.
This analysis revealed that the relation between standard
length and time to emerge from the shelter for ?sh from
the high-predation sites was positive but not signi?cant
(F1,78Z 1.70, PZ 0.20). In contrast, the ?sh from the
low-predation sites showed a highly signi?cant positive
relation (F1,78Z 17.41, P! 0.0001; Fig. 1). A post hoc test
of parallelism suggested that the gradients of these two
regression lines were not signi?cantly different
(F1,150Z 0.54, PZ 0.46).
The ANCOVA revealed a signi?cant effect of stream
(F3,151Z 4.91, PZ 0.003) but no difference between
upstream and downstream sites (F1,151Z 2.08, PZ 0.15).
There was a signi?cant interaction between the two factors
(F3,151Z 3.92, PZ 0.01). Post hoc analysis (Fisher?s Pro-
tected Least Signi?cant Difference, PLSD) revealed that
?sh from the Rio Agua Salud (AS) emerged from the refuge
signi?cantly faster than ?sh from all other rivers. The timeto emerge from the refuge for ?sh from the Rio Limbo
(RL), Quebrada Juan Grande (QJG) and Rio Macho (RM)
were not signi?cantly different from one another. Once
standard length had been controlled for, ?sh from the RM
upstream site emerged more rapidly than RM downstream
?sh, whereas the reverse was true for QJG (Fig. 2). Fish
from upstream areas of both the AS and the RL tended to
emerge earlier than downstream ?sh but these differences
were not signi?cant (Fig. 2).
The standard length of test ?sh differed between the
four streams (F3,152Z 4.27, P! 0.01). There was a mar-
ginal difference between upstream and downstream sites
(F1,152Z 3.05, PZ 0.08). Fish from high-predation sites
tended to be smaller than ?sh from low-predation
sites. Post hoc analysis (Fisher?s PLSD) revealed that ?sh
from RM were signi?cantly larger than ?sh from all other
rivers (versus QJG, RL and AS: PZ 0.013, 0.001 and 0.006,
0
0.5
1
1.5
2
2.5
3
1.5 2 2.5 3 3.5 4 4.5 5
Standard length (cm)
Lo
g 
(t
im
e 
to
 e
m
er
ge
)
Figure 1. The relation between standard length and time to emerge
from the shelter for fish from upstream (,; - - -; low-predation) and
downstream (A; d; high-predation) sites. Regressions: upstream:
yZ 0.5633x  0.2402; downstream: yZ 0.2194xC 0.9789.
2.5
2
1
1.5
0.5
0
AS RL RM QJG
Population
Lo
g 
(t
im
e 
to
 e
m
er
ge
)
Figure 2. The meanG SE time to emerge from the shelter adjusted
for standard length for fish from downstream (,; high-predation)
and upstream (&; low-predation) sites in each of the four streams.
BROWN & BRAITHWAITE: BOLDNESS AND BODY SIZE 1AS: Rio Aqua Salud; RL: Rio Limbo; RM: Rio Macho; QJG: Quebrada
Juan Grande.
1328respectively). There were no signi?cant differences in the
standard length of the ?sh at any of the other sites.
DISCUSSION
The boldness scores of B. episcopi (as de?ned as the time to
emerge from a shelter) were overwhelmingly affected by
their body size. As predicted by the metabolic hypothesis,
whereby juvenile ?sh have a higher metabolic rate and
fewer body fat reserves, smaller ?sh emerged from shelter
sooner than larger ?sh. It is well established that meta-
bolic rate increases with the natural log of body mass, but
metabolic rate per g of body weight decreases with
increasing body size. In teleost ?sh, the relation between
the natural log of resting oxygen consumption and the
natural log of body mass gives a scaling exponent of 0.79
(Clarke & Johnston 1999). As in all other animals,
metabolic rate per g of body mass and body size has
a negative exponential allometric relation in ?sh. It is
likely, therefore, that small ?sh are compelled to emerge
sooner than larger ?sh regardless of the level of predation
pressure at each location owing to energy constraints.
Although the effect of body size on boldness did not differ
signi?cantly between upstream and downstream sites, this
relation was far stronger in the upstream localities where
there are fewer predators. In the downstream locations
there may be a trade-off between the direct effects of
predation pressure and the metabolic requirements of
juveniles, thus ?attening the regression by forcing juve-
niles to stay in the refuge for longer (Fig. 1).
Alternatively, since age is correlated with standard
length, older ?sh may be more wary in general, having
learnt from experience that early emergence increases the
threat of predation. An alternative hypothesis is that all
bold ?sh have long since been killed by predators and only
the more wary ?sh survive to old age. Although these
hypotheses seem unlikely, it should be borne in mind that
there is little evidence to suggest that fooddeprivation leads
to earlier emergence times in ?sh (e.g. Dowling & Godin
2002). Nevertheless, activity levels do increase early on
during food deprivation periods in cyprinids (Wieser et al.
1992) suggesting a shift to increased foraging behaviour.
Once body length was controlled for, differences in
emergence times between upstream and downstream sites
varied signi?cantly between streams. In one stream,
upstream ?sh emerged earlier than downstream ?sh,
whereas in another the reverse was true. Fauna surveys
conducted at each location suggest that the predation
pressure is probably greatest in the Rio Limbo and the
Quebrada Juan Grande downstream sites and in these
locations B. episcopi comprises only a small proportion of
the total ?sh diversity (Table 1). These differences do not
translate directly into the emergence behaviour of the ?sh
since there is no relation between the relative abundance
of predators and the time to emerge at each site. Perhaps
equally importantly, the Rio Limbo and in particular the
Rio Macho upstream sites have a signi?cant number of
heterospeci?c competitors in the form of the killi?sh
ANIMAL BEHAVIOUR, 68, 6R. brunneus. Upstream sites may be relatively poor sites
in terms of productivity (Reznick et al. 2001) and thereforecompetition may play an important role in shaping the
behaviour of the ?sh in these regions. We suspect that
foraging competition and food availability may be just as
important as predation pressure in determining the time
to emerge from shelter, but these factors are dif?cult to
quantify in the wild.
In the present study there may be other, as yet un-
explored, ecological variables such as varying degrees of
juvenile cannibalism at each site that may result in varying
levels of boldness. For instance, Tully & Huntingford
(1987) found that boldness was partially heritable in
sticklebacks but was also affected by the level of chasing
by fathers. Wilson et al. (1994) suggested that the natural
environment creates and maintains individual differences
in boldness, but prolonged periods of captivity may cause
behavioural convergence. It is unlikely that the brief
period of captivity altered the behaviour of the ?sh in this
instance. Nevertheless, there may be differences in the way
the ?sh respond to the handling stress and the manner in
which they settle in the laboratory. Testing the procedure
in the ?eld would be one way of overcoming some of these
unknown variables.
In conclusion, balancing predation risk with foraging
rate is a well-studied phenomenon in ?sh (reviewed in
Lima & Dill 1990). There are numerous models examining
whether juvenile ?sh choose habitats that maximize
growth rate or minimize predation risk or whether they
attempt to minimize the ratio of risk to growth (Dahlgren
& Eggleston 2000; Skalski & Gilliam 2002). Our data
suggest that small ?sh are driven to emerge from hiding
to maximize food intake and thus growth rate rather than
to reduce predation pressure. While there is some evi-
dence that the relation is in?uenced by environmental
variables such as foraging competition and predation
pressure, this still requires further investigation.
Acknowledgments
This research was funded by NERC grant number (NER/A/
S/01/00608). We thank the Smithsonian Tropical Research
Institute for their continued support and, in particular,
Biff Birmingham for his help in identifying ?sh species.
We also thank Michael Jennions for his aid in identifying
?eld sites and for comments on the manuscript. Finally,
thanks to Felicity Brown for all her help collecting ?sh and
other ?eld data.
References
Benus, R., Bohus, B., Koolhas, J. & van Oortmerssen, G. 1991.
Heritable variation for aggression as a reflection of individual
coping strategies. Experientia, 47, 1008?1019.
Brick, O. & Jakobsson, S. 2002. Individual variation in risk taking:
the effect of a predatory threat on fighting behavior in Nannacara
anomala. Behavioral Ecology, 13, 439?442.
Budaev, S. V. 1997a. Alternative styles in the European wrasse,
Symphodus ocellatus: boldness-related schooling tendency. Environ-
mental Biologyof Fishes,49, 71?78.doi:10.1023/A:1007380212475.
Budaev, S. V. 1997b. ?Personality? in the guppy (Poecilia reticulata):
a correlational study of exploratory behaviour and social tendency.
Journal of Comparative Psychology, 111, 399?411.
Clarke, A. & Johnston, N. M. 1999. Scaling of metabolic rate with
body mass and temperature in teleost fish. Journal of Animal
Ecology, 68, 893?905. doi:10.1046/j.1365-2656.1999.00337.x.
Coleman, K. & Wilson, D. S. 1998. Shyness and boldness in
pumpkinseed sunfish: individual differences are context-specific.
Animal Behaviour, 56, 927?936.
Dahlgren, C. P. & Eggleston, D. B. 2000. Ecological processes
underlying ontogenetic habitat shifts in a coral reef fish. Ecology,
81, 2227?2240.
Dowling, L. M. & Godin, J. G. J. 2002. Refuge use in a killifish:
influence of body size and nutritional state. Canadian Journal of
Zoology, 80, 782?788.
Fraser, D. F., Gilliam, J. F., Daley, M. J., Le, A. N. & Skalsk, G. T.
2001. Explaining leptokurtic movement distributions: intrapopu-
lation variation in boldness and exploration. American Naturalist,
158, 124?135.
Godin, J. G. J. 1997. Evading predators. In: Behavoural Ecology of
Teleost Fishes (Ed. by J. G. J. Godin), pp. 191?236. Oxford: Oxford
University Press.
Godin, J. G. J. & Davis, S. A. 1995. Who dares, benefits: predator
approach behavior in the guppy (Poecilia reticulata) deters
Reznick, D. & Endler, J. A. 1981. The impact of predation on life
history evolution in Trinidadian guppies (Poecilia reticulata).
Evolution, 36, 160?177.
Reznick, D., Meyer, A. & Frear, D. 1993. Life-history of Brachyraphis
rhabdophora (Pisces, Poeciliidae). Copeia, 1993, 103?111.
Reznick, D., Butler, M. J. & Rodd, H. 2001. Life-history evolution in
guppies. VII. The comparative ecology of high and low predation
environments. American Naturalist, 157, 126?140.
Sih, A. 1997. To hide or not to hide? Refuge use in a fluctuating
environment. Trends in Ecology and Evolution, 12, 375?376.
doi:10.1016/S0169-5347(97)87376-4.
Skalski, G. T. & Gilliam, J. F. 2002. Feeding under predation hazard:
testing models of adaptive behavior with stream fish. American
Naturalist, 160, 158?172.
Sneddon, L. U. 2003. The bold and the shy: individual differences
in rainbow trout. Journal of Fish Biology, 62, 971?975.
doi:10.1046/j.1095-8649.2003.00084.x.
Sogard, S. M. 1997. Size-selective mortality in the juvenile stage of
teleost fishes: a review. Bulletin of Marine Science, 60, 1129?1157.
Tully, J. J. & Huntingford, F. A. 1987. Parental care and the
development of adaptive variation in anti-predator responses in
BROWN & BRAITHWAITE: BOLDNESS AND BODY SIZE 1329predator pursuit. Proceedings of the Royal Society of London,
Series B, 259, 193?200.
Godin, J. G. J. & Dugatkin, L. A. 1996. Female mating preference
for bold males in the guppy, Poecilia reticulata. Procedings of the
National Academy of Sciences, U.S.A., 93, 10262?10267.
Jennions, M. D. & Telford, S. R. 2002. Life-history phenotypes
in populations of Brachyrhaphis episcopi (Poeciliidae) with
different predator communities. Oecologia, 132, 44?50.
doi:10.1007/s00442-002-0942-4.
Krause, J., Loader, S. P., McDermott, J. & Ruxton, G. D. 1998.
Refuge use by fish as a function of body-length-related metabolic
expenditure and predation risk. Proceedings of the Royal Society of
London, Series B, 265, 2373?2379.
Lima, S. L. & Dill, L. M. 1990. Behavioral decisions made under the
risk of predation: a review and prospectus. Canadian Journal of
Zoology, 68, 619?640.
Murphy, K. E. & Pitcher, T. J. 1997. Predator attack motivation
influences the inspection behaviour of European minnows. Journal
of Fish Biology, 50, 407?417. doi:10.1006/jfbi.1996.0301.sticklebacks. Animal Behaviour, 35, 1570?1572.
Wieser, W., Krummschnabel, G. & Ojwang-Okwor, J. P. 1992. The
energetics of starvation and growth after refeeding in juveniles of
three cyprinid species. Environmental Biology of Fishes, 33, 63?71.
Wilson, D. S. 1998. Adaptive individual differences within single
populations. Philisophical Transactions of the Royal Society of
London, Series B, 353, 199?205.
Wilson, D. S., Coleman, K., Clark, A. B. & Biederman, L. 1993. Shy
bold continuum in pumpkinseed sunfish (Lepomis gibbosus): an
ecological study of a psychological trait. Journal Of Comparative
Psychology, 107, 250?260.
Wilson, D. S., Clark, A. B., Coleman, K. & Dearstyne, T. 1994.
Shyness and boldness in humans and other animals.
Trends in Ecology and Evolution, 9, 442?446. doi:10.1016/
0169-5347(94)90134-1.
Wooton, R. J. 1994. Energy allocation in the threespine stickleback.
In: The Evolutionary Biology of the Threespine Stickleback (Ed. by
M. A. Bell & S. A. Foster), pp. 116?143. Oxford: Oxford University
Press.