social context modulates predator evasion strategy in...

RESEARCH PAPER

Social Context Modulates Predator Evasion Strategy In GuppiesEva K. Fischer*, Adina J. Schwartz†, Kim L. Hoke* & Daphne Soares‡

* Department of Biology, Colorado State University, Fort Collins, CO, USA

† Biology Department, University of Maryland, College Park, MD, USA

‡ Department of Biological Science, New Jersey Institute of Technology, University Heights, Newark, NJ, USA

Correspondence

Eva K. Fischer, Department of Biology,

Colorado State University, 1878 Campus

Delivery, Fort Collins, CO 80523, USA.

E-mail: [email protected]

Received: August 20, 2014

Initial acceptance: November 1, 2014

Final acceptance: November 8, 2014

(S. Foster)

doi: 10.1111/eth.12345

Keywords: Poecilia reticulata, behavioral

strategy, predation, social context

Abstract

Social context is a powerful mediator of behavioral decisions across animal

taxa, as the presence of conspecifics comes with both costs and benefits.

In risky situations, the safety conferred by the presence of conspecifics can

outweigh the costs of competition for resources. How the costs and bene-

fits of grouping influence decisions among alternative antipredator behav-

iors remains largely unexplored. We took advantage of the Trinidadian

guppy (Poecilia reticulata) to examine the influence of social context on

alternative behavioral responses to threats. We compared the frequency

of active (startle) versus passive (freeze) responses to sudden acoustic

stimuli in the presence and absence of conspecifics. We found that fish

were relatively less likely to startle and more likely to freeze when in a

group than when alone, indicating that immediate social context modu-

lates predator evasion strategy in guppies. We suggest that these social

context-dependent changes reflect trade-offs between survival and energy

expenditure. To our knowledge, an effect of immediate social environ-

ment on startle probability has not been previously demonstrated in a tel-

eost.

Introduction

Social context is a powerful mediator of behavioral

decisions across animal taxa, as the presence of con-

specifics influences the relative value of diverse

behaviors. The mere presence of conspecifics alters

behavior when conspecifics represent competition for

resources, such as food (e.g., Pitcher & Parish 1993)

and mates (e.g., Farr & Herrnkind 1974; Doutrelant

2001; Sadowski et al. 2002). But competitors can turn

into allies under certain environmental conditions. In

risky situations, the safety afforded by the presence of

conspecifics can outweigh costs, such as intraspecific

competition or increased conspicuousness to preda-

tors (Abrahams & Dill 1989; Lima & Dill 1990; Krause

& Ruxton 2002). Being in a group confers safety by a

number of mechanisms, including risk dilution, pred-

ator confusion, and joint predator inspection (e.g.,

Magurran & Pitcher 1987; Abrahams & Dill 1989;

Lima & Dill 1990; Milinski 1993). In addition, being in

a group conserves energy by increasing vigilance at

the group level, such that individuals can be less

vigilant and spend more time engaging in profitable

behaviors such as foraging and courting (reviewed in

Krause & Ruxton 2002). Despite many examples of

social context altering behavior and of the antipreda-

tor benefits of grouping, few studies have examined

how social context modulates predator evasion tac-

tics.

Animals have various tactics for predator evasion,

ranging from active (fleeing) to passive (freezing, hid-

ing) responses. The relative efficacy of these alterna-

tive tactics may depend on social context. Nearly all

animal taxa display a stereotyped startle response fol-

lowing sudden, unexpected stimuli (Eaton & Hackett

1984; Eaton et al. 1991; Koch 1999). Active escape

responses are effective but costly because they (1) typ-

ically involve fast, sudden movements that are ener-

getically expensive and inefficient (e.g., Webb 1984;

Jayne & Lauder 1993; Hughes & Kelly 1996), (2) carry

the indirect cost of interrupting and/or interfering

with profitable behaviors (e.g., Ydenberg & Dill 1986;

Milinski 1993), and (3) may be undesirable if preda-

tors have not yet detected prey animals, as movement

Ethology 121 (2014) 1–8 © 2014 Blackwell Verlag GmbH 1

Ethology

ethologyinternational journal of behavioural biology

—in particular fast movement—can make prey more

conspicuous (e.g. Krause & Godin 1995). Although

passive tactics, such as freezing and/or hiding, may

also interrupt profitable behaviors they have the

advantage of allowing animals to remain undetected,

as well as conserve energy. However, passive tactics

do not carry the animal away from the threat and are

thus less effective if prey animals are detected (Stau-

dinger et al. 2011). In sum, there are trade-offs

between active and passive predator avoidance tactics,

and we expected behavioral decisions among tactics

to balance these trade-offs in a context-dependent

manner.

Active escape responses are particularly amenable

for study in fish, as they are characterized by highly

stereotyped c-startle escape responses present in nearly

all fish species (Roberts 1992). This behavior typically

involves two stages: In the first stage, the body bends

away from the stimulus, creating the arched posture

that gives the behavior its name. In the second phase,

the body straightens, creating a propulsive force that

moves the animal away from the threatening stimulus

(Eaton et al. 1977, 2001). Startle responses are medi-

ated primarily by two large neurons in the brainstem,

the Mauthner cells. A startle response occurs only

when an action potential is fired in one of the Mauth-

ner cells (Korn & Faber 2005). The Mauthner cells

receive inputs from the auditory, visual, and lateral

line systems, and startle responses can be elicited by a

single stimulus type (Eaton et al. 1977, 1991; Zottoli

1977), or by several, simultaneous stimulus types, as is

probably the case during actual predator encounters

(Eaton & Hackett 1984). Visual, mechanical, and audi-

tory stimuli are commonly used to elicit startle escape

responses. Auditory stimuli are advantageous as they

are easily controlled and manipulated (Preuss 2006;

Kastelein et al. 2008; Neumeister et al. 2008) and

because auditory inputs to the Mauthner cells come

directly from the 8th nerve (auditory nerve) (e.g.,

Eaton & Popper 1995; Zottoli & Faber 2000), making

this stimulus modality independent of the acoustic

details of the stimulus (Preuss 2006; Kastelein et al.

2008; Neumeister et al. 2008). Although the threshold

for startle initiation can be modulated, once a startle

response is initiated the behavior is almost always

completed (Eaton & Hackett 1984). While the startle

response is highly stereotyped overall, inter- and intra-

specific variation exists in various response parame-

ters, including response latency, acceleration,

maximum velocity, body angle, and response distance

(Eaton et al. 1977; Ghalambor & Reznick 2004; Neu-

meister et al. 2010). Grouping behavior has been

shown to influence response distance and response

latency in a number of fish species (e.g., Eaton et al.

1977; Seghers 1981; Abrahams 1995; Domenici &

Batty 1997; Semeniuk & Dill 2004), but to our knowl-

edge, the effects of immediate social context on startle

probability have not been explored.

We examined the influence of immediate social

context on alternative predator evasion tactics in Trin-

idadian guppies (Poecilia reticulata). Trinidadian gup-

pies are small, freshwater teleosts, native to the island

nation of Trinidad and Tobago as well as the adjacent

South American mainland. The Trinidadian guppy is a

well-established model system in ecology, evolution,

and ethology due to its ability to rapidly adapt to envi-

ronmental challenges. Guppies are known to alter

courtship (e.g. Farr 1975; Rodd & Sokolowski 1995),

foraging (e.g. Zandon�a et al. 2011; Elvidge et al.

2014), and antipredator (e.g., Seghers 1974; Breden

et al. 1987; Botham et al. 2008; Torres-Dowdall et al.

2012) behaviors in an adaptive manner in response to

changing predation pressure on evolutionary and

developmental timescales, as well as in response to

immediate changes in predation threat (e.g., Seghers

1974; Abrahams & Dill 1989; Houde 1997; Huizinga

et al. 2009). In response to a threat, guppies may

mount an active escape response, freeze/hide, or

increase shoaling (general affiliative) behavior

(reviewed in Magurran 2005), but it is not clear what

cues guppies use to decide among strategies and

whether social context plays a role.

In this study, we examined the influence of immedi-

ate social context on responses to acoustic stimuli in

guppies. We compared the probability of startle versus

freeze responses of individual guppies in the presence

and absence of conspecifics. To control for potential

confounding effects of behavioral differences between

social contexts, we examined the effect of an individu-

als’ behavior prior to stimulus presentation and the

behavior of other group members following stimulus

presentation on response strategy. In addition, we

examined responses over a range of stimulus intensi-

ties to determine whether changes in response proba-

bility across social contexts were consistent among

amplitudes. Given that the presence of conspecifics has

consequences for both energetics and conspicuous-

ness, we predicted that social context could shift the

balance between active versus passive evasion tactics.

Methods

Animals

The adult male fish in this study were bred in our fish

facility at Colorado State University before transportation

Ethology 121 (2014) 1–8 © 2014 Blackwell Verlag GmbH2

Social Context Modulates Predator Evasion Strategy E. K. Fischer, A. J. Schwartz, K. L. Hoke & D. Soares

to the Marine Biological Laboratory in Woods Hole,

Massachusetts, where we conducted all behavioral

experiments. We established laboratory populations

at Colorado State University from wild-caught, gravid

females collected from the Guanapo high-predation

locality in the Northern Range Mountains of Trinidad

in 2009. To establish family lines, we separated first-

generation laboratory-born fish by sex and kept them

in isolated tanks under identical conditions. Once

mature, we uniquely crossed first-generation fish to

generate the second generation of laboratory-born

fish for this study. We housed adult fish in single-sex,

5 l group tanks in a recirculating water system con-

taining conditioned water with a pH, hardness, tem-

perature, and chemistry similar to natural streams.

We kept fish on a 12:12 h light cycle and fed them a

measured diet twice daily, once in the morning (Tetr-

aminTM tropical fish flake paste) and once in the after-

noon (hatched Artemia cysts).

We transported fish used in the experiment to the

Woods Hole Marine Biological Laboratory, where we

housed fish in 3.2 l group tanks containing condi-

tioned water. We kept fish on a 12:12 h light cycle

and fed them twice daily (TetraminTM tropical fish

flake paste). We photographed all fish prior to the

experiment and used unique male color patterns to

track individuals. All male fish were mature at the

time of the experiment. All animal husbandry proto-

cols were approved by the Colorado State University

Animal Care and Use Committee (protocol #12-

3818A), and all behavioral methods were approved

by the Marine Biological Laboratory Animal Care and

Use Committee (protocol # 11-44).

Behavior

Visual, mechanical, and acoustic stimuli are com-

monly used to elicit startle escape responses in gup-

pies and other fish (e.g. Ghalambor & Reznick 2004;

Preuss 2006; Dadda et al. 2010). Acoustic stimuli are

easily controlled and manipulated, and stimulus

intensity is directly correlated with Mauthner cell

excitability and startle response probability (Preuss

2006; Kastelein et al. 2008; Neumeister et al. 2008).

Moreover, auditory inputs to the Mauthner cells

come directly from the 8th nerve (e.g. Eaton & Popper

1995; Zottoli & Faber 2000), and thus these inputs are

not sensitive to detailed stimulus characteristics. We

chose acoustic stimuli for this study because this

allowed us to modify stimulus intensity and ensure

intermediate response probabilities, allowing us to

detect subtle shifts in response probability among

groups. Group differences cannot be accurately

quantified when stimulus intensity is so high that

response probability in all groups is close to one, or

conversely, when stimulus intensity is so low that

response probability in all groups is close to zero (e.g.

Neumeister et al. 2010).

All behavioral trials were conducted in a circular,

12-cm diameter arena filled with conditioned water.

The arena was lit from below, and we used a filter

paper screen to evenly diffuse the light and increase

visual contrast. We recorded behavioral responses

using a high-speed camera (X-PRI model, Del Imagin-

ing, Cheshire, CT, USA) positioned above the tank

and connected to a computer running AOS Imaging

studio light software (AOS technology, Baden Daett-

wil, Switzerland). We delivered acoustic stimuli from

a speaker (RadioShack, Fort Worth, TX, USA) posi-

tioned 5 cm from the side of the tank. Acoustic stim-

uli were made in Matlab (Mathworks, Natick, MA,

USA) and were 2 ms 200 Hz sine wave tones at one

of four amplitudes, representing increasing stimulus

intensity (0.2, 0.4, 0.6, 0.8 volts). We conducted pre-

trials at a range of stimulus amplitudes to ensure that

startle probabilities were intermediate at the ampli-

tudes we chose (>20% and <80%).

Individual fish were tested in 3–6 acoustic startle

trials, with one trial (either social or isolated condi-

tion) conducted every 1–2 d. Each trial consisted of a

train of 20 stimuli with a unique combination of

interstimulus intervals and stimulus amplitudes. We

generated four unique stimulus trains. In each stimu-

lus train, the number of stimuli at each amplitude was

equal (i.e., five stimuli at each of four amplitudes),

but we randomized amplitude order to control for

order effects. Similarly, we randomized interstimulus

intervals between 2 and 8 min to prevent the fish

from anticipating or habituating to stimulus presenta-

tion. Because individuals experienced a different stim-

ulus train during each trial, this allowed us to

minimize habituation to stimulus presentation. The

order in which fish experienced the different trial ver-

sions and social conditions was randomized to control

for order effects.

Fish were placed in the arena either individually

(isolate condition) or in a group of three (social condi-

tion) and allowed to acclimate for 1 h, after which the

trial began. Social groups were selected randomly and

altered among trials to control for potential group

level effects (i.e., changes in one individuals behavior

due to the behavior of other group members). We

identified fish prior to the trial using unique color

markings and tracked individuals visually during the

trial. We recorded behavioral responses at 500

frames/second and subsequently analyzed all videos


E. K. Fischer, A. J. Schwartz, K. L. Hoke & D. Soares Social Context Modulates Predator Evasion Strategy

using Proanalyst software (Xcitex, Woburn, MA,

USA). We scored responses as startle, freeze, or no

response. To be categorized as a startle, fish had to

exhibit the characteristic c-bend following stimulus

presentation. In addition, we recorded fish behavior

just prior to stimulus presentation as either moving or

stationary. We could not accurately distinguish freeze

responses from no response when fish were already

stationary prior to stimulus presentation, so trials in

which fish were stationary and did not startle were

scored as no response. Fish were typically moving

rather than stationary prior to stimulus presentation,

and the time stationary did not differ based on social

condition; fish were moving at stimulus presentation

in 77% of all social trials and 80% of all individual tri-

als. Context-dependent differences in the proportion

of freezing thus were not an artifact of differences in

prior behavior, as fish in each experimental condition

were equally likely to be stationary when the stimulus

sounded. Given the difficultly in detecting a freezing

response in stationary fish, we did not assess the influ-

ence of prior behavior on probability of freezing and

only report the effect of prior behavior on startle prob-

ability.

Statistical Analysis

We tested the effects of social condition and stimulus

amplitude on startle and freeze probabilities using

logistic regression analyses. We modeled startle and

freeze probabilities separately as binary responses

with a logit link function. Social condition, stimulus

amplitude, and their interaction were included in the

model as fixed effects. To test for the influence of

other group members’ behavior, we performed an

additional analysis of startle probability that included

the behavior of other group members (whether one

or more group members startled versus no group

members startled) as a predictor of startle probability

in the social condition. We included stimulus number

within trial as a covariate in all analyses to control for

changes in response probability over time. Behavior

prior to stimulus presentation (moving or stationary)

was included as a covariate for startle but not freeze

responses, as described above. We included fish iden-

tity as a random effect to control for individual differ-

ences in overall responsivity. All post hoc comparisons

were Tukey’s adjusted to control for multiple hypoth-

esis testing.

Results

Fish responded (startled or froze) on average 64.5%

of the time, with individual response probabilities

ranging from 53.3% to 77.7%. Independent of social

context, fish were more likely to startle (71.4% of all

responses) than freeze (28.6% of all responses). Over

the course of the trial, startle probability decreased by

an average of 18% (F19,1302 = 2.20, p = 0.0021) and

freeze probability increased by an average of 7%

(F19,1302 = 2.01, p = 0.0060). Although startle proba-

bility decreased, it remained on average high (>50%)

indicating that fish did not completely habituate to

stimulus presentation. We included stimulus number

in all statistical analyses to control for differences in

response probability over the course of the trial. We

also included fish identity as a random effect to con-

trol for differences in individual response probability

(covariance parameter estimates: startle:

0.7767 � 0.366, freeze: 0.5952 � 0.361). Raw counts

and percentages for behavioral responses by social

condition and stimulus amplitude are shown in

Table 1.

Social condition strongly influenced response type.

Fish were relatively more likely to startle

(F1,1302 = 54.88, p < 0.0001) and less likely to freeze

Table 1: Raw counts and percentages for behavioral responses by social condition are shown for each stimulus amplitude and averaged across stim-

ulus amplitudes

Behavioral response

Stimulus amplitude (volts)

0.2 0.4 0.6 0.8 Average

Iso Social Iso Social Iso Social Iso Social Iso Social

Startle 129

44.0%

43

28.7%

161

54.9%

56

37.3%

184

62.2%

57

38.0%

192

64.9%

53

35.3%

666

56.6%

209

34.8%

Freeze 73

24.9%

50

33.3%

54

18.4%

43

28.7%

39

13.3%

30

20.0%

41

13.9%

23

15.3%

207

17.6%

146

24.3%

Neither 91

31.1%

57

38.0%

78

26.6%

51

34.0%

71

24.1%

63

42.0%

63

21.3%

74

49.3%

303

25.8%

245

40.8%

Iso, isolate condition, social, social condition.



(F1,1303 = 19.96, p < 0.0001) when they were isolated

than when they were in a social group (Fig. 1). Startle

probability was not affected by the behavior of other

individuals in the social condition (F1,575 = 2.76,

p = 0.0969). In fact, fish were slightly less likely to

startle (45.51% of responses) than not startle

(54.49% of responses) when other group members

startled.

Startle probability increased with increasing stimu-

lus amplitude (Fig. 2; F3,1302 = 7.57, p < 0.0001),

while freeze probability decreased with increasing

stimulus amplitude (Fig. 2; F3,1303 = 5.86,

p = 0.0006), and this effect did not differ by social

condition (social condition *amplitude, startle:

F3,1302 = 1.97, p = 0.1163; freeze: F3,1302 = 0.13,

p = 0.9446). Post hoc tests revealed that startle proba-

bility increased from 0.2 to 0.4 amplitude and then

plateaued at an apparent maximum. Conversely,

freeze probability decreased from 0.2 to 0.6 amplitude

and then reached a minimum (Fig. 2). Also indepen-

dent of social condition, fish were more likely to star-

tle if they were moving, rather than stationary, prior

to stimulus onset (Fig. 3; F1,1302 = 69.12, p < 0.0001).

Discussion

Social context changed threat responsivity and modu-

lated predator evasion strategy in guppies. Guppies

were less likely to startle and more likely to freeze in

response to a sudden acoustic stimulus when they

were in a group as compared to when they were

alone. As expected, startle probability increased and

freezing probability decreased with increasing stimu-

lus amplitude, but this effect was independent of

social condition. Similarly, the influence of prior

behavior on startle probability was independent of

social condition, and fish in both social contexts were

more likely to startle when moving prior to the stimu-

lus. The behavior of other group members did not

significantly affect startle probability. To our knowl-

edge, an effect of immediate social environment on

startle probability has not been previously demon-

strated in a teleost. We suggest these patterns result

from trade-offs between predator evasion and energy

expenditure.

Grouping behavior is a common predator evasion

tactic that can also conserve energy. The presence of

conspecifics confers safety by creating a dilution

effect, such that the probability of attack is less for any

given individual, and by increasing vigilance at the

group level. Grouping is energetically favorable, as it

reduces the need for costly active escape responses

and enables individuals to spend more time engaging

in other profitable behaviors (reviewed in Lima & Dill

1990). We found that guppies were relatively less

likely to startle (i.e., mount an active, high-energy

escape response) when they were in a group than

0.0

0.2

0.4

0.6

0.8

1.0

Individual Social

Prop

ortio

n of

resp

onse

s

FreezeNo response

Startle

Fig. 1: Effect of social context on response probability. Stacked bars

represent proportion of stimulus presentations after which fish startled,

froze, or did neither in the individual as compared to the social condi-

tion. Response type differed significantly based on social context. Fish

were relatively more likely to startle and less likely to freeze when they

were isolated versus in a social group. Error bars representing standard

error are not included on the graph because they are too small to be

seen.

0.2

0.3

0.4

0.5

0.2 0.4 0.6 0.8Amplitude

Res

pons

e pr

obab

ility

StartleFreeze

Fig. 2: Effect of stimulus amplitude on response probability. Startle

probability increased with increasing stimulus amplitude, while freeze

probability decreased with increasing stimulus amplitude. We show

average values as this effect did not differ based on social condition.

Error bars represent standard error.



when isolated. A decrease in startle responses in con-

cert with a relative increase in freezing behavior in

the social condition could conserve energy and main-

tain shoal cohesion, both of which are beneficial if

grouping does indeed increase survival and decrease

energetic costs in the presence of predators.

One limitation of our experimental design is that

we cannot distinguish whether decreased startle

behavior in the social condition is a response to the

proximity of conspecifics or is a more generalized

response to any nearby object. Decreased startling and

increased freezing in the presence of conspecifics and/

or objects could reflect a sit-and-hide strategy in

response to a threat (Templeton 2004), which would

conserve energy and reduce detection probability, as

discussed above. Alternatively, decreased startle prob-

ability could serve to reduce the likelihood of collision

with neighboring fish and/or objects. Proximity to

other individuals and/or the side of the tank does not

appear to prevent startle responses in guppies (E. K.

Fischer, personal observation), but further work is

necessary to explicitly test this idea. Our data support

the hypothesis that guppies may be utilizing alterna-

tive behavioral strategies in the presence of conspecif-

ics, although we acknowledge that this behavioral

switch could apply more generally in any structurally

complex environment that provides shelter and

reduces fish conspicuousness.

While social context influenced response tactics,

stimulus intensity modulated response probability

independently of social condition. As stimulus inten-

sity increased, startle probability increased and freeze

probability decreased. Similar increases in startle

probability at higher stimulus intensities have been

observed in other teleosts (e.g., Neumeister et al.

2010), but the mechanistic basis for this behavioral

pattern remains unclear. Given that the intensity of

behavioral responses should scale with the level of

danger (Ydenberg & Dill 1986; Helfman 1989; Lima &

Bednekoff 1999), increased startle probability is often

considered indicative of an increase in the perceived

danger of a threat (reviewed in Domenici 2010).

Alternatively, increased startle probability could have

a purely physiological basis, for example, if increased

stimulus intensity increases sensory neuron input to

brainstem escape circuitry and Mauthner cell excit-

ability (Neumeister et al. 2008; Mirjany & Faber

2011). Whichever the mechanism by which stimulus

amplitude influences startle probability, we empha-

size that the reduction in startle probability associated

with the presence of conspecifics was consistent across

stimulus intensities, suggesting a global shift in startle

response threshold in the presence of conspecifics.

The influence of behavior prior to stimulus presen-

tation on response probability was also independent

of social condition. Fish that were moving prior to the

stimulus were more likely to startle in response to a

stimulus. Animals engaging in behaviors that distract

them and/or make them more conspicuous may be

more likely to mount an active escape response

because they are more vulnerable to attack (Lima &

Dill 1990; Milinski 1993). Future studies could explic-

itly test whether the increase in startle probability we

observe in moving fish does indeed have an adaptive

value, for example, by examining the relationship

between alternative behavioral tactics and survival

probability in staged predator encounters.

Conclusions

We found that immediate social context influenced

predator evasion strategy in guppies. The presence of

conspecifics shifted antipredator tactic from active

toward passive responses, demonstrating that imme-

diate social context can shift the balance between

alternative evasion tactics. Stimulus intensity and

behavior prior to stimulus presentation altered

behavioral responses independently of social context.

We suggest that these behavioral changes balance

0.0

0.2

0.4

0.6

0.8

1.0

Startle Neither

MovingStationary

Fig. 3: Influence of behavior prior to stimulus presentation on startle

probability. Stacked bars represent proportion of stimulus presenta-

tions when fish were moving versus stationary prior to the stimulus split

by response type. Fish were more likely to startle if they were moving

prior to stimulus presentation, independent of social condition. Fish that

did not respond (neither startled nor froze) in response to the stimulus

were equally likely to be moving or stationary prior to stimulus presen-

tation. We show average values, as this effect did not differ based on

social condition. Error bars representing standard error are not included

on the graph because they are too small to be seen.



trade-offs between predator evasion and energy

expenditure. Given that natural populations of gup-

pies adapted to different predation regimes differ in

shoaling propensity and startle probability, these

trade-offs may also influence the evolution of behav-

ioral differences among populations in the wild.

Acknowledgements

We thank A. Streets for help with data collection, A.

Scott-Johnston, and L. Rose for help with behavior

quantification, G. Haspel for assistance with equip-

ment construction, the members of the Guppy Lab for

fish care, and two anonymous reviewers for helpful

comments on a previous version of the manuscript.

We gratefully acknowledge support from the Marine

Biological Laboratory Associates Endowed Fellowship

and Colwin Endowed Summer Research Award (to

K.L.H. & D.S.) and NSF DEB-0846175 (to C.K. Gha-

lambor).

Literature Cited

Abrahams, M. V. 1995: The interaction between antipre-

dator behaviour and antipredator morphology: experi-

ments with fathead minnows and brook sticklebacks.

Can. J. Zool. 73, 2209—2215.

Abrahams, M. V. & Dill, L. M. 1989: A determination of

the energetic equivalence of the risk of predation. Ecol-

ogy 70, 999—1007.

Botham, M. S., Hayward, R. K., Morrell, L. J., Croft, D. P.,

Ward, J. R., Ramnarine, I. & Krause, J. 2008: Risk-sensi-

tive antipredator behavior in the Trinidadian guppy,

Poecilia reticulata. Ecology 89, 3174—3185.

Breden, F., Scott, M. & Michel, E. 1987: Genetic differenti-

ation for antipredator behavior in the Trinidad Guppy,

Poecilia reticulata. Anim. Behav. 35, 618—620.

Dadda, M., Koolhaas, W. H. & Domenici, P. 2010:

Behavioural asymmetry affects escape performance in a

teleost fish. Biol. Lett. 6, 414—417.

Domenici, P. 2010: Context-dependent variability in the

components of fish escape response: integrating loco-

motor performance and behavior. J. Exp. Zool. A Ecol.

Genet. Physiol. 313, 59—79.

Domenici, P. & Batty, R.S. 1997: Escape behaviour of soli-

tary herring (Clupea harengus) and comparisons with

schooling individuals. Mar. Biol. 128, 29—38. Springer.

Doutrelant, C. 2001: The effect of an audience on intra-

sexual communication in male Siamese fighting fish,

Betta splendens. Behav. Ecol. 12, 283—286.

Eaton, R. C. & Hackett, J. T. 1984: The Role of the Mauth-

ner Cell in Fast-Starts Involving Escape in Teleost

Fishes. In: Neural mechanisms of startle behavior.

(Eaton, R. C., ed). Plenum Press, New York, pp. 213—

266.

Eaton, R. C. & Popper, A. N. 1995: The Octavolateralis Sys-

tem and Mauthner cell - Interactions and Questions.

Brain Behav. Evol. 46, 124—130.

Eaton, R. C., Bombardieri, R. A. & Meyer, D. L. 1977:

Mauthner-initiated startle in teleost fish. J. Exp. Biol.

66, 65—81.

Eaton, R. C., Didomenico, R. & Nissanov, J. 1991: Role of

the Mauthner cell in sensorimotor integration by the

brain stem escape network. Brain Behav. Evol. 37, 272

—285.

Eaton, R. C., Lee, R. & Foreman, M. 2001: The Mauthner

cell and other identified neurons of the brainstem

escape network of fish. Prog. Neurobiol. 63, 467—485.

Elvidge, C.K., Ramnarine, I. & Brown, G.E. 2014: Com-

pensatory foraging in Trinidadian guppies: effects of

acute and chronic predation threats. Curr. Zool.

(in press).

Farr, J. 1975: The role of predation in the evolution of

social behavior of natural populations of the guppy, Poe-

cilia reticulata. Evolution 29, 151—158.

Farr, J. A. III & Herrnkind, W. F. 1974: A quantitative

analysis of social interaction of the guppy, Poecilia reticu-

lata, as a function of population density. Anim. Behav.

22, 582—591.

Ghalambor, C. K. & Reznick, D. N. 2004: Constraints on

adaptive evolution: the functional trade-off between

reproduction and fast-start swimming performance in

the Trinidadian guppy (Poecilia reticulata). Am. Nat. 164,

38—50.

Helfman, G. S. 1989: Threat-sensitive predator avoidance

in damselfish-trumpetfish interactions. Behav. Ecol. So-

ciobiol. 24, 47—58.

Houde, A. E. 1997: Sex, Color, and Mate Choice in Gup-

pies. Princeton University Press, Princeton.

Hughes, N. F. & Kelly, L. H. 1996: A hydrodynamic model

for estimating the energetic cost of swimming maneu-

vers from a description of their geometry and dynamics.

Can. J. Fish Aquat. Sci. 53, 2484—2493.

Huizinga, M., Ghalambor, C. K. & Reznick, D. N. 2009:

The genetic and environmental basis of adaptive differ-

ences in shoaling behaviour among populations of

Trinidadian guppies, Poecilia reticulata. J. Evol. Biol. 22,

1860—1866.

Jayne, B. C. & Lauder, G. V. 1993: Red and white muscle

activity and kinematics of the escape response of the

bluegill sunfish during swimming. J. Comp. Physiol. A.

173, 495—508.

Kastelein, R. A., Heul, S. V. D., Verboom, W. C., Jennings,

N., Veen, J. V. D. & de Haan, D. 2008: Startle response

of captive North Sea fish species to underwater tones

between 0.1 and 64 kHz. Mar. Environ. Res. 65, 369—

377.



Koch, M. 1999: The neurobiology of startle. Prog. Neurobi-

ol. 59, 107—128.

Korn, H. & Faber, D. S. 2005: The Mauthner cell half a

century later: a neurobiological model for decision-

making? Neuron 47, 13—28.

Krause, J. & Godin, J.-G. J. 1995: Predator preferences for

attacking particular prey group sizes: consequences for

predator hunting success and prey predation risk. Anim.

Behav. 50, 465—473.

Krause, J. & Ruxton, G. D. 2002: Living in Groups. Oxford

University Press, New York.

Lima, S. L. & Bednekoff, P. A. 1999: Temporal variation in

danger drives antipredator behavior: the predation risk

allocation hypothesis. Am. Nat. 153, 649—659.

Lima, S. L. & Dill, L. M. 1990: Behavioral decisions made

under the risk of predation: a review and prospectus.

Can. J. Zool. 68, 619—640.

Magurran, A. E. 2005: Evolutionary Ecology of the Trin-

idadian guppy. Oxford University Press, USA.

Magurran, A. & Pitcher, T. 1987: Provenance, shoal size

and the sociobiology of predator-evasion behaviour in

minnow shoals. Proc. Roy. Soc. B 229, 439—465.

Milinski, M. 1993: Predation risk and feeding behaviour.

In: Behavior of teleost fishes. (Pitcher, T. J., ed). Chap-

man & Hall, London, pp. 285—306.

Mirjany, M. & Faber, D. S. 2011: Characteristics of the

anterior lateral line nerve input to the Mauthner cell.

J. Exp. Biol. 214, 3368—3377.

Neumeister, H., Szabo, T. M. & Preuss, T. 2008: Behavioral

and physiological characterization of sensorimotor gat-

ing in the goldfish startle response. J. Neurophysiol. 99,

1493—1502.

Neumeister, H., Whitaker, K. W., Hofmann, H. A. & Pre-

uss, T. 2010: Social and ecological regulation of a deci-

sion-making circuit. J. Neurophysiol. 104, 3180—3188.

Pitcher, T. J. & Parish, J. K. 1993: Functions of shoaling

behaviour in teleosts. In: Behavior of teleost fishes.

(Pitcher, T. J., ed). Chapman & Hall, London, pp. 363—

412.

Preuss, T. 2006: Neural representation of object approach

in a decision-making motor circuit. J. Neurosci. 26,

3454—3464.

Roberts, B. 1992: Neural mechanisms underlying escape

behaviour in fishes. Rev. Fish Biol. Fisheries 2, 243—

266.

Rodd, F. H. & Sokolowski, M. B. 1995: Complex origins of

variation in the sexual behavior of male Trinidadian

guppies, Poecilia reticulata - interactions between social-

environment, heredity, body-size and age. Anim.

Behav. 49, 1139—1159.

Sadowski, J. A., Grace, J. L. & Moore, A. J. 2002: Complex

courtship behavior in the striped ground cricket, Allon-

emobius socius (Orthoptera: Gryllidae): does social envi-

ronment affect male and female behavior? J. Insect

Behav. 15, 69—84.

Seghers, B. 1974: Schooling behavior in the guppy (Poecilia

reticulata): an evolutionary response to predation. Evo-

lution 28, 486—489.

Seghers, B. H. 1981: Facultative schooling behavior in the

spottail shiner (Notropis hudsonius): possible costs and

benefits. Environ. Biol. Fishes 6, 21—24.

Semeniuk, C. A. D. & Dill, L. M. 2004: Cost/benefit analy-

sis of group and solitary resting in the cowtail stingray,

Pastinachus sephen. Behav. Ecol. 16, 417—426.

Staudinger, M. D., Hanlon, R. T. & Juanes, F. 2011: Pri-

mary and secondary defences of squid to cruising and

ambush fish predators: variable tactics and their survival

value. Anim. Behav. 81, 585—594.

Templeton, C. N. 2004: Multiple selection pressures influ-

ence Trinidadian guppy (Poecilia reticulata) antipredator

behavior. Behav. Ecol. 15, 673—678.

Torres-Dowdall, J., Handelsman, C. A., Reznick, D. N. &

Ghalambor, C. K. 2012: Local adaptation and the evolu-

tion of phenotypic plasticity in the Trinidadian guppy

(Poecilia reticulata). Evolution 66, 3432—3443.

Webb, P. W. 1984: Body form, locomotion and foraging in

aquatic vertebrates. Amer. Zool. 24, 107—120.

Ydenberg, R. C. & Dill, L. M. 1986: The economics of flee-

ing from predators. Adv. Stud. Behav. 36, 229—249.

Zandon�a, E., Auer, S. K., Kilham, S. S., Howard, J. L.,

L�opez-Sepulcre, A., O’Connor, M. P., et al. 2011: Diet

quality and prey selectivity correlate with life histories

and predation regime in Trinidadian guppies. Funct.

Ecol. 25, 964—973.

Zottoli, S. J. 1977: Correlation of the startle reflex and

Mauthner cell auditory responses in unrestrained gold-

fish. J. Exp. Biol. 66, 243—254.

Zottoli, S. J. & Faber, D. S. 2000: The Mauthner cell: what

has it taught us? Neuroscientist 6, 26—38.



social context modulates predator evasion strategy in...

Documents