co-evolution of offspring antipredator competence and parental

14
ADVANCES IN CICHLID RESEARCH Co-evolution of offspring antipredator competence and parental brood defense in convict cichlids Brian D. Wisenden Anthony D. Stumbo Patrick A. Self Jennifer L. Snekser Daniel C. McEwen Patricia A. Wisenden Miles H. A. Keenleyside Murray Itzkowitz Ellen Brisch Received: 23 December 2013 / Accepted: 19 May 2014 / Published online: 14 July 2014 Ó Springer International Publishing Switzerland 2014 Abstract Convict cichlids are Neotropical freshwa- ter fish with biparental brood defense of their free- swimming young. In this study, we correlate ontoge- netic changes in swimming performance measured by maximum velocity and burst speeds with skeletal ossification of larvae and consider how these data provide insights into the proximate mechanisms for the evolution of parental care. In lab tests, swimming velocity and acceleration of the young increased nonlinearly with body length with a rapid improve- ment at 7 mm in standard length (SL). The timing of abrupt improvement in swimming velocity and accel- eration coincided with the timing of skeletal ossifica- tion from cartilage to bone. Radii of defended broods in the field reflected the interaction of parental defense competence and offspring antipredator competence. In the Rı ´o Cabuyo, Costa Rica, brood radii increased steadily until the young were 6.45 mm SL, and then plateaued. Taken together, these data indicate inter- dependent and co-evolved traits that encompass larval ontogenetic development, antipredator performance, habitat-specific predation regime, and parental care. Keywords Skeletal ossification Swimming speed Antipredator behavior Parental care Introduction The link between morphology and performance has long been of interest to evolutionary ecologists because natural selection acts on variation in perfor- mance (Arnold, 1983; Kingsolver & Huey, 2003). For example, sprint speed of hatchling and juvenile Urosaurus ornatus lizards is correlated to stride length (not body size per se) and also to survival (Miles, 2004). Similarly, burst speed increases with snout- vent length in neonate garter snakes, (Thamnophis sirtalis fitchii), and both absolute burst speed and burst speed corrected for length (residuals of burst speed on snout-vent length) are correlated with survival (Jayne & Bennett, 1990). Garter snake consumption rate of size-matched tadpoles of Pacific tree frogs, Pseudacris regilla, is inversely correlated with burst swimming speed of the tadpoles (Watkins, 1996). Guest editors: S. Koblmu ¨ ller, R. C. Albertson, M. J. Genner, K. M. Sefc & T. Takahashi / Advances in Cichlid Research: Behavior, Ecology and Evolutionary Biology B. D. Wisenden (&) A. D. Stumbo P. A. Self D. C. McEwen P. A. Wisenden E. Brisch Biosciences Department, Minnesota State University Moorhead, 1104 7th Avenue South, Moorhead, MN 56563, USA e-mail: [email protected] J. L. Snekser M. H. A. Keenleyside Zoology Department, University of Western Ontario, London, ON, Canada M. Itzkowitz Biological Sciences Department, Lehigh University, Bethlehem, PA, USA 123 Hydrobiologia (2015) 748:259–272 DOI 10.1007/s10750-014-1917-2

Upload: trankhue

Post on 14-Feb-2017

216 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Co-evolution of offspring antipredator competence and parental

ADVANCES IN CICHLID RESEARCH

Co-evolution of offspring antipredator competenceand parental brood defense in convict cichlids

Brian D. Wisenden • Anthony D. Stumbo • Patrick A. Self •

Jennifer L. Snekser • Daniel C. McEwen • Patricia A. Wisenden •

Miles H. A. Keenleyside • Murray Itzkowitz • Ellen Brisch

Received: 23 December 2013 / Accepted: 19 May 2014 / Published online: 14 July 2014

� Springer International Publishing Switzerland 2014

Abstract Convict cichlids are Neotropical freshwa-

ter fish with biparental brood defense of their free-

swimming young. In this study, we correlate ontoge-

netic changes in swimming performance measured by

maximum velocity and burst speeds with skeletal

ossification of larvae and consider how these data

provide insights into the proximate mechanisms for

the evolution of parental care. In lab tests, swimming

velocity and acceleration of the young increased

nonlinearly with body length with a rapid improve-

ment at 7 mm in standard length (SL). The timing of

abrupt improvement in swimming velocity and accel-

eration coincided with the timing of skeletal ossifica-

tion from cartilage to bone. Radii of defended broods

in the field reflected the interaction of parental defense

competence and offspring antipredator competence. In

the Rıo Cabuyo, Costa Rica, brood radii increased

steadily until the young were 6.45 mm SL, and then

plateaued. Taken together, these data indicate inter-

dependent and co-evolved traits that encompass larval

ontogenetic development, antipredator performance,

habitat-specific predation regime, and parental care.

Keywords Skeletal ossification � Swimming speed �Antipredator behavior � Parental care

Introduction

The link between morphology and performance has

long been of interest to evolutionary ecologists

because natural selection acts on variation in perfor-

mance (Arnold, 1983; Kingsolver & Huey, 2003). For

example, sprint speed of hatchling and juvenile

Urosaurus ornatus lizards is correlated to stride length

(not body size per se) and also to survival (Miles,

2004). Similarly, burst speed increases with snout-

vent length in neonate garter snakes, (Thamnophis

sirtalis fitchii), and both absolute burst speed and burst

speed corrected for length (residuals of burst speed on

snout-vent length) are correlated with survival (Jayne

& Bennett, 1990). Garter snake consumption rate of

size-matched tadpoles of Pacific tree frogs, Pseudacris

regilla, is inversely correlated with burst swimming

speed of the tadpoles (Watkins, 1996).

Guest editors: S. Koblmuller, R. C. Albertson, M. J. Genner, K.

M. Sefc & T. Takahashi / Advances in Cichlid Research:

Behavior, Ecology and Evolutionary Biology

B. D. Wisenden (&) � A. D. Stumbo � P. A. Self �D. C. McEwen � P. A. Wisenden � E. Brisch

Biosciences Department, Minnesota State University

Moorhead, 1104 7th Avenue South, Moorhead,

MN 56563, USA

e-mail: [email protected]

J. L. Snekser � M. H. A. Keenleyside

Zoology Department, University of Western Ontario,

London, ON, Canada

M. Itzkowitz

Biological Sciences Department, Lehigh University,

Bethlehem, PA, USA

123

Hydrobiologia (2015) 748:259–272

DOI 10.1007/s10750-014-1917-2

Page 2: Co-evolution of offspring antipredator competence and parental

In this paper, we consider the links between

morphology and antipredator performance in larval

convict cichlids, a cichlid fish with biparental care of

its young. The convict cichlid (Amatitlania siquia

Schmitter-Soto, 2007) is a small fish endemic to

Central America (Bussing, 2002). They are substrate

spawners that deposit their eggs on the roof of a cavity

formed under a solid object on the substratum. They

have biparental care of their eggs and free-swimming

young over a period of four to six weeks (Wisenden,

1995). During this time, parents defend their young

against a range of brood predators (Wisenden, 1994a).

Young convict cichlids are about 4.5 mm in standard

length (SL) when they first emerge from their lair.

Both parents contribute to brood defense, although

there is some division of labor between the sexes in the

pre-hatch period, where males patrol the territory

perimeter repelling intruders while females do most of

the egg-directed care such as fanning and cleaning

(Itzkowitz et al., 2001). Once the young are free-

swimming, both parents guard a mobile territory

around the young as the family roams about during the

day. This allows the young to forage on a large area of

the substrate. In the field, convict cichlid broods form a

cohesive two-dimensional disk hovering above the

substrate. Thus, brood radius is an important variable

of both parental defense and predation risk to the

young. Families return to their lair each night for

refuge from nocturnal predators. Parental protection of

the young at study sites in Costa Rican streams

continues until the young reach a SL of 10–12 mm

(4–6 weeks post-hatch) (Wisenden, 1995).

Reproductive success in convict cichlids is deter-

mined by success against brood predators (Wisenden

1994a, 1995). Evolution of parental brood defense by

convict cichlids is therefore a problem of antipredator

behavior that is shared by the adults repelling brood

intruders, and the larvae that must evade direct attacks

when brood predators inevitably penetrate parental

defenses. Thus, offspring survival exerts selection both

on parental defense behavior and morphological

development that contributes to swimming perfor-

mance of the young. Supporting evidence comes from

three lines of data. First, large parents are more

successful than small ones at preventing fry mortality

(Wisenden, 1994a; Gagliardi-Seeley & Itzkowitz,

2006). Second, despite widely varying starting num-

bers, the number of young in guarded broods at four

different field sites in Costa Rican streams all

converged on a mean number of 27 young at indepen-

dence (Wisenden, 1994a). This suggests that there is a

limit to the radius that parents can effectively defend

and that 27 large, fully-developed young is the number

of young that fully occupy that area. Moreover,

experimentally augmented broods sustain compensa-

tory predation mortality (Wisenden & Keenleyside,

1995). The third piece of evidence is that the growth

rate of fry is inversely proportional to the number of

young in a brood (Wisenden, 1994a, 2001). The likely

mechanism for inhibited growth rates in large broods is

crowding when large numbers of young are forced to

remain within the small area that parents can effec-

tively defend. Taken together, these data suggest that

efficiency of brood defense declines rapidly with brood

radius. We predict that brood radius should reflect the

equilibrium between predator density and the ability of

parents to defend an area and the ability of the young to

evade an attack. In this study, we quantify this dynamic

process for natural populations in Costa Rican streams.

Interacting with effectiveness of parental defense is

the ability of the young to escape a direct attack. Larval

convict cichlids are poorly developed and swim weakly

at the time of emergence from the lair but become better

at evading attack as they grow and develop (Wisenden

& Keenleyside, 1992, 1994). So pronounced is this rate

of change in antipredator competence that it is exploited

by parents engaging in alloparental care. Convict

cichlids preferentially adopt young from other broods

that are similar in size or smaller than their own broods

(Wisenden & Keenleyside, 1992). When predators

attack, the relatively weak-swimming adopted fry are

the preferred targets of brood predators, thereby

deflecting predation pressure away from the young

genetically related to the adults providing the care

(Wisenden & Keenleyside, 1994).

In the current study, we build on these earlier data to

examine the relationship among parental defense behav-

ior and ontogeny of antipredator competence of larvae

through a series of field and laboratory experiments.

Morphology and the ontogeny of swimming

performance of larval convict cichlids

In Costa Rican streams, the rate of brood reduction due

to predation is much greater during early development

(5–7 mm SL) than in later development (8–10 mm SL)

(Wisenden, 1994a). Experimentally contrived groups

260 Hydrobiologia (2015) 748:259–272

123

Page 3: Co-evolution of offspring antipredator competence and parental

of young of mixed sizes comprising young of 5, 6, 7,

and 8 mm SL showed that 5 and 6 mm young were

eaten significantly more often than 7 and 8 mm young

when exposed in lab aquaria to juvenile convict

cichlids or juveniles of the sympatric piscivore

Parachromis dovii (Wisenden & Keenleyside, 1992).

Differential predation on small versus large sizes of

young was demonstrated again in a field manipulation

experiment (Wisenden & Keenleyside, 1994). That

study demonstrated fitness benefits of adopting unre-

lated young that are smaller than the genetically related

young, and that this difference seems to be maximal

between young that are smaller and larger than 7.5 mm

SL. Swimming performance of larval convict cichlids

as a function of length has not been quantified and was

the first goal of this study. We quantified swimming

performance in terms of maximum velocity and

maximum acceleration of rapid starts. Previously

published empirical data from the lab and field led us

to anticipate a rapid improvement in antipredator

escape performance starting at about 7 mm SL (Wi-

senden & Keenleyside, 1992, 1994, 1995).

Proximate anatomic structures that contribute to

rapid starts in larval fish include neurologic, muscular,

and skeletal changes (Fuiman & Magurran, 1994). We

quantified the timing of skeletal ossification because

the techniques for doing so are well established (Song

& Parenti, 1995), and because ossification of the axial

and appendicular skeleton is necessary to translate the

forces of muscular contraction into propulsive forces

generated by fins (Hale, 1999). By scoring the timing

of cartilage-bone transition of individual skeletal

elements, we could then test for a correlation between

skeletal ontogeny and changes in swimming perfor-

mance. Because swimming performance was pre-

dicted to improve non-monotonically, we predicted

that if skeletal ossification is indeed a major contrib-

utor to swimming performance then ossification

should also proceed nonlinearly and coincide with

marked improvement in swimming ability.

Parental brood defense in response to the ontogeny

of larval antipredator competence

Swimming performance of larvae has a direct effect on

the parental brood defense required for their protec-

tion. Shoals of convict cichlids take the shape of a

shallow disk hovering above the substratum. Larvae

with very poor swimming ability (i.e., limited skeletal

ossification) must remain in a small shoal directly

below the parents, where parents are most effective at

repelling intruders. When attacked, the young dart

toward the brood center and downward into the

substratum. As larval swimming performance

improves, or in habitats where density of brood

predators is low, the young can afford to stray from

the brood center where foraging opportunities are

better (i.e., increase brood radius). To develop a model

of brood defense, we looked to previously unpublished

parental defense data from field sites in Costa Rica to

define the problem of defending a brood as a series of

concentric zones of increasing risk of predation

around centrally located parents. We then used field

data from the Rıo Cabuyo to test predictions about the

rate of expansion of brood radius with respect to the

timing of skeletal ossification. We predicted that brood

radii in Rıo Cabuyo would correlate with skeletal

ossification and swimming performance of larvae.

Methods

Antipredator competence

A breeding colony of convict cichlids at the aquatic

research facility at Minnesota State University Moor-

head (MSUM) was established with wild-caught fish

from Quebrada Amores, which is a tributary of the Rıo

Cabuyo, and from the Rıo Potrero, which is the

adjacent river to the Rıo Cabuyo in the same

watershed. The Rıo Potrero and Rıo Cabuyo both

flow into the Rıo Tempisque, Guanacaste, Costa Rica.

Fish used in the first part of this experiment were F3

offspring from a random mixture of these fish.

Individual young convict cichlids were placed in a

circular plastic arena (diam = 25.5 cm) containing

dechlorinated tap water to a depth of about 2 cm. We

tested young at 1-mm intervals of SL. Sample sizes

were: 5 mm (n = 23), 6 mm (n = 17), 7 mm

(n = 28), 8 mm (n = 13), and 9 mm (n = 12). Startle

responses were induced using a standardized mechan-

ical stimulus created by dropping a weight on the

bench surface outside but adjacent to the arena and

recorded by an overhead digital camcorder at 30

frames per second. Frame-by-frame video play-back

of swimming behavior was analyzed for maximum

velocity (m/s) and acceleration (m/s/s) of the startle

Hydrobiologia (2015) 748:259–272 261

123

Page 4: Co-evolution of offspring antipredator competence and parental

response using LoggerPro� software. Arena diameter

was used as the size reference to calibrate mm/pixel

ratios for calculating velocity and acceleration.

Skeletal ossification

Fish from MSUM Costa Rican lab stock were bred in

five separate breeder tanks. Five offspring from each

brood were euthanized and preserved in formalin at

1-mm size intervals from 5-9 mm SL. Offspring were

cleared and stained with Alcian Blue to reveal

cartilaginous tissues and Alizarin Red to reveal

ossified skeletal structures (Song & Parenti, 1995).

Cleared and stained offspring were stored in glycerol

with phenol crystals as a preservative. High-resolution

images of stained offspring were taken with a digital

camera (Olympus Q-color3) mounted on a compound

microscope (Olympus CX31). We assessed calcifica-

tion of the following skeletal structures: caudal fin,

hypourals (bones that connect the caudal fin to the

posterior tip of the vertebral column), dorsal fin rays

and their pterygiophores (small bones at base of fin

that support and anchor fin), and the anal fin rays and

their pterygiophores. We focused on these structures

because we hypothesized that calcification of fin

support would confer greater swimming force. A scale

of 0–4 was used to assess the level of ossification for

each skeletal element: 0 = all blue, i.e., completely

cartilage; 1 = mostly blue; 2 = approximately equal

amounts of blue and red; 3 = mostly red; 4 = all red,

i.e., completely bone. A total of 111 different skeletal

elements and an additional 17 fin groupings (=128

skeletal data) were scored for each of 115 fish. Thus,

the resulting data matrix contained 14,720 observa-

tions on skeletal calcification; however, fish that

showed no variance in ossification scores, i.e., fish of

5 mm SL that had not yet begun the ossification

process, were removed from the analysis. This left 92

total fish and (11,776 data) to analyze and to test for

changes in ossification as a function of fish length. We

used two complementary procedures to describe

ossification. First, we calculated the average ossifica-

tion score for each size class and then subtracted the

ossification for the proceeding size class. For example,

we found the average ossification to be 3.95 for the

7 mm fish for the Cbase element and 3.91 for the same

element in the 6 mm fish. The change (D) in ossifi-

cation score then would be 0.04 (i.e., 3.95–3.91) in

transitioning between 6 and 7 mm. Given all 5 mm

fish had no ossification, the change from 5 to 6 mm

reflects the average ossification score in 6 mm fish.

Second, we used nonmetric multidimensional scaling

(NMDS) to reduce the matrix of skeletal elements

(124) by fish (87) to two dimensions. We were

required to further trim our data matrix from 92 to 87

fish because ossification scores were identical for five

separate pairs of fish and NMDS cannot operate on

identical rows in matrices. These redundant records

were collapsed into five single records given fish were

the same size. After reducing the ossification matrix,

we then used an ANOVA to test the two reduced axes

against fish length while controlling for a brood effect.

We utilized a modified Gower’s similarity coefficient

appropriate for ordinal type data to generate our

distance matrix for the NMDS (Podani, 1999). Where

the ANOVA showed fish size to be significant, we

used a Tukey test for post-hoc pair-wise comparisons.

Attack radius by parental fish in Rıo Cabuyo, Costa

Rica

Attack radius is the distance from the brood center

(where parents typically reside) to the position of an

intruder. Because attacks occurred in all 360�, radius is

a useful descriptor of both risk to larvae and parental

defense effort. Parental convict cichlids directed their

attacks mainly against juvenile and adult conspecifics,

the substrate sifter Amphilophus longimanus, and the

piscivorous Parachromis dovii. Incidental attacks also

occurred against the characin Astyanax fasciatus and

the poeciliid Poecilia gillii. Baseline data on attack

frequency and attack radii come from unpublished

data collected from four study sites during the dry

seasons (December to May) of 1989–1990 and

1990–1991. The study sites were located in and near

Lomas Barbudal Biological Reserve, Guanacaste,

Costa Rica (10�30021.5600N, 85�22014.6400W, elev.

43 m), and described by Wisenden & Keenleyside

(1995) and Wisenden (1995). Each study site was a

defined section of stream in the Rıo Cabuyo or its

tributary Quebrada Amores. Within each stream, there

was one site that was a large deep pool (pool habitat)

and another site that was a series of small intercon-

nected shallow pools (stream habitat). Habitat affects

predation pressure (pools [ streams), male mate

desertion (streams [ pools), and overall brood suc-

cess (streams [ pools) (Wisenden, 1994a, b). Streams

were shallow and water transparency was excellent;

262 Hydrobiologia (2015) 748:259–272

123

Page 5: Co-evolution of offspring antipredator competence and parental

thus, parental defense behaviors were recorded from

above the surface with the aid of polarized sun-

glasses after calibrating distance estimates against a

meter stick. Parental defense behavior was recorded

from a distance of several meters for 10 min for the

one parent, followed immediately by 10 min of

parental behavior of the other parent. We conducted

840 focal follows on 198 different broods, typically

with one parental male and female per brood. Some

brood pairs were observed on more than one occasion.

We further separated our focal follows into two

separate responses: (1) attacks against cichlids and (2)

attacks against non-cichlid fishes. This effectively

doubled our records from 840 to 1,680 entries. Some

of these records were incomplete (e.g., estimates for

parent lengths, brood radii, etc., were not made).

Records without a full complement of data were

eliminated prior to analysis. Lengths and counts of

young were conducted at 5-day intervals for each

brood. Entire broods were encircled with a seine net,

the male was captured with a hand net, all of the

offspring were captured by hand net with the aid of a

scuba mask and the female was then captured by

chasing her into the seine net. On shore, a subsample

of 15 young was measured to the nearest 0.5 mm SL

with a ruler, the remaining young counted, and two

dorsal spines were excised from each parent in a

unique combination such that individuals could be

recognized if they respawned later that season (Wi-

senden 1994a, 1995). We also sketched the markings

of each parent to allow for convenient identification of

the breeding pair on subsequent samplings within the

same brood cycle. The family was carefully released

back into the stream at the site of capture by first

releasing the parents (the usually male fled and hid, but

the female always searched for her young and

defended the area; see Wisenden et al. 2008). The

young were returned via a clear plastic tube fashioned

from two 2-L pop bottles with the tops and bottoms

removed and stacked together. The young settled to

the bottom and made eye contact with the searching

female. When the female approached the tube we

raised the tube upward to allow the family to reunite.

Brood radii in the Rıo Cabuyo 2008

Brood radius is the distance from the brood center to

the outer limit of the area occupied by the brood. These

data were collected from the Rıo Cabuyo in January

2008. We surveyed about 1 km of river starting at the

deep pool below the road crossing (Cabuyo Pool site

used in 1990 and 1991) up to the pool with the

waterfall near the upstream boundary of Lomas

Barbudal Biological Reserve. For each brood, we

measured brood depth and brood diameter to the

nearest cm with a meter stick, SL of a subsample of 10

captured young to the nearest 0.5 mm using a ruler and

counted the total number of young in the brood. To

ensure accurate brood counts, we caught all but

several of the young with hand nets and brought them

on shore to measure and count. The remaining few

young in the water could be counted with certainty

with the aid of a SCUBA mask. Young were returned

to their parents immediately after sampling using the

transparent plastic tube described above.

Results

Antipredator swimming performance

Size of young had a significant effect on maximum

swimming velocity (ANOVA F4,88 = 36.96, P\0.001) and acceleration (ANOVA F4,88 = 48.85,

P\ 0.001) of the startle response (Fig. 1). Post-hoc

pair-wise comparisons (Tukey test) among size groups

showed that 5 mm = 6 mm \7 mm \8 mm =

9 mm for both velocity and acceleration (P\0.05).

Skeletal ossification

The skeleton of larvae ossified from cartilage (blue) to

bone (red) at 1-mm intervals from 5 mm SL to 9 mm

SL (Fig. 2A). Magnified images of cleared and stained

embryos allowed for unambiguous scoring of individ-

ual skeletal elements (Fig. 2B). The ANOVA per-

formed on the scores generated by the NMDS showed

the first axis to have significant groups for both size

(F3.79 = 25.14, P \ 0.001) and brood (F4,79 = 37.77,

P \ 0.001), but that variance partitioned from the

sums of squares indicated that the majority of variance

for the first axis could be attributed to brood source

(49%). Fish size explained 24% with the remaining

variance left unexplained (26%). The second axis also

showed fish size (F3.79 = 199.12, P \ 0.001) and

brood (F3.79 = 15.19, P \ 0.001) were both signifi-

cant, but here the primary factor associated with

structuring variance was fish size (81%). Brood

Hydrobiologia (2015) 748:259–272 263

123

Page 6: Co-evolution of offspring antipredator competence and parental

explained only a minor proportion (8%) with 11%

unexplained. The NMDS ordination plot showed good

discrimination of cichlid young of different lengths on

the basis of skeletal ossification data (Fig. 3). Post-hoc

pair-wise comparisons (Tukey test) among size groups

showed that 6 mm \ 7 mm \ 8 mm \ 9 mm for

scores along the second NMDS axis (P \ 0.001),

where the largest change in skeletal ossification

occurred between sizes 6 and 7 mm SL (Fig. 3).

Individual skeletal elements that ossified the most

between 5 and 6 mm SL were the skeleton of the

caudal fin and supporting elements of the hypourals

that anchor the caudal fins to the caudal tip of the

vertebral column (Table 1). Between 6 and 7 mm SL,

where the most pronounced improvements in swim-

ming performance occurred, the medial pterygio-

phores (middle rays that support the fin membrane) of

the dorsal and anal fin showed the greatest change.

Over the interval from 7 to 8 mm SL, skeletal

ossification continued on various elements without

any cluster of bones ossifying as a group. Between 8

and 9 mm SL, the distal pterygiophores that support

the outer edges of the dorsal and anal fins showed the

greatest change, indicating completion of the ossifi-

cation process (Table 1).

Brood defense radii in Costa Rican streams

A total of 840, 10-min recordings were made on 198

different broods, representing a comprehensive sam-

ple of all four sites (two in pool habitat, two in stream

habitat), and all brood ages and sizes (Table 2). Mean

attack radius over two breeding seasons in four sites by

males and females against six types of intruders was

33.56 ± 0.48 cm (Fig. 4). Attacks occurred on aver-

age 6.93 ± 0.24 times per 10-min observation period

(Fig. 4). Attack frequency and distance varied with

0

0.1

0.2

0.3

0.4

0.5

Max

imum

Vel

ocity

(m

/s)

Young SL (mm)

a

0

1

2

3

4

5 6 7 8 9

5 6 7 8 9

Max

imum

Acc

eler

atio

n (m

•s-2

)

Young SL (mm)

a

b

c c

a a

b

c

c

Fig. 1 Maximum velocity (upper panel) and maximum accel-

eration (lower panel) for young Costa Rican fish sized 5, 6, 7, 8

and 9 mm in standard length. Letters above bars indicate the

results of post-hoc pair-wise Bonferroni tests. Shared letters

indicate no significant difference (P [ 0.05)

5 mm 8 mm

6mm 9 mm

7 mm

(B)

Hypourals

Prox Med Dist

Anal pterygiophores

Prox Med Dist

Dorsal pterygiophores

Dorsal fin rays

Anal fin rays

Caudal fin rays

(A)

Fig. 2 Color images of cleared and stained embryos A at 1-mm

intervals of development and B showing individual skeletal

elements composed of cartilage in blue and bone in red

264 Hydrobiologia (2015) 748:259–272

123

Page 7: Co-evolution of offspring antipredator competence and parental

species of intruder (Table 3). The characin Astyanax

fasciatus and the poeciliid Poecilia gillii were attacked

only when they intruded directly into the mobile

territory of the brood. Brood radius within 10 cm is

physically occupied by the parents, therefore, mini-

mum parental attack distance was 10 cm. Because

attack frequencies were counts, we used a generalized

linear model to test for environmental effects on the

number of attacks. We had no a priori reason to

suspect overdispersion and used Akaikie Information

Criteria (AIC) model selection to choose the better

fitting of a Poisson model or a negative binomial

model for the case of overdispersion (Burnham &

Anderson, 2002). AIC indicated that the negative

binomial distribution fit much better than did the

Poisson distribution (DAIC = 1,104, w = 0). The

DAIC gives the difference in AIC score for a particular

model relative to the most parsimonious model and

generally a DAIC \ 2 indicates comparable models.

The AIC weight (w) gives the probability that upon

resampling, a particular model would end up being

ranked as the most parsimonious. Our result indicates

it is virtually impossible that the Poisson model would

ever be supported over the negative binomial. Conse-

quently, this indicates that there is overdispersion

associated with the counts. We used a set of predictor

variables that included random effects of year and

brood along with a covariate for time. After

controlling for these variables, our model resulted in

all terms we included were significant except for size

of the young. In order of importance of generating

high frequency of attacks were sex of the parent,

taxonomy of the target fish, brood depth, site, number

of young, and parent size. Males had 2.7 times as many

attacks as females did, and cichlids were attacked 1.8

times more than noncichlids (P \ 0.05). Effects for

the other predictors were significant, but weak with

higher attack frequencies for deeper broods, pool sites

as opposed to stream sites, larger as opposed to smaller

broods, and larger rather than smaller parents. The

coefficient of determination for the global model

estimated on 1,308 complete records using null and

residual deviance indicated our model explained a

significant proportion of variation (v62 = 179, R2 =

0.30, P \ 0.001).

We were able to use the same set of predictors to test

attack radius, but because the response was a continuous

variable, we were able to use a linear modeling

framework. The global model, generated on 943 focal

follows, explained a significant portion of the variance

in measured attack radius (F169,773 = 4.25, R2 = 0.368,

P \ 0.001). After controlling for year, time, and brood,

attack radius responded strongly with taxonomic group

(cichlids = 38.7 ± 0.56 cm; noncichlids = 22.2 ±

0.57 cm), which captured[35% of variability remain-

ing after controlling for covariates. Parent size, young

number, and brood depth also were statistically signif-

icant model terms (P \ 0.05) but accounted for a very

small proportion of the observed variance in attack radii,

explaining 0.1, 1.4, and 0.7% of the variance, respec-

tively, (None of sex, site, or size of the young were

significant (P [ 0.05)).

Taken together, parental attack radius data indicate

a zonal pattern of parental defense (Fig. 4). Very few

attacks occurred in the area immediately below the

parents, with a radius of less than 10 cm, because the

parents physically occupy this whole area. The next

concentric zone, with a radius between 10 and 20 cm,

is of intermediate exposure to brood predators. To

evade predator attack in this zone, young must be able

to dart 10–20 cm to the brood center. Relatively few

attacks occur within 10–20 cm likely because of

parental intimidation. The next zone, with radii

between 20 and 30 cm, is where attack frequency

begins to increase rapidly. To attack a potential brood

predator at this radius parents must momentarily

vacate the brood center leaving the brood vulnerable to

Fig. 3 Plot of NMDS1 versus NMDS 2 reveals major patterns

of variation in skeletal ossification for larval convict cichlids

ranging in size from 6 to 9 mm in SL. Minimum convex

polygons are drawn around points for fish of similar size. The

average score for each size group along NMDS 2 is shown along

with the 95% confidence intervals. Tukey HSD post-hoc test

showed each group to be significantly different (P \ 0.001)

Hydrobiologia (2015) 748:259–272 265

123

Page 8: Co-evolution of offspring antipredator competence and parental

Table 1 Change in mean calcification score for each 1-mm interval for larval convict cichlids for each of 129 skeletal elements in

larval convict cichlids over 1-mm intervals

5–6 mm SL 6–7 mm SL 7–8 mm SL 8–9 mm SL

Bone D Bone D Bone D Bone D

Cbase 3.91 DMED1 2.28 APROX3 0.91 ADIST10 1.50

Cmed 3.86 DMED4 2.21 H4 dist 0.91 ADIST11 1.31

H2 prox 3.14 DMED2 2.20 HL2prox 0.87 ADIST7 1.24

H3 prox 2.91 DMED3 2.16 APROX2 0.83 ADIST8 1.19

Dspiny rays 2.73 AMED7 2.03 H6 prox 0.83 DMED14 1.19

H4 prox 2.64 DMED5 2.03 DDIST3 0.78 ADIST5 1.18

H5 prox 2.59 AMED6 2.03 APROX4 0.78 ADIST9 1.13

Aspiny 2.50 DMED6 1.95 APROX1 0.78 ADIST6 1.10

H5 med 2.36 AMED1 1.95 AMED2 0.78 ADIST4 1.08

H4 med 2.32 AMED5 1.93 AMED4 0.78 DDIST2 1.08

Cdist 2.32 AMED3 1.91 H1 prox 0.74 DDIST4 1.05

H6 med 2.27 AMED8 1.91 H1 dist 0.74 DMED13 1.03

Dsoft rays 2.27 AMED4 1.88 APROX5 0.74 DMED11 0.99

H3 med 2.23 HU 3prox 1.88 AMED3 0.74 DDIST5 0.98

H6 prox 2.18 AMED2 1.86 AMED9 0.74 DDIST3 0.97

H2 med 2.09 Asoft 1.74 DDIST2 0.74 DDIST9 0.95

Asoft 1.91 DMED7 1.69 DDIST6 0.74 DDIST1 0.95

H6 dist 1.73 ADIST2 1.64 APROX9 0.74 DDIST11 0.93

HL 2med 1.68 DDIST1 1.64 AMED11 0.74 DDIST7 0.93

H1 prox 1.64 ADIST1 1.64 AMED14 0.74 DMED12 0.92

H5 dist 1.59 AMED9 1.61 AMED5 0.74 DMED9 0.91

H1 med 1.55 DMED8 1.61 DDIST1 0.70 DMED10 0.90

H4 dist 1.50 ADIST3 1.52 H1 med 0.70 DDIST6 0.89

H2 dist 1.45 Aspiny 1.50 H5 dist 0.70 DDIST12 0.86

H3 dist 1.32 DDIST2 1.47 APROX7 0.70 DDIST8 0.83

AMED1 1.23 H1 prox 1.45 APROX8 0.70 DMED8 0.80

ADIST1 1.23 ADIST4 1.44 DMED2 0.65 ADIST3 0.79

HL 2prox 1.18 DDIST3 1.43 AMED1 0.65 DDIST10 0.77

APROX1 1.18 DDIST4 1.43 AMED6 0.65 ADIST12 0.75

APROX2 1.18 ADIST7 1.42 AMED7 0.65 DMED7 0.74

AMED2 1.18 ADIST6 1.38 HU 3dist 0.65 DMED6 0.70

ADIST2 1.18 AMED10 1.35 HL 2dist 0.65 DMED15 0.63

AMED3 1.05 DMED9 1.35 DDIST5 0.65 DMED3 0.62

ADIST3 1.05 HL 2prox 1.34 APROX6 0.65 ADIST14 0.59

H1 dist 0.91 HU 3med 1.28 AMED10 0.65 AMED10 0.58

APROX3 0.91 ADIST5 1.28 DMED3 0.65 DMED4 0.57

ADIST4 0.77 APROX1 1.21 AMED8 0.65 DDIST14 0.56

APROX4 0.73 Dsoft rays 1.21 AMED13 0.61 DPROX11 0.56

AMED4 0.73 Dspiny rays 1.14 H3 med 0.61 DPROX12 0.56

HL 2dist 0.68 DDIST5 1.12 H3 dist 0.61 DMED5 0.55

HU 3med 0.64 APROX2 1.12 HL 2med 0.61 DMED1 0.54

APROX5 0.64 DMED10 1.09 H4 med 0.57 DMED2 0.54

ADIST5 0.64 H3 prox 1.00 H2 dist 0.57 ADIST1 0.53

266 Hydrobiologia (2015) 748:259–272

123

Page 9: Co-evolution of offspring antipredator competence and parental

Table 1 continued

5–6 mm SL 6–7 mm SL 7–8 mm SL 8–9 mm SL

Bone D Bone D Bone D Bone D

AMED5 0.59 APROX3 1.00 ADIST8 0.57 ADIST2 0.53

DMED1 0.45 AMED11 1.00 APROXS10 0.57 AMED15 0.51

DMED2 0.45 ADIST8 1.00 H6 med 0.57 ADIST13 0.51

APROX6 0.41 H5 prox 0.97 DMED1 0.57 AMED8 0.46

HU 3prox 0.36 DMED11 0.96 DMED5 0.57 AMED9 0.44

DMED3 0.36 AMED12 0.96 HI 3prox 0.54 APROX13 0.42

AMED6 0.36 H6 prox 0.95 H2 med 0.52 AMED7 0.40

ADIST6 0.36 APROX4 0.92 H5 med 0.52 DPROX9 0.40

DMED4 0.27 DDIST6 0.91 D PROX 1 0.52 DPROX10 0.39

DMED5 0.27 H5 med 0.90 D PROX2 0.52 DDIST13 0.39

AMED7 0.23 H4 prox 0.89 H6 dist 0.52 APROX12 0.37

ADIST7 0.23 DMED12 0.87 ADIST6 0.48 DPROX13 0.36

DDIST1 0.18 ADIST9 0.87 ADIST9 0.48 ADIST15 0.36

DDIST2 0.18 D PROX 1 0.87 H4 prox 0.48 DPROX8 0.33

APROX7 0.18 D PROX2 0.87 DMED4 0.48 APROX14 0.33

DMED6 0.14 D PROX3 0.87 DMED6 0.48 H1 med 0.32

DDIST3 0.14 D PROX4 0.87 DMED7 0.48 DDIST15 0.32

DDIST4 0.14 H6 med 0.86 H5 prox 0.43 H1 dist 0.30

DDIST5 0.14 H4 med 0.86 DMED10 0.43 H5 dist 0.30

AMED8 0.14 APROX5 0.84 ADIST3 0.43 AMED12 0.28

HU 3dist 0.09 D PROX5 0.83 ADIST5 0.43 AMED14 0.27

DMED7 0.09 H1 med 0.80 D PROX3 0.43 DPROX14 0.27

DDIST6 0.09 H1 dist 0.79 AMED15 0.43 AMED6 0.27

DDIST7 0.09 HU 3dist 0.78 D PROX 6 0.43 HU 3dist 0.27

ADIST8 0.09 APROX7 0.77 DPROX7 0.43 AMED11 0.26

D PROX 1 0.05 HL 2dist 0.75 DMED9 0.43 AMED3 0.25

D PROX2 0.05 DDIST8 0.74 DDIST4 0.43 APROXS10 0.24

D PROX3 0.05 APROX8 0.74 APROX11 0.43 AMED4 0.24

D PROX4 0.05 APROX6 0.72 ADIST2 0.43 HL 2dist 0.23

DMED8 0.05 D PROX 6 0.70 ADIST4 0.43 H2 dist 0.22

DMED9 0.05 ADIST10 0.70 DMED8 0.39 APROX11 0.21

D PROX5 0.00 H2 med 0.69 DDIST7 0.39 APROX15 0.19

D PROX 6 0.00 DDIST7 0.69 HU 3med 0.39 AMED13 0.19

DPROX7 0.00 HL 2med 0.67 Dsoft rays 0.39 H4 dist 0.18

DPROX8 0.00 H2 prox 0.65 DMED13 0.39 H6 dist 0.18

DPROX9 0.00 H3 med 0.64 AMED12 0.39 DPROX7 0.17

DPROX10 0.00 H5 dist 0.63 ADIST1 0.39 AMED5 0.16

DPROX11 0.00 H6 dist 0.62 D PROX4 0.35 ADIST16 0.16

DPROX12 0.00 DPROX7 0.61 D PROX5 0.35 H3 med 0.15

DPROX13 0.00 APROX9 0.61 DPROX8 0.35 H3 dist 0.15

DPROX14 0.00 DDIST9 0.57 DDIST12 0.35 APROX1 0.14

DPROX15 0.00 AMED13 0.57 DDIST13 0.35 D PROX5 0.14

DMED10 0.00 H3 dist 0.55 APROX13 0.35 APROX7 0.14

Hydrobiologia (2015) 748:259–272 267

123

Page 10: Co-evolution of offspring antipredator competence and parental

attack from the opposite direction. For this reason,

parents generally orient at 180� to each other to be

ready to defend against intruders approaching from

any direction. The next zone, with radii between 30

and 50 cm, is where most parental attacks occurred.

These attacks require parents to vacate the brood,

Table 1 continued

5–6 mm SL 6–7 mm SL 7–8 mm SL 8–9 mm SL

Bone D Bone D Bone D Bone D

DMED11 0.00 DMED13 0.52 ADIST12 0.35 D PROX 6 0.13

DMED12 0.00 H2 dist 0.50 ADIST13 0.35 AMED1 0.12

DMED13 0.00 DPROX8 0.48 DPROX9 0.30 AMED2 0.12

DMED14 0.00 ADIST11 0.48 APROX12 0.30 HU 3prox 0.11

DMED15 0.00 H4 dist 0.46 ADIST14 0.30 H5 med 0.11

DDIST8 0.00 DDIST10 0.39 Asoft 0.26 D PROX4 0.11

DDIST9 0.00 DPROX9 0.35 DPROX10 0.26 H6 med 0.09

DDIST10 0.00 DPROX10 0.35 DMED11 0.26 Cdist 0.08

DDIST11 0.00 DMED14 0.35 ADIST7 0.26 APROX2 0.08

DDIST12 0.00 APROXS10 0.35 DDIST10 0.26 AMED16 0.07

DDIST13 0.00 DMED15 0.30 DDIST11 0.26 H2 med 0.06

DDIST14 0.00 APROX11 0.30 DMED12 0.26 HU 3med 0.06

DDIST15 0.00 AMED14 0.30 ADIST11 0.26 DPROX15 0.06

APROX8 0.00 DPROX11 0.26 DDIST8 0.22 APROX6 0.06

APROX9 0.00 DPROX12 0.26 DDIST9 0.22 APROX16 0.05

APROXS10 0.00 AMED15 0.26 ADIST15 0.22 H4 med 0.05

APROX11 0.00 AMED16 0.26 Cdist 0.22 HL 2prox 0.03

APROX12 0.00 DPROX13 0.22 APROX14 0.17 Dsoft rays 0.03

APROX13 0.00 APROX12 0.22 DDIST14 0.17 D PROX3 0.02

APROX14 0.00 ADIST12 0.22 ADIST10 0.17 H1 prox 0.02

APROX15 0.00 DPROX14 0.17 H2 prox 0.17 APROX3 0.02

APROX16 0.00 DPROX15 0.17 Dspiny rays 0.13 H4 prox 0.00

AMED9 0.00 DDIST11 0.17 DPROX11 0.13 H5 prox 0.00

AMED10 0.00 APROX13 0.13 DPROX12 0.13 Cbase 0.00

AMED11 0.00 APROX14 0.13 Cbase 0.04 Cmed 0.00

AMED12 0.00 APROX15 0.13 Cmed 0.04 Dspiny rays 0.00

AMED13 0.00 APROX16 0.13 DMED14 0.04 HL 2med -0.01

AMED14 0.00 Cdist 0.12 APROX15 0.04 APROX4 -0.01

AMED15 0.00 Cmed 0.09 H3 prox 0.04 Asoft -0.02

AMED16 0.00 ADIST13 0.09 DPROX13 0.00 Aspiny -0.05

ADIST9 0.00 Cbase 0.05 DDIST15 0.00 H6 prox -0.06

ADIST10 0.00 DDIST12 0 Aspiny 0.00 D PROX 1 -0.07

ADIST11 0.00 DDIST13 0 ADIST16 0.00 D PROX2 -0.07

ADIST12 0.00 DDIST14 0 DPROX14 -0.13 APROX8 -0.07

ADIST13 0.00 DDIST15 0 DPROX15 -0.13 APROX5 -0.11

ADIST14 0.00 ADIST14 0 APROX16 -0.13 H3 prox -0.11

ADIST15 0.00 ADIST15 0 AMED16 -0.17 APROX9 -0.14

ADIST16 0.00 ADIST16 0 DMED15 -0.30 H2 prox -0.17

Fin codes: C, caudal; A, anal; D, dorsal; H, hypourals; prox, proximate; med, medial; dist, distal; numbers indicate series number

268 Hydrobiologia (2015) 748:259–272

123

Page 11: Co-evolution of offspring antipredator competence and parental

intercept and pursue intruders to drive them from the

vicinity before quickly retreating back to the brood

center. Parental reaction time at these distances is too

long to afford protection for the young. It seems that

for this reason, few young strayed more than 30 cm

from the brood center (Fig. 5).

Brood radii in Rıo Cabuyo 2008

Brood radius increased with Young SL independent of

the number of young in the brood (model

radius = SL ? Number ? SL * Number, F3,32 = 9.08,

P \ 0.001; test of coefficients SL: t = 2.19,

P = 0.036; Number t = 0.21, P = 0.834; SL * Num-

ber t = 0.428, P = 0.672; Fig. 5). The relationship

Table 2 Distribution of sample sizes for parental brood

defense behaviors in 1990, 1991

Study site No of

different

pairs

Size of young (mm) for

1990 ? 1991

1990 1991 Egg/

wriggler

\5 6–8 9–10 [10

Amores

Pool

45 20 52 61 85 51 18

Cabuyo

Pool

36 19 31 64 64 48 16

Amores

Stream

18 20 8 17 40 52 47

Cabuyo

Stream

17 23 8 29 55 49 41

0

20

40

60

80

100

120

140

160

180

0-9 10-19 20-29 30-39 40-49 50-59 60-69 70-79 80-89 90

Fre

quen

cy

Parental attack radius (cm)

JuvCC

AdCC

Long

Dovii

Average

Par

enta

l occ

upat

ion

Low

ris

k of

pre

datio

n

Sho

rt e

scap

e di

stan

ce

Inte

rmed

iate

ris

k

Long

esc

ape

dist

ance

H

igh

risk

Parental intercept and chase Very high risk

1.50

%

4.72

%

15.5

2%

30.2

8%

27.2

9%

10.4

2%

5.55

%

2.70

%

1.27

%

0.75

%

Fig. 4 A Parental attack

frequency as a function of

distance between predator

and brood center by species

of intruder and overall total

mean attack distance per

10-min observation period

(n = 840). Sample sizes and

mean values are given in

Table 2. JuvCC juvenile

convict cichlid, AdCC adult

convict cichlid, Long A.

longimanus, Dovii P. dovii.

Percentages above bars

represent the over

percentage of total attacks

by parents

Table 3 Mean (and standard error, SE) attack frequency per

10-min and attack radius (cm) for parental convict cichlids

defending broods in Rıo Cabuyo and Quebrada Amores, 1990

and 1991. Intruder species are Astyanax fasciatus, Poecilia

gillii, juvenile convict cichlids, adult convict cichlids, Am-

philophus longimanus, and Parachromis dovii

Total Astyanax P. gillii Juv CC Adult CC A. long. P. dovii

Attack frequency

X 6.93 1.75 0.50 1.50 2.02 0.2323 0.34

SE 0.24 0.12 0.05 0.12 0.09 0.02 0.027

Attack distance

X 33.56 17.00 21.39 36.40 40.30 31.17 40.62

SE 0.48 0.47 0.82 0.90 0.59 1.38 1.45

N 764 366 188 463 525 122 220

Hydrobiologia (2015) 748:259–272 269

123

Page 12: Co-evolution of offspring antipredator competence and parental

between brood radius and Young SL was not linear. A

significant change point (Siegel & Castellan, 1988)

occurred in the slope of the trend line defining the effect

of SL the change point occurred at 6.45 mm SL

(z = 4.63, P \ 0.001). Individual regressions on pre-

and post-change point data showed a significant slope

of brood radius on Young SL only before 6.45 mm SL

(F1,13 = 7.3, P = 0.019) but not after 6.45 mm SL

(F1,13 = 0.05, P = 0.835). Non-linear increase in

maximum velocity, maximum acceleration, and brood

radius all coincided with the timing of larval ossifica-

tion (Fig. 6).

Discussion

Taken together, field and laboratory data show a

correlation among the morphological and functional

aspects of larval antipredator competence, and the

evolution and expression of biparental brood defense

behavior in convict cichlids.

The overall nonlinear pattern of burst speed with

respect to body length observed in convict cichlids is

similar to the pattern seen in zebrafish (Fuiman &

Webb, 1988). The main determinant of predator

evasion is ontogenetic changes (Wisenden & Keen-

leyside, 1992, 1994) because other aspects of predator

detection and evasion contribute little to escape

potential (Fuiman, 1994; Fuiman et al., 2006). Skeletal

ossification is a convenient and easily quantifiable

measure of ontogeny. The ossification of the skeleton

from cartilage to bone confers greater efficiency of the

forces of muscular contraction into forces of propul-

sion from the fin areas. However, skeletal ossification

is not the sole determinant of locomotory perfor-

mance. Muscle mass also increases dramatically over

the larva-juvenile transition as does fin area and

overall body form, all of which contribute to locomo-

tory performance (Fuiman & Magurran, 1994;

McHenry & Lauder, 2006). The abrupt improvement

in swimming performance of convict cichlid larvae

occurred shortly before 7 mm SL, which coincides

with an abrupt acceleration in, and completion of,

skeletal ossification. Note that ossification scores of

some 7 mm fish clustered with 6 mm fish and some

clustered with 8 mm fish (Fig. 3).

Parental defense data from the field parsed the

young’s living space into zones of risk created by

0

10

20

30

4 6 8 10 12 14

Bro

od R

adiu

s (c

m)

Young SL

Low Risk

Intermediate Risk

High Risk

Fig. 5 Radius of natural broods in the Rıo Cabuyo as a function

of Young SL. The non-linear nature of this relationship is

highlighted by a switch-point in slope at 6.45 mm (solid symbols

and solid line, before 6.45 mm; open symbols and dashed line,

after 6.45 mm)

0

2

4

6

8

10

12

14

16

18

Bro

od r

adiu

s (c

m)

Max Velocity (m/s)

0

2

4

6

8

10

12

14

16

18

0.1 0.2 0.3 0.4 1.0 1.5 2.0 2.5 3.0

Bro

od r

adiu

s (c

m)

Max Acceleration (m/s2)

5 5

6 6

7 7 8 8 9

9

Fig. 6 Parental defense

attack radii as a function of

larval maximum velocity

(left panel) and maximum

acceleration (right panel)

270 Hydrobiologia (2015) 748:259–272

123

Page 13: Co-evolution of offspring antipredator competence and parental

differential tolerance for each species of intruder.

Low-risk intruders such as Poecilia and Astyanax are

allowed closer to the brood while medium-risk

predators such as juvenile convict cichlids receive

little tolerance and high-risk predators such as the

piscivorous P. dovii receive none at all. Parental brood

defense behavior recorded in the Rıo Cabuyo system

was broadly similar to that observed by Alonzo et al.

(2001) in Laguna de Xiloa. Like fish in the Rıo

Cabuyo, fish in Laguna de Xiloa increased attack rates

for broods with high numbers of fry, and attack

intensity (in terms of frequency, not distance to

intruder) in accordance with the predation threat

different species of intruder posed to the young

(Alonzo et al., 2001). In contrast to the Xiloa system,

our data did not find an effect of size of young on

attack rates.

Detailed data on parental defense suggest that the

economy of brood defense imposes a limit on defense

radii. Secondly, defense radius is graded in accordance

with the degree of threat represented by each species

of intruder. Thirdly, attack radii are constrained by

parental body size and swimming ability, which in

turn constrains the ecological space in which young

evolved optimal solutions to the foraging/predation

risk trade-off. In sum, these data indicate that parental

care behavior co-evolved in a three-way interaction

with swimming performance of the young and habitat-

specific predation pressure on the young. We predict

that convict cichlids in other systems will have

evolved different equilibria for these traits.

Brood diameter increased as the free-swimming

young grew from 4.5 to approximately 6.45 mm SL.

Expansion of brood diameter during this interval was

most likely a reflection of foraging activity of the

young and not a response to changes in parental

defense because attack radius was determined by

number of young, not size (age) of young. Experi-

mentally enlarged broods show suppressed rates of

daily growth increments relative to unmanipulated

control broods (Wisenden & Keenleyside, 1995),

suggesting that brood radius is a function of the

trade-off between the benefit of foraging and the cost

of predation risk.

Parental defense behavior based on antipredator

competence of the young set the stage for alloparental

care by creating young that differ markedly in

antipredator competence. Convict cichlids routinely

adopt conspecific young from neighboring broods. In

Costa Rican streams, at least 29% of broods contains

unrelated young at some point during the period of

care based on the presence of different size groups

(Wisenden & Keenleyside, 1992) while genetic anal-

yses reveal the actual rate of brood mixing is well

above 50% (Lee-Jenkins et al. unpublished data).

While their young are less than 7 mm SL, parents

judiciously adopt only young that are similar in size or

smaller than their own young (Wisenden & Keenley-

side, 1992). When predators succeed in attacking the

brood, it is the relatively small and weak adopted

young that bear significantly greater predation mor-

tality, sparing the young related to the adopting

parents (Wisenden & Keenleyside, 1994). Moreover,

the antipredatory competence switch-point that we

identified matches the size at which parents no longer

discriminate adoptions on the basis of size, and also

the size at which differential predation occurs in

adopted broods (Wisenden & Keenleyside, 1992,

1994). In broods experimentally manipulated to

comprise large host young and small adopted young,

large young (related to the parents) had significantly

higher survival over 10 days if smaller young were

added to their brood, a rate significantly higher that

what would occur by dilution of predation alone

(Wisenden & Keenleyside, 1994).

These data have bearing on the egg size—egg

number trade-off. Selection for earlier onset of

antipredator competence should select for an increase

in egg size and either reduced fecundity or larger adult

body size. The current body size and egg size/egg

number in this species (or any cichlid) reflects an

evolutionary equilibrium between these competing

life history traits (Kolm et al., 2006a, b). While this

link has always been intuitive, these data empirically

demonstrate the linkage between the ontogeny of

antipredator competence and the evolution of parental

care in cichlids.

Acknowledgments The 1990–1991 field data were collected

with the able assistance of Jeff Christie, Gillian De Gannes, and

Randy Jimenez. Logistical support in 1990 and 1991 provided

by Gordon and Jutta Frankie and Amigos de Lomas Barbudal,

Stewart Family (Quebrada Amores), and el Servicios de las

Parques Nacionales, Costa Rica. 1990 and 1991 data were

funded by an NSERC operating grant to Miles Keenleyside.

Data in 2008 were collected under MINAE Pasaporte de

Investigaciones numero 00058. Special thanks to Celso

Alvarado and Henry Ramirez of the Ministerio del Ambiente

y Energıa, Sistema Nacional de Areas de Conservacion Arenal

Tempisque (ACT MINAE). Ossification and startle responses

Hydrobiologia (2015) 748:259–272 271

123

Page 14: Co-evolution of offspring antipredator competence and parental

were enabled by research grants from the College of Social and

Natural Sciences, MSUM to Brian Wisenden, and the Dille

Fund for Excellence (MSUM) to Brian Wisenden and Ellen

Brisch.

References

Alonzo, J. J., K. R. McKaye & E. van den Berghe, 2001.

Parental defense of young by the convict cichlid, Archo-

centrus nigrofasciatus, in Lake Xiloa, Nicaragua. Journal

of Aquariculture and Aquatic Sciences 9: 208–228.

Arnold, S. J., 1983. Morphology, performance and fitness.

American Zoologist 23: 347–361.

Burnham, K. P. & D. R. Anderson, 2002. Model Selection and

Multimodel Inference: A Practical Information-Theoretic

Approach. Springer, New York.

Bussing, W. A., 2002. Peces de las Aguas Continentals de Costa

Rica. Editorial de la Universidad de Costa Rica, San Jose.

Fuiman, L. A., 1994. The interplay of ontogeny and scaling in

the interactions of fish larvae and their predators. Journal of

Fish Biology 45: 55–79.

Fuiman, L. A. & A. E. Magurran, 1994. Development of pred-

ator defences in fishes. Reviews in Fish Biology and

Fisheries 4: 145–183.

Fuiman, L. A. & P. W. Webb, 1988. Ontogeny of routine

activity and performance in zebra danios (Teleostei:

Cyprinidae). Animal Behaviour 36: 250–261.

Fuiman, L. A., K. A. Rose, J. H. Cowan & E. P. Smith, 2006.

Survival skills required for predator evasion by fish larvae

and their relation to laboratory measures of performance.

Animal Behaviour 71: 1389–1399.

Gagliardi-Seeley, J. L. & M. Itzkowitz, 2006. Male size predicts

the ability to defend offspring in the biparental convict

cichlid Archocentrus nigrofasciatus. Journal of Fish Biol-

ogy 69: 1239–1244.

Hale, M. E., 1999. Locomotor mechanics during early life his-

tory: effects of size and ontogeny on fast-start performance

of salmonid fishes. Journal of Experimental Biology 202:

1465–1479.

Itzkowitz, M., N. Santangelo & M. Richter, 2001. Parental

division of labour and the shift from minimal to maximal

role specialization: an experiment using a biparental fish.

Animal Behaviour 61: 1237–1245.

Jayne, B. C. & A. F. Bennett, 1990. Selection on locomotor

performance capacity in a natural population of garter

snakes. Evolution 44: 1204–1229.

Kingsolver, J. G. & R. B. Huey, 2003. Introduction: the evolu-

tion of morphology, performance and fitness. Integrative

and Comparative Biology 43: 361–366.

Kolm, N., N. B. Goodwin, S. Balshine & J. D. Reynolds, 2006a.

Life history evolution in cichlids 1: revisiting the evolution

of life histories in relation to parental care. Journal of

Evolutionary Biology 19: 66–75.

Kolm, N., N. B. Goodwin, S. Balshine & J. D. Reynolds, 2006b.

Life history evolution in cichlids 2: directional evolution of

the trade-off between egg number and egg size. Journal of

Evolutionary Biology 19: 76–84.

McHenry, M. J. & G. V. Lauder, 2006. Ontogeny of form and

function: locomotory morphology and drag in zebrafish

(Danio rerio). Journal of Morphology 267: 1099–1109.

Miles, D. B., 2004. The race goes to the swift: fitness conse-

quences of variation in sprint performance in juvenile liz-

ards. Evolutionary Ecology Research 6: 63–75.

Podani, J., 1999. Extending Gower’s general coefficient of

similarity for ordinal characters. Taxon 48: 331–340.

Schmitter-Soto, J. J., 2007. A systematic revision of the genus

Archocentrus (Perciformes: Cichlidae), with the descrip-

tion of two new genera and six new species. Zootaxa 1603:

1–76.

Siegel, S. & N. J. Castellan Jr, 1988. Nonparametric Statistics

for the Behavioral Sciences, 2nd ed. McGraw-Hill, New

York.

Song, J. & L. R. Parenti, 1995. Clearing and staining whole fish

specimens for simultaneous demonstration of bone, carti-

lage, and nerves. Copeia 1995: 114–118.

Watkins, T. B., 1996. Predator-mediated selection on burst

swimming performance in tadpoles of the Pacific tree frog,

Pseudacris regilla. Physiological Zoology 69: 154–167.

Wisenden, B. D., 1994a. Factors affecting reproductive success

of convict cichlids in Costa Rican streams. Canadian

Journal of Zoology 72: 2177–2185.

Wisenden, B. D., 1994b. Factors affecting male mate desertion

in the biparental cichlid fish (Cichlasoma nigrofasciatum)

in Costa Rica. Behavioral Ecology 5: 439–447.

Wisenden, B. D., 1995. Reproductive behaviour in free-ranging

convict cichlids. Environmental Biology of Fishes 43:

121–134.

Wisenden, B. D., 2001. Brood defense and optimal brood size in

convict cichlids Cichlasoma (Archocentrus) nigrofascia-

tum, a species with biparental care. Journal of Aquaricul-

ture & Aquatic Sciences 9: 303–320.

Wisenden, B. D. & M. H. A. Keenleyside, 1992. Intraspecific

brood adoption in convict cichlids: a mutual benefit.

Behavioral Ecology and Sociobiology 31: 263–269.

Wisenden, B. D. & M. H. A. Keenleyside, 1994. The dilution

effect and differential predation following brood adoption

in free-ranging convict cichlids (Cichlasoma nigrofascia-

tum). Ethology 96: 203–212.

Wisenden, B. D. & M. H. A. Keenleyside, 1995. Brood size and

the economy of brood defence: testing Lack’s hypothesis in

a biparental cichlid fish. Environmental Biology of Fishes

43: 145–151.

Wisenden, B. D., J. L. Snekser, A. D. Stumbo & J. M. Leese,

2008. Parental defense of an empty nest after catastrophic

brood loss. Animal Behaviour 76: 2059–2067.

272 Hydrobiologia (2015) 748:259–272

123